# Inflection > Inflection is a boutique venture capital fund investing in non-consensus engineering moonshots at the idea stage. Website: https://inflection.fund Contact: contact@inflection.fund Keywords: venture capital, engineering moonshots, pre-seed fund, seed fund, early stage investing, deep tech, future of computation, cryptography, physical AI, robotics, manufacturing, energy, resilience, aerospace ## About We build hard futures alongside the entrepreneurs we back: hard technologies that are difficult to engineer, often fusing bits and atoms; hard markets that are unfashionable; hard to underwrite companies where no playbook exists. We write first checks between $0.5M and $2M initially and deploy with low velocity and high conviction. Our customers are heretical founders with european roots doing their life's work, like us. Our platform rests on deep, trusted relationships that formed over decades. "Where did you go to school" or "who else is investing" aren't sentences you'll hear from us. We share our thoughts candidly to unlock growth. We support our founders with all we've got, sometimes by stepping out of their way. Limited Partners are our partners and co-owners. Our capital base is purely private, such that we can think and act independently. We value outcomes over management fees. We report with conservatism and are fond of radical transparency. We expect loyalty. ## Philosophy We are living through an era of hyper volatility that is leaving us with generational opportunities for creators and investors alike. Staying true to venture’s origins, we seek exposure to high convex, big-if true ventures. We perceive high write off quotas as a feature, not a flaw. We optimize for asymmetric, uncapped upside, not downside protection. Our investment approach is technology agnostic. We consider AI as force of concentration and equalization in pure information domains. Hence we are keen to explore problems rooted in the physical world where friction and chaos reign over the perfect illusion of the virtual world. We seek exposure to areas where bits meet atoms; where moats and network effects can be created around sensing, data and hardware. Some of our top of mind themes are the future of computation, resilience and aerospace, amongst others. Greatness cannot be planned. Engineering breakthroughs are the result of serendipitous exploration. The patterns of progress are shaped by outsiders with exceptional analogous reasoning capabilities or experts with highly unusual approaches to problem solving. We work with particular founder archetypes. Pathologically problem obsessed and resilient with scar tissue. No ego, no bullshit. Put on this planet to build what they’re building. We back them with all we’ve got, not just capital. Industry networks, tools, data, but most importantly: personal commitment. We show up when things get difficult. We don’t mind meeting on a Christmas to get a round done. What was once a niche industry run by misfits with ambitious ideas and artistic craft turned into a soulless startup factory for incrementalism. We are building a different kind of firm. ## Investment Focus - **Stages:** Idea, Pre-Seed, Seed - **Check Size:** $500K - $2M - **Geography:** Global with a preference for talent with European roots - **Sectors:** Future of Computation, Photonics, Semiconductors, Cryptography, Physical AI, Robotics, Manufacturing, Energy, Resilience, Aerospace ## Team - **Alexander Lange** — Founding GP — Grew up in a working class family, graduated top 5% in German Law (nation wide), parallel degree in Economics, KAS fellow. 3y as operator in growth, product and cyber at Google and early stage fintech startup Pepperbill (acquired). 5y VC with Earlybird and Index Ventures before starting Inflection 2019. Spent 3y raising Fund II of $40M through a pandemic as a Solo GP. - **Jonatan Luther-Bergquist** — GP — Engineering Physicist & Computational Science; MS from TUM, EPFL, Uppsala. 3y at startup building cryptography software for UN with 500k users in emerging economies. 3y in IT at BCG building large energy trading infra. Marathon in 2.40h. - **Robert S.** — Director — Serial entrepreneur in consumer electronics and software before founding a real estate investment and development firm which he led for over two decades. Co-founded Inflection in early 2019, focusing on fund administration, operations, and stakeholder management. Emory University graduate, based in Boston (or Maui). - **Rebecca M.** — VP Finance — Nearly three decades in real estate investment and development, leading due diligence, budgeting, and marketing. Co-founded Inflection in 2019 and handles investor relations, accounting, and controlling. University of the South graduate, based in Boston. - **Jon Levin** — Counsel — Nearly four decades practicing commercial law, real estate development, and venture capital transactions. Founding partner of boutique law firm Garrity, Levin and Muir LLP. Co-founded Inflection and handles compliance, governance, and legal matters. Based in Boston. ## Portfolio - **Stealth** (2026): Programmable fab for high precision optics - **Stealth** (2026): Agentic energy grid orchestration - **Hanseatic** (2025): Programmable merchant for energy assets: [Website](https://hanseatic.net/) - **Hedy** (2025): Presence security: [Website](https://hedycyber.com/) - **Levtek** (2025): Modular robots for industrial workers: [Website](https://www.levtek.io/) - **Ark** (2025): Autonomous fleet control: [Website](https://ark-robotics.com/) - **Nordic Air Defence** (2025): Irondome for the northern hemisphere: [Website](https://www.nordicairdefence.com/) - **Deep Earth** (2024): Subsurface mapping: [Website](https://www.deepearth.tech/) - **Lodestar** (2024): AI fighter pilots for the space domain: [Website](https://lodestar.space/) - **Ubitium** (2024): Universal edge processing: [Website](https://www.ubitium.com/) - **Fabric** (2023): Hardware acceleration for encrypted AI: [Website](https://fabriccryptography.com/) - **Radical** (2023): Stratospheric compute platform: [Website](https://radicalaero.com) - **Tune Insight** (2023): Privacy preserving ML: [Website](https://tuneinsight.com) - **Worldcoin** (2023): Financial network for verified humans: [Website](https://world.org/) - **Aptos** (2022): Edge compute for the space economy: [Website](https://www.aptosorbital.com/) - **Arx** (2022): Digital-to-physical twins: [Website](https://arx.org) - **Hologram** (2022): Gateway to the metaverse: [Website](https://hologram.xyz) - **Senken** (2022): Open carbon markets: [Website](https://senken.io) - **Defined** (2021): Open data supply: [Website](https://defined.fi) - **Flashbots** (2021): Private and secure transactions: [Website](https://flashbots.net) - **Moonpay** (2021): Crypto payment rails: [Website](https://www.moonpay.com/) - **Foundation** (2020): Hardware for the sovereign individual: [Website](https://foundationdevices.com/) - **Anytype** (2019): Local first AI: [Website](https://anytype.io/) - **Centrifuge** (2019): Real world assets on chain: [Website](https://centrifuge.io/) - **Molecule** (2019): A new era of drug development: [Website](https://molecule.to) ## Writings --- # Machine perception: sensing as AI's next frontier *Author: Alexander Lange | Section: Research* URL: https://inflection.fund/writings/machine-perception-sensing-ai-frontier Evolution spent 3.7 billion years perfecting perception and only a sliver of that on reasoning. Yet today's AI efforts are focused on reasoning and blind to most of what could be sensed. Machines perceive the world through a tiny slice of modalities (mostly cameras and microphones), ignoring chemistry, smell, electric and magnetic fields, radiant heat, polarisation and force that biology exploits routinely. We believe large-scale opportunities can be unlocked by building "sensory intelligence": co-designing novel sensors with AI so machines can perceive reality in fundamentally new ways and develop capabilities that haven't been possible before. This post explores (1) sensing in nature, (2) the state of machine sensing, (3) the opportunities ahead, and (4) a rough sketch of venture-compatible company shapes. ## Contents - [Sensing in nature and machines](#sensing-in-nature-and-machines) - [Sensing opportunities](#sensing-opportunities) - [Quantum sensing](#opportunity-quantum-sensing) - [Electronic nose](#opportunity-electronic-nose) - [Fine-force touch](#opportunity-fine-force-touch) - [Electroreception](#opportunity-electroreception) - [Company shapes](#company-shapes) - [Non-consensus beliefs](#non-consensus-beliefs) ## Sensing in nature and machines Nature's senses sort by the physical stimulus being measured. A bee's ultraviolet vision and a pit viper's heat pit are both photoreception, sampling different bands of one spectrum. Seven stimulus families cover almost everything alive, plus a plant-specific cluster. The oldest senses measure the most fundamental things — molecules, fields; image-forming vision and hearing are late, expensive refinements. Machine perception first evolved in the 1940s through [RADAR](https://en.wikipedia.org/wiki/Radar) (radio detection and ranging) and [SONAR](https://en.wikipedia.org/wiki/Sonar) (sound navigation and ranging); the [CCD (charge-coupled device)](https://en.wikipedia.org/wiki/Charge-coupled_device) image sensor in 1969; [LIDAR](https://en.wikipedia.org/wiki/Lidar) (light detection and ranging) and [thermal imaging](https://en.wikipedia.org/wiki/Thermographic_camera) in the 1980s; smartphone [sensor fusion](https://en.wikipedia.org/wiki/Sensor_fusion) (camera, microphone, GPS, inertial unit, magnetometer, barometer) in the 2000s; the [deep-learning vision breakthrough](https://en.wikipedia.org/wiki/AlexNet) in 2012; and [digital olfaction](https://en.wikipedia.org/wiki/Electronic_nose) and commercial [quantum sensors](https://en.wikipedia.org/wiki/Quantum_sensor) in the 2020s. Two senses — vision and hearing — plus ranging and navigation instruments dominate; smell, taste and touch barely register. The chart below sorts each capability by how far it reaches — what humans can feel, what only other life can, and what only machines can — with two five-point reads on the machine side: **maturity today** and **opportunity ahead** (untapped headroom). The gap between them is where the value is. ::embed{src=/embeds/sensing-nature-to-machines.html height=1360} **Two things jump out.** (1) Every human sense is a fainter copy of some animal's; the AI we train on human-legible data inherits that same narrow slice. (2) Quantum sensing and chemoreception (sensing molecules — the "e-nose") are the lowest-maturity, highest-opportunity frontier. Those are where we focus. ## Sensing opportunities This section breaks down the opportunity sets from the top down. First, an overview of the four most promising sensing technologies by their 5–10-year readiness and their novelty (the degree to which they enable new behaviours). Then we double-click into each of the four most promising capabilities — quantum sensing, the electronic nose, touch and electroreception — covering the leading technical approaches, the bottlenecks, and the application potential, including market-size estimates. ### Capabilities as opportunities The chart below reflects the opportunity sets with the highest asymmetry: low maturity today, large market opportunity in the future. The market sizes are approximations derived from various market reports, linked at the bottom. Importantly, these approximations are derived from integrated *capabilities* — they do not reflect the expected market of the underlying sensing technology alone. ::embed{src=/embeds/sensing-opportunities-map.html height=720} ### Opportunity: quantum sensing **What it is.** Quantum sensing uses single quantum systems — atoms, ions, electron spins, photons — as measurement instruments. These systems are so fragile that the smallest change in their environment shifts their quantum state in a precisely readable way. The fragility that makes quantum *computers* hard to build is what makes quantum *sensors* work: the disturbance is the measurable signal. The prize is sensitivity orders of magnitude beyond classical sensors, across magnetic and electric fields, gravity, acceleration, rotation, time and temperature. A first generation — atomic clocks and [SQUIDs](https://en.wikipedia.org/wiki/SQUID) (superconducting quantum interference devices, magnetic-field sensors that require cryogenic cooling) — existed for decades. The next generation works at room temperature and can be shrunk to chip size. Four method families are the most promising approaches today. No single method dominates; each use case picks its own quantum system, depending on context. ::embed{src=/embeds/quantum-sensing-approaches.html height=1000} **Where the opportunity sits.** The sensor hardware itself is a small market — McKinsey estimates roughly $1–6B by 2040. Value accrues in the capabilities each sensor unlocks (navigation ~$7.4B by 2034, brain–computer interfaces ~$14B by 2035). See the section on company shapes below. *Sources — science & approach:* [Degen, Reinhard & Cappellaro, *Quantum sensing*, Rev. Mod. Phys. 2017](https://arxiv.org/abs/1611.02427) · [*High-sensitivity nanoscale NV quantum sensors*, Communications Materials 2025](https://www.nature.com/articles/s43246-025-00770-x) · [*A portable OPM-MEG platform*, Imaging Neuroscience 2024](https://direct.mit.edu/imag/article/doi/10.1162/imag_a_00283/124093) · [*Quantum-enhanced navigation with atom interferometry*, arXiv 2025](https://arxiv.org/abs/2504.11119) · [*Optical clock resolving gravity across a millimetre*, Nature 2022](https://arxiv.org/abs/2109.12238) *Sources — market sizes (one representative forecast each; firms differ):* [quantum-sensing market — McKinsey 2024](https://www.mckinsey.com/capabilities/tech-and-ai/our-insights/tech-forward/quantum-sensing-poised-to-realize-immense-potential-in-many-sectors) · [quantum navigation & sensing outlook — McKinsey 2021](https://www.mckinsey.com/industries/industrials/our-insights/shaping-the-long-race-in-quantum-communication-and-quantum-sensing) ### Opportunity: electronic nose **What it is.** An electronic nose (e-nose) is machine olfaction — the artificial sense of smell. It copies the architecture of the biological nose through an array of cross-reactive sensors, each responding a little differently to many molecules, with AI reading the *pattern* across the whole array. This is combinatorial coding — the same trick the nose uses, where ~400 receptor types encode millions of smells. The target is [VOCs](https://en.wikipedia.org/wiki/Volatile_organic_compound) (volatile organic compounds — carbon-based molecules light enough to evaporate into the air we breathe), the chemical signatures that leak from a food, a mate, a disease or a buried explosive. Conventional lab instruments like [GC-MS](https://en.wikipedia.org/wiki/Gas_chromatography%E2%80%93mass_spectrometry) (gas chromatography–mass spectrometry, the current standard) name every constituent molecule, slowly and expensively. An e-nose reads the overall scent *character* in real time and learns to tell foreground from background — closer to how a dog works than a mass spectrometer. Two barriers long held it back: the *limit of detection* (reacting to single molecules) and the *limit of recognition* (decoding a smell in a noisy, shifting plume). Both are now falling — though there is still a long way to go (see bottlenecks). ::embed{src=/embeds/electronic-nose-approaches.html height=1000} **Bottlenecks.** The gating problem is data: olfaction has no [ImageNet](https://en.wikipedia.org/wiki/ImageNet) (the structured visual database that trains and benchmarks AI models) equivalent yet. Vision has PNG, audio has WAV, language has tokens — olfaction has no digital standard that captures a smell. Representations are fragmented, which limits benchmarking. And chemical sensors age, each type drifting differently, so a model trained today degrades tomorrow unless it keeps recalibrating. The challenges are enormous and likely to take the better half of a decade to overcome. **Where the opportunity sits.** The e-nose sensor market is real but mid-sized — about $30B in 2025, projected to ~$77B by 2032 at roughly 14% CAGR (forecasts vary widely by firm). As with quantum sensing, the leverage is one layer up, in the served markets each nose competes in — not e-nose revenue itself: food and pharma quality control (food-safety testing ~$56B by 2035), CBRN (chemical, biological, radiological, nuclear) threat detection (~$30B), and non-invasive cancer screening from breath and urine (multi-cancer early detection ~$6.8B). The moat is not the sensor, which commoditises, but the labelled **scent library** — the proprietary dataset of clinical and field samples that trains the classifier and compounds with every deployment. *Further reading — science & approach maturity:* [Mershin et al., *Machine Olfaction and Embedded AI*, arXiv 2025](https://arxiv.org/abs/2510.19660) · [*Advanced electronic noses for future robotic olfaction*, npj Robotics 2025](https://www.nature.com/articles/s44182-025-00071-y) · [France & Daescu, *AI and Olfaction: A Survey*, ChemRxiv 2025](https://chemrxiv.org/doi/full/10.26434/chemrxiv-2025-6xw1s) *Sources — market sizes (one representative forecast each; firms differ):* [e-nose market — Mershin et al.](https://arxiv.org/abs/2510.19660) · [multi-cancer early detection — Nova One Advisor](https://www.novaoneadvisor.com/report/multi-cancer-early-detection-market) · [CBRNE defence — GM Insights](https://www.gminsights.com/industry-analysis/cbrne-defense-market) · [food-safety testing — Expert Market Research](https://www.expertmarketresearch.com/reports/food-safety-testing-market) ### Opportunity: fine-force touch **What it is.** Fine-force touch is machine tactile sensing — an electronic skin that measures the mechanical contact between a robot and the world: force (how hard it presses), shear and slip (whether an object is sliding out of the grip), local shape, texture and hardness. Vision is too limited here, which is why touch is the critical bottleneck to dexterous manipulation: at the moment of a grasp the hand occludes the very thing it is holding, and force, slip and compliance are not optical quantities. Vision tells a robot *where* something is; touch tells it *how hard, whether it is slipping, and what it is made of*. Touch lagged vision for the same reason smell did: no ImageNet-scale training data, and no skin that was cheap, durable and large-area at once. Two shifts are breaking that. First, **vision-based tactile sensors** (optical tactile sensing) turn touch into an *image* an ordinary neural network can read, giving sub-millimetre contact geometry from a standard camera. Second, coverage is going whole-hand: the F-TAC hand (*Nature Machine Intelligence*, 2025) embeds high-resolution touch across ~70% of the hand's surface at 0.1 mm resolution and beats non-tactile baselines across 600 real-world grasping trials. ::embed{src=/embeds/fine-force-touch-approaches.html height=1000} **Where the opportunity sits.** The tactile-sensor market is mid-sized — roughly $14.5B (2025) rising to ~$47.5B by 2035. As with olfaction, the leverage is one layer up, in the platforms touch unlocks: dexterous manipulation for humanoid robots (~$38B by 2035), surgical and medical haptics (surgical-robotics ~$46B by 2035), and consumer and XR haptics (~$7.1B). The moat is the **contact dataset and grasping policy** — the proprietary flywheel of real manipulation episodes that trains the model. Whoever collects the most contact data wins, which is exactly the shape of startups now teaching dexterous hands from sensor-glove human demonstrations. *Sources — science & approach:* [*Tactile Robotics: An Outlook*, arXiv 2025](https://arxiv.org/html/2508.11261v1) · [*Embedding high-resolution touch across robotic hands* (F-TAC Hand), Nature Machine Intelligence 2025](https://www.nature.com/articles/s42256-025-01053-3) · [*Biomimetic multimodal tactile sensing*, Nature Sensors 2025](https://www.nature.com/articles/s44460-025-00006-y) · [Visuotactile field guide](https://visuotactile.com/) · [tactile-sensor comparison — SVRC](https://www.roboticscenter.ai/learn/tactile-sensor-comparison) *Sources — market sizes (one representative forecast each; firms differ):* [humanoid-robot market — Goldman Sachs](https://www.goldmansachs.com/insights/articles/the-global-market-for-robots-could-reach-38-billion-by-2035) · [tactile-sensor market — SNS Insider](https://www.snsinsider.com/reports/tactile-sensor-market-7592) · [surgical-robotics market — Precedence Research](https://www.precedenceresearch.com/surgical-robotics-market) · [haptics technology — IDTechEx](https://www.idtechex.com/en/research-article/haptics-technology-market-to-grow-to-us-7-1b-by-2035/31731) ### Opportunity: electroreception **What it is.** Electroreception is sensing electric and bioelectric fields — the faint voltages that every nerve, muscle and heartbeat produces, and the fields around any live wire or charged object. It is a sense humans lack entirely, but sharks, rays and the platypus rely on it: a shark's [ampullae of Lorenzini](https://en.wikipedia.org/wiki/Ampullae_of_Lorenzini) detect the bioelectric field of hidden prey down to a few billionths of a volt. For machines it opens two otherwise-closed domains: the body's own electrical activity — the firing of nerves, the heart and the muscles, normally reached only with gels, needles or implants — and the electrical state of infrastructure — the voltage and faults inside a live grid, read without touching it. Sensors sort by what field they read and how close they must get. The set below is focused, not exhaustive. ::embed{src=/embeds/electroreception-approaches.html height=1000} **Where the opportunity sits.** Electroreception is under-invested relative to olfaction and touch, and asymmetric — the body's electrical signals are continuous, information-dense and today mostly unread outside a clinic. The machine analogues tap large end markets: bioelectronic medicine and electroceuticals (~$48B by 2035), non-contact and wearable biopotential monitoring (EEG/EMG equipment ~$7.9B by 2035), and industrial electric-field sensing (~$4.8B by 2032). The moat is the same shape as in the other modalities — the labelled physiological dataset and the model that reads it, not the electrode, which commoditises. *Sources — science & approach:* [*Bioelectronic Medicine and Neural Interfaces*, J. Bio-X Research 2025](https://spj.science.org/doi/10.34133/jbioxresearch.0064) · [*Next-generation bioelectronic medicine: non-invasive closed-loop neuromodulation*, Bioelectronic Medicine 2024](https://link.springer.com/article/10.1186/s42234-024-00163-4) · [Chi et al., *Dry-Contact and Noncontact Biopotential Electrodes: A Review*, IEEE Rev. Biomed. Eng.](http://www.isn.ucsd.edu/pubs/rbme10.pdf) · [*Bioinspired soft electroreceptors for artificial precontact somatosensation*, Sci. Advances 2022](https://www.ncbi.nlm.nih.gov/pmc/articles/PMC9140963/) *Sources — market sizes (one representative forecast each; firms differ):* [bioelectronic-medicine market — Metatech Insights](https://www.metatechinsights.com/industry-insights/bioelectronic-medicine-market-3260) · [EEG & EMG equipment — Market Research Future](https://www.marketresearchfuture.com/reports/eeg-emg-equipment-market-39204) · [electric-field sensor — Verified Market Research](https://www.verifiedmarketresearch.com/product/electric-field-sensor-market/) ## Company shapes **Leverage principles.** We are eager to back companies that use many or all of the principles below to build durable businesses. 1. **Prevention beats cure; continuous beats episodic.** Most problems are cheap to fix early and catastrophic to fix late, so a sensor that catches a problem while it is still small is more valuable the steeper the cost-over-time curve (stage-1 cancer vs. stage 4; the first cracks in a bridge; an annual check-up vs. a continuous breath monitor). 2. **Resolving scarcity bottlenecks.** Some perceptions are locked inside a rare, expensive, trained human or animal. Sensing plus AI turns that capability into something cheap, everywhere and always-on (trained dogs and bomb-disposal experts are rare). 3. **Making the invisible priceable to create markets.** You cannot trade, insure or regulate what you cannot measure. When a sensor makes a hidden quantity continuously measurable, it can be priced (methane leaks became a tradable, finable emission once satellites and ground sensors measured them). 4. **Reaching the inaccessible.** Some places are physically or economically closed to existing sensors: inside a living body, behind a wall, under rubble, underground, in jammed airspace. A modality that reaches them opens a domain that was simply shut. 5. **Fusion compounds.** Two sensors that fail in different ways are worth more than the sum of their parts, because each covers the other's blind spot (vision + lidar in self-driving). **The durable sensing company is not a sensor company.** In every historical case the raw transducer commoditises; value accrues to whoever owns the layers above it. ### On data loops The most valuable companies of the smartphone era ran on a sensor that shipped in every phone for free (GPS, the microphone). None of them made the sensor. Each owned the **proprietary data loop** on top of it. Note what they are, though: they mostly sell you an *outcome* — a ride, a song ID, a match — not the engine itself. The flywheel is the **moat that defends a vertical**; it is not the same thing as being a platform. The sensing prize is to run this engine *and* extend to selling the capability. ### On business models A sensing company is a stack. The bottom layer — the sensor — is the one layer that commoditises. The value and the defensibility sit in the three layers above it, and in the **flywheel** that connects them: every deployment produces data no one else has, which trains a better model, which wins more deployments. Own more than the sensor. If the above is the "engine", consider the "chassis" — the business models taking shape around it. Sensing companies fall into three archetypes: 1. **Component manufacturers** stay durable only behind scale and process moats (Sony, Illumina). For most startups this path leads to the Velodyne trap (merged with Ouster, ~$3B market cap): hardware-first sensing gets commoditised to death — not venture-compatible. 2. **Vertically integrated businesses** "sell the outcome" by bundling sensor + model + data — in self-driving (Tesla, Waymo), diagnosis (Cala Health, ~$272M raised) or brain–computer interfaces (Neuralink, ~$9B valuation). TAM is bounded by the vertical. Venture-compatible. 3. **Platform businesses** "sell the data engine" by co-designing sensors and models, winning a lighthouse market, then selling the capability broadly across domains. Osmo (ca. $130M raised) intends to build the "ImageNet for smell" and sell it across domains; Tacta Systems (ca. $75M raised) pursues the same for full-body robotic touch. Venture-compatible. ## Non-consensus beliefs **AI is bottlenecked by perception, not reasoning — and the field is spending on the wrong constraint.** The consensus unlock is bigger models and more compute. We believe the next jump in AI capability comes from *widening the senses* — the way nature built capability out across every animal and plant. Evolution spent 3.7 billion years on perception and a sliver on cognition; today's AI inverts that ratio and is starting to hit the ceiling of it. The scarce input is not more of the same data, it is *new modalities* of data — chemistry, fields, force, neural current — that no model has ever seen. **The sensor itself is usually a red herring.** The deep-tech reflex is to pay a premium for defensible transducer IP — the novel e-skin, the exotic chemical sensor. We think that, for nearly every modality, the transducer commoditises the moment it works. The durable company points a *commodity* readout — a camera, a phone, a cheap electrode — at a quantity nobody was capturing, and owns the data loop on top. A GelSight is a camera doing touch; Shazam was a microphone; Uber was the GPS already in every pocket. Backing novel-transducer IP is backing the one layer that evaporates. Quantum is the exception that proves the rule: there the physics of capture is genuinely hard and interpretation is trivial, so the transducer stays the moat. **If you're building a vertically integrated or platform company on next-generation sensing, let's talk.** --- # On fundraising *Published: 2026-09-07 | Author: Alexander Lange | Section: Markets* URL: https://inflection.fund/writings/on-fundraising For an untrained mind fundraising is torture. As a founder you're exposing your life's work to the judgement of a stranger you're depending on. Naturally, you are experiencing a power asymmetry to your disadvantage. For the VC its one out of 12x founder meetings that week and one position out of many. For you its your life's work tied to your identity. You might get a straight "no", some hard truths or you might get ghosted without much empathy. It can be a traumatizing experience. But it is also a necessity and an opportunity for you to grow into your best self. In this post I'm sharing some tools, interventions and anecdotes that helped me stay sane for more than a decade in venture (on both sides of the fundraising table). The ideas laid out here shall serve as mere inspiration to develop your own tools and routines; it will require hard work and discipline beyond reading a short blog post. ## growth mindset Fundraising is something every founder has to go through in some form or another. Some seem to have a natural talent or a very high pain tolerance for it, others don't - yet! You can and will improve if you try. Fundraising is a skill that can be learned, not a condition you're born with. Further, that skill might be acquired more easily and faster than say becoming an engineer. Benchmarking your progress not against those who are better at it but against yourself is critical, especially over your first few raises. You could record pitch calls with granola and ask your favorite AI models for critical feedback on how you did and what you could improve. Writing down some quick notes on how you *felt* during and after the pitch can help you figure out your trajectory. ## negativity bias What makes fundraising hard emotionally and mentally is that your brain prioritizes bad news over good. The no to yes ratio will be 10:1 or worse. However, research shows that it takes five positive events to psychologically outweigh one negative event. Your brain is working in the wrong direction for good reason though. For our ancestors, failing to notice a threat (like a predator) meant death, while missing out on an opportunity (like a piece of fruit) just meant being hungry for a bit longer. The brains that overemphasized danger survived to pass on their genes. Brain imaging shows that the brain's electrical activity spikes much higher in response to negative stimuli than positive stimuli. We literally process bad events with more cognitive bandwidth. Awareness is helpful but we need interventions whenever we deal with emotions. (1) Batching negative inflow compresses the no : yes ratio. Label investor email threads and designate one time window a day to read them. Your brain will count 3x fast nos as one. (2) You can separate "data" from "danger" by documenting the nos in a spreadsheet with a single useful piece of feedback. (3) Cognitive offsetting: when you get a "no", consciously force yourself to list 5x things that went right that day (e.g. a great engineering push; a new customer; a happy customer quote; a strong team meeting etc.). That way you can artificially balance your brain's scales. (4) Reframing a "no" as something positive. Getting a no from a stranger can free up equity and a board seat for the "right investor" who "gets you" and who cares deeply about your vision and success. (5) Gamifying rejection is powerful and fun. Shift the reward mechanism of your brain. Instead of tracking the "yes" matches, set a goal for collecting 10x fast rejections. When a "no" comes it, check a box and reward yourself for achieving it. Finally, (6) a biological intervention: box breathing (inhale 4sec, hold 4sec, exhale 4sec, hold 4sec - repeat). This signals to your nervous system that there is no physical predator in the room and that you're safe. ## community Thousands of founders are going through the same experience as you are, right now. Many more have been through it over the last few centuries and many more will go through in the future. **You are not alone.** Some of them are very open to share their experiences. Invite them over a coffee, lunch or walk to exchange experience. Start by sharing your own experiences, thoughts and feelings. The more you share, the more you will get back from the other side - good old reciprocity at play. No surprise we germans have a saying for that: "*Geteiltes Leid ist halbes Leid*" - freely translated into "a sorrow shared is a sorrow halved". Do this a few times to get more perspectives. It will make you feel in good company. You're not alone. ## fear release Fear is an emotion that draws its overwhelming power from the unknown; from the blurry idea of what might be if we don't succeed. Intuitively, when faced with fear we look the other way and force our mind to think of something else. If we do this for too long the emotion of fear rises within us until a point where it starts to paralyze us. I suggest the opposite. Take a piece of paper or a journal and write down your fears as concretely as possible. What is a concrete situation you're afraid of? What clothes do you wear? How does the room / environment look like? Who is present? What are their facial expressions? What is their body language? What exactly do they say to you? Which words of judgement are you afraid of? "*You are not good enough. We won't back you and nobody else will either. Your company will fail. You will fail." Then observe your body. How does that *feel*? Write down your feelings as concrete body sensations, e.g. a burning heat in your chest; a pressure building up from within; dizziness in your head; emptiness or the "hole in your stomach"; tears in your eyes. What does this sensation remind you of? Have you been there before? How do you know this sensation? What is your earliest memory of it? Writing all of this down will help you manage and release your fear. *Why* does it work? Writing requires structure and language which is processed in a different part of your brain than fear, hence your brain activity shifts emotion to logic. Your brain treats an unwritten fear like an open browser tab, constantly burning background energy. Externalising fear to a piece of paper creates "cognitive distance". Further reading: [Fear-setting: A brief writing intervention increases motivation to reach personal goals and positive affect.](https://link.springer.com/article/10.1007/s10902-024-00767-2) ## equanimity The more attached you are to the success of your fundraise (aka the judgement of others), the bigger will be the pressure you put on yourself. There is a fine line between a strong motivation and a depression derived from that pressure. *Equanimity* describes a calm, balanced, neutral state of mind at any circumstance. Developing it can occupy you for a life time. Some of the tools that can support you on that journey are meditation, breathing techniques or psycho-active substances. A state of mind is nothing static. It shifts from time to time but you can develop interventions and tools to steer into a certain direction. This is obviously a huge topic that cannot be addressed properly in a short blog post. If you're curious about the topic I recommend reading [The Art of Living: Vipassana Meditation as Taught by S.N. Goenka](https://www.goodreads.com/book/show/18934657-the-art-of-living) or [Meditations by Marcus Aurelius. ](https://www.goodreads.com/book/show/61435653-meditations-by-marcus-aurelius-and-gregory-hays?ref=nav_sb_ss_4_27) ## manifestation Once you released the dark version of your future it is time to build up the positive one. Manifestation is the practice of turning a goal or desire into reality by aligning your thoughts, mindset and actions with it. When you focus intensely on a specific goal, you program your Reticular Activating System (RAS - basically a brain filter) to spot opportunities, resources and connections related to that goal. Believing that a positive outcome is possible naturally shifts your mood and confidence. This shift makes you more likely to take positive risks, speak up, and put yourself in a position to succeed. There are some easy interventions you can leverage. (1) Prime and sleep. Your brain is highly suggestible right before you fall asleep. Spend 2 minutes before closing your eyes on visualizing a specific, successful moment in the near future - like shaking hands on a deal or seeing a green checkbox on a dashboard. Focus on the *feeling* of relief, excitement and joy. (2) Journaling. Write down 1x specific opportunity you'd like to go for on a given day and visualize the outcome. At the end of your entry write down a positive affirmation like "I learn from every rejection and get sharper at every pitch." Further reading: [The neuroscience of manifestation: what the brain really does.](https://rowancenterla.com/the-neuroscience-of-manifestation-what-the-brain-really-does/) ## frame control Being the (early stage) founder pitching venture investors who manage millions or billions of dollars makes you put yourself into a position of relative weakness. This perceived power asymmetry is expressed in social protocols and our environments as a "frame". You take the elevator to a fancy office where you're greeted by a receptionist. They might let you wait for a couple of minutes. There is fancy art at the walls. You are a guest in foreign territory etc. There are ways to "break the frame" and invert the the power asymmetry to your advantage. You can reframe your situation: *Capital is not scarce, there are 100s if not 1,000s of VCs who could finance my company. My genuine talent, character and network are much more scarce than capital.* *If you don't fund me, somebody else will. You will regret not funding us. There are endless ways to "break the frame" socially. E.g. you can cut the meeting short by a few minutes saying that you have another meeting close by. This signals time scarcity on your end and demand from others. You can control the conversation by leaving some intentional, longer pauses before answering questions. You get the idea. If you're curious about frame control I recommend reading the book [Pitch Anything](https://www.goodreads.com/en/book/show/10321016-pitch-anything) . It's somewhat controversial and sometimes painfully awkward to read but I found it entertaining and quite informative about the hidden social rules governing our interactions. ## process engineering To get into the mind of your future investor it is helpful to design a process around your fundraise. You put together some documents (pitch deck, data room). You prepare your narrative and the points you'd like to get across - a lot has been written about that, so I won't repeat it. The hard part is to align investors on a reasonable, shared time line to optimize your optionality. Here are some things that worked for our portfolio companies in the past: - set an internal deadline for when you'd like to close and communicate it to investors clear enough to create urgency (next couple of weeks, we kicked off the process); once you hit concrete milestones you can increase the pressure (we have a term sheets but didn't counter sign. We stick to a disciplined process to find the best possible partner.) - build a funnel of critical mass. fundraising is a numbers game - you only need one yes out of 100 (or more) nos. Brainstorm with your business angels and early investors who might be the best possible partners for you - which firm, which partner and why. Cluster your funnel by priority. Start reaching out to your inner circle, to people where you feel comfortable with a bad pitch, ask for feedback, iterate. Go for prio 1 once you feel ready. - display velocity by sharing bits of information that indicate progress at every touchpoint you have with an investor. E.g. yesterday we signed 2x new customers for $X revenue; this week we will hear back from gov program X for a contract / grant (even if you already have certainty that it will come through); we hired elite talent Y for our critical open role of A etc. Make the investors *feel* your velocity. - amplify and back channel through your network. Your existing investors, angels and close business partners might have touch points with some of your respective VCs. Ask them to pro-actively to sing your praises in the right context and with the right timing. Close the loop by asking for any resonance your network might have gotten. - stay honest and real. Some VCs are pattern matching founder archetypes like "the traumatized misfit" or the "20yo renegade who dropped out of an ivy league engineering school". Don't try to fit in, be your authentic self. If the investor doesn't vibe with you she is not the right fit. Present facts in a positive light but don't lie. Investors are allergic to founders making things up or lying about issued term sheets. If you make bold claims, expect the investor to ask for proof. Being called out will erode trust and kill the relationship. - learn systematically. Actively request feedback from every investor you talk to. Summarise learnings, missed opportunities, misunderstandings in an "objection handling doc". Go through it before every important meeting. ## fun Most of us are much better at what they do if they have some fun on the way. Make sure that fundraising meetings are somewhat entertaining for yourself and the investor (who happens to be a human, after all). Personally, some of the my most productive fundraising meetings were "walk in the park" and "having ice cream" meetings. A friend of mine can't take people seriously who eat a banana, so he brings one to important meetings and offers it to his meeting partner, genius. If you think through your fundraise and are struggling to get into the right mindset, reach out. Feel free to share what worked for you in the comments. --- # Hard Futures *Published: 2026-06-03 | Author: Alexander Lange | Section: News* URL: https://inflection.fund/writings/hard-futures We are renaming our SVRGN bog to Hard Futures. The content is re-orgnised to optimise for exploration and search. **Capital flows downhill.** It pools where the TAM is clear and the next round is already circled. In 2026, 90% of all venture dollars are deployed against AI and 50% of LP $ are raised by the largest 10 venture firms. When consensus grows, capital concentrates and prices inflate. Returns compress. The crowd has never been larger or more expensive to stand in. This dynamic opens outsized opportunities for those operating outside of it. We work uphill. We invest into hard futures. We look for exposure to companies that contribute to bringing a hard future to reality: **Hard technology**. We back businesses that fuse bits and atoms: compute in the stratosphere, silicon that reprograms itself, defense for assets in orbit, software led optical factories, space laser ranging. Durable moats form where code has to move matter. That is where economies of scale meet data flywheels and where durable businesses compound. **Hard markets**. We go where consensus capital won't — unfashionable geographies (like Europe), contested domains, the earliest stages, sectors traditionally resistent to innovation. Difficult terrain thins the crowd. A thin crowd leaves low prices and asymmetric, high convex opportunity. Low entry valuations are earned by being there before anyone else was looking. **Hard to underwrite.** When a founder describes bodyguard satellites, self-reprogramming silicon or an interactive map of the subsurface, there is no comparable or playbook. We read that difficulty as the signal. The harder a thing is to underwrite, the fewer people are bidding, and the further the price drifts from the future it implies. That gap is the alpha. Then there is the second half of the name. **Future as distribution - the spread of outcomes.** Consensus capital prices the expected case and crowds the center of the distribution curve. We position for the tails - the low probability, high magnitude futures the crowd assigns zero odds and therefore gives away (almost) for free. That asymmetry is convexity that compounds across our portfolios. **Future as a time horizon.** We back capabilities that do not yet exist at scale and that are at the cusp of transitioning from R&D to commercialization. The diffusion gap between military and commercial deployment has compressed from centuries to merely a decade, thereby fitting inside a single fund's life of 10-15 years. We underwrite hard problems and route productive capital towards the futures that the market has mis-priced. The difficulty and time horizons keep the crowd out and prices low. --- # Observations from Earth *Published: 2026-03-18 | Author: Jonatan Luther-Bergquist | Section: Research* URL: https://inflection.fund/writings/observations-from-earth ## Backstory I grew up swimming in the Seas of Kattegat and Skagerrak. Snorkeling before I could swim on my own, exploring the green murkiness of what felt like an infinite, secret world. I felt at home among the cliffs, the cold salt water spray, and the uninhabited islands as the only barrier to the real Ocean. The real Ocean featuring 30m waves, unknown isolation, creatures and distances unheard of. This world belonged to me and I to it. It was my own Narnia, just to dream of this. What made it magical was how restricted it was. To enter this world you needed to prepare, to dare and then to be comfortable about not knowing and not being in your own element. Your perspective in the water shifts from eye-level to floor level, assuming there are no waves, but in return you are flying on top of a pillar of glassy substance that carries you wherever you go. It’s flexible enough that if you chose to you can split through it and let negative buoyancy work. Once you pass the equilibrium between compression and trapped air, you are pulled down and don’t need to fight anymore to gain speed. Aerodynamics are now hydrodynamics and the way you hold yourself has immediate feedback on your speed. When you dive, you don’t see the same way, colors are distorted and light reflects in weird places, if there is any at all. Looking up you see a mirror of the world you normally belong to. It’s bright and warm, yet the pressure under water, and constant touch makes air seem so lightweight and fleeting. Just getting out to the place where there might be something interesting to see underwater meant either swimming or navigating on a boat, which in the Swedish Archipelago is quite treacherous (pre widely available GPS). Water is, as we know, conductive. Not just to electricity, but also to waves in the electromagnetic spectrum. Or more precisely, water molecules get easily excited by incoming photons. This makes a little water transparent and, slightly more, very opaque. In fact, so opaque they use it to great success to contain radiation in nuclear reactors, and also why you can’t easily detect submarines. Of course, as a 6 year old, detecting submarines wasn’t my main priority. I just wanted to learn the secret passage to the sea. So I decided to become a marine biologist and started learning everything I could about dolphins. This didn’t exactly pan out, as I later went into a much drier (pun) field of engineering physics. Fast forward about 20 years and I was snorkeling in one of the most frequent shark-attack waters on earth to spot wild dolphins. But that’s another story. We moved away from the sea, I started diving in pools and flooded silver mines (probably in hindsight not super healthy), until I moved back to the sea on my own as a 16 y.o.. I felt entirely at home in the water, so much so that my new family thought they’d lost me the first day at the water, during a precursor for the autumn storms that were to thrash the North Atlantic coast in Bretagne. I was just exploring the moving sand and rolling waves after a decade of calm waters. Finally coming up for air and putting my hand to my head, just to say “tout va bien”. ### 60,000 Nautical Miles under the sea In the years away from the sea, my favorite book growing up was “20,000 Leagues under the sea” by Jules Verne, which tells the story of a secretive submarine captain taking a professor captive and exploring the worlds oceanic environments. Unbeknownst to me, I was moving not far from Jules Vernes native Nantes about 180 years after he was born on a tiny island in the Loire. He excelled at _mémoire_ (remembering things by heart and reciting them, which was a core skill of that time), latin, and geography. Captain Nemo and Professor Aronnax explore creatures and treasures hidden in the sea unheard of until then. Likely based on myths and seafarer stories from the docks. It was written in 1870, before submarines were really a thing, and when diving equipment was limited to diving bells and surface air supplied through tubes to massive metal helmets with glass windows. The vessel that they travel in, the Nautilus, was an electric sub, cigar shaped and about 70m long x 8m in diameter. Verne intentionally or accidentally actually didn’t violate too many physical laws or engineering principles when designing it. It’s surprisingly feasible. It would have a displacement of about 1,500 m3 sea water meaning it would weigh around 1.5e6 kg. Nautilus traverses the Mediterranean (~2,000 km in Verne’s geography) in about forty‑eight hours, implying an average speed on the order of 20–25 knots depending on exact endpoints. It’s fully electric and probably has to produce around 5MW in shaft power at that speed. It's said to run on some sodium-mercury batteries, with days and days of range without charging. These batteries or engines didn’t exist at the time of Verne’s writing, one might add. But to produce such power, you’d likely need O(200m3) worth of batteries, and about 200t worth of batteries. Which could plausibly fit inside a 70x8m cylinder and could be carried weight-wise! You’d need about 300-600Wh/kg on a system level, and about 10x what that type of batteries could do at the time. Now the fun part is that a nuclear submarine has virtually unlimited range and about 3-5x the shaft power of Nautilus. We’ve outdone Verne’s imagination. The sea monsters are human made, once again. ## Maritime surveillance economics If you skipped the preamble about 19th century sci-fi and dolphins, welcome back! You're alright. Recent conversations I’ve had, as well as naval warfare events I’m planning (as you do), as well as recent events in the Strait of Hormuz, plus pending events in the GIUK gap, Baltic Sea infractions, a pending South China Sea conflict, etc. had me think: **What is the most cost-effective way to detect what’s going on everywhere on the oceans, and maybe below?** ### Why this matters Maritime monitoring and surveillance was around ~$24-26B in 2024, but it only takes fairly little disturbance at sea to significantly disrupt entire industries. A few pirates with drones. A couple of jerky misjudgments on demand/supply of household electronics during a pandemic. A single nuclear sub. A few “fishing” vessels dragging their anchors a few hundred times over fiber optic cables. That last one isn’t hypothetical. [There’s currently something like 1.2 million km of submarine telecom cables](https://submarine-cable-map-2026.telegeography.com/) in the water, and [another $13B worth planned for 2025-2027](https://www.lightreading.com/cable-technology/2024-in-review-submarine-cables-become-a-battleground). This is driven by the hyperscalers and Meta mostly, with insane communications demand. The bandwidth is simply unbeatable, and we’re getting better and better at it. There are people working on building underwater power generators to avoid having to install land-based repeaters, and it looks like a large market. (If you are working on this or something related, come talk to us!) Beyond cables, historically the sea has been a fairly high leverage way to conquer new areas. This funnily enough also has a physics explanation, as transporting heavy, bulky things on water requires much less energy than on roads or rails. In fact [more than 80%](https://unctad.org/news/shipping-data-unctad-releases-new-seaborne-trade-statistics) of the worlds goods use maritime transport today! With developing economies and Asia growing massively in that domain, while the rest are declining. ### Who’s building The US just backed a maritime-only focused VC called “[Mare Liberum](https://www.mareliberum.com/)” after the 1609 treatise by Hugo Grotius arguing for free access to the South China Seas. Back then over the Portuguese, not presumably over the Chinese themselves. On the defense side, there’s a growing wave of naval startups - Saronic (~$700M raised), Kraken Technology Group (UK/NATO contracts), UForce (whose Magura drones have been instrumental in taking out the Russian navy in the Black Sea), plus HavocAI, Vatn Systems, Toloka, Nautrik, Polar Mist, Ray Systems, Regent Craft, and others. Helsing and Anduril have their own products too. Lots of capital deployed, unclear who’s made money yet. The proof point that these matter: [at a recent NATO exercise, REPMUS, Ukrainian USVs were pitted against allied naval vessels](https://www.faz.net/aktuell/politik/ukraine/nato-manoever-vor-portugal-ukraine-versenkt-alliierte-fregatte-accg-200633625.html) armed to the teeth. NATO forces didn’t even have time to notice the opponent was there before they were hit. A real confrontation would not have ended well for NATO. On the civilian/dual-use side, companies like Hawkeye360, Seasats, and Saildrone started with commercial missions but have been pulled toward defense. Newer ones are going after seabed mapping, hull cleaning, or offshore platform maintenance. Saildrone set up [its EU headquarters in Denmark](https://www.saildrone.com/news/saildrone-expands-to-europe-copenhagen-denmark-headquarters) after the Danes invested ~$25M. The dual-use trajectories are converging. ### Our approach So assuming there is demand for persistent monitoring of maritime environments, we need to optimize for one variable: **cost per km2-hr of monitoring**. In our findings, these vary with over 3 orders of magnitude across platforms. Mainly due to operating constraints and clear superior reach of certain technologies. Manned frigates are just insanely expensive to build and to man and operate. It feels like absolute insanity that these are still being built to the extent they are, and still people complain about our ship building capacity. At the same time, we need something to carry all our drones and USVs and fancy unmanned systems as they won’t have the range needed to fight somewhere in the middle of the pacific. We built an interactive page to test our own assumptions: [Maritime Economics](https://maritime-economics.inflection.fund). Check it out, and if you’re extra nerdy, go to the [physics section](https://maritime-economics.inflection.fund/physics.html). ![](https://cdn.sanity.io/images/9ycxs2qn/production/e254c4bd308c1aa70df39b894653d2c13bba906a-2460x1456.png?w=450) ## The big physical problem Is that the seas are huge. Like 361 million square km or so big _on the surface_. Naval forces probably cover <1% of that at any given time. Which gives boats and sneaky dolphins way too much space for mischief and play. Enter AIS (automatic identification system), which is an identification system for ships. Ironically it’s not so automatic as it should be, as it can easily be turned off. And even if something is detected, the sea is so huge that doing something about it, intercepting or so, becomes an even greater challenge. So the sea is vast, isn’t nice to electromagnetic fields, and generally deteriorates most mechanical systems. In summary, this is why it’s hard. There’s an information theoretic “principle” which becomes apparent when you try to design a system that looks everywhere at once at high resolution. You can’t look everywhere at once, with high resolution. The explanation is on the physics page of the [Maritime Economics dashboard](https://maritime-economics.inflection.fund/physics.html#infocapacity) we built! ![](https://cdn.sanity.io/images/9ycxs2qn/production/2df0b5950183b8ebb063e3200bdcfc4114edff6f-2262x1296.png?w=450) ### What about radar? Radar at sea becomes an even bigger pain than normal radar, because all the water starts making the radar waves bounce, leading to low signal-to-clutter (SCR) ratios. It’s “clutter-dominated” (insert joke about my tidiness) meaning the amount of background noise is determining if you find something rather than receiver sensitivity, on a whole. Unless it’s a very calm day at sea, it becomes impossible after some distance to detect anything with a low false positive rate. Finding a periscope at sea is like finding a sparrow in a thunderstorm. The radar range equation governs all active microwave sensor performance. The range scales as R^4, which means that detecting something 2x further away requires 16x the power. So just making bigger antennas quickly becomes impractical. The maximum detection range R\_max for a monostatic radar is: $$R_{max} = \left[ \frac{P_s \cdot G^2 \cdot \lambda^2 \cdot \sigma}{(4\pi)^3 \cdot P_{E_{min}} \cdot L_{total}} \right]^{1/4}$$ Where: - **Ps** = Transmitted peak power (W) - typical maritime surveillance radar: 25 kW to 1 MW - **G** = Antenna gain (dimensionless, typically expressed in dBi) - **λ** = Wavelength (m), inversely related to frequency: λ = c/f - **σ** = Radar cross section of target (m²) - **P\_Emin** = Minimum detectable received power (W), set by receiver noise floor - **L\_total** = Total system losses (atmospheric attenuation + hardware losses) The R^4 dependence is the defining constraint: doubling detection range requires a 16x increase in transmitted power or antenna aperture. This creates a fundamental size/weight/power tradeoff for airborne and space-based platforms. Switching from Gaussian to K-distribution statistics - which occurs at high resolution or heavy sea states - degrades detection performance by up to **12 dB** when using standard pulse-to-pulse integration. This means systems designed for calm-sea performance can catastrophically underperform in heavy weather. There are interesting ways around some these limitations, and ways that engineers try to limit the radar cross section (RCS) of their vessels. My favorite type of radar is probably so-called over-the-horizon radar (**OTHR**), which uses the earths ionosphere to bounce radar waves off of to extend the range and avoid some sea surface scattering. This means ionospheric variability - diurnal, seasonal, solar cycle, and storm-induced - creates **significant temporal uncertainty** in system performance. Frequency management systems must continuously adjust operating frequency (typically sweeping 4-30 MHz) based on real-time ionospheric sounding. It also means OTHR have blind spots due to the multi-hop nature. There’s a non weather-dependent type where we instead bounce polarized RF off of the sea surface, called **Surface Wave Radar** (HFSWR), which is useful for persistent coastal detection. On a whole, though, these radars are massive and enormously expensive. For reference, a single OTHR cost OOM hundreds of $M, in capex and annual opex in the tens of millions. And their coverage is good but resolution is not great, e.g., for small vessels. Canada recently [bought one $4B OTHR system](https://dsm.forecastinternational.com/2025/03/20/canada-buying-4-billion-arctic-over-the-horizon-radar-from-australia/) from Australia to cover the Arctic… ### What about optical sensors? About 67% of the sea is covered in clouds at any given point in time. So pure optical is hard/inefficient. This is why **SAR** (Synthetic Aperture Radar) has become the real workhorse for maritime surveillance from space - it works through clouds, at night, and produces images good enough to detect ships. Companies like ICEYE, Capella, and Umbra are deploying constellations of small SAR satellites at ~$3M per unit. At 0.25-1m resolution you can identify vessel type, and wide-area scan modes cover 100x100 km per scene. Sentinel-1 does 400 km swaths at 40m resolution. The tradeoff between resolution and swath width is physics, not engineering - same information-theoretic constraint from above. For the visible/IR spectrum, things get more interesting but also more limited. Water vapor absorbs certain IR wavelengths aggressively, which makes it difficult. Interestingly, MWIR (3-5 um) actually outperforms LWIR (8-12 um) in humid maritime conditions, because of precisely this. And it works best for things that have a thermal signature, like ship engines. Using the [Johnson criteria](https://en.wikipedia.org/wiki/Johnson%E2%80%99s_criteria) and NVTherm modeling, we can figure out **practical detection ranges for maritime vessel targets** using MWIR (640x512, 330 mm lens): - Large vessel (heat signature, 10 m target extent): Detection to 15-30 km - Small vessel (5 m target extent): Detection to 8-15 km - Human in water: Detection to 2-4 km - **Range degrades sharply in fog, rain, or high humidity** - often to <1 km in bad conditions Thermal is interesting, but of course there are countermeasures. Submarines running hot engines cool down their exhausts before ejecting them so no hot bubbles rise to the surface. From space, optical constellations like Planet (~200 satellites, 3-5m resolution, up to 12 passes/day) and Maxar WorldView Legion (30 cm resolution, 15 revisits/day) can image ships in clear weather. But “clear weather” is the catch - two thirds of the time you’re looking at clouds. This is why SAR, not optical, is what actually finds dark ships today. The intelligence companies doing AIS anomaly detection (Windward, Global Fishing Watch) correlate SAR detections with missing AIS signals to flag vessels that have gone dark. It’s clever, but it’s only as persistent as the satellite revisit rate allows. We’re still limited by the information theoretic principle of swath vs. resolution tradeoff from above. If we’re closer to the earth, we can have higher resolution but can’t get good coverage. If we’re using satellites, high resolution forces narrow swaths. No amount of engineering budget changes this - it’s the space-bandwidth product doing what the space-bandwidth product does. ### Using water as a lens Importantly, since radar doesn’t work under water, we need its slightly cooler cousin who only listens to vinyl (analog only baby), **sonar**. It’s effectively like radar, but instead of RF emissions, you emit sound (vibrations). Since speed of sound in a medium depends on the temperature and pressure of the medium, and those two things differ in water depending on depth, there’s a phenomenon called SOFAR, or SOund Fixing And Ranging. It’s essentially using a depth range where the sound bounces cylindrically instead of spherically, leading to a ~30dB advantage. It’s usually 600m-1000m but depends on the environment. There it’s easier to detect subs or things under water, but also much easier to communicate between vessels. ![Adapted from Munk et al., 1995](https://substack-post-media.s3.amazonaws.com/public/images/108fec50-6f01-4948-b70f-fb8e9909b9fc_400x192.gif) Still, sound degrades very quickly under water, which means detection at distance becomes a real problem, especially when someone tries to avoid detection by going electric or generally masking their sound signature. So if you’ve played around in the sonar detection section of the Maritime Economics page, you’ve probably found it pretty hard to detect a sub. ![](https://cdn.sanity.io/images/9ycxs2qn/production/dcacf928a019f3970ef0cf126ef2bb49c176ff68-2346x863.png?w=450) Still, in the SOFAR, sound can travel thousands of km underwater, whereas radar is horizon limited, at least. Unless it’s OTHR of course. Active sonar transmits a pulse and listens for the echo. The sonar equation for active detection: $$SL - 2 \cdot TL + TS - (NL - DI) \geq DT$$ The critical difference: **2 ⋅ TL** - the signal must travel to the target AND back, doubling the transmission loss in dB. This means active sonar detection range is always shorter than passive for the same source level. Active sonar also faces reverberation - acoustic energy scattered by the ocean surface, bottom, and volume. In shallow water, reverberation often exceeds ambient noise, making active sonar **reverberation-limited** rather than noise-limited. ### The fancy methods If we think specifically about detecting modern submarines, there are a bunch of fun ways people have thought about to detect them. Creativity is the limit here, really. A submarine is something very very heavy, moving quite a lot of water, making some sounds, and probably disturbing whatever is around it. How can we detect it? Modern diesel-electric and AIP submarines have reduced source levels by 30-40 dB compared to Cold War boats: ![](https://cdn.sanity.io/images/9ycxs2qn/production/65c43f3c8134eb7d9e88676a42463dc18c1a8dea-1440x480.png?w=450) A 30 dB reduction in source level translates to a 1,000x reduction in acoustic power. Detection range scales roughly as $$10^{\Delta SL/20}$$ for cylindrical spreading - so a 30 dB quieter submarine is detectable at roughly 1/30th the range… Not good for detection. Subs are heavy! So that means we could sense an increase in gravitational force near it, right? This is called **Gravity Gradiometry.** A sub weighing 2,000-8,000 tonnes _does_ create a gravitational anomaly, but unfortunately that anomaly is about 1,000x smaller than the sensitivity of modern gravimeters. So either subs need to get a lot bigger, or we need to make quantum gravimeters a reality. Subs are metallic chunks. That should be magnetic, right? A submarine’s steel hull _does_ create a magnetic dipole anomaly in Earth’s field. The dipole field falls as 1/R³: $$B_{anomaly} \sim \frac{\mu_0 \cdot m}{4\pi R^3}$$ For a submarine with magnetic moment ~10^6 A\*m^2 and a magnetometer sensitivity of ~0.01 nT we get: - Detection range: ~500m-1km laterally - Requires direct overflight at low altitude (<300m) - Used by P-8A, P-3C, and ASW helicopters for final localization I’ve heard that there are places dedicated solely to the demagnetization of submarines though, so this might not work as well. ![Degaussing — USS Jimmy Carter submarine undergoing its first deperming treatment at Naval Base Kitsap's magnetic silencing facility.](https://substack-post-media.s3.amazonaws.com/public/images/126aec5f-6cdd-4e48-9720-8f7f88bb66df_1960x3008.jpeg) This could be avoided if submarines were built outside of a magnetic field completely, like in space, [but that has different difficulties attached to it.](https://www.reddit.com/r/videos/comments/1b7beqn/xkcd_would_a_submarine_work_as_a_spaceship/) Even so, when the metal moves through Earth’s magnetic field, it accumulates a temporary magnetic component that changes with direction. Modern submarines also have built-in degaussers, which are essentially coils built into the hull to counteract magnetic fields. Submarines should make waves underwater, right? A submarine moving at depth creates detectable surface signatures: 1. **Kelvin wake**: V-shaped surface wave pattern (half-angle 19.5 degrees) from pressure wave at hull 2. **Internal waves**: Displacement of density layers, potentially visible in SAR as modulated surface roughness 3. **Bernoulli hump**: Local sea surface height increase above the submarine (~mm scale) SAR satellites can theoretically detect internal waves from submarines at periscope depth (30-50m). Operational capability is claimed by several programs but unconfirmed in open literature. The physics is marginal - the signal is comparable to natural internal wave amplitudes. When it moves it should stir up organic matter from depths, right? In waters with high dinoflagellate concentrations, a submarine’s passage mechanically stimulates bioluminescence. This creates a glowing wake visible to sensitive airborne EO sensors at night. Detection ranges: 1-5 km from directly above. Limited to specific ocean regions and seasons. Not operationally reliable. Same applies to thermal mixing from different temperatures at different depths. Not a big enough temp diff to measure. All of these non-acoustic methods share the same problem: they’re either too short-range to search with (MAD), too insensitive to work (gravity), or too dependent on perfect conditions (wake, bioluminescence). Acoustics will remain the primary way to find subs for at least another decade. None of these methods alone solves the problem. So instead of asking which sensor is best, the right question is: what does each one cost per unit of ocean monitored? ## So what does it cost? All of these physics constraints translate into wildly different price tags for the same output: knowing what’s going on in a patch of ocean. We normalized everything to **cost per km2-hour of persistent coverage** and the range is absurd. More than three orders of magnitude: ![](https://cdn.sanity.io/images/9ycxs2qn/production/7f5f73fe97999aabb77eb529ad7950e6869b7de6-1428x940.png?w=450) Reading this table top-to-bottom reveals that it isn’t a market failure we don’t have persistent surveillance - it’s all physics. Cheap persistent coverage tells you very little (OTHR: “something is somewhere”). Expensive close-range platforms tell you everything (boarding a vessel) but cover almost nothing. The design question isn’t “which platform” but “how to stack them.” The optimal architecture is likely nested: wide-area cueing from satellites and OTHR triggers HALE UAVs, or HAPS or SAR tasking, which cues USVs or MPA to investigate, which cues a surface vessel to intercept, which could cue UUVs. Each layer is using physics it’s good at and handing off to the next. The alternative - patrolling a million km2 with frigates - costs roughly **$5B/year** in operating costs for the US surface fleet alone, and still leaves >99% of the ocean unobserved. ### The small boat market that isn’t small Austin Gray, a former Navy officer and now co-founder of Blue Water Autonomy, wrote [a piece](https://austinegray.substack.com/) that’s making some very good points. His thesis: the small USV market is tiny (~$120M/year) and already saturated. He counts 20 chokepoints, estimates ~1,190 USVs needed (330 sentries + 860 speedboats), subtracts the ~600 already under contract, and concludes there’s only ~$117M/year left if you refresh every 5 years. He’s right about the **sea denial** market. If you’re building speedboats to blow up (a la Ukraine), MAPC already has the Replicator production contract at 32/month, Saronic has hundreds of millions in bookings, and there isn’t room for a fifth entrant. That market is spoken for. Assuming those companies are able to fulfil the requirements of working products... But he misses the **monitoring** market because he counts USVs like inventory rather than like patrols. A renewably-powered sentry USV (Saildrone, SeaSats, OceanAero) cruises at 2-3 knots. It covers about 50-70 km/day. To persistently monitor the GIUK Gap (~300 km wide, 200 km deep), you need about 48 USVs on station **simultaneously**. But they rotate - transit to station, time on station, transit back, maintenance. At 67% utilization, you need ~72 in the fleet just for one gap. Gray’s model puts maybe 10-15 there. Scale this across the six NATO priority theaters (GIUK, Baltic, Mediterranean, Black Sea, North Sea, Arctic) and you get ~336 sentry USVs for NATO alone. Add the Indo-Pacific, Gulf, and allied waters and you’re at 650-700 globally. That’s 2x Gray’s estimate, and the refresh rate isn’t every 5-10 years like hardware - it’s continuous. Sensors need upgrading every 3-5 years, hulls take a beating, and you lose 5-10% annually to weather and the occasional unfriendly interaction. Effective annual fleet replacement: ~22%. So the monitoring hardware market is maybe $77M/year globally. Still not huge. But that’s the wrong frame. ### Don’t sell boats. Sell awareness. The right comparison isn’t “what does a boat cost” but “what does it cost to know what’s happening in this piece of ocean.” Today, NATO pays for this knowledge with frigates. A Constellation-class frigate costs [$63-130M/year](https://www.cbo.gov/publication/56675) to operate. Its radar covers maybe 1,250 km2 around it. For persistent coverage of the Baltic cable corridors (~10,000 km2), you’d need 8 frigates on continuous rotation. That’s $500M-1B/year. To monitor cables. 50 sentry USVs covering the same area as a service? $25-40M/year. One-twentieth the cost, genuinely persistent (24/7/365), and you don’t need to find 2,000 sailors to man them. The pricing implication is clear: **don’t price against other USVs. Price against the frigate.** If your monitoring service costs $15,000-25,000/day for 5,000 km2 of persistent coverage, you’re charging 10-15% of what a frigate costs for coverage a frigate literally cannot provide. The customer saves money AND gets better coverage. Those are the kinds of trades that close. This reframes the addressable market from Gray’s $117M/year hardware business to a **$200-350M/year recurring service business** anchored against NATO maritime patrol budgets. Different game entirely. ## Does NATO have the budget? Yes. And it’s getting bigger fast. NATO defense spending hit 2.76% of GDP in 2025 - first time all 32 allies met the old 2% floor. At the [June 2025 Hague Summit](https://www.nato.int/en/what-we-do/introduction-to-nato/defence-expenditures-and-natos-5-commitment), they set a new target: **5% of GDP by 2035**. That’s not a typo. Five percent. For context, the US is around 3.4% and Europe was at 1.4% as recently as 2014. European allies + Canada collectively spent [$482B in 2024](https://www.nato.int/content/dam/nato/webready/documents/finance/def-exp-2025-en.pdf). EU equipment investment grew 42% in a single year to EUR 106B. Germany went from a geopolitical punchline to spending $107B on defense. Norway is dropping $55B through 2036 on 5 new frigates, 5 submarines, and 28 other vessels. The UK’s Royal Navy equipment budget jumped 41%. The money is there. The question is where it goes. Most of it is still going to traditional platforms. Germany’s F126 frigate program is EUR 10B for 8 ships. France just greenlit a EUR 10.25B aircraft carrier. These are necessary but they’re not solving the monitoring problem. You can’t have a EUR 1.25B frigate babysitting a fiber optic cable 24/7. The encouraging signal is that NATO is already building autonomous monitoring at small scale. [Task Force X Baltic](https://www.act.nato.int/article/tfxb-future-nato-maritime-vigilance/) launched in late 2024 with 8 nations contributing USVs. By mid-2025 they had 50-60 deployed with a target of ~100. [Baltic Sentry](https://news.usni.org/2025/01/15/nato-launches-baltic-sentry-mission-in-baltic-sea) launched in January 2025 specifically for undersea cable protection. The US stood up its third USV squadron. [NATO DIANA](https://www.diana.nato.int/) selected 150 innovators with maritime autonomy as an explicit challenge area. The spending wave from 2.76% to 5% GDP creates roughly **$100-150B/year** in new allied defense spending over the next decade. If even 1% of that increment flows to autonomous maritime monitoring, that’s $1-1.5B/year in new market by 2030. ### The awkward part about selling to the US Navy There’s a catch, and Gray & Gray document it painfully well in “[Startups & Sea Power](https://austinegray.substack.com/)“: the US Navy is structurally hostile to startups. Only 4% of top defense startup revenue comes from the Navy (vs. 31% Army, 37% Air Force). The Navy has spent $7B on flexible OTA contracts vs. the Army’s $53B. 71% of NAVAIR contracts are sole-sourced. Only 8% of NAVSEA contracts are even theoretically accessible to a company without an existing relationship and a security clearance. Navy FY25 budget for USVs? $172M. For shipbuilding? $36B. That’s a 200:1 ratio of old to new. So if you’re building a maritime monitoring company, the US Navy is probably your worst first customer. Not because they don’t need it - they clearly do - but because the institution is optimized to buy expensive, exquisite, manned platforms from incumbents who’ve been doing this for 50 years. European NATO allies are the better wedge. They’re scaling budgets 40-70% from a lower base, they have acute monitoring gaps (ask anyone in the Baltic states about cable security), and their procurement systems - while not perfect - don’t have the same institutional antibodies to new entrants. Denmark invested $25M in Saildrone’s European HQ. Norway is planning 28 new vessels. The demand signal is there. If you get in front of it, you might not need the US Navy at all for the first few years. ## What it means The ocean is vast, hostile to sensors, and increasingly contested. The physics is hard but well understood. The economics are clear: manned platforms are 10-100x more expensive per unit of awareness than unmanned alternatives. The budgets are growing. The technology works - Ukraine proved that small USVs change the calculus at sea, and Saildrone proved you can keep a drone on station for months. The opportunity isn’t in building another speedboat. It’s in building the company that makes “what’s happening in this patch of ocean” a commodity service priced at a fraction of what a frigate costs. The winner will own the operational complexity - keeping 50-300 USVs at sea, processing the data, delivering actionable intelligence - and price it as surveillance-as-a-service against naval patrol budgets. We’re keeping a close eye on this space. If you’re building something here, or thinking about it, and have a product or technology, you should [reach out](https://inflection.fund). We’re also still building the interactive tools - the [Maritime Economics dashboard](https://maritime-economics.inflection.fund) and [physics page](https://maritime-economics.inflection.fund/physics.html) are live and probably have bugs. Have a poke around and let us know what you think. ![Whiskey on the rocks — By Marinmuseum, digitaltmuseum.se, CC BY 4.0](https://cdn.sanity.io/images/9ycxs2qn/production/48943908e44caa3d12de9ba67fab49eac68dd6b6-3840x2444.png?w=450) --- # AI Enhanced Second Brain *Published: 2026-03-03 | Author: Alexander Lange | Section: Building* URL: https://inflection.fund/writings/ai-enhanced-second-brain The second brain is not a new idea. Da Vinci kept notebooks. Luhmann built a 90,000-card Zettelkasten. Tiago Forte turned it into a method: capture, organize, distill, express. The shared insight was that your future self is the customer of your current thinking. And without a system, most of that thinking evaporates. What changed with the introduction of MCP servers and AI tools is that the system can now "think back". Note: If you're new to the most powerful way to organize your knowledge base, [check out the book Building a Second Brain](https://www.buildingasecondbrain.com/book). You can find a [quick summary video here](https://www.youtube.com/watch?v=g6GbJpVppqo). ![](https://substack-post-media.s3.amazonaws.com/public/images/0025b41e-1893-4556-b48e-f1c4d6b1df0e_625x625.jpeg) ## The Problem With Every Note-Taking System Humans think in associations, not hierarchies. You connect a semiconductor supply chain memo to a defense strategy note to a book passage about national sovereignty - because they share structural logic. Traditional tools force this associative thinking into trees. The graph gets flattened into a filing cabinet. The second brain movement solved the capture problem. Millions of people now have vaults, notebooks, and databases full of highlights, clippings, and notes. **But retrieval stayed manual**. **You had to remember what you wrote, where you put it, and why it mattered.** The more you captured, the harder it got to find anything. Past a few hundred notes, most vaults become graveyards - well-organized, rarely revisited. ![](https://substack-post-media.s3.amazonaws.com/public/images/f347fe70-9836-4b52-9d32-0df2799896dc_2752x1526.png) Three things had to converge to change this: - **Graph-native knowledge tools** (Obsidian, Anytype, Roam) that treat links as first-class citizens, not afterthoughts - **Embedding models** that find connections you never explicitly made, based on meaning rather than keywords - **Agentic AI** that can traverse, query, and synthesize across the entire graph in a single reasoning chain The knowledge tool stores structure. The embedding model finds hidden similarity. The agent reasons over both. Together, they turn a passive archive into an active thinking partner. ## What This Looks Like in Practice One implementation of the above is [Cornelius, an open source library you can check out here](https://github.com/Abilityai/cornelius). A Claude Code agent configured to operate my Obsidian vault (or Anytype or Roam vault) as a second brain. Think of it as a layer cake: me at the top, Claude Code as the general-purpose AI, and Cornelius as a specialized layer underneath with purpose-built tools for knowledge work. It manages notes, extracts insights from what I read, discovers connections between ideas, and synthesizes across months of accumulated thinking. ![](https://substack-post-media.s3.amazonaws.com/public/images/a3ce3c09-76eb-4691-8cf4-b9e544bd0cc2_1166x950.png) The system currently runs across 298 notes with 2,354 edges in the knowledge graph - roughly half explicit wiki-links I created manually, half semantic connections the AI discovered on its own. This is a real time glimpse into my second brain: ![](https://substack-post-media.s3.amazonaws.com/public/images/2b16401e-dc0b-4f28-bd7b-664df8295afa_1408x1324.png) ![](https://substack-post-media.s3.amazonaws.com/public/images/87a8a6d2-6764-4b8f-8b9f-a79ba302f344_2866x1600.png) This stack enables a few things that haven't been possible before. Here is one out of hundreds of examples from the last few months: **1/ Cross-domain connection discovery.** Three portfolio companies - one doing encrypted computation, one building cryptographic silicon, one working on undetectable networking - were written up in separate memos years apart. The system found they form a single architecture. Fully homomorphic encryption is too slow for real workloads, which creates the need for hardware acceleration, but encrypted data still travels on observable networks where metadata alone reveals everything, which creates the need for dark networking. Each company's binding constraint is the next company's thesis. The AI traced a constraint cascade across several unrelated investment memos and surfaced an emergent "zero trust compute stack". Obviously we were aware of those connections when we took the investment decisions but still I was very surprised to see a machine "reason" through seemingly unrelated notes where such connections weren't made explicit at all. **2/ Synthesized recall across months of work.** One question - "_what do we know about post quantum cryptography_" - pulled together a technical notes on the state of quantum computing, several deal calls with companies working on quantum hardware and software, and book passages about national security infrastructure. Finding those building blocks of knowledge and connect them through years of time and siloed information scattered across different emails, folders, gdrive and CRM tools would have been impossible before. The system reconstructed my accumulated position on the topic in seconds. **3/ Portfolio-level pattern recognition.** Once the system identified the zero trust stack, it kept going. It found that the constraint cascade pattern - where one company's limitation creates the next company's opportunity - repeated across other portfolio clusters. It surfaced a bridge between the encrypted compute thesis and some of my public market semiconductor positions. That emergent portfolio coherence was invisible until the tool traversed the full graph. **4/ Unsolicited expert network mapping.** While researching the encrypted computation space, the system returned the obvious domain experts in my network. Then it added people I didn't ask about. A professor whose papers were already in my research library, and an engineering lead my partner had met at a conference months earlier who had worked on exactly the problem space I was diving into. The AI mapped my network against my research question and found relevant nodes I had forgotten existed. ## The Tool Unification Problem Knowledge does not live in one app. My research is in Obsidian, Anytype and Gdrive. My relationships and dealflow are in a CRM. My venture portfolio positions sit in Carta. My public market positions in a tracker. The breakthrough here is MCP - Model Context Protocol - which lets the AI agent query across all of them in a single reasoning chain. Any tool with an MCP server becomes part of the second brain. A query can run across our CRM, board decks, founder conversation notes and the knowledge graph simultaneously. Instead of having ten tools I have one intelligence layer that reasons across ten tools. Context switching and knowledge synthesis is done in one pass. ## Why the Compounding Effect Matters A traditional note-taking system has linear returns. Note number 300 is about as useful as note number 30. You still have to find it, read it, and connect it yourself. An AI-enhanced system has compounding returns. Every note you add increases the surface area for future connections. Note 300 can be cross-referenced against all 299 that came before it - semantically, and across domains you would never think to check. **This changes the economics of note-taking.** The quality of notes still matters enormously (garbage in, garbage out) but the cross referencing of other notes became exponentially better. Write more, organize less. Capture your thinking in your own words, link where connections are obvious, and let the AI find the rest. In the old world you needed to "build a perfect system." Now you only need to "feed a good-enough system consistently." ## What This Means for Knowledge Workers What I'm trying to describe is not just a little productivity hack. It is a structural shift in how individuals can memorize reason at scale. Before long we will see the same patterns emerge on an organizational level too. A venture investor can maintain deep, synthesized positions across dozens of companies and sectors without losing threads. A researcher can surface cross-disciplinary connections that would take months of literature review. A writer can draw on years of accumulated thinking without re-reading everything they have ever written. A founder can maintain institutional memory as the company scales past what fits in one person's head. The key design choices, for anyone building their own version: - **Pick a graph-native tool.** Obsidian, Logseq, Anytype (portfolio), Roam - the specific tool matters less than its ability to represent links between ideas as first-class objects. - **Write atomic notes in your own words.** One idea per note, in your voice. Copy-pasted highlights are raw material, not thinking. The AI needs your reasoning patterns, not someone else's. - **Link aggressively.** Every explicit link you create strengthens the graph the AI traverses. Over-linking is almost impossible. - **Point an AI agent at the result.** Use MCP servers, vector search, or agentic workflows to turn the vault from a reference library into a reasoning partner. The second brain was always a powerful idea I found deeply fascinating. The missing piece was the ability to think with it rather than just store static knowledge in it. That piece is here now, and the gap between people who build these systems and people who do not will compound faster than most realize. _Be aware that many of the technologies mentioned in this post are highly experimental. We set up our system with the support of seasoned engineers and best practices for security and data protection in mind._ How are you building and maintaining your second brain? --- # When LEO Fails: The case for stratospheric infrastructure *Published: 2026-02-19 | Author: Alexander Lange | Section: Research* URL: https://inflection.fund/writings/when-leo-fails-stratospheric-infrastructure Low Earth Orbit is the most consequential shared commons humanity has ever built on, yet we are treating it like a landfill. GPS, broadband, weather forecasting, military ISR, financial timing, precision agriculture all share one thing: they all depend on a thin band of space between 300 and 2,000 kilometers above Earth. If that band degrades, they all go down together. The question is not whether LEO will face a crisis but what form it will take - and whether we have a fallback. This post explores the problem space and potential mitigators. ![](https://substack-post-media.s3.amazonaws.com/public/images/5dc4737c-887c-4a77-a1a2-3f2891c4e868_2278x1094.png) ## The Fragility of Orbit LEO faces several risks, some more detrimental and more probable than others. Let's explore them one by one. ### Kessler Syndrome In 1978, NASA scientist Donald Kessler described a feedback loop: collisions in orbit create debris, which creates more collisions, which creates more debris, until an orbital band becomes unusable. For decades this was treated as a distant concern. It is now an operational reality. **The numbers are past the threshold**. A March 2025 analysis by Lewis and Kessler using the latest population data found that the current number of intact objects [exceeds the runaway threshold](http://\(https://conference.sdo.esoc.esa.int/proceedings/sdc9/paper/305/SDC9-paper305.pdf\)) at nearly all altitudes between 520 km and 1,000 km. Planned mega-constellation deployments will push even more altitude bands beyond stability limits. **The operational evidence.** SpaceX's Starlink constellation - 9,300+ satellites as of late 2025 - performed [300,000 collision avoidance maneuvers in 2025](https://www.space.com/spacex-starlink-50000-collision-avoidance-maneuvers-space-safety) alone. That is roughly 820 maneuvers per day. And each new satellite added to orbit increases collision probability with every existing satellite. **The CRASH Clock tells the story**. Researchers developed a metric called the [Conjunction Risk Assessment for Space Highways](https://www.livescience.com/space/space-exploration/orbiting-satellites-could-start-crashing-into-one-another-in-less-than-3-days-theoretical-new-crash-clock-reveals) - which tells us how quickly satellites would start colliding if they lost the ability to avoid each other. In 2018, it stood at 164 days. By January 2026, it had collapsed to 3.8 days. The safety margin eroded by 97% in eight years. [IEEE Spectrum's analysis](https://spectrum.ieee.org/kessler-syndrome-crash-clock) notes an important nuance: the CRASH Clock measures time to the **first** collision, **not a full Kessler cascade**. But a first collision is exactly how a cascade starts - and 3.8 days without active avoidance is the margin that stands between operational orbit and the beginning of a chain reaction. ![](https://substack-post-media.s3.amazonaws.com/public/images/aa1b1184-8a03-43d9-8675-e4eba4b505e1_1516x890.png) [ESA's 2025 Space Environment Report](https://www.esa.int/Space_Safety/Space_Debris/ESA_Space_Environment_Report_2025) tracks over 40,000 objects, with models estimating 1.2 million objects larger than 1 cm and 130 million larger than 1 mm. The [2009 collision between Iridium 33 and Cosmos 2251](https://en.wikipedia.org/wiki/2009_satellite_collision) - a single event - created over 1,800 trackable debris fragments, many of which will remain in orbit for decades. The collision doubled the risk in the 700-800 km altitude band. Now imagine this with 15,000-18,000 satellites in LEO by end of 2026, and China filing for a 200,000-satellite constellation. ![](https://substack-post-media.s3.amazonaws.com/public/images/fe30e541-6210-4470-93e1-d225f086d3a3_1512x1212.png) ### Space War Kessler Syndrome is an accident. Space war is a choice - and the arsenals are ready. **Kinetic anti-satellite weapons (ASAT):** Russia's November 2021 test [destroyed its own Cosmos 1408 satellite](https://www.spacecom.mil/Newsroom/News/Article-Display/Article/2842957/russian-direct-ascent-anti-satellite-missile-test-creates-significant-long-last/), generating 1,500+ trackable debris fragments at 480 km altitude - directly threatening the ISS and Starlink's orbital band. China demonstrated kinetic ASAT capability in 2007 by destroying a weather satellite, creating debris still tracked today. The US, India, and likely others have operational capability. **Directed energy weapons:** China has likely fielded ground-based lasers capable of dazzling or damaging low-orbit satellite sensors. Russia has deployed the [Peresvet system](https://sgp.fas.org/crs/natsec/IF11882.pdf) with mobile ICBM units. The U.S. Space Force is [deploying electronic satellite jammers](https://www.washingtontimes.com/news/2026/feb/11/us-racing-build-space-weapons-counter-anti-satellite-power-china/) and accelerating counterspace weapons development. Every major debris removal laser has inherent dual-use capability - the ablation mechanism works identically on debris and functional satellites. **Cyber warfare:** The 2022 [Viasat/KA-SAT attack](https://www.viasat.com/perspectives/corporate/2022/ka-sat-network-cyber-attack-overview/) at the start of Russia's invasion of Ukraine disabled satellite broadband across Europe in hours - a single cyberattack that affected tens of thousands of terminals. The [CSIS 2025 Space Threat Assessment](https://www.csis.org/analysis/space-threat-assessment-2025) reports that cyberattacks, jamming, and spoofing have "become commonplace and rarely trigger an escalatory or retaliatory response." GPS jamming is now routine across the Baltic, Middle East, and South Asia. **Nuclear EMP:** Russia is reportedly developing an [orbital nuclear anti-satellite capability](https://www.swfound.org/publications-and-reports/faq-what-we-know-about-russias-alleged-nuclear-anti-satellite-weapon). A nuclear detonation in LEO would not just destroy nearby satellites - it would create an electromagnetic pulse and a persistent radiation belt (similar to the 1962 Starfish Prime test) that could degrade electronics on every satellite passing through the affected region for months or years. The Secure World Foundation's [2025 Global Counterspace Capabilities report](https://www.swfound.org/publications-and-reports/2025-global-counterspace-capabilities-report) grew from 148 pages in 2018 to 316 pages in 2025, tracking 12 nations with active counterspace programs. The domain is militarizing faster than governance can respond. **The cascading scenario:** A kinetic ASAT strike during a Taiwan Strait conflict could deliberately trigger a Kessler cascade in specific orbital bands. Even a "limited" exchange - targeting 10-20 reconnaissance satellites - would generate thousands of debris fragments, each capable of destroying other satellites. The debris does not distinguish between military and civilian infrastructure. GPS, Starlink, weather satellites, and the ISS all share the same orbital neighborhood. ### Solar Weather The Sun is indifferent to human infrastructure planning. In February 2022, a moderate geomagnetic storm during the rising phase of Solar Cycle 25 [increased atmospheric drag in LEO](https://link.springer.com/article/10.1186/s40623-024-02124-2), causing 38-40 newly launched Starlink satellites to lose altitude and burn up before reaching their operational orbit. We are currently near the maximum of Solar Cycle 25. Research published in [Frontiers in Astronomy and Space Sciences (2025)](https://www.frontiersin.org/journals/astronomy-and-space-sciences/articles/10.3389/fspas.2025.1572313/full) shows that reentry rates for LEO satellites have increased significantly during this solar maximum, with atmospheric drag becoming harder to predict as solar activity fluctuates. **The Carrington scenario:** A Carrington-class event (the 1859 solar superstorm, the strongest recorded) directed at Earth would be categorically different from a moderate storm. Simulations suggest it could disable or degrade a significant fraction of LEO satellites through radiation damage to electronics, dramatically increased atmospheric drag, and disruption of ground control links. [New simulations](https://www.livescience.com/space/the-sun/the-next-carrington-level-solar-superstorm-could-wipe-out-all-our-satellites-new-simulations-reveal) warn that a Carrington-level event could "wipe out all our satellites." The probability is not negligible. Solar physicists estimate a 1-12% probability per decade of a Carrington-class event. Over a 50-year infrastructure planning horizon, the cumulative probability becomes significant. ### Economic and Regulatory Failure LEO's commons problem is economics. **The bankruptcy scenario:** Starlink reached profitability in 2024 - barely, with $72.7 million net profit on $2.7 billion revenue. Project Kuiper has invested $10B+ and must launch 1,600 satellites by July 2026 just to [retain FCC license rights](https://circleid.com/posts/western-leo-satellite-internet-update-oneweb-telesat-kuiper-iris). China's constellations are state-backed. OneWeb went bankrupt once already. If a mega-constellation operator fails, thousands of uncontrolled satellites remain in orbit with no entity responsible for their disposal. At a 1-2% annual failure rate, even a healthy Starlink generates 20-40 derelict satellites per year that cannot self-deorbit. **No enforceable global space traffic management exists.** ITU spectrum allocation operates first-come-first-served. China's 200,000-satellite filing is a spectrum land-grab, not a real deployment plan. The FCC regulates US operators but has no jurisdiction over Chinese or Russian constellations. Active debris removal technology exists (ClearSpace, Astroscale) but remains entirely government-funded. There is no commercial market because there is no mandatory liability framework. A [proposed EU Space Act](https://www.twobirds.com/en/insights/2026/space-and-satellite-wrap-up---legal-and-regulatory-developments-in-2025) (June 2025) covers only EU-licensed operators. **Insurance barely covers the risk.** Only about [6% of satellites carry in-orbit insurance](https://www.internationalinsurance.org/insights_cyber_the_space_debris_dilemma). The space insurance market cannot price catastrophic correlated events (Kessler cascade, solar superstorm) because they would simultaneously affect all insured assets. This is not insurable risk. It is too systemic. A [PNAS analysis](https://www.pnas.org/doi/10.1073/pnas.1921260117) found that a harmonized orbital-use fee of ~$235,000 per satellite-year would correct incentives and increase long-run industry value from ~$600B to ~$3T. No such mechanism exists. ### The Correlation Problem This is the factor that transforms LEO vulnerability from a sectoral concern into a civilizational one. All LEO-dependent services share the same physical environment. A debris cascade, solar storm, or ASAT campaign does not selectively degrade broadband while leaving GPS intact. It does not spare weather satellites while destroying ISR platforms. The failure mode is correlated across every service category. A modern economy losing GPS simultaneously with satellite broadband and weather forecasting faces compounding failures that no sector-specific contingency plan addresses. Financial markets lose timing synchronization. Aviation loses precision approach capability. Precision agriculture loses guidance. Military forces lose ISR and secure communications simultaneously. Maritime shipping loses AIS and navigation. **LEO is a single point of failure for civilization-scale infrastructure.** No equivalent single-point-of-failure exists in terrestrial infrastructure - no one earthquake can simultaneously disable the internet, GPS, weather forecasting, and military communications globally. ![](https://substack-post-media.s3.amazonaws.com/public/images/b62e5149-b20c-4082-8c7b-c32abd064707_1514x946.png) ## The Stratospheric Alternative At 20 kilometers altitude, above weather and conventional air traffic but far below the orbital debris field, there exists a layer of the atmosphere that is largely empty, physically benign, and operationally proven. **The stratosphere is a hedge against LEO's correlated failure modes - and for several applications, it is the superior solution regardless.** ### The Physics Case The comparison between a satellite at 500 km and a stratospheric platform at 20 km comes down to distance - and the physics of distance are unforgiving. **Signal strength:** Free Space Path Loss (FSPL) at 10 GHz from 20 km altitude is 138 dB; from 500 km it is 166 dB. The 28 dB difference sounds small until you remember that decibels are logarithmic- every 10 dB represents a 10x change in power. So 28 dB = 10^(28/10) = ~630x more signal power reaching a ground receiver from a HAPS than from a LEO satellite. A stratospheric platform talking to a standard 4G handset has physics working in its favour by nearly three orders of magnitude. **Latency:** Round-trip latency from 20 km is approximately 0.13 milliseconds. From 500 km LEO orbit, it is 3.33 milliseconds. From GEO, it is 600+ milliseconds. **Imaging resolution:** The Rayleigh criterion for diffraction-limited optics means that achieving 10 cm ground resolution from 500 km requires a 3.05 meter mirror. From 20 km, the same resolution needs a 12.2 cm lens. 25x times smaller. This is why satellites top out around 30 cm commercial resolution while HAPS can achieve sub-10 cm with modest optics. Direct-to-hand-held-device connectivity: The April 2025 [HAPS Alliance whitepaper](https://hapsalliance.org/wp-content/uploads/formidable/12/2025_HAPSAlliance_Reference_Architecture_Advantages_Satellite_Connectivity_TWG_Whitepaper.pdf) establishes that HAPS are compatible with existing 4G/5G handsets without modification. Satellite direct-to-device (Starlink, T-Mobile partnership) requires specialized protocols and delivers a fraction of the throughput. In early 2025, Aalto's Zephyr demonstrated the [first wireless connection from a fixed-wing HAPS](https://www.aaltohaps.com/zephyr-sets-world-record-for-longest-continuous-flight-flying-67-days-in-stratosphere/) at 60,000 feet to a standard 4G mobile device on the ground. ![dB is a logarithmic scale. 166 vs 138 = 28 dB difference = 10^(2.8) = ~630x power difference. The numbers look close; the physics are not.](https://substack-post-media.s3.amazonaws.com/public/images/5868c46e-2586-4be2-99da-ba639a9cb342_1576x922.png) ![](https://substack-post-media.s3.amazonaws.com/public/images/9bbce62e-bfd7-468e-9bcb-b0d32d8ee416_1590x1374.png) ### Technology Maturity In April 2025, Airbus subsidiary AALTO's Zephyr completed [67 days, 6 hours, and 52 minutes of continuous stratospheric flight](https://www.aaltohaps.com/zephyr-sets-world-record-for-longest-continuous-flight-flying-67-days-in-stratosphere/) - a world record. Solar-powered, battery-sustained through the night, operating at 60,000+ feet. It is an operational endurance demonstration exceeding most military deployment cycles. **The HAPS ecosystem in 2025-2026:** **[Radical (Seattle, US)](https://www.radicalaero.com/):** A fixed wing form factor; solar powered; flew their full systems maiden flight in autumn 20254; (an Inflection portfolio company) **Aalto/Zephyr:** [67-day record](https://www.aaltohaps.com/zephyr-sets-world-record-for-longest-continuous-flight-flying-67-days-in-stratosphere/). Commercial entry-into-service in Japan planned for 2026, backed by $100M consortium (NTT DOCOMO, Space Compass, Mizuho Bank, DBJ) **Kea Aerospace (New Zealand)**: [Historic first stratospheric flight](https://www.keaaerospace.com/news-and-events/media-release-kea-aerospace-achieves-historic-stratospheric-flight/), February 8, 2025. Atmos Mk1b reached 56,284 feet, flew 8 hours 20 minutes **Sceye (US)**: [NASA/USGS partnership](https://news.satnews.com/2024/10/30/sceye-partners-with-nasa-usgs-to-address-climate-change-from-the-stratosphere/) for methane monitoring. [SoftBank investment](https://www.softbank.jp/en/corp/news/press/sbkk/2025/20250626_01/). Pre-commercial HAPS services in Japan 2026 **BAE Systems PHASA-35:** Presented at [DSEI 2025](https://www.baesystems.com/en/product/phasa-35) as a defense product for persistent ISR and communications. Months-long endurance without landing **India DRDO AS-HAPS:** First flight trials May 2025 at 17 km altitude. [Defence Acquisition Council approved](https://swarajyamag.com/defence/high-altitude-pseudo-satellite-gets-dac-nod-heres-what-it-is-and-why-india-needs-it) **Russia Barrage-1**:Tested February 2026: - a stratospheric aerostat developed explicitly as a Starlink alternative after SpaceX restricted access to Russian military user. The proliferation is telling. When Russia builds a battlefield stratospheric relay within weeks of losing Starlink access, the dependency - and the remediation speed - are clear. ### The Resilience Case When LEO degrades - through debris cascade, ASAT strike, or solar storm - the stratosphere remains intact. A HAPS at 20km sits below the orbital debris field, is shielded from radiation by the atmosphere, presents no viable ASAT target, and can be replaced in days rather than years. It is an independent infrastructure layer whose failure modes are uncorrelated with orbit. The applications that matter most in a LEO-denial scenario are precisely where HAPS excels: For connectivity, a single platform covers a 50km+ radius with direct-to-device service to standard handsets, deployable from a portable ground station in hours - Japan's SoftBank/Sceye program is explicitly designed for earthquake disaster recovery, and Taiwan demonstrated 14+ days of balloon-based emergency comms in 2025. For navigation, a HAPS pseudolite at 20km delivers a signal roughly one million times stronger than GPS at 20,200km, making it much harder to jam while providing sub-meter accuracy over sovereign territory that can be repositioned in hours. **For ISR and earth observation**, sub-10cm resolution with indefinite dwell time and real-time video replaces the satellite model of daily revisits at 30cm with minutes of dwell per pass - the difference between a photograph and continuous surveillance. [Defense analysts](https://www.bridge-connect.com/post/haps-in-defence-the-new-stratospheric-domain-for-isr-communications) are converging on a three-tier aerial architecture - aircraft, stratospheric HAPS, and space - treating the stratosphere as an existential building block. ### Building the Mitigation Stack The five failure modes described above are not inevitable. They are engineering problems. Three companies in Inflection's portfolio are building specific capabilities to extend LEO's operational life and mitigate its risks. If the CRASH Clock reads 3.8 days, the first question is: can we track what we need to avoid? Foundational builds the precision tracking layer for space situational awareness using Space Laser Ranging (SLR). Where radar tracking provides meter-level accuracy, SLR provides sub-centimeter-level precision - 1,000x more accurate. The difference matters when you are trying to predict whether two objects at 7.5 km/s will miss each other by 10 meters or collide. Tracking threats is necessary but not sufficient. Something must physically intervene. [Lodestar](https://www.lodestar.space/) is building autonomous satellite defense capabilities in contested space environments. Their system provides onboard decision-making - using game theory to predict adversary intent, detect threats, and execute orbital responses without ground control. [Radical](https://www.radicalaero.com/) is building autonomous solar-powered aircraft that fly perpetually in the stratosphere at 20km altitude. An aircraft can deliver what a constellation of thousands of satellites struggles to match. Two verticals: (1) connectivity - beaming 5G-speed internet directly to handheld devices over a 50km+ radius, targeting 3 billion underserved people, and (2) persistent imagery - real-time sub-10cm resolution earth observation including live video, physically impossible from orbit due to lens size and revisit constraints. **If you're building related solution to mitigate infrastructure fall outs due to LEO failure, please get in touch.** --- # The End of Software As We Know It *Published: 2026-02-11 | Author: Alexander Lange | Section: Research* URL: https://inflection.fund/writings/the-end-of-software-as-we-know-it Our team has been building on Claude for over a year. Since the launch of Claude Code it became far more than a coding assistant, it turned into a new type of computer. We describe objectives in plain language. It reads our files, queries our databases, writes code, generates charts, drafts memos, and returns finished output - **all from a terminal**. No GUI. No buttons. No application windows. As SemiAnalysis put it: > "Claude Code is the inflection point for AI Agents and is a glimpse into the future of how AI will function." - [Semi Analysis, Claude Code is the Inflection Point](https://newsletter.semianalysis.com/p/claude-code-is-the-inflection-point) We agree. 4% of all GitHub public commits are now authored by Claude Code. At the current trajectory, that number reaches 20%+ by year-end. Something fundamental is shifting - and coding is just the beachhead. What we are experiencing is a canary in the coal mine for a profound **reorganization of the software stack**. The boundary between application and operating system is dissolving. The interface between human and machine is collapsing from visual to verbal. And the implications extend far beyond software development into every information business. ![transition: https://www.cosmos.so/e/1023157549](https://substack-post-media.s3.amazonaws.com/public/images/62e5c831-a35e-410d-925d-a9031214617e_1920x1080.jpeg) ## The Old Era The technology stack that dominated computing for four decades has four layers. ![](https://substack-post-media.s3.amazonaws.com/public/images/79116f3a-eae1-46db-bd5e-1d25d20f7b25_1480x1142.png) For forty years, humans adapted to this stack. We learned keyboard shortcuts. We memorized menu hierarchies. We followed workflows that developers prescribed for us with marginal customization opportunity. We created accounts, accepted terms of service, and organized our work around the constraints of each application. **Value concentrated at the application layer** because building complex software was expensive and switching costs were high. Once you stored your leads in Salesforce, your documents in Google Drive, your creative assets in Adobe, and your messages in Slack - that data became hard to move. Software companies optimized for this lock-in. It produced a generation of businesses trading at premium multiples on recurring revenue and feature moats. The **human brain was the only integration** layer. We were the ones copying data between applications, context-switching between fifteen tabs, manually synthesizing information scattered across silos. The applications didn't talk to each other. We did the talking for them. ## The New Era The stack is collapsing from four layers to three. ![](https://substack-post-media.s3.amazonaws.com/public/images/0a63a6bc-57fa-4c5b-a9f9-317cb7624df5_1428x1058.png) Hardware remains at the base - unchanged, still the physics constraint. Operating system still manages resources but now includes the model runtime. AI and interface merge into a single layer. The AI is simultaneously the reasoning engine and the interface. Natural language in, generative output out. **Applications vanish as a standalone layer.** Their logic gets absorbed into the AI layer. Lead scoring, image editing, financial modeling, document formatting, data visualization - all of this becomes commodity inference. The AI doesn't "use" applications. It replaces the need for them by synthesizing directly from data. The output is format-agnostic. The same reasoning engine can produce a chart, a piece of code, an audio summary, a video, a formatted document, or a fully functional piece of software. Applications were containers - standardized, rigid, same for everyone. The AI layer is a synthesizer. Think of the difference between IKEA furniture and having a carpenter in your house who builds exactly what you need from raw materials, on demand. This is what we experience daily. When we ask Claude Code to analyze a dataset, it doesn't open Excel. It writes a script, runs the analysis, generates the visualization, and returns the result. When we need a report, it doesn't open Google Docs. It reads the source material, reasons about it, and produces the output in whatever format we need. The application layer is absent. ## What it means in the big picture of info tech Zoom out far enough and you see this shift as the fourth inflection point in how humans handle information. ![](https://substack-post-media.s3.amazonaws.com/public/images/a0c2a851-549a-46ff-8499-33d231418958_1258x766.png) Each leap solved the previous era's bottleneck but created a new one. Printing solved storage. The internet solved distribution but drowned us in noise. AI solves synthesis. But has several bottlenecks. The most imminent is context: to reason, machines need access to everything. **Current application design is fundamentally at odds with this requirement**. Information is organized in vertical silos by function. Google Drive stores documents. Superhuman handles email. Signal carries messages. Calendar manages time. Attio tracks business relationships. Substack publishes writing. Each application is a walled garden with its own login, its own data model, its own way of holding your information hostage. A human can context-switch between fifteen apps and mentally stitch the picture together. An AI needs unified access to reason across all of them. The most obvious quick fix is connecting these silos through machine-readable protocols. APIs provide standardized doors into each silo - they've existed for decades but adoption remains uneven. MCPs (Model Context Protocols) are a newer standard that lets AI models connect to external tools and data sources through a universal adapter format. Together, they are the duct tape holding the transition together. They let AI reach into existing silos without tearing them down. This works. For now. ## What It Means for Software Businesses The instinct of every software incumbent is to bolt AI onto the existing product. Preserve the interface, preserve the subscription, preserve the revenue model. Adobe added Firefly inside Photoshop. Figma embedded AI design assistants into its canvas. Microsoft launched Copilot across Office 365. Salesforce built Agentforce on top of its CRM. If the interface collapses into language and output is generated directly by AI, adding AI features to your interface doesn't save the interface. It's like adding a GPS to a horse-drawn carriage after cars arrive. It won't save software as we know it. Not every software company is equally exposed. The **critical variable is whether the moat exists beyond software layer**. Some examples: Uber looks like a software company but is actually a physical marketplace - 5 million drivers, regulatory licenses in 10,000+ cities, real-time liquidity density that took a decade and billions in subsidies to build. The app is the thinnest layer. Making software free makes Uber cheaper to operate, not easier to disrupt. Wolters Kluwer sells regulatory truth with legal standing. AI makes their data more queryable - it doesn't replace it. Google's moats sit below the software layer in cloud infrastructure, search index, and proprietary data assets. Salesforce is the exception that proves the rule - it's moving down the stack with MuleSoft (API infrastructure), Data Cloud (derived data assets), and Agentforce (governed gateway for AI agents accessing enterprise data). Public markets are pricing this in - aggressively. ![](https://substack-post-media.s3.amazonaws.com/public/images/c2d2c8eb-8615-440b-a098-4a68ad2bda77_1410x326.png) **The paradox is striking**. Earnings are growing. Margins are stable or expanding. Yet stocks are cratering. Markets are repricing the terminal value of software moats. Current earnings are fine - but future earnings are worth less because the moat is dissolving. **Are markets overshooting?** Possibly, in some cases. AI adoption won't happen overnight. In regulated industries - healthcare, accounting, law - institutional inertia is enormous. Employees need retraining. Compliance workflows are sticky. Liability frameworks haven't adapted. Legacy tech is protected by friction, not innovation. This buys time but the terminal direction is clear - the life cycle ends. **In private markets the repricing hasn't even started**. Most VCs have significant portfolio exposure to enterprise SaaS. The existential threat to interface-centric software isn't reflected in book values. Adding AI features on top of a traditional SaaS product is cosmetic - it doesn't address the structural shift. Most VCs didn't hear the shot. Their LPs are looking at over-bloated book values that will quietly collapse over the next few years as markdowns catch up with reality. ## What's Next APIs and MCPs are transitional. They solve the access problem by breaking open data silos, but they create a new one: when one AI provider connects to your email, files, CRM, calendar, and messages, you've built a single point of failure. Compromise one integration layer and the attacker gets everything. It's a security nightmare. The long-term stack needs to be fundamentally different. Some developments we follow closely: Fully homomorphic encryption allows computation on data that stays encrypted throughout - the AI reasons without ever seeing the plaintext. Zero-knowledge proofs let you verify a computation was done correctly without revealing the inputs. Apple's Private Cloud Compute extends on-device security into the cloud through stateless, zero-trust enclaves where data is processed and immediately deleted. Local-first architectures keep data on your device and sync peer-to-peer, eliminating the central server entirely. On-device models handle sensitive queries without data ever leaving your hardware. None of these are mature yet but they are laying the plumbing of the next era. Plumbing is harder to replace than applications. Inflection has barely any SaaS exposure and concentrated on companies that address critical bottlenecks of the emerging stack. We invest thematically and focus rigidly on where the new architecture breaks down. Some examples: Trust and verifiability: \[\[Anytype\]\] (local-first, encrypted collaboration where users own their data), [Fabric](https://www.fabriccryptography.com/) (custom silicon accelerating encrypted AI workloads), [Ubitium](https://www.ubitium.com/) (reprogrammable silicon for edge AI workloads). Data bottlenecks: [Deep Earth](https://www.deepearth.tech/) (subsurface mapping / geospacial AI). Connectivity resilience: [Hedy](https://hedycyber.com/) (an alternative networking stack that makes connected devices invisible and provides failover in contested environments. Physical-world integration: [Ark](https://ark-robotics.com/) (robotic fleet control - vertically integrated yet modular), [NAD](https://www.nordicairdefence.com/) (drone interception network - vertically integrated yet modular), [Levtek](https://www.levtek.io/) (modular industrial robotics), Stealth (positioning data through space laser ranging - physical sensor networks addressing data bottlenecks). None of these are pure software plays. None can be replicated by an LLM with API access. When software production costs go to zero, these companies become more valuable - they are anti-fragile in this era. If you're building at the bottlenecks of the new stack, we'd like to talk. And if you disagree with any of the above - we'd like to hear that too. --- # Building for Venture Capital *Published: 2026-01-27 | Author: Alex Patow | Section: Building* URL: https://inflection.fund/writings/building-for-venture-capital I've spent the last five years building technology for VC funds, first at EQT as part of the [Motherbrain](https://eqtgroup.com/about/motherbrain) platform, now the past two years at [Inflection](https://inflection.fund/) where I've built our core infrastructure with the help and guidance of and . Along the way, I've noticed a recurring pattern: most engineers joining VC come from the outside, and there's no real onboarding guide for the role. They're talented people who haven't worked in venture capital before, don't yet know how funds operate, and often spend their early months figuring out the basics: what data matters, what tools exist, where to focus first. I was one of them back in the day. So I wrote the guide I wish I'd had: [Building for Venture Capital](https://buildingfor.vc). ![](https://substack-post-media.s3.amazonaws.com/public/images/6c3af24a-f24e-45da-85b5-68ce9c66e21f_1024x1024.png) ### What It Covers The guide is organized into three parts: **[Part 1: Understanding VC](https://buildingfor.vc/guide/part-1-understanding-vc/what-is-a-vc-fund)** How venture capital funds actually work, from the perspective of someone building technology for them. Fund structures, LP relationships, investment processes, and the common mistakes technical hires make when they don't understand the domain. **[Part 2: The VC Tech Stack](https://buildingfor.vc/guide/part-2-tech-stack/introduction)** A tour of the tools and systems VC funds use: research platforms, sourcing tools, CRM and deal flow, fund operations, portfolio support, fundraising, and external presence. For each category, I cover what the tool does, when to build vs. buy, and real examples from working funds. **[Part 3: Technical Foundations](https://buildingfor.vc/guide/part-3-technical-foundations/choosing-your-stack)** The deeper technical work: choosing your technology stack, working with data providers, data modeling for VC, entity resolution, data quality, warehousing, integrations, security and compliance, and emerging trends like MCP and AI agent orchestration. There's also a [resources section](https://buildingfor.vc/resources) with curated newsletters, books, research papers, and a prompt bank for common VC workflows. ### Why Open Source? At Inflection, we believe the VC industry benefits when funds share knowledge about how technology is shaping the way we operate. Too often, this expertise stays siloed within individual firms. By publishing this guide, we want to create more conversation about how new technologies are changing what's possible for VC funds of all sizes. The whole thing is on [GitHub](https://github.com/alexpatow/building-for-vc). If something's wrong or missing, you can open an issue or submit a PR. If you've got prompts, resources, or perspectives that would help others, contributions are more than welcome. ### Acknowledgments This wouldn't exist without the people I've worked with at Inflection and EQT who shaped how I think about this work. Special thanks to the [external contributors](https://buildingfor.vc/guide/contributors) who shared their perspectives. Check it out at [buildingfor.vc](https://buildingfor.vc). Feedback welcome. // AP --- # Heresy and the Venture Industrial Complex *Published: 2025-12-10 | Author: Alexander Lange | Section: Markets* URL: https://inflection.fund/writings/heresy-and-the-venture-industrial-complex The venture industry was born out of curiosity. Over time it traded serendipity for playbooks. The result is an increasingly homogenous and ultra-concentrated ecosystem that is routing capital towards consensus ideas instead of funding heretical ones with vast potential. This post is an attempt to contextualise such developments and explore their implications for stakeholders. Towards the end I'm sharing some ideas around the concept of _heresy_ and how it might still be applied in venture going forward. ![Citizens of Earth, a conceptual installation designed by architect Marc Thorpe](https://substack-post-media.s3.amazonaws.com/public/images/4f9fc640-4bd9-4389-bd73-f49e3068a73e_1200x800.jpeg) ## The ocean of incremental sameness The venture industry sits at the intersection of technology, politics and culture. Over the last few decades all of those areas have become more homogenous, more compliant, more complacent and more boring. Let's go through some examples. **Music** used to evolve alongside other cultural phenomena in a specific, local context and _Zeitgeist_. It polarised between old and young. It needed to overcome repression by establishment institutions before it could spread into the world. From RUN DMC's _Raising Hell_ Album (1986) by Bob Gendron: > _Few albums change the world. Raising Hell did so in the face of bitter resistance and prejudice. (…) Still, nothing and no one could stem the tide unleashed by Raising Hell. Not the regressive Parents Resource Music Center whose co-founder maintained that "angry, disillusioned, unloved kids unite behind heavy metal and rap music, and the music says it's OK to beat people"; not irresponsible city mayors who threatened to ban Run-DMC concerts; not pundits who sought to preserve the status quo by falling back on the type of ignorant contentions that nearly muted Rock 'n' Roll in the late 1950s. Raising Hell disrupted tradition, dared listeners to reconsider what they thought they knew and provided a platform for art, sound and discourse that lent power, strength and identity to marginalised voices._ When was the last time you listed to _fundamentally new, radically different_ music? Contrast this to the emergence of hyper-popular artists and types of music who master the industry and fan-interactions to an unprecedented level of perfection. Who _doesn't_ like Taylor Swift? **Furniture and interior design** seem to have converged immensely. No matter if you book an AirBNB in Berlin, Cape Town or San Francisco, chances are very high that you find the same plants, design furniture, exposed walls and minimalist art decoration in all of them. The post–Cold War period saw a notable institutional convergence in **political systems**. Several countries adopted multi‑party elections, similar constitutions, independent central banks, regulatory agencies, and market‑oriented policies. The European Union pushed for regional integration. China innovated around an authoritarian-capitalist system and more recently the US started to resemble such a system increasingly: both rely on powerful executives - China through plans and directives, the US through executive orders. Both moved from laissez-faire towards state-shaping of strategic sectors. Both are fusing national security with their economies through an interlock. Europe is years behind but started the fusion process as well. Apparently, **technology startups** and the venture ecosystem around them are not immune to the trend of convergence as we will explore more in the next section. **Why?** What is driving all this? Complex systems and (single factor) causalities don't match well. Therefore, I won't make any bold claims on what the definitive drivers are. Rather I share some speculations on what might be _contributin_g ones. **1/ Incentives**: through the internet we learned to track consumer preferences and responsiveness to any kind of content. Music, design, movies, political views. In fact consumer preferences can be predicted and manipulated at scale. This is flipping the script for creators. Instead of taking high conviction bets on creating something from scratch and iterating over it with a local community they can just _listen to their audience_. In order to boost their reach and distribution they rather create something incremental that is consumed by the masses than something genuine that is consumed by a tiny minority but holds the potential to set a trend. Faster horses instead of cars. **2/ Mimetics:** Mimetics studies how ideas, behaviours, and aspirations spread by imitation. People copy what others signal as desirable. In markets, this copying creates clusters of shared beliefs about what "good" looks like. When many chase the same signals of success, competition narrows around a few accepted patterns. Rivalry intensifies and differentiation collapses. There are fewer outliers because deviation carries social and professional risk. Mimetics are massively amplified by social media and global connectivity. While memes have newfound potential to reach a wide audience much faster than in the past, they are also much more short-lived. We get bored faster, or get used to new ideas faster, to phrase it positively. **3/ AI training and governance:** AI model training is an increasingly homogenous process where incrementally different LLMs are fed with mostly the same data. The selected input data is subject to selection bias in terms of religious beliefs, political views, cultural norms and legislation amongst others. This type of monopoly control over the mind is amplified by the terms of service applied by the big tech platforms deploying said models. My friend Erik Voorhees had a few things to say about this in his incredible piece [The Separation of Mind and State](https://moneyandstate.com/blog/the-separation-of-mind-and-state). Here is the introduction: > _Hundreds of years ago, through tremendous sacrifice, the institution of religion was incrementally removed from the government's sphere of authority. Today, religious or not, we recognize the importance of that separation of church and state.The original cypherpunks sought to separate language from state through encryption—for a while declared a "munition," where certain types of math were [banned from export](https://en.wikipedia.org/wiki/Crypto_Wars). Their descendants, the early Bitcoin pioneers, sought to similarly separate money and state.But if monopoly control over god or language or money should be granted to no one, then at the dawn of powerful machine intelligence, we should ask ourselves, what of monopoly control over mind?_ This is not something AI researchers aren't aware of, so we see people working on custom data sets, expert-curation, and alternative data types in order to make models spikier, or even just have more common sense. What happens when the venture ecosystem loses its diversity and becomes one with the vast ocean of incremental sameness? ## The rise of consensus capital and the venture industrial complex The characteristics of venture capital changed profoundly over the last few decades: **1/ Fund sizes ballooned, capital concentrated:** In the early days, fund sizes used to be around $100M with return expectations of 5x and more for the top performing ones. Small funds took a high number of absurdly ambitious moon shot bets with extreme variance in outcomes. Today, large, multi stage funds are dominating while [emerging managers struggle to survive](https://svrgn.substack.com/p/ventures-consolidation-and-the-case). In 2024, a16z raised $7.2 billion—over 11% of all US VC. The top 30 firms captured 75% of the venture market. With large fund sizes come different expectations of return predictability and certainty. Venture became inherently risk averse. **2/ State-economy fusion:** with billion dollar fund sizes the game changed. Instead of funding high risk, high reward experiments, the venture industry started to merge with the state by getting increasingly involved in lobbying. This might have started with a16z's initiatives around operation choke point, a coordinated attempt of the US government to shut down the crypto industry without legislation by leveraging de-banking and de-platforming practices. It is ironic that venture's push back against executive over-reach to protect an inherently libertarian industry spilled over into a full blown fusion of private venture with government. In "[Honey, We Need To Talk About Venture Capital](https://butthistimeitsdifferent.substack.com/p/honey-we-need-to-talk-about-venture?r=4eaetr&utm_medium=ios&triedRedirect=true)" captures the essence of the above developments with more nuance and countless examples. Her conclusion for the role of what she describes as _little venture_ is the following: > _If you're a venture fund that isn't doing any of this (no policy alignment, no industrial positioning, no narrative or regulatory leverage), then you are not in the same business as Big Venture. You're in venture capital, the original game of chasing founders, markets, and luck. Said another way, you're in the nostalgia business. While Big Venture manufactures outcomes, you're literally just buying lottery tickets, and getting paid 2% a year to do that. (…) Big Venture shapes the game and collects steady wins. Venture capital thinks it's finding alpha, but really it's just getting played._ In a similar vain my friend and colleague wrote in his piece [Consensus Capital](https://stateofthefuture.substack.com/p/consensus-capital): > _Being right in isolation, insofar as it ever was actually true, is a dead strategy. If you have an insight but cannot coordinate follow-on capital, you lose. If the state floods your sector with subsidies favouring different players, you lose. If megafunds deploy $500 million rounds and dilute your position to irrelevance, you lose. The era of the lone wolf VC who finds the overlooked founder and waits for the world to catch up has largely passed. (…) The best VC today is not the one who sees the future first. Or writes the best blog. It is the one who convenes the consortium that funds it. The future that is, not the blog._ ## The dark side of consensus investing I share many of the observations made in the above articles but I disagree with the conclusions for several reasons. Consensus investing is unlikely to be the dominant, alpha-generating strategy in venture over extended time horizons. **1/ Central planning has a history of failure:** despite a massive deployment of resources, centrally planned technologies historically rarely worked. In [Why Greatness cannot be planned](https://www.goodreads.com/book/show/25670869-why-greatness-cannot-be-planned?ac=1&from_search=true&qid=EeEKbBC6WX&rank=1) by Kenneth O. Stanley and Joel Lehman there are countless examples explaining why the world's most significant discoveries and innovations arose from serendipity, exploration and open ended searches, rather than from pursuing pre-defined goals. E.g. **novelty search in robotics**: Agents learn to move or "walk" better when optimized for behavioral novelty rather than a direct "walk farther/faster" objective. By rewarding new, interesting behaviours, they accumulate the prerequisites to walk. Top down objectives can be a false compass. **2/ Financial Bubbles:** When large funds must deploy vast sums, they herd into consensus categories and companies. Prices are pushed above fundamentals. Escalating valuations driven by cross over funding rounds look like guaranteed demand. They are not. Public-market discipline eventually re-prices growth stories at some point. Looking at Palantir's 600x PE ratio that might not be tomorrow though. Extreme concentration isn't limited to private markets as late cycle dynamics favour big techs structurally. **3/ Policy capture is brittle:** Aligning with procurement and subsidies can raise certainty but it also concentrates political risk and single buyer exposure. Trump Jr.'s Vulcan Elements and associated 1789 fund portfolio companies won several $600M+ government contracts. This looks like an unfair advantage (in the literal sense, looking up the definition of "corruption"). However, things can change through election reversals, program cancellations, government shut downs and shifting security doctrines. Long term capture in frontier tech seems easier said than done long term. **4/ Founder incentives:** As soon as you can put a label on it the alpha is gone. With labels and established categories comes fierce competition. Talent wars break out. Margins collapse. Such an environment is not an ideal place for breakthrough innovation. As my colleague David Peterson was writing i[n Startup Stagnation](https://newsletter.angularventures.com/p/startup-stagnation?_bhlid=11d671c494fa10862975b1ca5c511034b832dea3&last_resource_guid=Post%3A6187919b-020c-4776-afe0-1830e9a950fc): > _The advantage of an overcrowded consensus is simple: the field is wide open for things that are actually different. As consensus ideas soak up attention and capital, the founders building in weird markets have zero competition. They have time to be illegible, to iterate without the spotlight and to build credible monopolies. The cover bands will keep playing the hits. But somewhere, quietly, the new music is being written._ This anecdotal wisdom is supported by some data. A recent piece of MIT Research found that [Disagreement Predicts Startup Success: Evidence from Venture Competition](https://lucagius.github.io/files/Disagreement_judges.pdf?utm_source=www.newsletter.datadrivenvc.io&utm_medium=newsletter&utm_campaign=this-factor-predicts-startup-success)s. The paper finds that higher dispersion in judges' scores strongly correlates with better outcomes—more funding, higher revenues, and greater exit likelihood. The mechanism is that unique and hard-to-evaluate ideas naturally spark disagreement; common opinions rarely confer advantage. The big caveat of this research is a selection bias as the underlying data has been gathered only between 2011-2020. I would love to see someone continue this research over extended time periods and throughout the era of consensus capital. ## Where the heretics go If vast parts of the market are soaked up by consensus capital - where do _heretics_ go? Now is a good time to define what I mean by heretics in this context. I would describe them as people who prioritise first-principles over playbooks, accept delayed validation and higher variance to unlock non-obvious breakthroughs. ![Francisco Goya - The Inquisition Tribunal, 1812-1819](https://substack-post-media.s3.amazonaws.com/public/images/14fcd8b8-6893-421e-ad18-d16b3f56e96c_1920x1170.jpeg) **1/ Scarcity bottlenecks:** The consensus herd typically looks at a problem and backs obvious solutions. AI training and inference needs more compute, so they back data centers and accompanying infrastructure. Heretics think 2, 3 steps ahead by anticipating shifts in future scarcity bottlenecks. In an interview with Floodgate Founder put it this way: > _(…) maybe genuine heresy requires all three. You need the perceptual difference to notice something meaningful, the relevant experience to know when that perception matters, and the courage to act on it. Which might explain why heresy is so rare. (…) Every major technology shift creates a new scarcity, and that scarcity is never the enabling technology itself. When computation got cheap, software became scarce, because it was the layer that made computers useful. When communication got cheap, organizing layers that connected people to content, commerce and community became scarce. Now AI is making cognition abundant._ What might be the next scarcity bottlenecks? In high stakes, high friction fields (nuclear reactor design; drug molecule design; autonomous robotics) cheap cognition amplifies the risk of failure. We need to be able to **verify and trust** the results put out by machines. Provenance and auditability (logs, reproducibility) will play important roles in those areas and very few people are working on it. Our portfolios Fabric and Flashbots are some of them. AI's next frontier is energy. Physical delivery energy markets are the achilles heel of AI scalability going forward. They are opaque, inefficient and unreliable. Fixing them is a vast, structural challenge some heretics are working on already. **2/ unregulated domains:** Regulatory capture is only feasible in areas where legislation is in place. _Heretics_ go where there is no regulation in place that could be captured by the venture industrial complex. The space domain comes to mind. Space security services as provided by our portfolio Lodestar or stratospheric satellites for remote sensing, connectivity and imagery built by our portfolio Radical come to mind. Ocean autonomy, deep sea mining or geo engineering fall into the same category. Network states might be an alternative to over-regulation and create a space for innovation. **3/ short consensus**: bold heretics might consider to short consensus narratives and ventures. Are we really expecting vibe coding startups to build sustainable moats against established dev tooling and infra companies? Do we really believe that Perplexity can win against Google's full stack infrastructure moats and integrated services landscape? Do we really think that AI native marketing and customer care products can win against Salesforce's agent platform? Besides buying some deep out of the money put options on some trends, heretics might consider to use prediction markets as a tool of choice to short private market consensus in some areas. This is an area we started exploring outside of our venture funds. **4/ other weirdness:** we still believe there will be undiscovered talent (e.g., consensus today is 20-something "based" drop-out with fierce velocity, but who would back a [biology high-school teacher](https://en.wikipedia.org/wiki/Project_Hail_Mary)?) and that there will be ideas that don't work with incentive structures of big venture firms. Which venture fund would back a company in an industry that historically distributes a large chunk of their earnings as dividends? Common beliefs need to be questioned but people are usually not promoted if they bring in 19 strange ideas underpinned by non-common beliefs. [Studies on peer review](https://ui.adsabs.harvard.edu/abs/2014arXiv1406.5520M/abstract) and grants in the scientific community show that the truly novel, orthogonal ideas (like mRNA therapies) rarely make it to publication or receive awards. We need more improbable bets with a high variance in outcomes. Some of them will be massive. **At Inflection, we continue to back companies at the heretic frontier. If you're in our camp and build something weird - reach out.** --- # Depth vs. Breadth *Published: 2025-12-05 | Author: Jonatan Luther-Bergquist | Section: Markets* URL: https://inflection.fund/writings/depth-vs-breadth There is a great number of points in every person's life when a line of subjective inquiry ends with the thought "I don't know enough about this to go one level deeper." And if you have an inquisitive 4 year-old then it happens multiple times a day. Tonight's example: > Why won't you fall out of a rollercoaster? > Well, you are usually strapped in. > > And why not, more reasons? > I guess if you're going really fast in a loop then you're pressed against the seat anyway by the centripetal force and won't fall down. > > But how can it hold you? > Hm, the seats are really durable and they are held by really strong metal bars that also support the wagon. > > Is it harder than teeth? > Uhhh, I think so? In some ways… > > And what about dinosaur teeth? > Good question… Because of the nature of our work, I'm usually on the kid's side of this dialogue with founders or researchers. (Hopefully with slightly less non sequiturs involving dinosaur teeth.) We try to understand the tech as quickly as possible, have good conversations about the business at hand, follow our own curiosity and the specific complexities of the area. Ideally, we have a prepared mind and can skip the first couple of layers of questions and dive into the more interesting stuff faster. In some cases we've written about the topic publicly before, or we're close to companies who have indicated where the interesting problems of that space exist. But there comes a time when the response amounts to "Good question…" in some more or less strategically phrased way. ![Nano Banana Pro](https://substack-post-media.s3.amazonaws.com/public/images/602dc2bd-3884-4e04-8bcd-a4059a66c305_1024x768.png) ### Why does the depth matter? We care about how deep we can go because it shows us something about the founder's way of thinking. To give some examples, it shows us: a. How much the other person thought about their business, which is a proxy for: i. How obsessed are they? ii. Are their level of curiosity greater than ours? iii. Are they stuck in a particular way of thinking or can they change perspectives? b. What the founder's angle-of-ideas (see illustration below) for the future of the company looks like, which shows: i. Priorities and certainty in predictions ii. What dimensions of the plan are fixed vs. flexible? iii. Understanding of getting big, ambitious projects done c. How the person deals with not knowing, this can cause: i. Avoidance, misdirection, conversational "tactics" ii. Nervousness and awkward silence (side note: it doesn't make us uncomfortable) iii. Making stuff up, acting like you know when you don't iv. (this is the worst) Actually thinking they know when they don't v. Acceptance of lack of understanding as a common part of going deep into something ![Author's illustration. Imperfect 2D representation of the idea of Angle of ideas. Sort of light a local lightcone](https://substack-post-media.s3.amazonaws.com/public/images/8d9c119b-9c9a-4a01-bc49-0b98c1b33dc5_1375x728.png) ### Bad reasons and people who use them The most frustrating situations we face aren't the ones where the people we talk to simply don't know and we wouldn't expect them to know more, or even where people make stuff up (doesn't happen so often). Many business conversations end up becoming rhetorical battles. Some companies try to fight it systematically (see [Bridgewater's real-time meeting scoring system](https://www.bastianmoritz.com/blog/bridgewater-associates-believability-weighted-system-for-algorithmic-decision-making/), and [Amazon's memo practice](https://www.sixpagermemo.com/blog/what-is-an-amazon-six-pager)), but before there's much of a company we need to go for the individuals shaping the culture. If someone is overusing techniques from "How to make friends and influence people" in early conversations, then we have a hard time believing they'll hire the right people and instill them with a culture that cares about progress over politics. The counter argument to this is someone who seems overly sales-y to us, might actually be great sales people in certain settings. Since both hiring and fundraising are sales games, besides selling the product, it might be the right person for _that_ company. But authenticity needs to be present. My absolute worst pet peeve is the "you don't need to know that, and so I didn't look into it"-style response, unless it comes with very good reasoning for why we're asking the wrong question (it happens, see dinosaur teeth). Arrogance isn't necessarily bad in a founder; to some degree you need to believe in yourself to an extraordinary amount to start a company; but if it's paired with hiding insecurities and going on the defensive when faced with some limit of their own knowledge, then that rather shows a lack of open-mindedness and curiosity. The feeling we get is "trust me, bro" and when we don't trust, the counter is "Oh, so you don't have conviction". Or inversely, we hear of other investors committing on very short timelines, and the founders saying "They had very high conviction". But on what exactly? ### Conviction capital To us, having high conviction doesn't mean that we think the company has a particularly high probability of success, conviction means that we've done the work to understand what we are underwriting, and importantly what we're not. Conviction is multi-dimensional, at the earliest stages, it's 80%+ dimensions in the founder traits, but what might not be obvious is that every aspect of the company is a reflection of the founder. The founder is the soil and the company is a flower whose petals and roots we want to observe to understand what brought such a thing to life. We need to be the truth-seeking biologists and horticulturists, not florists; we need to be Georg Mendel taking careful note of cause and effect rather than a superficial observer. I'm saying this on an individual company-level, as well as a market level. Only by studying something carefully can we know it. This assumes that we believe in something being knowable, of course. The obvious conclusion to seeing investing as a knowledge-enhancing activity is that in order to do great investments, we need to operate on the fringes of what is known. Some things should be more certain, like the physics governing the fundamental technology, or that someone will want to buy it, should it work (aka. "Big if true"). Other things can be more uncertain. What we don't want, are the companies moving entirely within the scope of human knowledge—the playbook companies (_note: my own em dash)._ We are not interested because those companies will be funded by other pots of money, so our dollars are not really value-adding. ![Author's illustration. The little bump above the dotted line is the textbook "The Cell" and the one above that is all the implications of Maxwell's equations.](https://substack-post-media.s3.amazonaws.com/public/images/1bf444dc-247e-4805-a83c-7c28e9514e34_859x723.png) We build conviction around a founder, and their ability to be the best at building exactly that company they want to build. This conviction-building requires us to understand everything around the company possible, up to some level where the uncertainties on the predictions become too vague to have an impact. This includes technology, market dynamics, other players, funding landscape, etc. It also includes personal judgement on the founder as a person, spending time with them, getting to know them over calls, walks, coffees, and awkward silences. These are all variables in the conviction equation. --- # The Agent Economy: A New Computational Paradigm *Published: 2025-05-06 | Author: Alex Patow | Section: Markets* URL: https://inflection.fund/writings/the-agent-economy-a-new-computational-paradigm ## **A Brief History of Agents** The concept of autonomous agents has deep philosophical roots, beginning with [Alan Turing's 1950 paper on machine intelligence](https://courses.cs.umbc.edu/471/papers/turing.pdf) and [John McCarthy's 1959 "advice taker" concept](http://jmc.stanford.edu/articles/mcc59/mcc59.pdf) that explored computer systems capable of using logic to deduce new information and take actions. The theoretical foundation for modern agent systems emerged in the 1970s with [Carl Hewitt's Actor Model](https://publications.csail.mit.edu/lcs/pubs/pdf/MIT-LCS-TR-194.pdf), which proposed computational entities that could communicate asynchronously. The 1990s saw practical implementations through expert systems and agent communication languages like [Knowledge Query and Manipulation Language](https://cdn.aaai.org/Workshops/1994/WS-94-02/WS94-02-007.pdf) (KQML). Microsoft's Office Assistant ("Clippy") represented an early consumer application attempt, despite its limitations. ![](https://cdn.sanity.io/images/9ycxs2qn/production/3a1ab69642b168700043a42955183297325a4dd2-910x604.png?w=450) *Clippy: The O.G. Agent?* Academic research on multi-agent systems continued through the 1990s-2010s with frameworks like [Swarm](https://www.swarm.org/wiki/Swarm:Documentation_main_page), [Repast](https://repast.github.io/), and [JADE](https://jade.tilab.com/). While these platforms pioneered standardized communication protocols and modular architectures, most applications remained in simulations rather than real-world deployment, due to factors such as market readiness, limited computing resources of the era, and insufficient reasoning capabilities within the agents themselves. The 2000s brought web services and service-oriented architecture with protocols like SOAP and REST, breaking down monolithic applications into independently deployable services and setting the stage for distributed computational models. Voice assistants emerged in the 2010s with Siri (2011), Google Assistant (2016), and Amazon Alexa (2014), offering natural language interfaces but with agency limited to pre-programmed functions. The 2020s marked a fundamental shift with transformer-based large language models, beginning with the “[Attention is All You Need](https://arxiv.org/abs/1706.03762)” paper (2017). [Early experiments](https://arxiv.org/pdf/2304.03442) like [AutoGPT](https://github.com/Significant-Gravitas/AutoGPT) and [BabyAGI](https://github.com/yoheinakajima/babyagi/tree/main) in 2023 demonstrated potential for LLMs as reasoning engines, but were primarily focused on prompt chaining and task management rather than true agent functionality. By 2024, dedicated agent frameworks emerged ([LangGraph](https://www.langchain.com/langgraph), [CrewAI](https://www.crewai.com/), and [AutoGen](https://microsoft.github.io/autogen/stable//index.html)), alongside Anthropic's [Model Context Protocol (MCP)](https://www.anthropic.com/news/model-context-protocol), a proposed standard for managing access to additional context, moving beyond simple prompt chaining toward more sophisticated agent functionality. ### **Beyond Isolated Agents: Communication Standards Emerge (2025)** *Interest around Google’s A2A is already seeing an incredible growth rate (reaching 10k stars on Github in 5 days vs. 12 weeks for MCP, indicating substantial community interest).* The introduction of agent communication protocols like the [Agent-to-Agent (A2A)](https://developers.googleblog.com/en/a2a-a-new-era-of-agent-interoperability/) standard in 2025 represents an important milestone in this evolutionary path. These emerging protocols are facilitating the transition from isolated agents to an integrated agent ecosystem for several reasons: 1. **Industry Collaboration**: The support of over 50 major technology companies for standards like A2A demonstrates widespread industry recognition of the agent paradigm and a collective commitment to interoperability. 2. **Addressing Integration Challenges**: Enterprise software has long been plagued by integration problems. Open communication standards provide standardized ways for autonomous agents to work across organizational boundaries and software systems. 3. **Enabling Composition Without Central Control**: By establishing protocols for agents to discover and leverage each other's capabilities, these standards enable the composition of increasingly complex agent behaviors without requiring centralized coordination. Just as earlier protocols like SOAP and REST enabled the API economy, emerging agent communication standards will likely form the foundation for the agent economy, allowing agents to discover each other, coordinate activities, and work together across organizational boundaries in ways that were previously impossible. ## **Inflections** A question that we ask ourselves at Inflection every day is *"What are the profound shifts in technology, science and markets that are enabling novel behaviors at a vast scale?”* ### **Cultural Inflections** ![](https://cdn.sanity.io/images/9ycxs2qn/production/23194bc7dcd4b43343a88355da46d6499a641ac7-1910x1028.png?w=450) *Public sentiment around AI usage ([Stanford](https://hai.stanford.edu/ai-index/2025-ai-index-report/public-opinion))* Global sentiment data reveals a profound shift in how AI is perceived. From 2022-2024, the Ipsos survey shows most countries increasingly believe AI products and services offer more benefits than drawbacks, with nations like France and Germany seeing 10% increases in positive sentiment. This transition from skepticism to acceptance represents a foundational inflection enabling novel behaviors at scale. This inflection is most evident in enterprise adoption of autonomous AI agents. Johnson & Johnson now employs agents to optimize pharmaceutical solvent switches in drug discovery, while Moody's has developed 35 interconnected agents with distinct "personalities" that can reach different analytical conclusions on complex financial matters ([WSJ Article](https://www.wsj.com/articles/how-are-companies-using-ai-agents-heres-a-look-at-five-early-users-of-the-bots-26f87845)). Similarly, Inflection's own team of agents assist with [deal sourcing](https://svrgn.substack.com/i/147091826/our-first-year) and help with [research](https://svrgn.substack.com/p/introducing-kepler-inflections-home). These examples demonstrate how organizations are evolving from viewing AI as mere tools to treating them as semi-autonomous team members with defined roles and responsibilities, creating human-AI collaborative ecosystems that will restructure knowledge work across industries. ### **Technical Inflections** The technological foundation for the agent economy rests on several key inflections: ![](https://cdn.sanity.io/images/9ycxs2qn/production/ea5893217627abc08cb2ecf15266533653128cdb-1890x1066.png?w=450) *Increasing parameter counts ([Our World in Data](https://ourworldindata.org/grapher/artificial-intelligence-parameter-count))* **1)** We've witnessed an exponential growth in model capability, with systems now exceeding 1 trillion parameters, delivering unprecedented performance in generating text, images, audio, and video. This scale-up has been accompanied by democratization through open source, allowing secure deployment of powerful models virtually anywhere. ![](https://cdn.sanity.io/images/9ycxs2qn/production/eb5434b608a2fcaf1cbb0bdee997e30292add7fd-2337x1305.png?w=450) *The rise of small language models ([Objectbox.io](https://objectbox.io/the-rise-of-small-language-models/))* **2)** Interestingly, 2024 has seen the emergence of a countertrend with "Small Language Models" (SLMs). Despite being larger than nearly all pre-2020 models, these SLMs represent a pivot away from ever-increasing parameter counts. ![](https://cdn.sanity.io/images/9ycxs2qn/production/32374c0584ceaabf0b75fc1f7848df7d3ece2566-1388x772.png?w=450) *Recent small model performance increases ([Artificial Intelligence Index Report](https://hai-production.s3.amazonaws.com/files/hai_ai_index_report_2025.pdf))* This trend toward optimization rather than raw scale enables local-first agents that can run directly on consumer devices. The performance trajectory of these smaller models is impressive: some 8B parameter models now achieve over 60% on challenging benchmarks like Massive Multitask Language Understanding (MMLU), approaching the capabilities previously exclusive to flagship models but with dramatically reduced computational requirements. **3)** The agent paradigm has been further accelerated by specialized architectures like Large Action Models (LAMs). Unlike traditional LLMs optimized for text generation, LAMs are specifically designed for tool use and action execution. This architectural specialization allows LAMs to outperform general-purpose LLMs of comparable size when it comes to completing multi-step tasks, making them ideal for autonomous agent applications. The work by [Salesforce on xLAM](https://www.salesforce.com/blog/xlam-large-action-models/) exemplifies this trend toward purpose-built model architectures for specific agent capabilities. *Inference speeds with specialized ASICs ([Groq](https://groq.com/inference/))* **4)** Complementing these model advancements is the rapid evolution of inference hardware. Companies like [Groq](https://groq.com/) and [Cerebras](https://cerebras.ai/) have developed Application-Specific Integrated Circuits (ASICs) achieving inference speeds up to 3x faster than traditional GPUs. This hardware acceleration fundamentally changes the economics and responsiveness of agent interactions, enabling real-time performance even with complex reasoning chains. These technical inflections: trillion-parameter models, efficient SLMs, specialized action-oriented architectures, and accelerated inference hardware, combined with standardized A2A communication **create the perfect storm of capabilities needed for the agent economy to flourish**. ## **The Agent Technology Stack: Current State and Future Evolution** The following analysis examines five critical components of the agent technology stack, contrasting their current state with projected developments over the next decade, culminating in our vision for 2035. This timeframe allows for the necessary technical advances and market adoption cycles to realize the full potential of these technologies. ### **1. Intent Expression Systems** **Current State**: Today's programming paradigm remains rooted in explicit instruction-giving. Specialized frameworks like [DSPy](https://dspy.ai/) and [LMQL](https://lmql.ai/) provide abstraction layers for LLM integration, while IDE-integrated assistants (such as [Cursor](https://www.cursor.com/), [Windsurf](https://windsurf.com/editor), and [Github Copilot](https://github.com/features/copilot?OCID=AIDcmmb150vbv1_SEM__k_Cj0KCQjw8cHABhC-ARIsAJnY12wfR0Qm7GQCzpm_6oYispyPf4bmX3SPqmPxQ3X9_Yy9hQnm8ahr2PoaAmDbEALw_wcB_k_&gbraid=0AAAAADcJh_tZT1IUwM65UPRvzdfqRY6kC)) supplement traditional programming. Tools such as [Lovable](https://lovable.dev/), [v0](https://v0.dev/), and [Bolt](https://bolt.new/) are pioneering natural language interfaces that enable software creation with minimal coding knowledge; they still primarily serve traditional software development workflows: generating, modifying, and maintaining conventional code bases. **Future Evolution:** By 2035, computation itself will be fundamentally reimagined around human intent rather than explicit programming. These new intent expression systems will transcend traditional software development and become universal interfaces for directing both digital and physical systems: - Scientific research will be guided through high-level experimental design. Systems will understand protocols, manage lab automation, and autonomously adapt procedures based on emerging results. - Creative professionals will express high-level goals while systems handle technical implementation: Architects will specify buildings that automatically conform to structural and zoning requirements. - Manufacturing systems will understand intent throughout the entire process. They will translate high-level product specifications into detailed hardware optimizations, bridging human goals and physical production. - Business processes will be defined through intuitive interfaces. Systems will orchestrate data collection, analysis, and process automation while maintaining compliance and operational constraints. These systems will understand more than syntax or semantics. They will incorporate deep domain expertise, including regulatory requirements, safety constraints, best practices, and complex interdependencies. This will enable true intent-to-execution workflows where human experts can focus entirely on their domain goals instead of computational implementation. The distinction between programmer and domain expert will vanish as these intent expression systems become the primary interface for directing both computational and physical work. This represents a fundamental shift: **computation will adapt to human thought patterns rather than humans adapting to computational thinking.** ### **2. Infrastructure Primitives** **Current State**: Multiple well-capitalized durable compute startups ([Temporal](https://temporal.io/), [Inngest](https://www.inngest.com/), [Modal](https://modal.com/), [Trigger.dev](http://Trigger.dev)) provide essential foundations for agent workloads by addressing fundamental limitations of traditional serverless functions. These platforms enable long-running processes, persistent state management, and workflow continuity. In parallel, specialized services like [OpenRouter](https://openrouter.ai/) have emerged to help developers overcome reliability challenges in LLM routing and model selection. Despite these advances, **current infrastructure solutions remain centralized** and don't adequately address the needs for fluid operation across different computing environments. **Future Evolution**: By 2035, agent infrastructure will evolve into a distributed system with three distinct capabilities: - **Adaptive Execution Location**: Agents will intelligently shift processing between edge devices and cloud resources based on contextual factors (network conditions, privacy requirements, battery life), maintaining state during transitions. - **Resource-Optimized Model Selection**: Infrastructure will dynamically allocate computation across on-device, edge, and cloud resources. This includes using lightweight models locally for privacy-sensitive tasks, edge computing for regional needs, and cloud resources for complex reasoning. The system will optimize performance, efficiency, and responsiveness based on task requirements. - **Decentralized Data Architecture**: Local-first synchronization engines ([AnyType](https://anytype.io/)*, [Zero](https://zero.rocicorp.dev/)) will enable peer-to-peer data transfer between devices without requiring cloud centralization, allowing agents to maintain data consistency even with intermittent connectivity. While realizing this vision will require solving complex challenges around maintaining consistency and reliability across highly distributed systems, the foundational building blocks are already emerging. This distributed infrastructure will support continuous agent operation across diverse computing environments, balancing privacy and security with performance while optimizing resource utilization for each specific task. ### **3. Memory Systems** **Current State**: Today's agents face fundamental memory limitations stemming from limited context windows. Current solutions include [Retrieval Augmented Generation (RAG) systems](https://arxiv.org/pdf/2005.11401) for incorporating external knowledge, and early implementations of hierarchical memory models ([Zep](https://www.getzep.com/), [Mem0](https://mem0.ai/)). These systems typically run on conventional hardware using standard DRAM and high-bandwidth memory (HBM) designed for general GPU/TPU workloads rather than agent-specific memory patterns. The combination of software architectures not optimized for associative recall and off-the-shelf memory hardware restricts agents' long-term coherence and reasoning capabilities. **Future Evolution**: By 2035, specialized hardware-accelerated memory systems will transform agent capabilities through: - **Neuromorphic Memory**: Hardware architectures inspired by the brain's neural structures that enable efficient associative memory and pattern recognition. Unlike traditional memory that requires exact addressing, neuromorphic systems can retrieve information based on similarity and context, enabling more human-like recall. - **In-Memory Computing**: Technology that performs calculations directly within memory units rather than transferring data to a separate processor. This eliminates the bottleneck of moving data between storage and computation, dramatically accelerating similarity searches and pattern matching operations essential for agent memory systems. - **Persistent Memory Technologies**: Advanced storage mediums that maintain state without power while offering speeds approaching RAM. These technologies bridge the gap between volatile memory and permanent storage, preserving agent context across sessions without time-consuming serialization and deserialization processes. Though significant engineering challenges remain in translating these biological inspirations into production-ready systems, the potential benefits justify continued investment and development. These memory advances will enable agents with more human-like memory characteristics: contextual recall, intelligent forgetting, and the ability to form connections between seemingly unrelated concepts, without requiring constant reloading of context or expensive computation to access relevant information. ### **4. Compute** **Current State**: Agent computation today runs across CPUs and specialized AI accelerators like GPUs and TPUs. While these architectures are highly optimized for their primary workloads (CPUs for general computation, AI accelerators for parallel matrix operations), none are specifically designed for the unique demands of agent systems. This leads to inefficiencies as agents frequently need to coordinate across these different compute architectures. **Future Evolution**: By 2035, specialized compute architectures designed specifically for agent workloads will emerge: - **Heterogeneous Compute Orchestration**: Systems that intelligently route different agent tasks to optimal hardware based on their computational profile, seamlessly bridging between AI operations and traditional code execution. - **Custom Inference ASICs**: Application-specific integrated circuits optimized for agent inference workloads running directly on edge and mobile devices, enabling sophisticated agent capabilities without cloud dependency. - **Agent Communication Processors**: Purpose-built silicon optimizing message parsing and routing between agents, dramatically reducing the overhead of inter-agent communication. These specialized computing architectures will substantially reduce the computational costs of agent operations while enabling more sophisticated capabilities, making always-on agents economically viable for a broader range of applications. ### **5. Networking** **Current State**: As described earlier, the Agent-to-Agent (A2A) protocol has established a standardized way for agents to discover capabilities, coordinate activities, and work together across organizational boundaries. While this represents a significant advance in agent interoperability, the current JSON-based approach creates substantial overhead at scale, with verbose text-based protocols consuming excessive network resources during multi-agent communication. **Future Evolution**: By 2035, agent networking will evolve from today's verbose text-based protocols to more efficient communication paradigms optimized for machine-to-machine interactions at scale: - **Optimized Binary Protocols**: Like the evolution from XML-SOAP to JSON-REST to gRPC in web services, agent communication will follow a similar optimization path, dramatically reducing bandwidth consumption and parsing overhead. - **Hardware-Accelerated Authentication**: Specialized circuitry for high-throughput credential verification will enable secure agent interactions across organizational boundaries without the computational overhead of constantly verifying credentials and evaluating complex permission trees. - **Standardized Intent Schemas**: A shared vocabulary of intentions and capabilities will enable more efficient negotiation between agents, allowing them to coordinate complex activities with minimal communication overhead. - **Privacy-Preserving Communication**: Secure information sharing mechanisms will enable agents to communicate effectively without exposing underlying sensitive data, balancing collaboration needs with privacy requirements. These advancements will make agent-to-agent communication not only more efficient but also more secure and trustworthy, enabling truly autonomous collaboration across organizational boundaries. ## **Investment Opportunities** The agent technology stack presents a wealth of investment opportunities across multiple layers. From a deep-tech, pre-seed investment perspective, here’s what excites us the most: ### **Infrastructure Optimization** - Purpose-built infrastructure optimized for agent workloads: will we see an agent-first cloud the same way EC2 got AWS started? ([Agentuity](https://agentuity.com/)) - Agent orchestration systems that distribute tasks optimally across different types of compute - Edge deployment orchestration systems that enable disconnected operation ### **Memory Systems** - Neuromorphic memory architectures for efficient associative recall - In-memory computing solutions eliminating data transfer bottlenecks - Persistent memory technologies preserving agent state without serialization overhead ### **Compute Optimization** - Dynamic model switching frameworks optimizing for specific workloads - Specialized inference chips for edge and mobile devices ([Ubitium](https://ubitium.com/)*) - Energy-efficient processors designed for agent operation patterns ### **Security and Information Sharing** - Technologies enabling secure agent-to-agent communication without exposing underlying data or meta-data ([Tune Insight](https://tuneinsight.com/)*, [Hedy](https://hedycyber.com/)) - Local computation frameworks that keep sensitive data within secure environments ([AnyType](https://anytype.io/)*) - Context-awareness "bubbles" constraining information to certain areas and actions - Privacy-preserving techniques like federated learning, secure enclaves, and homomorphic encryption ([Fabric](https://fabriccryptography.com/)*, [Blyss](https://blyss.dev/)) ### **Agent Discovery and Interoperability** - Agent registries and discovery protocols enabling agents to locate, assess, and invoke one another ([Synergetics](https://synergetics.ai/platform/agentregistry/)) - Intent extraction and summarization techniques enabling more efficient coordination - Distributed trust verification systems allowing secure cross-organizational interactions The most compelling opportunities lie at the intersection of these domains, where technological advances in one area can unlock capabilities across the entire stack. If you’re building something in this space, please let us know! *) Inflection portfolio company ## **Conclusion** The emergence of standardized agent communication protocols marks a fundamental inflection point in computing. Just as the Cambrian explosion was enabled by foundational biological building blocks, the agent economy is being unlocked by key technical advances across the entire stack: from intent expression systems to specialized silicon, from neuromorphic memory to agent-specific compute architectures. This shift from isolated AI systems to interconnected agent networks represents more than an iterative improvement. It signals the emergence of truly autonomous computational systems that can collaborate, reason, and evolve without constant human intervention. The economic and societal impact of this transition will likely exceed previous computing paradigms, as it fundamentally transforms how we express intent to machines and how machines coordinate with each other. The agent technology stack emerging today represents a complete reimagining of our computing infrastructure. Whether you're an investor, builder, or observer, now is the time to engage with the protocols and platforms that will define this next era of computation. --- *A tremendous **thank you** to Lele Cao from Microsoft for critical feedback and ideas. Our gratitude also goes to the many dozens of founders and researches who spent time with the Inflection team discussing the above themes over the last few years.* --- # Physical AI: From Vertical Services to a Self-Replicating Robot Economy *Published: 2025-04-22 | Author: Alexander Lange | Section: Research* URL: https://inflection.fund/writings/physical-ai-vertical-services-self-replicating-robot-economy This post explores **Physical AI**—the fusion of artificial intelligence with robotic embodiment. After a brief **history** of the theme we discuss its underlying **[inflections](https://svrgn.substack.com/p/the-anatomy-of-inflections)**, **bottlenecks**, the rise of the **physical AI stack** as well as potential **opportunities** at the short and long term horizon. # **A brief history** **Physical AI** refers to artificial intelligence embodied in **physical agents** (robots, smart machines) that can **sense and act** in the **physical world​**. Their intelligence arises from continual interaction with their environment via sensors and actuators besides data and algorithms in the cloud. The robotics revolution has long been in the making. Since the **1960s**, we've seen waves of interest in robotics: from Unimate's industrial arms to Rodney Brooks' embodied cognition principles to DARPA's autonomous vehicle challenges. Each era pushed boundaries—but also exposed limitations due to hardware cost, brittle AI, or unreliable sensing. **That is changing now**. The convergence of cheap, powerful edge compute, high-fidelity simulation, algorithmic breakthroughs (think foundation models, general purpose robot intelligence), and macroeconomic shifts (like aging populations, reshoring, labor shortages) enable the next wave of physical AI. We expect this wave to unfold over the next **5-10 years.** Autonomous drone swarms (from bee swarms to tank sized UGVs) for transport, warfare, surveillance all the way to industrial robots (micro factories, additive manufacturing, autonomous science) are underway today. The current stage of intelligent robotics is comparable to the **1950s** of the information age - when machines could only run one specialised algorithm at a time like decrypting messages or steering missiles and before CPUs for general compute were invented in the 1980s. While it took humanity decades to unlock general purpose compute we have reason to expect to achieve general purpose robotics faster thanks to compounding flywheel effects across the stack. _"The material economy could autonomously make and assemble the parts required for more key parts of the material economy — extracting materials, making parts, assembling robots, building entire new factories and power plants, and producing more chips to train AI to control the robots, too. **The result is an industrial base which grows itself, and which can keep growing over many doublings without being bottlenecked by human labour**, visibly transforming the world in the process."_ - Will MacAskill & Fin Moorhouse in Preparing for the Intelligence Explosion, March 2025 How much faster we will get from specialised services to a general purpose robotic economy is very hard to tell. In their [AI 2027 Report](https://ai-2027.com/), the AI Futures Project, a non profit think tank forecasting the future of AI expects **economic growth** to accelerate by about **1.5 orders of magnitude** within a few years **after super intelligence** based on some historical precedence and trends. They also recognise that: > _"Obviously, all of this is hard to predict. It's like asking the inventors of the steam engine to guess how long it takes for a modern car factory to produce its own weight in cars, and also to guess how long it would take until such a factory first exists."_ With that off our chest, let's explore what might be feasible today, what the underlying inflections and bottlenecks are before discussing some opportunities. # **Inflections** A question that we ask ourselves at Inflection every day is "_What are the profound shifts in technology, science and markets that are enabling novel behaviours at a vast scale?_" ## **Technology** The hardware cost curves and AI capabilities have finally aligned. Critical components like LiDAR and edge computing have **plummeted in price** – LiDAR sensors that cost $75,000 in 2015 now cost under $7,500 (90% drop), with some automotive LiDARs targeting <$500. On board processing units (like NVIDIA Jetson) declined from $3,000 to $399. Ubiquitous **connectivity (5G)** and IoT infrastructure further enable distributed robots. We expect the **developments in software to positively affect developments in hardware**, hence the hardware capability curve started to steepen. E.g. **generative design**, where AI models optimize mechanical parts based on goals like strength, weight, or thermal resistance. **Foundation Models** (like GPT-4 or Claude) are beginning to assist in hardware workflows—writing firmware, generating CAD models, or summarizing engineering specs to enable a shorter, cheaper and more iterative hardware loop. Katie Vasquez has been writing up a great summary piece going a bit deeper, [AI and Physics-Based Modeling: A Force Multiplier for the Future of Hardware](https://substack.com/inbox/post/160094595?r=1n41u&utm_medium=ios&triedRedirect=true). ![Over-simplified schematic illustration of how increased software capabilities (generative design tools, simulations, world models etc.) enable faster and cheaper hardware improvements, thereby steepening the "hardware capability" curve over time.](https://substack-post-media.s3.amazonaws.com/public/images/f73b1196-e699-4e32-9edb-c512ee413439_1600x679.png) Meanwhile, **algorithms** have leapt forward from basic perception to deep learning and large "world models" that give machines a form of common sense about physical environments, all the way to neuro inspired approaches to learning. The software toolchain for developing and testing hardware has matured, with high-fidelity simulators and better developer tools lowering the barrier to entry. We will dive into this more under "The Physical AI stack" below. ## **Macro** After decades of outsourcing productivity to the East, the West has a productivity issue and with that a (industrial, military and compute) **sovereignty issue**. Productivity levels are the input for industrial capacity. Industrial capacity is the input for military capacity. ![](https://substack-post-media.s3.amazonaws.com/public/images/f44f1ce4-504d-4d95-9bd2-1b92e2abbd31_1600x793.png) Due to declining populations across western countries (and most of the globe actually), additional productivity levels can only be rooted in technology growth. The "**New Labor Economy**" - manufacturing, logistics, healthcare, agriculture, construction, etc. – represent a **$100+ trillion market** (the "atoms" economy), vastly dwarfing the ~$11T digital economy of software and internet services​. In terms of industrial robotic installations **China** is leading and it controls the global supply of industrial robots with 47% (decreasing though for now). The **US** is leading in physical AI models and [woke up to the risk of missing the boat](https://semianalysis.com/2025/03/11/america-is-missing-the-new-labor-economy-robotics-part-1/). **Europe** needs to catch up on all front but has an exceptionally strong industrial base with deep process knowledge, especially in Germany. ![](https://substack-post-media.s3.amazonaws.com/public/images/eff744be-5442-4c02-9563-bff7984a85f8_1600x1063.png) > _"The impact of this in robotics will be exponential compared to their last strategic industry captures. These will be **robotics systems manufacturing more robotics systems**, and with each unit produced the cost will be driven down continuously and the quality will improve, only strengthening their production flywheel."_ [https://semianalysis.com/2025/03/11/america-is-missing-the-new-labor-economy-robotics-part-1/](https://semianalysis.com/2025/03/11/america-is-missing-the-new-labor-economy-robotics-part-1/) ## **Unblocking the future** Despite the strong tail winds mentioned above there are several non-technical bottlenecks robotics innovators are facing. Overcoming them will be cumbersome and time consuming, especially the ones outside the entrepreneur's sphere of influence. **Cultural resistance**: Robots often face resistance due to job displacement fears and cultural discomfort. Labor unions, policymakers, and conservative industries (like healthcare or education) often push back on automation. Even military adoption of autonomous systems faces deep ethical scrutiny. **Regulation:** liability is often unclear - is it the manufacturer, the operator or the software provider? The complexities faced by Tesla and Wayve around insurance and liabilities are similar in other verticals, depending on exposure and risk profiles. **Standards & Interoperability:** There is no equivalent of USB or TCP/IP for robotics as the industry relies on proprietary SDKs, control stacks or custom comms protocols. Components are rarely interoperable, software needs to be re-written from scratch for each hardware platform. **Distribution**: unlike for mobile apps or books there is no clear distribution layer for robotics in place. Industrial sales cycles are long, trust-intensive, and vertical-specific. **Deployment**: is complex and requires specialised, scarce knowledge around simulations and edge AI operations. There might be many more bottlenecks to be overcome. At the end of the day we believe that the inflections and tail winds will out-weigh the resistance in terms of culture and consequently regulation. All other challenges can and will be solved by entrepreneurs, not bureaucrats. # **Rise of the physical AI stack** **Physical AI Stack –** an overview from the foundational hardware at the bottom to high-level applications at the top: ![](https://substack-post-media.s3.amazonaws.com/public/images/e9e2a78e-8752-43b4-94ce-ea76a1a7e4c1_1432x1220.png) This is how **technology maturity** looks like today (2025) and how it might look like in about a decade from now (2035). Finding proxies for technology maturity and tipping points is hard. [TRL needs a refresh](https://substack.com/home/post/p-157405785) as pointed out very coherently by Nathan Mintz. We tried to synthesize maturity levels from proxy data on **manufacturing readiness, adoption levels, integration maturity, commercialisation** and **open standards adoption**. Open AI's _Deep Research_ _Model_ helped alongside feedback from various researchers and practitioners. To be enjoyed with a huge **pinch of salt**. Here are the definitions: 1. Concept Only: Idea under academic or speculative discussion. 2. Early Research: Limited experiments, no real-world implementation. 3. Prototype in Lab: Exists in testbeds or research demos. 4. Lab Validated: Components tested under realistic, but controlled conditions. 5. Pilot-Ready: Used in small-scale field trials, not broadly reliable. 6. Deployed in Niche Use-Cases: Market exists, but only in verticalized settings. 7. Interoperable + Scalable: Standardized and modular; integrates with existing infrastructure. 8. Ecosystem-Integrated: Supported by dev tools, APIs, and external vendors. 9. Industry-Standard: Robust, widely deployed, with supply chain maturity. 10. Invisible Infrastructure : Ubiquitous, trusted, plug-and-play (e.g. like Wi-Fi, USB). ![](https://substack-post-media.s3.amazonaws.com/public/images/c99aaf2c-0585-42de-ba23-884cb940c083_1600x1490.png) ## **Sensors & Actuators (2025: 7 → 2035: 9)** Mature industrial base: depth cameras (Intel RealSense), LiDAR (Ouster, Hesai), force-torque sensors, and servo motors are widespread and commoditized. Bottlenecks remain in **multi-modal fusion** (sensor latency and integration) and in **cost/power constraints** for mobile platforms. Actuators are robust in industrial robots (ABB, Fanuc), but underpowered or inefficient in other areas. ## **Edge Compute & Control (2025: 6 → 2035: 9)** Edge AI hardware (e.g., Jetson Orin, Google Coral) is powerful and affordable, but thermal, power, and ruggedization (hardening systems to operate reliably under harsh environments, e.g. extreme temperature, vibration, dust etc.) are still issues. Real-time control stacks (e.g., ROS2 + DDS) are improving but remain brittle at scale. Hard real-time industrial controllers are mature, but they lack flexibility for general-purpose autonomy. Balancing compute loads between onboard and cloud remains a challenge with regards to latency and reliability. Analogies can be drawn from Daniel Kahneman's [Thinking Fast and Slow](https://www.goodreads.com/book/show/13062120-thinking-fast-and-slow) where different systems are responsible for different types of decision making, just like in human brains and nervous systems. NVIDIA has been pioneering the [three computer framework](https://www.maginative.com/article/nvidias-bold-bet-on-physical-ai-takes-shape/#:~:text=At%20the%20core%20of%20this,algorithms%20in%20the%20physical%20world) spanning AI training, simulation and on board resources but many questions remain open. ## **AI Algorithms & Autonomy (2025: 5 → 2035: 8)** This layer is the most software centric. It caught a lot of attention and funding recently and spans key functions of perception, localization, decision / planning and learning / adaptation. The below is an attempt to break down a vast category with overlapping problem spaces. **Perception**: Interpreting raw sensor data to understand the environment. Perception tells the machine about the world as it currently is given its sensor experience. For example, computer vision models detect and classify objects from camera images (using CNNs or now even vision transformers), while SLAM (Simultaneous Localization and Mapping) algorithms build 3D maps from lidar or camera data. Modern robots often have a _perception stack_ that **fuses multiple sensors** to output a coherent representation of the world. **Localization**: Determining the robot's own position and orientation in the world. This can involve sensor fusion of IMU (inertial measurement unit), GPS and visual cues. Accurate localization underpins autonomy – whether it's a vacuum cleaner knowing its position in a home, or an autonomous car pinpointing itself on a map. **Decision and Planning**: Given a goal (like "move from A to B" or "pick up that object"), the AI must decide on a sequence of actions. This involves motion planning algorithms, e.g. to find a route for a self driving car as well as higher-level task planning, e.g. deciding in what order to pick items in an order fulfillment task. Low-level **control** algorithms ensure the planned actions are executed by the actuators. **Learning and Adaptation**: What sets "AI" apart is the ability to learn from data and improve. Many Physical AI systems use machine learning models – from vision to control – that are trained on data in simulations. **World models** are an attempt to leverage neural nets to model the physics of the world by simulating environments. They enable accurate predictions about how the world will change given a choice of actions. This is critical for proper learning and reveals a technology gap because most dominant approaches are policy or imitation based, neither of which doing anything that resembles planning. NVIDIA's Cosmos or Google's Genie teams (see a great [overview here](https://rohitbandaru.github.io/blog/World-Models/)) have been working on this problem for years. Another frontier are **embodied foundation models** focused on **general purpose AI for multi task robotics**, e.g. Google's RT-2, [Physical Intelligence's π0](https://www.physicalintelligence.company/blog/openpi)). They output task specific actions (like grasping objects) and serve as base models for downstream applications. In their [Welcome to the Era of Experience](https://storage.googleapis.com/deepmind-media/Era-of-Experience%20/The%20Era%20of%20Experience%20Paper.pdf) paper David Silver and Richard S. Sutton explain how agents will acquire superhuman abilities by **learning predominantly from** _**experience**_ (think reinforcement learning) as opposed to simulations or imitation learning which are limited by data bottlenecks and and don't really make machines _intelligent:_ > _"The era of human data offered an appealing solution. Massive corpuses of human data contain examples of natural language for a huge diversity of tasks. Agents trained on this data achieved a wide range of competencies compared to the more narrow successes of the era of simulation. (...) **However, something was lost in this transition: an agent's ability to self-discover its own knowledge.**_ In the era of experience, agents can inhibit **streams of experience** rather than short snippets of interaction. **Their actions, observations and rewards will be grounded in the environment rather than human dialogue.**" ![Welcome to the era of experience, Fig 1 (highlighted by the author)](https://storage.googleapis.com/deepmind-media/Era-of-Experience%20/The%20Era%20of%20Experience%20Paper.pdf) Following this line of thinking we expect to see entirely new, **experience based models** rise that depend less on static simulation or human interaction. One of them is called **decentralized sensory learning** and is inspired by living nervous systems. Every signal a "sensor cell" receives (think a change in temperature) is a _problem_ to be solved. When the machine acts in a way that makes those signals smaller by moving towards lower temperature for example, it learns. This approach is discussed in [A Foundational Theory for Decentralized Sensory Learning](https://arxiv.org/abs/2503.15130) and implemented by [Intuicell](https://intuicell.com/). Similarly, [Noumenal](https://www.noumenal.ai/) is combining an "experienced physics" approach with a library of behaviours robots can pick from (on edge or loaded from the cloud) to dynamically interact with their environment. Despite all this progress **generalization across form factors/tasks remains very challenging**. Unpredictable environments (mind the slippery floor!) are an unsolved problem. Cobot's Brad Porter wrote a great in depth piece for more context: [The Business of Robotics Foundation Models](https://medium.com/@bp_64302/this-business-of-robotics-foundation-models-cb4bdede1444) ## **DevOps & SimOps (2025: 4 → 2035: 8)** Robotics is still in its "DevOps infancy." Simulation engines like Isaac Gym, MuJoCo, Brax, and Unity Robotics are powerful, but **workflow tooling is fragmented**. No GitHub/Hugging Face-style hubs for simulation versioning, scenario benchmarking, or real-world model validation exist yet. The support layer to handle cloud management, Continuous Integration **analogous to MLOps** is lacking. Until recently the **sim-to-real gap** seemed to be a critical bottleneck to overcome as it has been more art than science: AI models have been first tested in simulation, then deployed on physical robots; real-word data and outcomes have been collected to refine the models and the simulator to improve the next cycle's performance. **sim → train → deploy → learn → improve sim.** This approach could turn into a self-reinforcing flywheel if managed well (think "**physics-as-a-software**" feedback loop). This problem might be solved by novel approaches to learning and adaptation as discussed above. However, sensory accuracy and messy environments (dust, fog, noise) might still be challenging to overcome. ## **Apps & Services (2025: 3 → 2035: 7)** There's no "**robot app store**" or **SDK** ecosystem akin to iOS/Android available yet. Most robots are closed systems or require firmware-level dev work. Modular skill deployment (e.g., "fold laundry," "pick lettuce") is rare. Commercial **APIs** exist (e.g., for drone fleet ops), but only in tightly verticalized domains. **The TLDR is** that the physical AI stack is still in its infancy. Basic hardware components like sensor and edge compute infrastructure made huge leaps over the last few years with AI algorithms and autonomy catching up quickly. Higher levels of abstraction like DevOps / SimOps or Apps & Services are under developed but expected to take off as the lower levels mature. # **Opportunities** Over the last few years physical AI is seeing a narrative shift based on the above inflections, media presence and significant funding rounds especially in the autonomy software category. Yet, there is a delta between insider and outsider perception creating an arbitrage opportunity for venture investors and entrepreneurs alike - at least for those who know what they are looking for. Jordan Nel put it well in [Robotics: a product selection problem](https://jordsnel.substack.com/p/robotics?r=1n41u&utm_medium=ios&triedRedirect=true): > _It seems the consensus outside-robotics take is "capex heavy, hard to scale, small TAM", the consensus inside take is "scaling laws hold, full autonomy, humanoids, El Segundo, lab-spinouts, 1:1 domain-transfer from LLM learnings"._ At Inflection we like to take high convex positions where we assign a higher probability of success to an opportunity than the market consensus while optimising for fat tail outcomes. We also take our **first-check mandate** into account which keeps us disciplined and focused on specific opportunities. ## **Robotic AGI** As discussed above (AI algorithms & autonomy) this is _the_ frontier in physical AI these days. It is dominated by rock star robotics entrepreneurs straight out of the leading labs with backing from large, multi stage funds. Therefore, a hyper competitive opportunity set very few micro funds like us should compete in. Besides strong market signals we currently don't think that this category will be particularly lucrative. Drawing **analogies** to the model wars in **LLM foundation models** we expect to see a very fragmented landscape of hierarchically organised models to power the robotic brains of the future. Depending on the task at hand a combination of different models will be used to optimise for various trade offs. Further, we struggle to imagine how those businesses can create real moats, parallel to Google's "[We have no moat, neither does OpenAI](https://semianalysis.com/2023/05/04/google-we-have-no-moat-and-neither/)". Defensibility could be increased through (1) control of **distribution** (e.g. marketplaces for collections of behaviours) or (2) **vertical integration of robotic AGI** (what NVIDIA seems to go after). We'd be very keen to explore novel approaches bringing the _experiential era_ to life. ## **Vertically integrated services** Based on the constraints discussed above we believe that **vertically integrated specialist companies** have more appeal in the near term. Here are some of the high level patterns we are looking for: **Hair on fire problem in a blue ocean / fragmented market**: The problem set at hand should be urgent and existential for customers with no or only insufficient alternative solutions available. **Critical industries** that are typically fragmented and didn't benefit much from automation over the last decades might hold more attractive opportunities for start-ups. Reducing human exposure to **hazardous or dangerous locations** increases urgency in general. Some examples: - _**Military:**_ [ARK](https://ark-robotics.com/)\* for fleet control in autonomous drone warfare; [NAD](https://www.nordicairdefence.com/)\* for autonomous counter UAV; [Laelaps](https://laelaps.ai/) for autonomous physical security; [Radical](https://www.radicalaero.com/)\* for cell towers and eyes in the stratosphere. - _**Construction**_**:** [Built Robotics](https://www.builtrobotics.com/) for solar construction, [Cosmic](https://www.cosmicrobotics.com/) for critical infra maintenance, [Shantui](https://www.shantui-global.com/product/bulldozer.htm) bulldozers, [MAX](https://www.reddit.com/r/STEW_ScTecEngWorld/comments/1ievokk/the_autonomous_rebartying_bot_by_japans_max/) for rebar tying. [Raise Robotis](https://raiserobotics.ai/) and [Monumental](https://www.monumental.co/) for on site construction. - _**Inspection, Maintenance**_**:** [Koks](https://robotics.koks.com/industries/food-and-food-processing-industry) for silo cleaning; [Nautica](https://nauticatechnologies.com/) for under water inspection. - _**Space**_**:** [Lodestar](https://lodestar.space/)\* for autonomous object manipulation to secure the space domain, [Motive Space Systems](https://motivss.com/) for in space servicing, assembly and manufacturing. - _**Science**_**:** [Trilobio](https://trio.bio) for whole-lab automation in syn bio, [LabLynx](https://lablynx.com) and [Sapio](https://sapiosciences.com) for Laboratory Information Management Systems). **Capability centric**: most customers aren't interested in buying robots (those who do are large industrials) but full blown capabilities. They want to buy a service that can be easily integrated into their operations. E.g. the **Military** has no interest in buying hardware components from X and software from Y to put them together. Instead they need fully fledged, working solutions. **Lower cost**: common pitfalls for robotics companies have been high R&D and up front CAPEX spending. As described in the Inflections section, those cycles started compressing significantly. Off the shelf hardware components, additive manufacturing techniques and intelligent simulation and CAD software drive down costs and increase velocity. **Simple hardware form factors** play a large role because hardware is expensive and complex hardware is _very_ expensive. Humanoids ([Figure](https://www.figure.ai/), Tesla) don't make sense for most use cases (high center of mass, wheels are simpler and cheaper than legs). From a customer's perspective the service should be significantly cheaper than the next best alternative. **Defensibility**: economies of scale can create strong moats but require deep integration and thereby time. Software and or data enabled network effects should be actively pursued as hardware alone will commoditise quickly. Robotics **data** is still a critical bottleneck to be overcome, e.g. environmental (presence sensing for collaborative robots, air quality, temperature, 3D spatial maps to avoid collisions or other accidents) or robot internal (joint angles, velocity, pressure of grippers, maintenance logs, balance etc.). ## **Micro Factories** Microfactories are highly automated, small-to-medium-scale manufacturing facilities designed to produce low volumes of products with high flexibility and efficiency. Unlike traditional factories that rely on mass production and large-scale infrastructure, microfactories use advanced technologies—such as robotics, artificial intelligence (AI), and digital fabrication tools (e.g., 3D printing, CNC machines)—to enable agile, on-demand, and often localized manufacturing. Micro factories can provide an alternative, **horizontal infrastructure to manufacture machines of all kinds** as they allow for adaptive process management and the fast redirection of materials and distribution if needed. Their modular and distributed design increases supply chain resilience and flexibility. This category is still very early in its development. Companies like [Bright Machines](https://www.brightmachines.com/) (end to end automation suite for manufacturing), [Isembard](https://www.isembard.com/) (franchise network for machine shops) or [Arrival](https://arrival.com/card/why-arrival-microfactory) (electric vehicles) are early pioneers. As indicated in the introduction we might see a "industrial explosion" accelerated by embodied AI feedback loops. > _"During World War II the United States and many other countries converted their civilian economies to total war economies. This meant **converting factories that produced cars into factories that produced planes and tanks**, redirecting raw materials from consumer products to military products, and rerouting transportation networks accordingly. (...) Roughly speaking, the plan is to convert existing factories to **mass-produce a variety of robots** (designed by superintelligences to be both better than existing robots and cheaper to produce) which then assist in the construction of newer, more efficient factories and laboratories, which produce larger quantities of more sophisticated robots, which produce even more advanced factories and laboratories, etc. until the combined robot economy spread across all the SEZs is as large as the human economy (and therefore needs to procure its own raw materials, energy, etc.)"_ Timelines remain very hard to predict but we expect fast adoption particularly in **NATO defence applications** where urgency is very high and local procurement laws often require "local manufacturing" of critical components. Think drone micro factories near front lines etc. ## **Picks and Shovels** The complementary approach to vertically integrated services would be horizontal services and components like **sensors, chips, simulators, data platforms or "SimOps"** infrastructure that those doing the deployment will need ([Cogniteam](https://www.cogniteam.com/) for robotics cloud), **Marketplaces for robotic data** sets and **behavioural libraries** might fall into this category. Specialised **edge AI** silicon ([Ubitium](https://www.ubitium.com/)\*) and resource orchetration or next generation **sensing** ([Xavveo](https://www.xavveo.com/) for synthetic aperture radar; [Singular Photonics](https://singularphotonics.com/) and [Pixel Photonics](https://www.pixelphotonics.com/) for single photon detection; [Qurv](https://www.qurv.tech/) for wide spectrum image sensing) etc. As the ecosystem matures, we expect some consolidation: shovel companies partnering with or being acquired by vertically integrated services companies who need their tech, and vice versa (a robot company open-sourcing some tools once they build their own, commoditizing that layer). \*) Inflection portfolio company **If you are working on related companies solving any of the hard problems mapped out in this piece, we'd love to hear from you!** * * * _A big \***thank you**\* to Prof. Jeff Beck from Noumenal, Felix Neubeck from Playfair, Viktor Luthman from Intuicell and Sophia Belser from Laelaps for critical feedback and ideas. My gratitude also goes to the many dozens of founders and researches who spent time with the Inflection team discussing the above themes over the last few years._ --- # Europe’s New Defense *Published: 2025-03-28 | Author: Jonatan Luther-Bergquist | Section: Research* URL: https://inflection.fund/writings/europes-new-defense *In October 2023, I put out New Frontiers in Defense Tech on Substack. It’s gotten rediscovered recently, and we wanted to update it with some events, thoughts and experiences since then. Needless to say, quite a few things have changed. But fundamentally, the conclusion remains the same. It’s just easier for everyone to see. We’re not a defense tech fund, we invest into ++[Sovereign Compute](https://svrgn.substack.com/p/thesis-20-sovereign-computation)++ companies.* *Note: in this document, we may use military terms, which is why we’ve included a primer on talking about defense missions at the very end. Feel free to refer to that whenever you hit an unknown term. It is based on a workshop held at the European Defense Tech Hackathon in Munich Feb 2025 by ++[Colin Macleod*](https://www.linkedin.com/in/colin-macleod-33a80714/)++ ## **Executive Summary** The war in Ukraine is not an isolated geopolitical flashpoint but a continuation of centuries of conflict rooted in the geography and power dynamics of Eastern Europe. Russia’s invasion must be understood in the context of a 500-year history of invasions, occupations, and shifting alliances. This historical perspective underlines the enduring strategic importance of buffer states like Ukraine and highlights the long-term fragility of peace on the European continent. In light of growing instability, Europe can no longer rely on U.S. military guarantees as it did throughout the post–World War II era. Transatlantic cracks, fueled by political uncertainty in Washington and a pivot toward Asia are forcing Europe to re-evaluate its security posture, defense capabilities, and industrial resilience. This post introduces the concept of **New Defense**, which is a generational shift in how Europe must build, fund, and deploy defense capabilities. New Defense refers to a new breed of technology-first, venture-backable defense startups that operate with speed, adaptability, and technical excellence. Unlike legacy defense primes, New Defense companies are agile, modular, and software-centric. Enabled by inflection points in sensor tech, autonomy, distributed manufacturing, and digital infrastructure, these startups represent the future of defense innovation. Structural challenges persist. Europe’s defense ecosystem is fragmented across national lines, leading to inefficient procurement practices, protectionist tendencies, and a limited market for early-stage companies. Legacy defense primes dominate contracting, while startups struggle with burdensome regulations, opaque requirements, and the infamous “valley of death” between prototyping and scaled deployment. Moreover, cultural and political divergences among EU members hinder the formation of a unified defense-industrial strategy. Yet significant change is underway. Defense spending across Europe is rising sharply, with over €1.3 trillion mobilized for defense, infrastructure, and strategic resilience. Public sentiment is shifting in favor of military investment, particularly in Eastern Europe (esp. those with a border to russia). Institutional support for cross-border collaboration and SME access is growing, exemplified by programs like EDIRPA. Simultaneously, battlefield dynamics in Ukraine are demonstrating how cheap, attritable systems (e.g., drones) can outperform legacy systems, provided they are developed and deployed with urgency. Against this backdrop, venture capital has a critical role to play. Traditionally government-financed projects are not able to move at the same speed as private capital paired with elite talent. Startups can move faster, innovate at the edge, and build systems that are good enough, cheap enough, and scalable enough to meet the evolving needs of 21st-century warfare. We identify three types of venture-backable defense companies: (1) vertically integrated system builders that replace primes with faster, modular platforms; (2) horizontal component companies that supply critical subsystems such as AI targeting, GNSS-denied navigation, and propulsion; and (3) deep tech dual-use companies that spin out frontier technologies with civilian and military applications. This thesis is also a call to action: Europe must embrace a defense innovation culture, one that rewards speed over perfection, iteration over inertia, and collaboration over siloed nationalism. Each nation can’t build their own top performing drone company. In the case of Ukraine, agile procurement models and field-led innovation have proven decisive. Similar models must be embraced across Europe. The convergence of geopolitical urgency, industrial potential, and technological progress offers a rare opportunity to reshape European defense. “New Defense” is not only a market trend, it is a strategic imperative for European sovereignty, peace, and prosperity. By building the technological and industrial foundations of a resilient defense ecosystem, Europe can ensure it is not only protected but also prepared to lead in an increasingly contested and multi-polar world. ## **A Brief History**  The war in Ukraine is more than a regional conflict. It is the newest episode of 500 years of wars rooted in geography and ideology. Russia was invaded from the west by the Poles in 1605, by the Swedes in 1707, by the French under Napoleon in 1812, by the Germans under Hitler in 1941. In return, Russia under Peter the Great acquired Latvia and Sweden in the Great Northern War (1700-1721), partitioned Poland three times (between 1772-1795), annexed Finland from Sweden (1809), re-occupied the Baltic states and Eastern Poland from 1949-1991. Ever since many Eastern European and Baltic countries declared independence. Ukraine was perceived as a somewhat neutral buffer country by the Russians until noises were made about further NATO expansion and the fall of the pro-Russian government led by Viktor Yanukovych in 2014 which led to Russia's interference ever since. Taking those last 500 years of history into account, further conflicts across eastern Europe need to be expected. It is in Europe’s interest to prevent such conflicts in the first place and ideally through diplomacy and deterrence. The latter can only be achieved through a solid military and industrial base in Europe.  If anything has become clear since Russia’s full-scale invasion, it is that new tactics, technology, and procurement innovation have changed the course of the war and that the post WW2 world order has ceased to exist. In observing not just this conflict but also the general instability of the geopolitical landscape, as well as our dwindling European economy, we’re increasingly certain that the private sector, specifically nimble startups, is needed. In turn, our governments, not just our defense departments, need to embrace the tools available to everyone else and rethink how procurement is done, on a European level. We’ve seen the earliest, most obvious signs of replacement of primes and disruption happen in the US with SpaceX, [++Anduril](https://research.contrary.com/company/anduril)++, and [++Palantir](https://palantir.com/)++, and more recently with [++Helsing](https://www.thetwentyminutevc.com/torsten-reil)++, ++[Quantum Systems](https://quantum-systems.com/)++, and ++[Auterion](https://auterion.com/)++ in Europe. They have shown that getting direct government contracts can be done by relatively young companies and that there is a willingness from investors, procurement agencies, and founders to collaborate. In this thesis article, we want to explore what the next generation of defense companies will look like, and how we think about investing in this category. We call these companies **New Defense**. ## **Inflections** As a venture firm we apply ++[inflection theory](https://svrgn.substack.com/p/the-anatomy-of-inflections)++ as a framework to identify venture opportunities. We define them as pronounced **shifts in science, technology, culture or (geo) politics that hold the potential to change human behaviour at scale**. Many of them occurred over the last 12 months and set in motion a profound shift in Europe’s self-perception, industry and military endeavours. **US military backing is coming to an end:** For decades, Europe has relied on a stable transatlantic alliance as the backbone of its security. The United States, through NATO, has been the ultimate guarantor of European defense, providing military infrastructure, nuclear deterrence, and rapid-response capabilities. But that world no longer exists. Today, cracks in the Atlantic bridge are widening. U.S. political uncertainty—exemplified by recent discussions of a potential NATO withdrawal—signals that Europe can no longer take American protection for granted. The fading bipartisan support for Ukraine, coupled with Washington’s shift in focus towards China, leaves European leaders facing an uncomfortable reality: we are on our own. **Europe is already at (hybrid) war:** While not officially at war outside of Ukraine, Europe is engaged in a broad conflict with Russia and its allies, China and Iran, fought through economic, cyber, and political means. Hybrid warfare pursues several goals. The destabilisation of political systems (e.g., election interference). The erosion of public trust in institutions (e.g., disinformation). Taking control over critical infrastructure (e.g., energy grids, financial systems) and the expansion of geopolitical influence.Unlike traditional wars, it operates below the threshold of open combat, exploiting societal vulnerabilities and institutional weaknesses.  ++[“[…] it’s crucial to explain to Europeans that this war isn’t solely about Ukraine’s future, but also about Europe’s ability to maintain secure borders and preserve peace across the continent.”](https://foreignpolicy.com/2024/12/20/europe-ukraine-russia-war-peace-trump-nato-peacekeeping-military/)++ - Foreign Policy - Will European Troops Enforce A Cease-Fire In Ukraine? **Europe started mobilising capital:** In recent weeks, in the wake of US retraction and instability in their leadership, Europe immediately reacted. In aggregate, this is over €1.3 trillion, not counting the UK increase in defense budgets. Admittedly, it’s not pure defense; it’s also for infrastructure like roads and railroads, so comparing it to the US defense budget of approximately €822B isn’t entirely fair. - ++**[EU Defense Spending**:](https://ubn.news/the-eu-has-agreed-to-increase-its-defense-budget-e800b-will-be-allocated-for-european-security-and-support-for-ukraine/)++ The European Union committed €800 billion to enhance defense capabilities and support Ukraine. This includes €150 billion in loans and €650 billion in flexible budgetary mechanisms for military spending - ++**[Germany's Infrastructure and Defense Fund](https://www.reuters.com/markets/europe/spending-u-turn-puts-germany-back-europes-driving-seat-2025-03-05/)**++: Germany proposed a €500 billion special fund for infrastructure and defense by reforming its constitutional "debt brake." This fund includes allocations for repairing roads, railways, and other essential infrastructure, with €100 billion earmarked for federal states - ++**[UK Defense spending increase](https://www.bbc.com/news/articles/clyrkkv4gd7o)++:** Zelenskyy visited Keir Starmer in no. 10 after the screaming match in the White House, and soon thereafter the PM announced an increase from 2.3% to 2.5% of UK GDP by 2027. And after that, it will increase to up to 3% according to Keir Starmer. This will largely come from reductions in foreign aid. For reference, the relative spending on defense has decreased by [++several multiples compared to healthcare and social](https://open.spotify.com/episode/5pGWwhwYpIFoF245kvxB61?si=AlSNPtMVTnOyj-E0iM624w)++ spending in the UK. **European resilience innovation culture started shifting:** The old continent became increasingly aware of its vulnerabilities since the invasion of Ukraine. Ever since, venture funds started to loosen the grips of their LPAs to allow for dual use and defence exposure. Entrepreneurs are flocking to industry events and hackathons such as the EDTH ones we are involved with. Public sentiment is shifting towards more support for defence budgets and initiatives depending on the country’s location - the further east, the more support, the further south / west the less. **The nature of warfare is changing rapidly:** Through Ukraine it became obvious how drastically the very nature of warfare is changing. The topic is vast and entire books have been written about it but the tldr; is: - **Strategy:** The cold war era was driven by a bi-polar rivalry which was resolved through nuclear deterrence and high precision warfare. Modern warfare will be multi polar, potentially holding higher global risks due to more complex game theoretical equilbira. Proxy wars have been contained through NATO alliances but today we see hybrid warfare tactics unfold instead. Offensive dominance (nuclear, aircraft carriers) is replaced by defensive deterrence via asymmetric capabilities (e.g. China’s DF-21 carrier killer missile). - **Technology:** The cold war era was driven by nuclear arms, centralised platforms (aircraft carriers) and human controlled weaponry. New defence is focused on hypersonic weapons, decentralised platforms (mesh networks, edge compute) and autonomous systems taking. - **Battlefield dynamics:** Instead of linear fronts in clearly defined areas we have to expect multi domain combat (cyber, space, urban, hybrid). Mass mobilisation will become less relevant than high precision strikes (hypersonics, AI targeting). - **Logistics:** Instead of bureaucratic, centralised procurement and supply chains, more agile acquisition tactics and distributed manufacturing (3D printed parts, expandable drone swarms) can be expected.  Such inflections disrupt the status quo and expose Europe to serious challenges.  ## **Challenges** ### European Military Culture Needs Rapid Innovation Cycles Drones are the obvious change to warfare, to the point that it is ++[being integrated into basic training in Ukraine](https://mod.gov.ua/en/news/the-drone-line-project-has-been-launched-in-the-defence-forces-of-ukraine-to-develop-uav-operating-units)++, and is what everyone mentions when they think about defense tech. Of course, they're very important, but we shouldn't overemphasize their role in a future battle of Europe. However, we can use them as a proxy for innovation capability and adaptability of our armed forces and looking at absolute procurement volumes of all drones in Europe.  In this aspect, non-Ukraine Europe is not looking too rosy. Individual European countries have procured some quad-copter drones, and some long-range ISR drones, but most of them are of the expensive MQ-9 Reaper type, which have been in use since the start of the GWOT. Built by General Atomics, a US company, we might add.  And even if countries have bought a few quad-copters, the types of tactics, techniques, and procedures currently employed in Ukraine are unlikely to have made it into the training of recruits or even doctrine. And this is even though we have known since 2022 of the immense advantage that cheap, attritable one-way drones can have. In summary, the challenge is twofold: capability priorities for procurement are not being updated rapidly enough, and modern technology doesn’t seep into practice within the military rapidly enough.  Other examples beyond drones where we need very rapid iterations from the field to the procurable capabilities and training of recruits are multi-sensor integration in the operational picture, electronic warfare, UGVs, and AI-in-the-OODA-loop. ### Europe Is Not a United States of Europe European countries are unlikely to purchase from a foreign, European company unless they absolutely have to.  This is why we have giants such as Rheinmetall (DE) produce the Leopard 2 tank, but also BAE Systems (UK) produce the Challenger 2 tank. Or Leonardo (IT) and Saab (SE), or Krauss-Maffei (DE) and Nexter (FR) before the merger.  Even though we are within Schengen, if one may speculate, which we may since it’s our article, this is, besides the historic context, partly due to political reasons wanting to generate tax revenues within the country, employ people, etc.; partly due to mistrust in foreign countries, no matter how long the EU has existed, French will distrust the Germans; and finally due to their own strategic concerns about controlling their own firepower. Furthermore, some European countries such as Germany and Poland have put heavy emphasis on the transatlantic link for security, whereas France has been advocating for stronger defence cooperation within the EU. The consequence, however, is detrimental to early-stage, small-scale companies that can’t afford to only scale in one specialty if there’s only one customer you can sell to per country. It further limits the market size and the chunkiness of the revenues. This is also why we have less attractive procurement conditions, i.e., it’s worse for both the procurer and the companies. Tender requirements (financial guarantees, past performance, security clearances) can inadvertently exclude young companies. This has led to what ++[Sifted described](http://sifted.eu)++ as procurement that “tends to favour traditional defence companies” over startups.  Startups can spend years on trials and prototype programs without securing a large production contract, which strains their finances. This is the infamous “valley of death” between a successful pilot and a scaled deployment, where a procurement agency’s inertia or budget constraints can stall a startup’s growth. This is not unique to Europe though. In the US, one can look at the graduation rate of SBIRs to contracts, which is somewhere around 0.5%. This leads to smaller contracts and more overhead for any individual company wishing to sell across multiple nations. They may also have slightly different requirements, leading to potential for exclusion. Another factor worth mentioning is the political incentives for individual procurement agencies to buy from a national company, stimulating local growth rather than European growth. **The only way to overcome this is innovation, through operational and product excellence.** [https://securityconference.org/publikationen/sonderausgaben/defense-sitters/procurement-processes/](https://securityconference.org/publikationen/sonderausgaben/defense-sitters/procurement-processes/) In case of urgency, fast beats perfect, as demonstrated by Ukraine in opening its many paths to procurement for private companies, and speeding up accreditation processes significantly. This also applied to Poland as they bought South Korean tanks, rather than German Leopards, citing velocity as a deciding factor. ## **Opportunities** ### **Technology Can Trump Exquisite Systems** We don’t have to have the capacity to produce 23.2B tons of naval ships per year like China does to be competitive, [++as seen in Ukraine’s naval warfare](https://www.forbes.com/sites/davidaxe/2024/05/21/ukrainian-missiles-are-blowing-up-the-black-sea-fleets-new-missile-corvettes-faster-than-russia-can-build-them/)++.  Raw industrial capacity is necessary for the huge volumes required of attritable systems such as drones, and Europe has more manufacturing capacity than the US (ref needed). Germany leads the way, and collectively, Europe can [++outproduce the US in steel, ships, civilian airplanes, and vehicles](https://foreignpolicy.com/2025/03/07/europe-heavy-industry-trump-us-competition/)++.  This is largely due to a few inflection points in tech, namely the massively decreasing cost in compute (Moore’s law), our ability to prototype and test systems faster (simulations, 3D printing, off-the-shelf components), and finally, general software practices impact on a very traditional engineering-driven industry (e.g., RL agents for targeting mechanisms using cheap optical sensors vs. hard-coding trajectory calculations based off of expensive multi-sensors). In New Defense, we see the following technological inflection points as most critical: - Commoditization of certain types of software (via coding agents) - Continued scaling of other types of software (RL/DL) - Embodied autonomy models increasing in performance - Production and acquisition volumes of UxSs rising rapidly - Hardware SWAP-C continuing to decrease according to Moore’s Law A common success factor in all defense tech startups we’ve seen is to reduce complexity on the hardware side, in favor of more advanced software. This allows us to come closer to the type of SaaS business scaling we saw in the 2010s but for fundamentally hardware-selling businesses. Over-the-air updates and *upgrades* become possible, where previously you had to replace the entire multi-million dollar tank, you can now download an upgrade for free. This works especially well for the air domain, and the close-range attack missions of drones, as the attritable systems need to be extremely cheap, and are mostly software-driven in progress. ### **European Manufacturing Productivity** As mentioned above, Europe has more productive industrial capacity than the US. After all, the industrial revolution happened in Europe. While a lot of the larger industries were relocated to China due to higher labor costs (resulting from the stronger social system), niche, high-value manufacturing such as high-precision machining, automotive components, lithography and photonics, have remained strong in Europe.  The US has a scale advantage here due to the market being larger for chemicals, heavy machinery, and electronics and the fragmentation in Europe still being an issue. Considering outcome quality, Europe is hard to beat. See Airbus in aerospace, BMW in automotive, and ASML for lithography. There’s room for improvement in the digitalization and software-driven platform approaches to manufacturing. [https://stat.unido.org/portal/storage/publication/yearbook/2023/Yearbook_2023_UNIDO_IndustrialStatistics_Yearbook_2023_NAE.pdf](https://stat.unido.org/portal/storage/publication/yearbook/2023/Yearbook_2023_UNIDO_IndustrialStatistics_Yearbook_2023_NAE.pdf) ### **Cross-Border Collaboration** Recently, the EU also approved **EDIRPA** (a €300 million initiative) to support joint procurement among countries for urgent capabilities (++[thesoufancenter.org](http://thesoufancenter.org)++). While these programs are new, the intent is to foster cross-border collaboration and enable smaller tech firms to contribute to multinational defense projects.  **Cross-border procurement** is a big focus: the EU issued recommendations to improve SME access across borders, aiming to “foster a more dynamic defence market” and help companies operate [Europe-wide ++pubaffairsbruxelles.eu](https://pubaffairsbruxelles.eu)++. If implemented effectively, such measures could let a startup developed in, say, France more easily sell to Germany or Poland without starting from scratch in each country. New Defense technology (as imagined by 4o) ## New Defense New Defense is the hope and necessity for European sovereignty. It’s imperative for a new wave of tech companies to build for our ability to defend our territory and allies. New Defense is not giving bullies and dictators any leverage. They are a new generation of technology-first defense companies. New Defense companies will need to exist alongside the old primes, component manufacturers, infrastructure providers, and procurement bodies. They will need to pair “Move fast and break things” with “Slow is smooth. Smooth is fast.” PhDs who can win a bar fight, and who can talk to everyone from ministers to welders. We will need to make the most out of the constraints and opportunities in Europe because Europe hasn’t changed fast traditionally. However, it’s in no way impossible; it’s just playing the startup game on hard mode. ## **Market Landscape** ### The EU NATO Market Is Large Enough Relative to peer competitors, NATO’s defence spending is overall low. This is likely to change over the next few years. That said, NATO countries commonly agree to the [++2% of GDP rule](https://www.nato.int/cps/en/natohq/topics_67655.htm)++, and in 2024, almost everyone is expected to reach that. Across the EU NATO, on average 20% of that 2% is spent on equipment procurement. This is a high-level view from NATO, estimated for 2024. In 2025, we already see a massive increase in GDP-indexed budgets, as mentioned above. [https://www.nato.int/nato_static_fl2014/assets/pdf/2024/6/pdf/240617-def-exp-2024-en.pdf](https://www.nato.int/nato_static_fl2014/assets/pdf/2024/6/pdf/240617-def-exp-2024-en.pdf) The United States is still ahead of NATO Europe and Canada in spending, but the rate of change is higher in Europe and Canada than in the US, indicating we’re catching up. [https://www.nato.int/nato_static_fl2014/assets/pdf/2024/6/pdf/240617-def-exp-2024-en.pdf](https://www.nato.int/nato_static_fl2014/assets/pdf/2024/6/pdf/240617-def-exp-2024-en.pdf) Much of this spending goes to primes, the companies that win contracts directly with the government. To understand what New Defense companies are going to be competing with, we need to understand the structure of primes today. ### **Structure of Defense Prime Contracts** Defense prime contracts are agreements between a government and a primary contractor, often a large defense company, to deliver specific military goods or services. For example, in 2022, the U.S. Department of Defense awarded contracts worth over $400 billion, with major contractors like Lockheed Martin, Boeing, and Raytheon Technologies receiving substantial portions.  The primary contractor, or "prime," is responsible for fulfilling the contract's terms, which may include manufacturing weapons, developing technology, or providing services. Primes often subcontract portions of the work to other companies, known as subcontractors. Anduril was recently awarded a [++$682M C-UAS contract over 10 years](https://www.anduril.com/article/anduril-awarded-10-year-642m-program-of-record-to-deliver-cuas-systems-for-u-s-marine-corps/)++ by the DoD, and has multiple others running already. ### **Economics of Defense Contractors** Prominent defense contractors like BAE Systems, Saab, and Rheinmetall have established themselves as key players in the defense industry due to their ability to build large, exquisite systems, bundling a lot of various technologies. They meet the complex demands of military contracts by subcontracting or purchasing certain areas of expertise. These companies benefit from: 1. **Stable Revenue Streams**: Defense contracts often span multiple years, providing a stable revenue stream for those who can get them. For instance, BAE Systems reported revenues of £21.3 billion in 2022, largely driven by long-term contracts. 2. **Research and Development**: Because of the scale of Primes, they can invest massive amounts into research and labs. This turns semi-academic, and while it allows them to do things startups can’t, it’s not the only way to do innovation. In 2022, Saab invested approximately 25% of its sales in R&D, focusing on advanced defense systems. 3. **Global Presence**: Many defense contractors operate globally, allowing them to tap into various markets. Rheinmetall, for example, has a strong presence in Europe and North America, contributing to its €6.4 billion revenue in 2022. 4. **Strategic Partnerships**: Collaborations with governments and other defense firms enhance their capabilities. BAE Systems, for instance, partners with the UK Ministry of Defence and other international bodies to develop cutting-edge defense solutions. 5. **Regulatory Environment**: The defense industry is heavily regulated, which can create barriers to entry for new competitors, benefiting established firms. This regulatory environment ensures that companies like BAE Systems, Saab, and Rheinmetall maintain a competitive edge. In the US, because of the massive consolidation in the primes in the last 30 years and presumably because of regulatory requirements, [++two-thirds of bids only have one bidder](https://www.economicliberties.us/our-work/courage-to-learn-defense-aerospace/)++! Less than 10% of bids had more than three bidders.  This obviously harms innovation and doesn’t simplify the entrance of startups. But it also means that these contracts are attractive from a strategic perspective, in the eyes of someone who can achieve the same capability with a lot less, it looks like an opportunity.  Given the required capability to protect a harbor, a current prime might offer an exquisite system of large complexity, funding years of R&D off of a large contract. Whereas a New Defense company might take off-the-shelf sensors, and then enhance the processing capabilities using advanced AI capabilities. This is likely cheaper and scales better across multiple harbors, domains, customers, and even nations than an exquisite system. **2024 Estimated Defense Spending Split Across EU NATO** A lot of the spending on defense goes to mundane but essential things, like boots, laptops and food. We’re really interested in the estimated $86B advanced equipment and $22-43B R&D budgets in EU NATO. We might also add that other countries like Ukraine, the UK, Norway, and Switzerland should be considered as part of the market as well, even though they are not in the EU. This would almost double the >$100B market, mostly due to the UK and Ukraine's focus on defense. ### **Agile Procurement Requires Agile Suppliers** We don’t only need to set the budget correctly, but we also need to ensure that it’s spent correctly and effectively. Granted, it’s in our interest as venture capitalists that startups can receive contracts from the government more easily, but it’s a pretty simple life-and-death equation of measuring the time of the feedback cycle from intelligence, observations and casualties in the battlefield due to changing operational requirements, to awarding a contract, and then to delivering on that. As an example, take the US DoD’s and American primes reaction to the IED threat in Afghanistan. They started a mine-resistant ambush-vehicle (MRAPs) program. It was a huge, clear problem at the time. It still took ++[20 months](https://apps.dtic.mil/sti/tr/pdf/ADA529404.pdf)++ between the first formal field request for MRAPs and validated requirements were obtained. This would have allowed Russia to take Kyiv in the first weeks. Instead, let’s look at Ukraine, a country historically mired by corruption, pre- and ++[post-Soviet](https://www.osw.waw.pl/en/publikacje/analyses/2025-01-28/ukraine-defence-procurement-agency-scandal)++, and how they are combatting that while allowing for rapid procurement processes. In Ukraine, individual units and brigades have been able to raise money in the form of donations, and procure items they need without going through much formal procurement procedures. A decentralized approach introduced inefficiencies in terms of pricing and potential for distraction of soldier resources. Ultimately though, it has allowed for a bottoms-up approach to requirements setting and validation, as well as having forced brigades to take control over their own supply chain, integration and even development in a new way. We have been working with the Third Army Corps of the Armed Forces of Ukraine (recently ++[promoted from Brigade](https://mil.in.ua/en/news/ukrainian-third-separate-assault-brigade-upgraded-to-army-corps/)++) and are continuously impressed by how they are able to engage high quality  engineering talent in their operations, modifying and building their own hardware and software while also engaging with many supplier companies. They are just one example from Ukraine. This model is further being formalized and embraced through the ++[DOT-Chain initiative launched in 2025](https://mod.gov.ua/en/news/starting-next-year-the-demand-data-for-drones-electronic-warfare-and-signals-intelligence-equipment-and-ground-robotic-systems-from-the-armed-forces-will-be-gathered-through-dot-chain-defense)++, which should streamline procurement of robotic, EW, and SIGINT systems. Each unit of a brigade declares they want to procure a certain item, which is then added to the DOT-Chain system. DOT, the state entity responsible for non-lethal acquisition, then adds orders. The hope is that this can rein in some of the inefficiencies and potentials for corruption while maintaining the flexibility and demand generation.  ## **What A Venture-Backable Defense Company Looks Like** At Inflection, we largely see three kinds of business models that we think have a shot at being very, very valuable companies, and thus venture-backable., Of course, there will be many other types of companies required for a sovereign Europe, which aren’t likely to become venture-scale. So, this section is specifically for VCs and startups who want to raise money from VCs. It’s also generally for the ultra-ambitious, the high-agency, high-risk taker entrepreneurs who want a shot at ++[building a Monument,](https://substack.com/@jlb39/note/c-101390518)++ not a B2B SaaS AI-slop company. ## Dual-use vs. Defense Companies People generally dislike the term dual-use because it’s associated with “ESG-washing” defense companies and comes from investors who can not invest in defense. But, looking at the business model itself, we at Inflection have nothing at all against dual-use companies, *provided they are prioritizing correctly!*  Many companies building for the military as a first customer could potentially sell to civilian customers as well, especially if the technology is a very deep innovation. In fact, many of the groundbreaking innovations in the past have come from defense R&D (GPS, the Internet, silicon chips). AI for GPS-less navigation can be used in urban environments for delivery UGVs, secure networking and communication solutions can be used for remote areas, underwater surveillance robots can be used for harbor inspections, etc.  That being said, for the sake of talking about the early days of a startup, where the focus is everything, we believe it’s only possible to have both civilian and military customers if you have low additional efforts on the civilian side (and defense customers are the focus customers anyway). For everyone else, we don’t think it’s possible to compete in defense without being 100% committed to getting to know your defense customers' requirements, building your product with military standards in mind, working hard to understand the procurement pipeline, and building deep relationships with defense actors. ## The Types of Defense Companies We’re Interested In Largely, we believe there will be three types of venture-backable defense companies: 1. Vertically integrators 2. Horizontal component manufacturers 3. Deep tech dual-use companies These types of business models are not easy to create, as they have all the requirements of a civilian startup, in addition to powerful incumbents, long sales cycles, chunky revenues, operational and information security, direct relationships to politicians, and more. Given these issues and the fact that the market is rather slow moving, we don’t expect there to be **more than a handful of very valuable ones in Europe in the next 10-15 years**. With the exception of deep tech dual-use companies who can move into civilian markets if necessary. #### Vertical Integrators Our definition of Vertical integrators are very inspired by Packy McCormick’s ++[Not Boring series on Vertical Integrators](https://www.notboring.co/p/vertical-integrators)++. If you haven’t read that, go ahead and do that. In short, a vertically integrated strategy is superior in hardware-driven business fields because of the stronger control over the stack, maintained end-customer contact, and in defense, they will be able to sell directly to the end-user. They compete directly with incumbents, but with cheaper, often better products. We’ve reached the point in military innovation where the existing products aren’t “good enough” anymore. For a tank gunner, not having a counter-drone solution isn’t good enough, it’s deadly. Similarly, the rate at which innovation on the battlefield happens in Ukraine is something which a traditional prime could not keep up with. Hence, to prepare for an imminent active conflict, they’re absolutely not good enough and will need to be replaced.  Vertical Integrators are companies that: 1. Integrate multiple cutting-edge-but-proven technologies. 2. Develop significant in-house capabilities across their stack. 3. Modularize commoditized components while controlling overall system integration. 4. Compete directly with incumbents. 5. Offer products that are better, faster, or cheaper (often all three). In the earlier stages, they may look like pure software companies who then start building their own hardware. See Helsing and Anduril, who both started out in pure software, acquiring the hardware off-the-shelf or from subcontractors, and are then moving on to build ++[their own](https://sifted.eu/articles/helsing-ai-attack-drones-factory-germany)++ ++[factories](https://defensescoop.com/2024/08/08/anduril-arsenal-1-facility-autonomous-weapons/)++. One might argue Helsing is not yet vertically integrated, but in our opinion they are on that path currently. Others go vertical pretty much from the start, like ++[SpaceX](https://spacex.com/)++, ++[Hadrian](https://www.hadrian.co/)++ and ++[UNION](https://union.tech/)++. Vertical integrators will be well positioned to sell systems directly to militaries, and thus getting prime contracts. However, we think they will be more likely to be reactive to the accelerating requirements setting-capability building cycle than current primes who are structured according to very long timelines of procurement with rigorous testing. We expect these companies to produce larger systems or sub-systems of the more “exquisite” level, rather than the attritable or even expendable kind, to begin with. After they’ve disrupted the expensive, high absolute unit margin systems they may move down the line to lower cost attritable ones. **Customers:** Governmental defense organizations **Competitors:** Existing primes, system integrators **Inputs/suppliers:** Data, niche components, raw materials **Moats:** Supply-side economies of scale, can bear regulatory overhead and to some extent brand **Exit path:** IPO or merger with another prime  ++**[Danger zone:](https://youtu.be/kyAn3fSs8_A?si=ZIfb1LQshEDJCXyO)**++ Vertical integrators should watch out once the economics of their target markets become so large and differentiated in the customer base that more and more of the component manufacturers start breaking out into specialization. Examples from the startup scene include [Nordic Air Defence](https://nordicairdefence.com), [Highcat](http://highcat.io), Kraken, Hypersonica #### Horizontal Component Companies There are deeply differentiated technologies that will become crucial to the New Defense sector, and the companies who supply them will have a chance to become very valuable. The types of companies we’ve seen in the past that are representative of this would be ++[Martin-Baker](https://simpleflying.com/2-companies-ejection-seats-us-uk-fighrer-jets/)++, a 75 year old company producing ejection seats. Together with ++[Collins Aerospace](https://www.collinsaerospace.com/)++ (now a subsidiary of Raytheon/RTX) they comprise 100% of the market for ejection seats. So highly consolidated. Today, ejection seats are likely to decrease in importance as we remove the pilots from the aircraft and put them on the ground, until someone decides to start ejecting the cameras, gimbals and compute units from loitering munitions for re-use.  For example, one opportunity for such a technology company  we see is the “minds and brains” of the unmanned systems. Much like Cruise or Waymo have been able to build the brains for autonomous driving, we expect a few companies to emerge to compete for the controlling AI capabilities for unmanned systems, in all domains, and for all missions. They might compete with the internal teams of vertical integrators and primes, at first, but assuming data and versatility is the bottleneck for performance, they are likely to win out.  Other examples of technologies here are:  - UxS fleet management software (e.g., ++[Ark](https://ark-robotics.com/)++) - Control systems (e.g., [Mutable tactics](https://mutabletactics.com/), [Farsight vision](https://farsightvision.com/), [Bavovna](https://bavovna.ai)) - Laser target designators or gimbals (e.g., [M-FLY](https://www.linkedin.com/company/m-flyorg/posts/?feedView=all)) - Energy storage and propulsion systems (e.g., [Greenjets](https://www.greenjets.co.uk/), [Max Joule Tech](https://maxjouletech.com/)) **Customers:** Primes, system integrator, and down the line vertical integrators **Competitors:** Internal teams at primes, other component manufacturers **Inputs/suppliers:** Data, niche components, raw materials **Moats:** Brand (known to be the best component), supply-side economies of scale, and potentially demand-side (network effects by collecting more, better data) **Exit path:** Most likely acquisition by a prime, or a large OEM **Danger zone:** Horizontal component manufacturers should watch out not to become middleware, as this could be commoditized, hence specialized technical differentiation and in-house knowledge is necessary. #### Deep Tech, Dual-Use  Defense spending on research has already spawned many “deep tech” companies and technologies, and one might make the comment that almost any sufficiently important innovation will make it into military usage, whether it’s materials, battery technology, LLMs, or cryptography. As the capabilities of adversaries increase across domains, so do the requirements on the defensive and deterring side. If China develops the possibility to take out satellites and disrupt earth observation intelligence gathering, the US needs to think about how to defend satellites, or create redundant systems (e.g., in the form of stratospheric airplanes). The same goes for energy supply. If we base our power mix on imported gas, we either need to protect those supply lines, shipping or pipe-line-based, or we need alternative power generation methods which are harder to attack (e.g., solar+SMRs). All of these are defense companies in the sense that the military is directly impacted by their existence, can employ them in an offensive scenario, and improve our deterrent effect.  Other examples of relevant technologies include:  - AGI/ASI (e.g., OpenAI, Deepseek labs) - Alternative positioning, navigation and timing (PNT) (e.g., [Tern AI](https://tern.ai), - Rapid-cycle pharma development infrastructure - Chips for edge compute (e.g., [Ubitium](https://ubitium.com)) - Secure communications protocols and infra - Robotics, world-understanding and autonomous navigation (e.g., [Radical Aero](https://radical.aero), [Lodestar Space](https://lodestar.space) , [SE3 labs](https://se3.ai)) **Customers:** Primes, system integrator, and down the line vertical integrators **Competitors:** Well-integrated, legacy tech, other deep tech companies **Inputs/suppliers:** Science, OEMs,  **Moats:** Case-by-case **Exit path:** IPO or acquisition by a competitor **Danger zone:** Optimizing along one vector may not be good enough to replace legacy solutions. The magnitude of “better” needs to be 10x+. If cost-effectiveness isn’t met, even 10x higher performance might be insufficient **Conclusion** At the moment, the defense tech market is certainly going through a boom, with renewed European focus and prioritization normalizing the topic and explicit need for defense tech. The public attention is likely temporary, while the spending and investments are longer-term due to the nature of how budgets are set. As usual in venture markets, something is overlooked for a long time, and then it isn’t. We’re perhaps at the inflection point of defense having decreasing alpha as the theme becomes more consensus. That doesn’t mean that those with networks and insight, and a strong thesis won’t have an advantage. The cycle time for the larger funds to amend their LPAs to allow for defense investing is now approaching the end. We expect more multi-stage funds to enter defense verticals explicitly. This is overall a good thing. It’s inherent to the ++[cycles of technological revolutions and financial capital](https://www.e-elgar.com/shop/gbp/technological-revolutions-and-financial-capital-9781840649222.html)++.  In the case of a cease-fire in Ukraine and Israel/Gaza, we don’t expect the relevance of defense tech as a sector to diminish. It’s one of the oldest industries in the world, after all. The difference is that the world is now sensitized to the fact that there are new threats we need to be able to respond to quickly. And rapid technology development and adoption is the only way to counter those threats. Engineered viruses, cyberattacks on logistics and energy infrastructure, drone swarms or bot-driven election influence campaigns are all of a sudden no longer science fiction but part of our history. For the stability and prosperity of our democracies, we need dedicated deterrent capabilities, but our key belief and reason for optimism around this sector is that they will spawn new civilian-use technologies that better humanity. “++**[Defense tech is peace tech](https://www.breakit.se/artikel/42570/statsministern-defence-tech-ar-peace-tech)**++”. *Thanks to Toby Stone, Benjamin Wolba, Eveline Beer and Larysa Visingeriyeva for comments and suggestions.* # Appendix. Primer On Talking About Defense Missions We’d like to share a hopefully helpful primer on how to speak about military operations. It’s based on a workshop held by Colin Macleod during our Munich Defense Tech Hackathon in February 2025. It was targeted at hackers so that they could more accurately describe what type of mission their solution should be for. In his and Primal Scream’s words: “*Just what is it that you want to do?*”. We can simplify this question into four dimensions: **Domain:** Where are we fighting? **Echelons:** At what distance? **Missions:** What are we trying to do? **Kill Chain**: What parts of the mission? ### **Domains** Domains are traditionally split into five areas: Land , Sea , Air , Space , and Cyber . However, this is not always a helpful classification, as it might be confused with, e.g., the Army, Navy, Air Force, Space Force, and Cyber Force. Those are service branches and not domains per se. The US Navy has fighter jets, the Space Force passes through the atmosphere on its way to space, and all are affected by cyber operations. So, the domains just tell us where a mission is taking place. ### **Echelons** Echelons describe the different areas of military operations. **Rear** area is where: - Forces required to support and sustain forces in the close area operate - Support areas are positioned **Close** area is where: - Decisive operations using maneuver and fire are conducted - Most maneuver forces are positioned **Deep** area is where: - Conditions for future success in close combat are set - Enemy lead command structures from follow-ons are separated - Enemy long-range fires, command and control, and sustainment are disintegrated ### **Missions** Missions are what the military is asked to do. According to NATO standards, these are generally split into three types: 1. Offensive, e.g.,: 1. Attack 2. Destroy 3. Seize 2. Defensive, e.g.,: 1. Defend 2. Delay 3. Contain 3. Others, e.g.,: 1. Reconnaissance 2. Withdraw 3. Retire As an example of a mission, Colin uses this strike mission on Kursk: ### **Stages of the US Military Kill Chain (F2T2EA)** The Kill Chain is not only a famous book by one of the co-founders of Anduril but also a concept that describes the structure of an attack in warfare. The US armed forces use the F2T2EA-version, and NATO has a similar one (F3EAD) 1. **Find**: Identify potential targets through surveillance, reconnaissance, or intelligence gathering. 2. **Fix**: Determine the exact location and coordinates of the target. 3. **Track**: Monitor the target's movement continuously until a decision is made to engage or disengage. 4. **Target**: Select the appropriate weapon or asset to use against the target based on desired effects and available resources. 5. **Engage**: Apply the chosen weapon or asset to the target. 6. **Assess**: Evaluate the effects of the attack, including any intelligence gathered at the location. An example of this would be how an anti-Shahed kill chain works: 1. Find: Learn that a Shahed is incoming using, e.g., acoustic sensors. Here, a battlefield management system would be informed about an incoming threat and its approximate vector, then tasking an air defense unit to fix the targets. 2. Fix: Establish where the Shaheds are located and what the trajectory is more exactly. This would be using a surveillance system such as a radar or electro-optical system. 3. Track: the surveillance systems track the Shaheds through the sky, potentially passing on new information back to the air defense unit through the BMS in order to dispatch a new unit to fix new Shaheds. 4. Target: Decide how to deal with the incoming Shaheds. For example, tasking the operator of interceptors with taking them down, or deciding the Shahed is not a threat to anyone and not do anything. 5. Engage: The Interceptor engages the Shahed kinetically. 6. Assess: Check whether the interceptors destroyed the Shahed. The surveillance system does a battle damage assessment. --- # Introducing Kepler: Inflection's Home for Research *Published: 2025-02-19 | Author: Alex Patow | Section: Building* URL: https://inflection.fund/writings/introducing-kepler-inflection-research-platform ![Kepler's Diamond Mine of Stars (credit: NASA/JPL)](https://substack-post-media.s3.amazonaws.com/public/images/4c0ecad9-0f1d-481c-a1d7-9f5801444f2e_5876x3306.jpeg) > The Kepler space telescope was NASA's first planet-hunting mission, assigned to search a portion of the Milky Way galaxy for Earth-sized planets orbiting stars outside our solar system. > > \- NASA ## Doing Our Homework At [Inflection](https://inflection.xyz/), we really enjoy "getting into the weeds" of emerging ideas and industries. It's part of our firm's DNA, what leads to outsized returns, and what makes the job of investing interesting! You can dive into our published research [here](https://svrgn.substack.com/archive). ### An Overlooked Need Part of our [thesis around Data-Driven VC](https://svrgn.substack.com/p/an-engineering-approach-to-venture) is that building the future of venture capital requires **tooling across the value chain**, not only in sourcing. We've seen how tools such as [Perplexity](https://www.perplexity.ai/) are fantastic for guiding research, and tools such as [Notion](https://notion.so/) are great for writing and collaboration, but through our experience noticed these drawbacks for investment research: - **Limited Collaboration:** AI research assistants like Perplexity and ChatGPT are fundamentally single-user tools. You can share links to chats you've had, but there's no way to "jump-in" and collaborate on exploring a topic together. - **Fragmented Workflow:** The current research process requires constant switching between multiple tools – AI assistants for initial exploration, research papers on arxiv.org for deep dives, and separate tools for synthesizing findings. This context-switching makes it harder to maintain a coherent investment thesis and wastes valuable time. - **Generic Intelligence:** Existing platforms are designed for general users doing broad research. Venture investing requires specialized capabilities – like mapping emerging technology landscapes, tracking shifts in market dynamics, and building deep expertise in specific domains. Getting meaningful insights from these existing tools requires careful prompt engineering or manual synthesis, which slows down the team. We weren't able to find an out-of-the-box tool that addresses these pain-points, so we decided to build something ourselves: Kepler. ## The Kepler Platform ### Mission Goals - **Automate Research Generation:** Streamline the process of collecting, synthesizing, and distributing research on emerging trends, enhancing the fund's ability to identify new opportunities quickly. - **Enhance Proactive Research:** Conduct more comprehensive research earlier in the investment process to reduce time between identification of a potential investment and making an investment decision. - **Expand [Pathfinder's](https://svrgn.substack.com/i/147091826/our-first-year) World View:** Add research on emerging industries identified in Kepler to Pathfinder (our sourcing engine) to increase the volume of high-quality information available to Pathfinder, enabling more informed assessments of companies and founders. ### Key Features #### Explorations !["Multiplayer" Exploration of the Green Hydrogen market](https://substack-post-media.s3.amazonaws.com/public/images/c7d8c367-c9b7-403a-b761-98ece7273809_5486x3280.png) At the heart of Kepler is a chatbot experience powered by agents specialized for VC research. Unlike traditional research tools, Kepler enables true collaboration, allowing investment team members to work simultaneously within the chat interface. The platform leverages existing context from research notes, uploaded documents, scraped websites, agent memories and tools to provide richer insights. Sources are cited so the team can verify the accuracy of information in the agent's responses. Users can "fork" explorations to pursue new directions while maintaining the original thread, and with a single click, synthesize their explorations into structured research notes. This seamless flow from exploration to documentation keeps insights from getting lost in the research process. #### Research Notes ![Editing a Research Note on Organ-on-a-Chip Technology](https://substack-post-media.s3.amazonaws.com/public/images/3dacc3f9-4bc2-422e-868b-fe115c025f76_5486x3280.png) Research Notes in Kepler combines a familiar, rich editing experience (similar to Notion) with the power of an AI assistant that has access to all of Kepler's accumulated knowledge. This creates a fluid experience where the assistant can reference any piece of research, data, or insight the team has previously captured while helping craft new analyses. #### **Additional Sources** Kepler's knowledge base grows in two ways: users can directly upload documents or scrape websites themselves, and the platform automatically captures new sources during exploration. When Kepler's agents cite a source, that source is automatically added to the knowledge base – whether it's a research paper, website, or industry report. This means the platform becomes more intelligent and context-aware with every interaction, building a rich corpus of institutional knowledge that grows both intentionally and organically through the team's research activities. ## Impact Like its namesake – NASA's space telescope that systematically hunted for habitable planets – our Kepler systematically hunts for emerging industries and trends that could power the next generation of transformative companies. While Kepler is just two weeks into its launch, we expect three key areas of impact: - Faster iterations on research through purpose-built agents that understand our firm and investing strategy - A more aligned investment team through collaborative research experiences - Growing corpus of institutional knowledge that compounds with every interaction ## Where We're Heading Kepler represents just the beginning of our journey to revolutionize research tools at Inflection. Our next phase of development will focus on: - Building a comprehensive knowledge graph to connect insights across our firm - Sharing our research through public-facing "Inflection Points" Interested in learning more about Kepler or our other analytics and automation tools? Please reach out to [ap@inflection.xyz](mailto:ap@inflection.xyz). Onwards & upwards Team Inflection --- # Crypto - Quo Vadis *Published: 2024-12-05 | Author: Alexander Lange | Section: Markets* URL: https://inflection.fund/writings/crypto-quo-vadis # **Status Quo** Where are we as an industry? A lot has been written about the [state of crypto](https://a16zcrypto.com/posts/article/state-of-crypto-report-2024/) to provide a bottom up view of what's going on in the industry in terms of adoption, developer momentum, regulation, use cases etc. But how does progress look like relative to other emerging industries? To get a sense for an industry's progress we could think of its inputs and outputs: Total value generated could be defined as a mix of quantitative metrics: market cap + annual revenue + active users for example. Resource input could be defined as a mix of metrics such as venture capital injections and talent working in the industry. The higher the ratio, the more resource efficient is the industry. For simplicity I expressed the market cap numbers in 100x $BN instead of trillions to give them reasonable weight relative to revenues and users. The utility ratio was derived by calculating (Market Cap + Annual Rev + Active Users)/(VC invest + talent). $$UtilityRatio= \cfrac{Value Output}{Resource Input}$$ ![](https://substack-post-media.s3.amazonaws.com/public/images/b0cd616c-90ec-44d2-a8b8-eb82f1466ba8_1626x502.png) This is an extremely over-simplified way of benchmarking progress across industries as it leaves aside any qualitative elements and treats all inputs equal. The utility of a technology is something very subjective though as we've seen when the debates of Bitcoin's usefulness relative to its energy demand unfolded - something we see with AI again now but somehow with much less emotion at play. Hence, I stuck to simple quantitative measures and equal weights. - **Caveat #1:** prices might sharply correct downwards anytime and volatility is higher than in other sectors - **Caveat #2:** A lot of the market cap value is derived through BTCs and ETH's _monetary premium_ that is fundamentally different from market caps derived from profit generating _companies_ - **Caveat #3:** all numbers are averages from the various spectra of data available \[1\]; crypto monthly active user numbers quoted from a16z's crypto report seem high - **Caveat #4**: there is a time lag between resource injections (capital, talent) and value generation expressed in market caps, revenues and users. Resources put to work at t0 might yield outcomes many years into the future. Despite all those flaws it seems that the crypto economy at least in tendency does pretty well in terms of utility value if benchmarked against other nascent tech industries. This goes against public sentiment. Still, there is huge potential for growth if the industry gets the following things right. # **Regulation** The end of operation choke point 2.0 is great but it can only be a starting point. So much injustice has been done to the crypto economy that it leaves the effectiveness of western constitutions and court systems in doubt. From the OFAC sanctions against Tornado Cash's smart contracts as 'property' (which got [successfully appealed](https://www.reuters.com/legal/court-overturns-us-sanctions-against-cryptocurrency-mixer-tornado-cash-2024-11-27/)) to the European Commission's [attempt to undermine online privacy](https://eur-lex.europa.eu/legal-content/EN/TXT/?uri=COM%3A2022%3A209%3AFIN&qid=1652451192472) in the name of child protection all the way to [countless, baseless subpoenas against many major actors](https://fortune.com/crypto/2024/03/20/sec-gary-gensler-ethereum-security-commodity-crypto-foundation/) in the space leading to no tangible outcomes. We live in a time where the rule of law is increasingly captured by political interests. This trend will be amplified by a climate of populism as most citizens in the west would rather like to see their particular interests being represented as opposed to benefiting from a maximally neutral and predictable system of law rooted in western constitutions. > _The two enemies of the people are criminals and government, so let us tie the second down with the chains of the Constitution so the second will not become the legalized version of the first._ Thomas Jefferson That said, it is refreshing to see governments change their minds on critical policy topics from time to time. We've seen Trump and many other Bitcoin sceptics turn around after a while. Be it because of new insights or shifting incentives. It seems that many officials have been perceiving Bitcoin as a threat to the US dollars dominance before they [realised that Bitcoin doesn't compete with fiat money](https://qz.com/bitcoin-gold-us-dollar-federal-reserve-chair-powell-1851713794) but with wealth preserving assets like gold, real estate, art etc. Their fear of losing control over a weaponised financial system might be more rooted in reality but they started seeing the benefits of neutral, open architectures that can still be sufficiently controlled at the interface level, entry and exit points. Therefore, we expect crypto regulations to soften in the west over the next few years. Zooming out a bit, it seems that the west is slowly forming a consensus around the fact that bureaucracy is killing productivity. And productivity is in high demand in a [de-globalising world](https://svrgn.substack.com/p/key-learnings). This is a development we'd [classify as an inflection](https://svrgn.substack.com/p/the-anatomy-of-inflections) - a profound shift in politics unleashing novel behaviours at scale. Imagination for what is to come might be drawn from Thatcherism or [Argentina's Milei pioneering radical government reform with early successes](https://www.economist.com/leaders/2024/11/28/javier-milei-my-contempt-for-the-state-is-infinite) surprising to most. Trump might follow suite in the US while Europe is busy picking up the pieces after the collapse of two of its leading governments in Germany and France. At Inflection we perceive crisis as opportunity and are optimistic that Europe will get its act together like it always did over hundreds of years of wars and revolutions. We expect over-regulated industries such as finance and technology in general to benefit from such developments, including the crypto economy. Reforming Western democracies by cutting through red tape and reducing spending may be painful in the short term but holds vast potential in the long term. # **Incentives** This brings us to incentives. The crypto sphere is great at quoting Charlie Munger with his famous words: > _Show me the incentives and I will show you the outcome._ In tendency, the industry is leaning towards early liquidity through token launches. Arguments underpinning that school of thought are manifold: (1) early liquidity allows retail investors to participate in a network's upside much earlier than in public stock markets. Crypto is more egalitarian and fair than most other industries. (2) Fast liquidity incentivises money and talent to flow towards the best ideas more efficiently. (3) Projects "owned" and controlled by the community of users should be more aligned with the public's interest. etc. In principle I agree with those ideas. Yet, the devil is in the details. As long as private markets exist in crypto we will see founders and early backers enriching themselves on the back of retail investors. Through many cycles we've seen the same play book unfold across ICOs, NFTs all the way to meme coins, just that the actors have changed and the mechanics are iterated on. A group of people is hacking together a product, launches a token, sells vast amounts thanks to short or no vesting before the project is de facto abandoned. This behaviour is zero sum. It slows down progress, caused a lot of the regulatory fall out and public trust issues. As a venture firm we think in decades and act in years but that is not true for many of our peers. Crypto VCs and their short term acting capital bases are fundamentally at odds with long term progress. More than a few times we have seen investors putting pressure on founders to launch a token as fast as possible - ignoring the long term effects on the organisation and its stakeholders. Market structure and a lack of large acquisitions or IPOs amplified such behaviour. **W**e have to do better than this. We encourage innovators to revisit best practices in private venture markets. There is a reason for founder shares re-vesting from financing round to financing round. There is a reason for tag-along and drag-along clauses. There is a reason for liquidation preferences and waterfalls. There is a reason to limit secondary sales for early investors and founders to certain amounts and pricing them based on private market valuations. That reason is long term stakeholder alignment to build meaningful things instead of the financial hedonism we often see in crypto. # **Infrastructure** Crypto infrastructure improved massively over the last 4 years and is likely to continue to do so. E.g., - Transaction Costs: L2 transaction costs reduced by 90%+ relative to main chain - Data storage: EIP-4844 (proto-danksharding) optimised data storage of additional 2MB per block - Interoperability: enhancements through cross chain bridges - UX: account abstraction, embedded wallets, smart accounts It seems that block space overall is much less of an issue going forward and that UX is improving slowly and steadily. Yet, all of this is still emerging and there are more challenges ahead that are less solved for the time being, e.g., - Privacy enhancements through ZKP, TEE and FHE implementations - Decentralisation of block building to maintain sovereignty - Deep integration of crypto rails into traditional financial infrastructure (or the other way around) - Security guardrails and UX for regular users - … Fundamentally, the industry's infrastructure improved by an order of magnitude relative to 4 years ago. It still isn't perfect for all imaginable use cases but each cost reduction and infra improvement unlocks economic feasibility of new use cases. # **Talent** We have seen a net outflow of talent in the industry since late 2021. what is perfectly in line with Chris Dixon's [price-innovation cycle](https://a16z.com/the-crypto-price-innovation-cycle/). The core idea being that rising prices create interest and awareness what leads to new ideas and ultimately talent and new companies in the space. What kept new talent from joining was negative public sentiment coupled with the fear of being punished by regulators but this bottle neck is overcome now. Not just the quantity but also the quality of talent joining the industry will be foundational for future progress. In past cycles we've mainly seen young, curious risk takers joining an infancy stage industry. This is very different today with trillions of dollars in market cap and a broad based institutionalisation of the industry. At Inflection our hope is to see more talent with inter-disciplinary backgrounds join. People who have seen and created greatness in other technology industries to break open current misconceptions and group think in crypto. # **Ideas** Closely related to talent are ideas. They are co-dependent in fact. When the crypto economy took off its actors aligned in many powerful visions of creating state free money for the world, a permission-less and open world computer, a decentralised web, an eternal library, open source Wall Street and markets as a mechanism to predict the future. When I joined the industry a decade ago those ideas felt far more powerful than anything else people have been working on - e-commerce, enterprise SaaS, dating or payment apps for example. These days most teams seem to work on incremental improvements (faster, cheaper, better UX) instead of unlocking fundamentally new behaviours. This was driven by a lack of infrastructure readiness as well as a maturing and institutionalising industry with less radical, less interesting ideas. Fixing MEV, bringing down transaction costs, tokenising existing assets, stable coins, AI powered meme coins feel very different from earlier ambitions. Further, the world changed as we accelerated in the exponential age. Opportunity costs to join the crypto economy exploded from an impact perspective. As a young, renegade genius would you like to contribute to a new financial system or populate Mars, defend western civilisation, make energy / compute / knowledge abundant or engineer life? Through the convergence of emerging technologies and break neck pace increases in AI capabilities the opportunity costs to join crypto for its diluting causes seem higher than ever. That said, at Inflection we believe that crypto technologies are a huge enabler of various new behaviours and that they will play a key role in [defending democratic civilisations](https://svrgn.substack.com/p/crypto-as-defence-tech) and contribute to the [agent economy by providing economic rails and attestation services to distinguish between humans and machines](https://www.augmenthack.xyz/) besides many other things. * * * All of the above ingredients make the recipe for building meaningful tools for humanity. It will lead to the formation of new companies and the launch of ground breaking products very few can dream up today. To explore them collaboratively is what excites us at Inflection. \[1\] Resources: [Nearly 190k people work in crypto, more than 50% located in the west](https://blockworks.co/news/190k-people-work-in-crypto) [New data shows rapidly expanding space workforce](https://payloadspace.com/new-data-shows-a-rapidly-expanding-space-workforce/) [54 New Artificial Intelligence Statistics (Dec 2024)](https://explodingtopics.com/blog/ai-statistics) [Top 500 AI companies](https://growjo.com/industry/AI) [Blockchain Technology Market Size, Trends](https://www.precedenceresearch.com/blockchain-technology-market) [The new space race - outlook and opportunities in 2024](https://www.taylorwessing.com/en/interface/2024/the-space-race/the-new-space-race-outlook-and-opportunities-in-2024?utm_source=chatgpt.com) --- # Ubitium – ubiquitous compute *Published: 2024-11-21 | Author: Jonatan Luther-Bergquist | Section: News* URL: https://inflection.fund/writings/ubitium-ubiquitous-compute If you have ever been stuck in an elevator for more than an hour, you probably had time to think about what goes on behind the metal plates. "Why did it stop!?" you might be asking yourself, stepwise going through your actions before it stopped, then going through the imagined machinery that _normally_ makes the elevator go to the right floor. Odds are, if you're reading this, you got out at some point. Maybe like me, with the help of a whole team of fire fighters and elevator service technicians, or maybe, with the help of the edge-AI and IoT if you were in an elevator conceptualized and built by Hyun Shin Cho. Hyun Shin spent a good portion of his working life not just in the elevator business, but in building compute platforms for making smart buildings and cities a reality by connecting infrastructure with elevators and escalators while at ThyssenKrupp Elevators. But his first job was actually interning for Martin Vorbach - one of the main contributors to FPGA technology. Martin and Peter Weber, a veteran in semiconductors who once led Intel's market expansion into Europe, had started a company together building on the FPGA IP developed by Martin. 20 odd years later the three of them are re-united in a singular mission, to enable a future with compute everywhere. ### Why is my toaster so dumb? _Or, what makes the production and integration of advanced compute so hard?_ ![Credit to Lumafield for scans of everyday things](https://substack-post-media.s3.amazonaws.com/public/images/ea559ad3-965c-41ed-a13f-9ce741495712_758x548.png) The current compute industry relies heavily on heterogeneous architectures that combine multiple specialized processors. If you break open an Airpod, it contains highly specialized circuitry. Just from first principles, it likely has the following internal components: 1. Bluetooth chip: The Apple W1 or H2 chip 2. Audio codec: A Maxim audio codec for audio processing 3. Accelerometers: from manufacturers like Bosch and STMicroelectronics for motion sensing and gesture controls 4. Power management: Low-dropout regulators (LDOs) e.g., from STMicroelectronics 5. Microphones: MEMS microphones are used for voice pickup and noise cancellation 6. Battery: A tiny lithium-ion battery (around 93 milliwatt capacity) per earbud 7. Optical sensors: Light sensors 8. Antennas: For connectivity 9. Speaker driver: A dynamic speaker unit (around 11mm diameter) produces audio output This becomes quite a complex development, sourcing and integration cycle even if you have a massive market and a chokehold on the world's semiconductor industry. Luckily airpods are a huge success for Apple. But the core challenges are the same for any toaster/headphone/car/lamp/etc. company: - **Integration Complexity**: Combining different processors requires complex hardware design and software development efforts. Each processor type has its own architecture, programming model, and toolchains - **Hardware Costs**: Multiple processors increase the Bill of Materials (BOM), manufacturing and supply chain complexity, as well as power consumption. This is especially problematic for edge devices, which often have stringent cost and energy budgets - **Limited Scalability**: As applications evolve, upgrading systems with specialized processors can be inflexible. Adapting to new workloads may require significant redesigns or completely new hardware The above reasons are part of why the IoT revolution didn't pan out as [McKinsey forecasted in 2015](https://www.mckinsey.com/~/media/McKinsey/Industries/Technology%20Media%20and%20Telecommunications/High%20Tech/Our%20Insights/The%20Internet%20of%20Things%20The%20value%20of%20digitizing%20the%20physical%20world/Unlocking_the_potential_of_the_Internet_of_Things_Executive_summary.ashx), why we don't have smart cities yet, and why firefighters had to come break open the elevator for me. Hyun Shin had experienced the reasons for this himself, as he designed edge-AI enabled product platforms using existing silicon. By investing heavily in compute everywhere, he thought he could compensate for the underlying issue of the hardware heterogeneity. Each product was unique, with massive discrepancies in standards and composition. Uniquely tailored PCBs with specialized chips, which require specialized hardware teams. There was enough compute, but integration was slow, connectivity was difficult and most of all, **it was too expensive**. Both in the specialized teams that needed to be built up to deal with firmware and codecs, but also in the actual BOM cost. The business case just doesn't make sense for most products today. ### Rethinking the basics of computing While Hyun Shin was innovating in the smart building space, Martin was brewing new ideas around how to remake the CPU. He was going as far back in his re-design as to the 1960's when an IBM engineer named Robert Tomasulo came up with an algorithm for processing data in parallel. He needed to remove internal dependencies in the processing to speed up execution so that there was less idle time. He did this by introducing register renaming (using placeholder values when the real data isn't available), reservation stations (all the data needed to perform one operation) and a common data bus (a central communication channel for connecting functional units). This worked great for many decades, in fact it, or variants of Tomasulo's Algorithm, are the standard since the 90s in how the x86 and x64 architectures do so-called out-of-order execution in the CPU. In the 20 years of designing coarse grain reconfigurable arrays (CGRAs), Martin had come up with a different way of doing exactly that, out-of-order execution, but better. Not everything in modern computing is out-of-order execution of course. There is a lot of data processing which should ideally be done as a dataflow loop, for example, most AI workloads. Luckily, the architecture Martin had in mind was a dual operation mode that would combine the flexibility of an FPGA with the simple, standardized programming mode of a CPU. This meant that not only could it replace the CPU, it could also remove a couple of the other components necessary for advanced capabilities! ### Lowering TCO and scaling compute Zooming back out, Ubitium is building for a future where we have advanced compute capacity everywhere. Software is pushing towards the same direction, with performant quantized, local models, toolchain abstractions, serverless and distributed compute orchestration. But hardware is growing more and more fragmented, with specialization as a second lever to improve performance and keep Moore's Law alive as we're approaching the limits of process innovation. While others are developing chiplets, 2D materials and more performant ways to do lithography with better mask coatings, etc., we can still do architectural innovation. In Ubitium's case, it's not innovation for the sake of have a different type of chip, but rather a remake of the development and deployment of compute. Their design can be a drop-in replacement for a whole host of other, specialized chips. Most importantly, this lowers the total cost of ownership (TCO), from the sourcing, through software development, to over-the-air-updates. In building on top of the established RISC-V ISA, they can bootstrap from an ecosystem of software, while remaining independent of any vendor lock-in. ![](https://substack-post-media.s3.amazonaws.com/public/images/f19d1f0c-96b8-418c-a7e9-541d67681ee0_1080x1080.jpeg) ### Inflection's investment thesis In summary, the current situation is: - edge compute is 100-1000x more fragmented and heterogenous in components, use cases and development frameworks than the data center compute market - this leads to prohibitively high cost and complexity with the device manufacturers in the supply chain, development cycles and maintenance - this in turn, leads to less functionality and fewer products with advanced capabilities, longer time to market, more specialized development teams and less flexibility on the software side With a re-programmable, universal processor built on RISC-V, we can replace a large number of the specialized components for the same end-device performance or better, at less cost and complexity. Moving the **hardware configuration into software** also keeps the light-cone of what the end-user should experience wide for longer. I.e., design decisions can happen later in the development cycle, with less lock-in! We could not imagine a better team for the job, ranging technical excellence, industry experience, a vision to remake a core component of modern computer architecture, and a deep understanding for the customer. [Martin](https://www.linkedin.com/in/martin-vorbach-95099b1/) started his first semiconductor company during his studies, and invented much of the modern FPGA, holding over 200 patents in the space. [Hyun Shin](https://www.linkedin.com/in/hyunshincho/) served as a UN peacekeeper after studying in Korea, Germany and US, leading Global Manufacturing controlling and then digital transformation at ThyssenKrupp Elevator. [Peter](https://www.linkedin.com/in/peter-w-weber-893348/) (chairman) was one of the first Intel employees in Europe and spent a whole career in global semiconductors. They've already expanded the team with stellar first collaborators but still have a few [vacancies](https://www.ubitium.com/career/). Thanks to the team for trusting us to partner with you on the journey, and to our co-investors Dmitry and Francesco from Runa Capital and Rudi of KBC Focus Fund. We look forward to a compute-abundant future! --- # The Anatomy of Inflections *Published: 2024-10-24 | Author: Alexander Lange | Section: Markets* URL: https://inflection.fund/writings/the-anatomy-of-inflections The venture industry is going through a [transformational phase](https://svrgn.substack.com/p/ventures-consolidation-and-the-case). After two decades of funding incrementalism in a "software is eating the world" paradigm we got used to mis-labelling enterprise SaaS, food delivery services and fintech as venture. We are seeing a mid 20th century science fiction renaissance altering life at a civilisational scale - spanning AI, future of compute, space and computational biology amongst others. Why is that happening now as opposed to the 1990s or 2050s? The answer has to do with scientific, technological, economic and geo political inflections. This post is an attempt to provide a basic framework of how to think about inflections and outlier companies built upon them. It will serve us as the foundation for open sourcing some of our research and tooling going forward. ![As dreamed up by Midjourney - A renaissance of mid 20th century science fiction. Minimalist art. Black, white and grey. high resolution, finest detail. 1970s style.](https://substack-post-media.s3.amazonaws.com/public/images/db369d45-fe1d-448f-bac8-74d4eb8516f5_1456x816.png) # **Inflection theory** In mathematics, an inflection is where a curve changes its direction of bending. We called our firm that way because we believe that human progress is nothing but a continuous flow of inflections in a metaphorical sense. > _A strategic inflection point is a time in the life of a business when its fundamentals are about to change. The change can mean an opportunity to rise to new heights. But it may just as likely signal the beginning of the end._ Andy Grove in his book Only the paranoid survive. There have been other attempts of defining inflection points in a broader context than business strategy. For example in [Pattern Breakers](https://www.goodreads.com/book/show/201757917-pattern-breakers?ac=1&from_search=true&qid=H1nuJ1ylxY&rank=1), Mike Maples Jr. & Peter Ziebelman systematically explore _inflection theory_ as a framework to spot outlier ideas. Their TLDR; is that pattern breaking companies are rooted in (1) one or more inflections which led the founders to develop (2) a unique insight that is then manifested in (3) a product and scaled through (4) a movement. We will only look into 1-3 briefly but focus on inflections only. At Inflection (we created the firm long before the book came out btw) we apply parts of this framework amongst many other things to assess companies and drive our research. ![](https://substack-post-media.s3.amazonaws.com/public/images/745385e7-0ad0-4083-9bfe-74943b04b205_1568x1506.png) ## Inflections Inflections can be described as **pronounced shifts in science, technology, culture or (geo) politics that hold the potential to change human behaviour at scale.** They are hard to see at the time of their arrival and very easy to spot in hindsight. Getting the timing right is existential for the creation of venture scale companies. If you're too early you end up with a science project (left). If you're too late you end up with intense competition and compressed margins (right). Most builders and venture investors are leaning towards the 'too early end of the spectrum'. Ideally, inflections are not very obvious and consensus but nuanced or over-looked by most to build ![from an internal slide - too early (left), too late (right)](https://substack-post-media.s3.amazonaws.com/public/images/38aa3475-1fbd-48a0-8cb7-e228992ae3db_3706x1733.png) **Inflection examples:** ![from an internal slide - inflection categories](https://substack-post-media.s3.amazonaws.com/public/images/546c744f-f603-4517-b999-aae36d99572a_1408x358.png) ## Insights Maples and Ziebelman go on to describe **insights** as a **non obvious or contrarian truth harnessing one or multiple inflections to create a breakthrough product**. They make the case for startups competing on being **different**, not better, faster or cheaper than incumbents. Changing the rules of the game to win. **Applying inflection theory to some companies:** ![from an internal slide](https://substack-post-media.s3.amazonaws.com/public/images/7413c924-5286-40d6-9ecd-5ce310755a23_1428x428.png) As venture investors we might stumble upon some insights occasionally but our primary work is to spot inflections. Ideally the ones that are **over-looked or under-estimated by most**. ## Products Products in this context are **manifestations of inflections and insights**. Typically, in early stage startup land the first products don't find product-market-fit and need various iterations, sometimes pivots. The companies innovating through them usually have been right about the underlying inflections and insights but wrong about the manifestation of them. A famous example is twitch, a company built on inflections around smartphone cameras and internet streaming becoming feasible but manifested in a first-person-infinite reality show called Justin TV. With this basic framework in mind we can now look a little bit deeper into what inflections are and what makes them so tricky to spot for founders and venture investors alike. # **Contextual awareness** The most powerful venture companies are rarely underpinned by just one inflection but by many. In their inspiring book [Why Greatness cannot be planned](https://www.goodreads.com/book/show/25670869-why-greatness-cannot-be-planned?ac=1&from_search=true&qid=3YyDgg0FTL&rank=1) (shout out to Alex Obadia from [Flashbots](https://www.flashbots.net/) for recommending it) AI researchers Kenneth O. Stanley and Joel Lehman think about it this way (strongly paraphrasing based on my flakey memory): all powerful ideas are already out there and want to be discovered. To reveal them we can use a tool called novelty search - an algorithm helping us to explore new frontiers in a serendipitous way. To understand what is novel we need a framework of reference because something can only be novel compared to something else, to something that already exists. Novelty cannot be defined for itself. **The same applies to inflections - they never come alone and can only be spotted and understood in the context of what is today.** Matt Cohler from Benchmark put it this way: > _Our job is not to see the future, it's to see the present very clearly._ A wonderful illustration of the networked nature of inflections is the interactive tech tree from [Calculating Empires](https://calculatingempires.net/). Watch the 5min audio tour, it's absolutely brilliant. (Shout out to Anton from [Anytype](https://anytype.io/) who shared it with me). ![source: https://calculatingempires.net](https://substack-post-media.s3.amazonaws.com/public/images/a6cae86a-3e02-4447-9eeb-27356c615d04_4770x2600.png) # **Time resolution** By nature, inflection's aren't dots. In [Technological Revolutions and Financial Capital: The Dynamics of Bubbles and Golden Ages](https://www.goodreads.com/book/show/60509.Technological_Revolutions_and_Financial_Capital?from_search=true&from_srp=true&qid=0rzpjYR0tn&rank=1) Carlota Perez established a model for disruptive innovations using S-Curves (left). Somewhat related [Gartner](https://en.wikipedia.org/wiki/Gartner_hype_cycle) (right) established a hype cycle model capturing similar ideas. ![](https://substack-post-media.s3.amazonaws.com/public/images/a30f5b5c-4cd1-480e-9dc1-1bdb92a0d680_1800x610.png) Those frameworks work pretty well to think through technology cycles. The right way to apply them would be to factor in **time resolution.** That is a tricky undertaking. ![](https://substack-post-media.s3.amazonaws.com/public/images/f08934e8-6f5b-489f-b481-203e8ff061e8_2456x3261.png) > _Let's observe the facts: Humans took more than a hundred thousand Earth years to progress from the Hunter-Gatherer Age to the Agricultural Age. To get from the Agricultural Age to the Industrial Age took a few thousand Earth years. But to go from the Industrial Age to the Atomic Age took only two hundred Earth years. Thereafter, in only a few Earth decades, they entered the Information Age. This civilisation possesses the terrifying ability to accelerate their progress. On Trisolaris, of the more than two hundred civilisations, including our own, none has ever experienced such accelerating development._ Trisolarian scientist in Liu Cixin's 3 Body Problem One might be right seeing inflections at the horizon early but that doesn't help to build a company with a 2-5 year time horizon to find product market fit if the inflection is too early in its s-curve journey. Often, crucial building blocks are still missing to unlock the next breakthrough. We couldn't have built the first computers without electricity or light bulbs (none of which have been invented with computers in mind though). Many novel ideas explored by previous innovators have been **directionally right but wrong in terms of timing**: Henry Ford envisioned [energy money](https://cointelegraph.com/news/100-years-ago-henry-ford-proposed-energy-currency-to-replace-gold) to "substitute gold with units of energy to end wars" 100 years before Bitcoin was invented. In the [1830s the first electric cars](https://en.wikipedia.org/wiki/History_of_the_electric_vehicle) were invented but never took off because of lacking battery density and price pressure from alternative design. The list goes on. A simple analogy to think of is a **microscope with variable resolution** to look at inflections of different scale, moving on vastly different time scales. The charts are AI generated and very rule-of thumb to illustrate the idea, simplicity over nuance. ![1/ lowest resolution](https://substack-post-media.s3.amazonaws.com/public/images/a4b4300b-c9e3-4e5d-bbba-a18ede7c3484_1997x1101.png) ![2/ medium resolution](https://substack-post-media.s3.amazonaws.com/public/images/7fe267fb-77b5-4e9f-aecd-c194c9e839d5_1697x1101.png) ![3/ high resolution](https://substack-post-media.s3.amazonaws.com/public/images/bc1ccda4-b19e-429c-a4a3-41271e218f61_1697x1101.png) From here we could dive deeper to explore the inflections underpinning the development of the mobile web like e.g. broad band connections, small and powerful enough chips and batteries to power them amongst many other things. You get the idea. # **Conclusion** To start (or fund) a company rooted in non-obvious inflections and unique, contrarian insights requires the right intuition with regards to time resolution and contextual awareness. Figuring them out by studying history and observing what is today will be time and energy well spent. We are looking forward to open sourcing some of our work on that soon and cannot wait to collaborate on it with you. --- # Venture’s Consolidation and the Case for Emerging Managers *Published: 2024-10-15 | Author: Alexander Lange | Section: Markets* URL: https://inflection.fund/writings/venture-consolidation-case-for-emerging-managers The [venture apocalypse](https://medium.com/angularventures/the-venture-apocalypse-316c37e241ef) is upon us. [Only two funds raised 44%](https://pitchbook.com/news/articles/general-catalyst-a16z-funds-lp-commitments) of the venture money in the US in 2024. The cash amounts deployed against emerging managers (firms with less than three funds) relative to more established firms [dropped from 19% in the 2010 to 2020 periode to 16% and falling](https://files.pitchbook.com/website/files/pdf/Q2_2024_PitchBook_Analyst_Note_Establishing_a_Case_for_Emerging_Managers.pdf#page=1). Lux Capital's predictions of [30-50% of emerging managers ceasing to exist](https://twitter.com/wolfejosh/status/1827822860733128897) over the next couple of years are over-shadowed only by claims of an outright [extinction of venture capital](https://newsletter.equal.vc/p/the-extinction-of-venture-capital). This post intends sheds some light on the long term trends in venture capital market concentration before making the case for emerging managers to uphold innovation. We are talking our own book here but we are right and here is why: ![](https://substack-post-media.s3.amazonaws.com/public/images/68ad5181-fc96-43e6-8044-43dc50dd659f_1920x1281.jpeg) ## What are we talking about? The above discussion refers to two related phenomena: market **consolidation** is the process in which companies in a particular industry merge or exit the market. It occurs in times of upheaval - recessions, conflicts, crisis - or as a result of over-heated markets. Market **concentration** on the other hand refers to the extent to which a small number of firms dominate a market with reduced competition. Concentration can be the result of a consolidation period and is expressed through the [HHI (Herfindahl-Hirschman Index) or concentration ratios](https://web-archive.oecd.org/temp/2022-03-02/473895-market-concentration.htm) . The tricky part is to define what exactly constitutes a market in terms of products / services and geographies. Considerations around substitutability of a product or service and price elasticity are key factors, see [SSNIP test](https://en.wikipedia.org/wiki/Small_but_significant_and_non-transitory_increase_in_price) for details. The top 5 concentration ratio in terms of 'funds raised' (US markets only) didn't change much since 2016 and has been hovering around the 30% mark what would be comparatively low if benchmarked against [other industrie](https://watermark.silverchair.com/rfz007.pdf?token=AQECAHi208BE49Ooan9kkhW_Ercy7Dm3ZL_9Cf3qfKAc485ysgAAA0owggNGBgkqhkiG9w0BBwagggM3MIIDMwIBADCCAywGCSqGSIb3DQEHATAeBglghkgBZQMEAS4wEQQM6GIrIQKJRPG9j2pxAgEQgIIC_dvpPHorWrDrLwNtah_p-TfZm0ZHW-6fjbokuSdCXGaJb9ErS3fJboCrQ9TM8yxSaxioM3HLP8Ifamax57KU4RwccuvGLOXpVevfOsRF64uVKF79feRvtqWmjKPUJ5iqvrFbd8C2oySHRsBzoHUgw08ngtHT0V-A0tP6jhAQAwwgOwJW2l0QoTAOSyodWnHG6wxNpymqKcEaRo0t9gphxdNSFNv0bUnOII2hrJ6C8Dzdvv1GDH4potr-SemT6z49M9K07-lscIbis3HSa0OTEhP2AYZbmsNrAyhAGVjMofMIELOKa5vg050MQWpWgrWBjjjYarpa5a1ME8iCCyMmgVrMwV4QgRO2d2xcNCHgcSSJJ5dud95efePIG5g7OP972m9Kr-s_3PWcuNsTYEgRm_6lMDF_1ZOr5TN1PCoCy1iLk-R6ZJH40Er8VIjLByv4xfqYJYvV6F8I8TAzBZBadQ23aVqH0u7Za4zxfXwdOFIEz9h2gyuL7aGHzICzteTY-vfeg75eoqwYhM4shOexk8CgB4oh3mJQdwoB7RvTIRmuJB6C1PQwxwd6_3NaDXufr7cpAtAmF74I0HuAdhmrZJCaEkckq1aUpvYZoaR4-P7e6F25h-1cyT0IXZEbjxK3kYvmho5eo4uBOwriv24-_fkbWrU9rlovdCPszeXV0P2FYc0m4AJ-79NPEtcVh70JZqXJPW4KHxyHjf94bf8XQkVMT3kEuFaz-oiRsALH8MebyPmmdyKvbsPEMkuRxeEsGQfSRh13rVxvZKPsuhMpPkmQTJdbkjkgPJ76yBkeqrjldGQGQZtcOZlceuZda7JsoC4cwhYfu_5F_ekpx9NJpULrwnFVk6tT1WcTDslWOqynrBhSLqqEyShM5X0tE5nYXnRCRCGWNkda-1-bK36B4LSNCCLM29RdJu-DgXIwPL6j_580eAhkeXX4xADzcx9n4aSY7fEVbFt75NhGLWee4OrTrqtEJ_9p2sOBjWNtcQBQIN9hXqkKcEXaM_SM8g)s. Be aware that the data below references the entire US venture industry - geographical and stage splits would paint a different situation. Especially early stage funding sources are much more fragmented than later stage ones. ![CR5 data US; Based on data from NVCA yearbooks.](https://substack-post-media.s3.amazonaws.com/public/images/d0e5ec85-b846-450b-9694-8d05e39de843_2126x1274.png) Other metrics helping us to grasp the level of concentration in venture might be the relative (1) and absolute (2) share of emerging manager funds raised over time as well as fund count. ![](https://substack-post-media.s3.amazonaws.com/public/images/9711fee7-6bc8-4b3e-8d4b-670775ed0b09_1990x1290.png) ![](https://substack-post-media.s3.amazonaws.com/public/images/d60f72ff-fc8b-481e-a178-ab4357d13700_2496x930.png) Now that we see the bigger picture it becomes clear that emerging managers won't disappear tomorrow and that many of the click-bait headlines and claims above are unrealistic. Yet, the multi decade trend towards market concentration is clear. ## Drivers and effects of market concentration What are the main drivers leading to high concentration and how does concentration affect markets? **Drivers:** As discussed above phases of consolidation are an input for the level of market concentration - often driven by externalities (crisis, recessions etc.). Looking at much larger time windows though mere phases of consolidation aren't a sufficient explanation. There must be more powerful forces at play, such as: **(1) economies of scale:** Large firms can produce at lower average costs, making it difficult for smaller competitors to match their pricing and efficiency, **(2) (data) network effects:** where the value of a product increases with more users a few firms tend to dominate (social media, AI, telecommunication), **(3) capital or asset accumulation:** concentrating through M&A by merging with competitors, **(4) distribution power:** Control over key distribution channels can limit competitors' ability to reach consumers effectively, increasing concentration and **(5) globalisation and market integration:** Larger firms benefit from globalization by spreading fixed costs across international markets, which can lead to increased concentration in domestic markets. Those drivers are hitting industries across the board: There ar[e sever](https://obamawhitehouse.archives.gov/sites/default/files/page/files/20160502_competition_issue_brief_updated_cea.pdf)a[l studi](https://academic.oup.com/qje/article/135/2/645/5721266)es indicating increased concentration across a large number of industries in the US - pharmaceuticals, agriculture, airlines, personal compute devices and the list goes on. Concentration increased i[n over 75 perce](https://academic.oup.com/rof/article/23/4/697/5477414?login=true)nt of U.S. industries since the late 1990s. **Effects:** Let's take a look at the effects of market concentration. Following [consensus](https://www.whitehouse.gov/cea/written-materials/2021/07/09/the-importance-of-competition-for-the-american-economy/) views the outcomes are (1) less innovation, (2) worse products at higher costs and (3) wage suppression, (4) regulatory capture and (5) higher barriers to entry for newcomers. Applied to the venture capital industry many stakeholder would be affected should concentration levels rise to the extreme: **1/ Many emerging managers** will be washed out the ecosystem through this wave of consolidation. And that is necessary as well as healthy. The last few years in venture have been outliers - excessive volumes have been raised and deployed in record time without producing much value. Seeing the days of 'easy VC' and the 'venture playbook' [come to an end](https://svrgn.substack.com/p/key-learnings) means a fresh start, a [clearing the air](https://techcrunch.com/2024/05/12/emerging-fund-venture-capital/) event for the industry. A net positive. However, if the pendulum swings too far and the consolidation trend accelerates mid and long term the effects will be more negative: **2/ Founders** would have less choice in picking their capital partners. If growth and pre IPO stages were dominated by only a handful of funds competition would fade in every category as large funds typically don't take conflicting positions (conflicts of interest). **3/ Limited partners** would be limited in their choices as access to large funds is typically more restricted than access to emerging managers. Further, larger funds come with very different risk-reward ratios and expected return multiples. Since the early 1990s emerging fund managers have mostly been outperforming established funds - by IRR and TVPI metrics. ![](https://substack-post-media.s3.amazonaws.com/public/images/198ef3c8-0a23-4d35-a62f-2d43ee734a15_1668x682.png) The median return of emerging fund portfolios slightly exceeded that of their established counterparts, while the top-quartile figure delivered by emerging managers significantly outperformed. The pattern speaks to the ability of emerging managers to generate outsized returns while the spectrum of returns is wider and therefore performance is less predictable. ![](https://substack-post-media.s3.amazonaws.com/public/images/a0c1e99c-7da6-4cdf-91f1-9dfa3587ca14_1668x616.png) Within both the established and emerging manager groups, specialists outperform their generalist peers: ![](https://substack-post-media.s3.amazonaws.com/public/images/5a1045c4-8eb3-495e-abfc-3d0d29a32508_1672x634.png) **4/ The public and civilisation at large** have an interest in seeing the most powerful ideas being funded to push the boundaries of the human race. Philanthropists and public money sources are attracted by the venture industry for good reason. Most breakthrough ideas are controversial, hard to spot and [non-consensus](https://svrgn.substack.com/p/diversity?utm_source=publication-search). In a highly concentrated venture ecosystem dominated by large funds and their 15 people investment committees such seemingly alien and weird ideas wouldn't be funded - reversion to the mean. It would slow down human progress. ## How to limit unhealthy market concentration? Fighting concentration in any industry is like fighting gravity. However, there are a few things stakeholders could undertake to limit concentration down the line. > Show me the incentives and I will show you the outcome. _Charlie Munger_ **1/ Regulation:** The debate on wether or not there should be political correctives to uphold more competitive markets is as old as the industrial age. The first anti trust laws root back to the Sherman Act in 1890. The libertarian camp argues that if concentration drives down innovation and increases prices the incentives for newcomers are compounding to unlock a disruptive new way of doing things. Just wait for the invisible hand to do its magic. Famous examples are Tesla in automotive, AirBnB in hospitality, Anduril in defence, SpaceX in space etc. This narrative is flawed because it does not lead to less concentration but more once the new kids on the block built up their next generation monopolies. The nanny state camp argues for governmental interference through tools such as merger regulation, restriction of agreements that distort competition or collusive behaviours and the prevention of dominant market position abuse. The current tools of anti trust regulators. This narrative is idealistic as none of such tools effectively prevented hyper concentration in any market (as the data and reports above indicate) due to an endless game of whac-a-mole driven by comparatively new phenomena like cross market network effects in social media for example (nobody saw it coming when Meta was allowed to acquire Instagram and What's App?). **2/ LP decision making:** If LPs came to their senses and overcame the irrational bias of preferring \[big name fund\] exposure over emerging manager exposure because of lowered risk perception the ecosystem would be in a better place. **3/ Founder decision making:** If founders were to run a diligence process on the fund they are about to collaborate with they might find that emerging funds - especially in early stages - have much better incentive alignment and as they need to prove themselves are more likely to put down the necessary leg work. **4/ Emerging manager collaboration:** collectively, emerging managers could start to collaborating closer to achieve some economies of scale. Shared back office infrastructure, software or services might be a starting point to improve unit economics in venture. Especially complementary micro funds might significantly benefit from closer collaboration from sharing research, dealflow or LP access amongst many other things. But this is stuff for another post. --- # Fabric *Published: 2024-08-21 | Author: Alexander Lange | Section: News* URL: https://inflection.fund/writings/fabric-cryptography-vpu-investment Without privacy or trust we cannot have a functioning civil society. Yet, those two ideals seem to be at odds with one another. If all information is transparent we have perfect trust but sacrifice privacy. The result is a society prone to mass manipulation, psy ops and suppression. If all information is private we have perfect privacy but can barely learn from one another or trust each other. FTX and Wirecard kept their books private. How can this dilemma be resolved? What if we could balance privacy and trust in more nuanced ways through technology? ![An artistic reference to moon math as illustrated by Alex Patow.](https://substack-post-media.s3.amazonaws.com/public/images/2587343c-21c0-49ff-a9db-04b68e8ec03f_1867x1050.png) Moon math to the rescue. Over the last few decades numerous cryptographic techniques have been developed to solve the conflict between privacy and trust. Zero Knowledge Proofs (ZKP) empower us to prove a fact without revealing its underlying information. [Fully Homomorphic Encryption (FHE)](https://spectrum.ieee.org/homomorphic-encryption) enables us to run computations on encrypted data sets. With multi party computation (MPC) we can jointly compute a function over their inputs while keeping those inputs private (MPC). The blockchain industry and militaries have been early pioneers of implementing those techniques at rather small scales. On a high level broader commercialisation requires (1) orders of magnitude higher performance in terms of proof generations, verifications and computations, (2) drastically lower costs per operation and (3) standards and tooling for seamless implementations. How do we get there? Enter [Fabric Cryptopgraphy](https://www.fabriccryptography.com/), a recent addition to the Inflection portfolio which just announced its [$33M Series A round](https://www.forbes.com/sites/davidprosser/2024/08/19/how-fabric-cryptography-plans-to-become-the-nvidia-of-privacy-tech/). We co-led their $6M Seed round in 2023 and participated in their Series A. In 2022 the Fabric founders came to the conclusion that cryptographic algorithms will need a native processing unit, just like AI has GPUs / TPUs or general purpose computers have CPUs. This realisation drove their development of verifiable processing units (VPU) - a general purpose processor for cryptography that comes with its own instruction set and software stack. An NVIDIA for cryptography that can accelerate ZKPs, FHE, MPC and many other classes of crypto 2.0 algorithms down the road. ![](https://substack-post-media.s3.amazonaws.com/public/images/5a929ae7-ea46-478e-818b-9357a69ae9de_1804x1030.png) ## Our Thesis We couldn't be more excited about the problem set Fabric is going after. Our digital-physical fabric is prone to numerous attacks - spanning impersonation, data theft and deep fakes. Such problems are gigantic in terms of societal and economic impact for themselves. In the context of military conflicts decided by autonomous weaponry they become existential. Think of AI as offence and [cryptography as defence](https://svrgn.substack.com/p/crypto-as-defence-tech). The rat race for programmable cryptography just started and will accelerate over the next decade. What distinguishes Fabric from others is the design of their architecture. An FPGA centric approach would have come with the highest degree of programmability and flexibility but lower performance. A fixed function ASIC would have optimised for one specific algorithm only to yield maximum performance but compromising flexibility. Instead, Fabric develops a general purpose crypto chip that optimises for performance _and_ flexibility with reasonable trade offs. We consider this a dominant strategy as crypto software is ever evolving and hyper dynamic. Algorithms that work in a certain way today might operate very differently in three months from now. To borrow an analogy from biology: in predictable environments with fixed rules and marginal changes specialisation wins due to pattern recognition. In messy environments with changing rules the most adaptive, generalist organisms win. New cryptography certainly falls into the second category for the foreseeable future. Their incredible team of world class semiconductor and cryptography talent is growing. If you consider a collaboration feel free to reach out or [check out open positions here](https://jobs.inflection.xyz/jobs?filter=eyJvcmdhbml6YXRpb24uaWQiOlsiMTEwNDU3OSJdfQ%3D%3D). --- # Data-Driven Investing as a Micro Fund *Published: 2024-08-02 | Author: Alex Patow | Section: Building* URL: https://inflection.fund/writings/data-driven-investing-as-a-micro-fund ![Topographic Martian Glaciers and "Brain Terrain"](https://substack-post-media.s3.amazonaws.com/public/images/9bc32918-4d2a-4648-ad9d-d1ca9e05d933_3507x2480.png) Last week, Inflection wrote a guest article on [Dr. Andre Retterath's](https://www.linkedin.com/in/andreretterath/) Data-Driven VC newsletter: [ ![](https://bucketeer-e05bbc84-baa3-437e-9518-adb32be77984.s3.amazonaws.com/public/images/b1d96ba8-7106-47ab-aa26-f9dd08922611_428x428.png)Data-driven VC 6 Lessons From Leveraging Data & AI as a Micro VC Fund With $50M AUM 👋 Hi, I'm Andre and welcome to my newsletter Data-driven VC which is all about becoming a better investor with data & AI. Every Tuesday, I publish "Insights" to digest the most relevant startup research & reports. Every Thursday, I publish "Essays" that cover hands-on insights about data-driven innovation & AI in VC, and every… Read more 2 years ago · 15 likes · Andre Retterath and Alex Patow ](https://www.newsletter.datadrivenvc.io/p/6-lessons-from-leveraging-data-and?utm_source=substack&utm_campaign=post_embed&utm_medium=web) Andre's done a tremendous job of building a community of investors, engineers, data scientists and generally curious people who are interested in learning more about leveraging data in VC (over 25,000 strong 🤯). You can discover more content on [his site](https://www.datadrivenvc.io/). As he eludes to in his intro, we didn't see eye-to-eye on one of his recent posts, ["Can We Fully Automate Startup Investing?"](https://www.newsletter.datadrivenvc.io/p/can-we-fully-automate-startup-investing?publication_id=1083211&post_id=146018062&triggerShare=true&isFreemail=true&r=2xlbtb&triedRedirect=true). This led to a great conversation and, ultimately, this article. Since publishing, there's been an outpouring of people interested in starting something at their firm or swapping "war stories" from their own data-driven initiatives. It's incredibly energizing. Below is the article in full. If you're curious for more, please [reach out](mailto:ap@inflection.xyz). * * * # Data-Driven Investing as a Micro Fund At Inflection, we firmly believe that the [future of VC is "data-driven"](https://svrgn.substack.com/p/an-engineering-approach-to-venture), especially for funds our size. We also equally value the human element of venture capital. There's no substitute for deep connections with founders, other VCs, and our own intuition. We believe this combination of data and human insight is where outsized returns become possible. Being data-driven, for us, means building tools that augment the entire team's work, not just that of investors. We focus on enhancing all aspects of our operations, not only sourcing investments. As the sole "data person" in our six-member team, my role involves wearing many hats, but ultimately, my mission is to scale the efficiency and impact of our team through strategic use of data and software. ## Second Wave of Data-Driven VC We're now entering our second wave of "data-driven VC". Early pioneers in this space (such as EQT, Earlybird, Moonfire, amongst others) had to build substantial infrastructure from scratch: - Complex pipelines to wrangle spotty and immature data sources - Training and hosting custom models - Creating platforms due to a lack of VC-specific tools (particularly CRM systems) While there's still value in prioritizing in-house development, it requires a large team to implement effectively (something that's not realistic for funds of our size). Fortunately, the industry has evolved significantly since the first wave: - Proliferation of (relatively) stable, cheap, and reliable Large Language Models (LLMs) - Decreasing cost and increasing availability of out-of-the-box data and tools, specifically targeted at our industry (we're big fans of [Specter](https://tryspecter.com/), [Gravity](https://launchgravity.com/), [People Data Labs](https://peopledatalabs.com/), and [Attio](https://attio.com/) for our needs) - Increasing engineering efficiency through [AI-assisted coding](https://github.com/features/copilot), [data transformation](https://getdbt.com/), and [cloud deployment tools](https://modal.com/) We expect these trends to accelerate, with advancing technologies enabling even smaller funds to compete effectively, shifting the key differentiator from data infrastructure to insight interpretation and action. To funds looking to become data-driven: the time is now. ## Our First Year Inflection's initial blog post on ["An Engineering Approach to Venture Capital"](https://svrgn.substack.com/p/an-engineering-approach-to-venture) was published just over a year ago. I started with the firm in November of last year. Since then, we've built some really cool stuff, we've also done some really "boring" work: - Launched our internal deal sourcing tool, Pathfinder - Utilizes internal and external data, LLM agents, and a graph database for sourcing opportunities in our sector - Automatically populates our CRM with sourced opportunities for the investment team's review - It's currently evaluating roughly 500 new companies and 4000 founders per week, a task that would be impossible for us without tooling - Brought in best-in-class tools for deal tracking and portfolio hiring - Reduced the time for creating valuation memos for our audit process by 80% - Achieved through LLM Agents and Retrieval-Augmented Generation (RAG) - Acquired signal and people data sources to support our projects - Refreshed our website - Streamlined our toolset by eliminating unused tools Like I said, not everything is glamorous, but it's important for being a holistic, data-driven fund. ![](https://substack-post-media.s3.amazonaws.com/public/images/9991aa53-b25c-4961-ade8-f42383ac5c68_649x499.png) ## Lessons Learned Although we are early on in our journey, we learned some lessons along the way: ### #1 Start Small, as Small Projects Snowball While building our data-driven sourcing tool (Pathfinder), we were asked by the operations team to help automate the creation of this year's valuation memos. These letters summarize the investments we've made, their current valuation, and the rationale for their valuations for the audit process. As there's one memo for each investment, it's typically a time consuming task to complete manually. By doing a small, one-week project we were able to validate Pathfinder's tool stack (specifically LLM Agents) in a controlled environment. These learnings significantly accelerated our progress on sourcing. Importantly, it also helped achieve buy-in from our operations team on the vision of a data-driven fund. We've found that starting with small, manageable projects can lead to unexpected benefits and growth for longer-term strategic projects. ### #2 Take Calculated Risks The first iteration of Pathfinder relied on a linear process where LLM agents gathered the information needed to evaluate potential investment opportunities as they arrived in the system. While functional, this approach had limitations in providing context and integrating existing results, often leading to confusing outcomes. Recognizing these limitations, we decided to take a calculated risk by rethinking our data architecture and transitioning to a graph database. We believed that modeling our database to mirror the entities and relationships investors use in assessing companies would lead to better results from our LLM agents. This required substantial rework, but the results have been significantly more aligned with our investment thesis and early-stage opportunities. Adopting a GraphRAG approach has enhanced the agents' contextual understanding (see [Neo4j's writeup on GraphRAG](https://neo4j.com/blog/graphrag-manifesto/), and stay tuned for more on [our blog](https://svrgn.substack.com/)). This decision exemplifies how embracing calculated risks, when guided by clear objectives, can drive innovation and unlock new value. ### #3 Embrace Nimbleness Some tools we build may only provide 3-6 months of "alpha" before external tools emerge and commoditize that part of our platform. When more mature, off-the-shelf solutions become available, we transition from in-house development to leveraging these external solutions. This approach allows us to redirect our focus and resources toward identifying and developing the next source of alpha. ### #4 Secure Buy-in from the Team Team support is crucial when implementing data-driven initiatives at a VC fund. Achieving reliable, scalable, and impactful results takes time, and not all efforts will succeed. Organizational buy-in enables experimentation and grants the independence needed for effective decision-making. At Inflection, the team had already explored various Python scripts and external tools to enhance workflows, from podcast transcription to research summarization and web scraping. Recognizing the potential leverage of a technical role in the post-LLM era, the investment team outlined initial ideas [on their blog](https://svrgn.substack.com/p/an-engineering-approach-to-venture). To retain this alignment and maintain momentum, we implemented a structured approach: 1. A week-long kick-off session to set principles and direction 2. Weekly product calls and bi-monthly in-person work days 3. Bi-annual presentations on innovations and roadmaps This structure fosters tight feedback loops and a culture of innovation. ### #5 Find a Product-Engineering Hybrid Look for someone who enjoys both product and engineering. In smaller funds, this role often falls to one person. Much of the job involves managing product, not just engineering in isolation. It's essential to have someone who can excel (and enjoys working!) in both areas to ensure what's being built matches the fund's needs. This should be a full-time role for someone passionate about building, not a part-time job for an investor. ### #6 Develop Guiding Principles, Not a Backlog Start by establishing guiding principles for the first 12 months, rather than specific projects. Let the projects emerge over time. For us, this meant creating an _architectural philosophy_ that addressed questions like how to approach "buy vs. build", how to balance "speed of iteration vs. quality of output", and whether to focus on "platforms" or "projects". This approach gives us the comfort to get started without locking us into specific projects or implementation details, allowing for flexibility as we learn and grow. ## What's Next for Inflection - Continuously improving our sourcing tool, Pathfinder - Adding new data sources (research papers, social media, etc.) and giving our agents ability to expand the graph through their "intuition" - Crafting a "magical" experience for the investment team to input and extract information from our knowledge graph (we love the work USV has done with their ["Librarian" tool](https://avc.xyz/the-usv-librarian)) - Generate novel ideas for companies and areas of exploration based off of research and our thesis - Launching tools for our founders that go beyond just hiring talent - Actively growing the community of data-driven VCs by: - Open sourcing our work - Collaborating with our community and new data providers - Presenting at industry events Hopefully this serves as a helpful guide for smaller funds looking to be "data-driven". We believe that by embracing these practices, micro funds can punch above their weight and compete effectively in today's fast-paced and competitive venture landscape. --- # European dynamism through sovereign compute in space *Published: 2024-07-26 | Author: Jonatan Luther-Bergquist | Section: Research* URL: https://inflection.fund/writings/european-dynamism-sovereign-compute-space We recently passed 10,000 satellites in orbit, most of which belong to one company, and most of which are in LEO, serving the earth economy. In the next decades, we see this diversifying, and the economics of space systems changing massively. [The total capacity is estimated to be 12.6 million satellites in LEO.](https://semanticscholar.org/paper/943c2d59684681d9dd9f1dd4e496a1519647223b) This will be one of the biggest shifts in exploration in human history. ![LAURENT GRASSO, STUDIES INTO THE PAST, Oil on wood, silver leaf, 24 x 21 cm | 9 7/16 x 8 1/4 inch, Unique](https://substack-post-media.s3.amazonaws.com/public/images/aedc9809-0665-4042-8004-829693fbb240_1198x1414.png) With this future in mind, [Neil Buchanan](https://www.linkedin.com/in/neil-buchanan/) joined Entrepreneur First, thinking he'd be surrounded by talented people tackling the most important and difficult problems of all. Instead he found mostly talented founders, mostly interested in B2B SaaS. Luckily one other person stood out - the industrial design and roboticist [Thomas Santini](https://www.linkedin.com/in/thomas-santini/). Neil's experience building Hyperloop and Thomas's product design and engineering skills made them a great match. They first joined forces to explore first 3D printing in space, but realized quickly that the terrestrial hardware wasn't suited for space, and the economics didn't yet make sense. So they looked at what would be required for assembling something constructed in space, and identified the robotic arm as a first step, and they formed [Lodestar Space](http://lodestarspace.com). ![Neil and Thomas preparing for take-off](https://substack-post-media.s3.amazonaws.com/public/images/6ff615b3-dbc3-4a7c-9274-df36381af262_5383x3959.jpeg) Lodestar isn't just about reaching for (and grabbing) the stars—it's about protecting our space infrastructure and scaling operations to allow new businesses to be built in space. In their own words, they're creating autonomous dexterity tools for the space domain. This requires them to innovate across the full stack of sovereign compute, cutting edge robotics systems, perception tools and some truly novel thinking around autonomous systems. Just out of stealth yesterday, we can finally announce their [$2.5M pre-seed round](https://techcrunch.com/2024/07/25/lodestars-robotic-arm-is-an-orbital-first-responder-for-satellites-in-need/) led by Inflection and Lunar ventures. Lodestar has achieved significant commercial traction, with tens of millions of dollars in LOIs, an ESA as well as UK Space Agency contracts secured. Also, a few months ago they tested their arm in a parabolic flight simulating microgravity (and we have the flight patch to prove it). ![Robotic arm tracking a (soft) cube in low gravity.](https://substack-post-media.s3.amazonaws.com/public/images/09fc3296-c399-49dd-a844-0c66e79c8ef8_2562x1440.png) But lets dig into the thesis for autonomous space operations, and why robotic arms are important. ## Operating in space > "We're bringing the ability to inspect, protect, and repair high value assets in space in a way that's truly scalable for the first time" says Thomas Santini, co-founder and CTO. "As the industry moves towards proliferated capability, building a physical layer for autonomous interaction is going to be a bedrock technology that others will be enabled by." ### Automating space operations Space is, without a doubt, one of the most hostile environments we've ever encountered as a species. Rapid temperature fluctuations spanning hundreds of degrees Celsius, the absence of atmospheric pressure, and extreme lighting conditions make it a challenge for mechanical engineers. Add to this the lack of gravity, and you have an environment that's decidedly unfriendly to human habitation. A big question of traveling to other planets is how to avoid humans going blind… There is some research working on genetically engineering humans to optimize for space environments, and at least [one startup](https://x.com/UrsaBio) working on space medicine. But the fundamental issues remain. Because of these very human considerations, every hour of astronaut labor has an estimated cost of [$130,000](https://www.theregister.com/2021/05/10/iss_private_prices/) aboard the ISS. Astronauts need a very specific temperature range, correct gas mixture of the air, food, exercise, entertainment and more. Every requirement adds cost. From the guiding principle of the three "D's" of automation: "Dull, Dirty or Dangerous", space operations fulfill them all (and comes at high cost). Automation in space isn't per se new, the first Canadarm's was delivered in 1981 to NASA and the program [cost about $600M](https://cc-space.weebly.com/canadarm.html). However, there are a few **inflection points** that lead us to believe that robots designed cheaply, from the ground up for space will thrive today: - Better compute per watt available for space applications, leads to: - Less issues with thermal management in vacuum - Improved intelligence on the edge - Lower launch weight required - Higher frequency of launches and lower cost per kg to orbit, leads to: - Faster rate of iteration for testing in space - Fewer restrictions on SWaP for the arm - More power availability through larger solar power arrays - Improved machine intelligence, through: - Better independent reasoning and task planning capabilities - Advanced computer vision and image recognition - Improvements in general control systems All of these factors combined are part of the technological "Why now", but in fact the commercial reasons are contributing at least as much. The fact that the customer base is growing by [11.5% CAGR and the space economy is already at $514B](https://www.coherentmi.com/de/industry-reports/global-space-economy-market) helps of course. ### Securing space infrastructure The challenges or operating in space don't stop at environmental factors. As we expand our presence in space, we face potential threats from other actors. Space forces worldwide are gearing up to secure the vital infrastructure we rely on—from navigation and communication systems to satellite observation platforms and strategic assets like ICBMs. The risk of disruption, whether from space-based or Earth-based attacks, is a pressing concern that demands our attention. Large parts of Ukraine and much of Eastern Europe continues to be [GPS-denied or spoofed](https://spoofing.skai-data-services.com/), causing disruptions to commercial flights and transports. China is planning [anti-satellite weapons](https://www.washingtontimes.com/news/2024/jan/4/china-space-warfare-includes-cyberattacks-jamming-/), and [Shijan 17 and 21](https://www.bloomberg.com/news/articles/2024-01-25/china-russia-disguise-threats-posed-by-satellites-us-space-force-says) are suspected as satellites to be used against adversaries. From a strategic perspective, securing our physical presence in space requires a three-pronged approach: 1\. Mobility as a defense mechanism: The ability to maneuver with high delta-v capabilities. 2\. Dynamic security: Implementing mobile security systems around space vehicles. 3\. Vigilant inspection: Developing the capacity to thoroughly examine other space assets to determine their nature and intentions. In addition to this, there are of course cyber warfare and software-led disruptions to protect against. Hence, being able to operate freely in space is a strategic imperative. Just check out the dedicated [Space-section at Breaking Defense](https://breakingdefense.com/category/space/). ### Making of the space economy Looking a bit more optimistically towards the future, how can we not just protect ourselves, but how do space capabilities move humanity as a whole, forwards? To truly enable the coming space age, we need to shift our paradigm from human-driven operations to intelligence-driven systems. This transition is not just a matter of preference; it's necessary due to the unique challenges of space exploration and utilization. While numerous companies are making strides in getting assets into space, we are still in the early stages of this frontier. It's reminiscent of the early days of the computer age—we've invented the transistors, but we're still working on building the operating systems and software that will truly harness their potential. Launch costs have been the primary focus for the last 20 years with SpaceX pioneering re-usable designs and how to do hardware engineering. After that, the next frontier in making space accessible is reducing operating costs and generating new value from space. This can be achieved through: 1\. Minimizing human involvement: This not only reduces the need for complex life support systems but also limits the costs associated with training and transporting human operators 2\. Extending the lifespan of space assets: Through refueling, servicing, and repair capabilities we keep existing satellites alive for longer. See [Northrop Grumman's life-extension mission of the IntelSat901 in 2020, costing $13M per year for an extension of 5 years](https://spaceflightnow.com/2020/04/21/aging-intelsat-satellite-resumes-operations-after-docking-of-robotic-servicer/) 3\. In-space manufacturing: From assembling large-scale structures like solar power plants to operating space laboratories for unique research opportunities and at a later stage full-scale production facilities ![DALL-E semiconductor facility in space being serviced by robotic arms](https://substack-post-media.s3.amazonaws.com/public/images/a69529fd-b298-44a2-9b57-ff2f54d17ee8_1024x1792.webp) ### The space in between The potential applications are vast and exciting. We're looking at a future where we could be growing crystalline structures more effectively in zero-gravity environments for drugs, mining asteroids for rare minerals, constructing habitats on the Moon, and preparing for extended operations on Mars. And all of them need autonomous, secure, physical operations, which is what Lodestar provides. But until sci-fi becomes reality, Lodestar has a pretty wide selection of "hair-on-fire" level problems they can solve for customers. Just adding an arm with a manipulator adds tremendous value today already. We couldn't be happier to collaborate with Neil, Thomas, Jakub and the rest of the team. They've already built a high-performing culture of ownership and are insanely mission-driven. The new offices have a big 🇬🇧 hanging on the wall. If you liked what you read here, check out their [open positions](https://job-boards.eu.greenhouse.io/lodestarspace). Especially the [Senior Space Systems engineer](https://job-boards.eu.greenhouse.io/lodestarspace/jobs/4367429101) is a high prio hire right now. To a sovereign compute future in space! --- # European Acceleration *Published: 2024-06-26 | Author: Alexander Lange | Section: Markets* URL: https://inflection.fund/writings/european-acceleration This post is based on a key note I gave at [Tech Open Air Berlin](https://event.toa.media/) this summer. ![](https://substack-post-media.s3.amazonaws.com/public/images/f4c034d8-a4fa-416a-a048-2d79b2df150b_3840x2160.jpeg) A few weeks ago [Emmanuel Macron was interviewed by The Economist](https://www.economist.com/leaders/2024/05/02/emmanuel-macrons-urgent-message-for-europe) and made some bold statements about the state of European civilisation and its risk of collapse. > '_A civilisation can die. Things can fall apart very quickly._'. While I cannot judge the likelihood of that happening on any time frame I do think that we are living through the EU's most fragile period since 1989. I also believe that technological progress, aka the acceleration of innovation, is one of the few valid answers to this problem. # Reasons To understand **why** I think so, we can ask ourselves: "What are the fundamental drivers of civilisational collapse?". No matter if we look at Rome or other ancient societies - be it the Maya, Inka or Chinese dynasties - the patterns rhyme: ![](https://substack-post-media.s3.amazonaws.com/public/images/b790d102-eec6-468f-aad3-a1829227a7d1_1786x952.png) **1/** Civilisations grow in complexity by creating bureaucracies with the goal of orchestrating labour and resources to the benefit of its population. **2/** Return on complexity (aka bureaucracy) is extremely high in the beginning, since the efforts are focused mostly on critical infrastructure such as security, nutrition, healthcare or education. **3/** Over time, the **bureaucracy is expanding** into less critical areas of life like such as banning plastic straws and memes or regulating the length and shape of cucumbers sold in the EU. The law of diminishing returns is at play: while complexity and societal cost are increasing, the returns are decreasing. The pressure to decrease bureaucracy is almost non-existent as the administrator class protects their own interests. **4/** At some point, the **returns on complexity diminish**. The bureaucracy turns into a burden for its population. Its enormous costs can only be covered through additional income, meaning higher taxes or steep growth compensating for the costs. **5/** Reforming complex societies to simplify them historically never worked because of the bureaucrat's incentives to grow and maintain power besides organisational inertia. To use [Balaji's words](https://x.com/balajis/status/1580864151340998657?lang=en): > It was easier to start Bitcoin than to reform the Fed. It will be easier to start a new country than to reform the FDA. If cost cutting or decreasing complexity through reform is not an option, we have to focus on growth. **Growth can be achieved through mainly three things**: **1/ population growth**: not a likely suspect given European population growth is slowing down since the 1950s. ![](https://substack-post-media.s3.amazonaws.com/public/images/1feefce4-269c-49be-8e7b-2784767396dc_1600x1312.png) **2/** **resource growth**: though the exploration or capture of new land or fossil fuels is also not a real option given the potential for geo-political tensions and atmospheric limitations of our planet. The space economy or novel types of energy sources like fusion seem more suitable but are rather related to point 3/. **3/ technology growth** (aka acceleration): seems to be the only way to grow a civilisation sustainably. This conclusion is in line with Mark Andreessen's statement taken from his [Techno Optimist Manifesto](https://a16z.com/the-techno-optimist-manifesto/): > The only perpetual source of growth is technology. For more context I highly recommend reading 's piece [A Beginner's Guide to sociopolitical collapse](https://www.elidourado.com/p/collapse?utm_source=post-email-title&publication_id=20834&post_id=144148793&utm_campaign=email-post-title&isFreemail=true&r=1n41u&triedRedirect=true&utm_medium=email) from where I borrowed some of the above. [ ![](https://bucketeer-e05bbc84-baa3-437e-9518-adb32be77984.s3.amazonaws.com/public/images/2866d0d3-028b-4548-b000-dd98b8d9ff25_256x256.png)Eli Dourado A beginner's guide to sociopolitical collapse "A society does not ever die 'from natural causes,' but always dies from suicide or murder—and nearly always from the former." — D. C. Somervell. À propos of nothing, I have found myself wondering recently what it would be like to live through a collapse. Would I see it coming? What would be the signs… Read more 2 years ago · 92 likes · Eli Dourado ](https://www.elidourado.com/p/collapse?utm_source=substack&utm_campaign=post_embed&utm_medium=web) # Realities Now that we understand the **why** let's take a look at the **current realities**. Where are we in the cycle of rising and falling civilisations? Which data points can give us some orientation? ![](https://substack-post-media.s3.amazonaws.com/public/images/bcce5f7e-2c36-42ba-91f5-55635dd82371_1780x984.png) Over many years Ray Dalio has been open sourcing his work to help us find an answer to this question. He looked at the driving forces making or breaking an empire spanning education, science, military, health and **reserve currency** status. On the left you can see a schematic visualisation of the of the rise and fall of the Netherlands, Great Britain, US and China over centuries driven by such forces. Eras of conflict highlighted in grey. As indicated by the circles at the very right end of the slide we are approaching a flipping of the reserve currency status once again. On the right side you can see a double click on each of the arches on the left. According to Dalio the US (and with it its western allies) are somewhere between Printing Money and Revolutions. If Dalio is by and large right in his assessment we'd probably see some early **indicators of conflict** and **peer competition** amongst large empires. We can only scratch the surface here. ![](https://substack-post-media.s3.amazonaws.com/public/images/ab5db69d-8dd6-4d0b-8157-b538dd9900cb_1792x986.png) We've been [writing about this before](https://svrgn.substack.com/p/key-learnings?utm_source=publication-search). We are in the early innings of a strong trend towards **de-globalisation** expressed by falling foreign direct investment flows since the financial crisis in 2008 (left chart) and the decreasing global share of trade relative to GDP (right chart). At the same time, **de-industrialisation** of the west might have peaked (hopefully). ![](https://substack-post-media.s3.amazonaws.com/public/images/4ef002f5-9282-4533-8ef7-fe1f6d0583b4_1776x902.png) In a de-globalising world, sovereignty becomes existential. Collectively, **we need to think deeply about our dependencies when it comes to energy, manufacturing, compute and military resources** (which are sort of an output of the aforementioned). We outsourced productivity to the east for decades and doubled down on **knowledge work, services** and of course - **bureaucracy.** ![](https://substack-post-media.s3.amazonaws.com/public/images/5e701be0-35f0-4e5c-afa9-e1b252049039_1750x954.png) # Routes **'Leading by regulation'** has become a normalised statement humorously displaying Europe's decadence and ignorance. But not everything is lost - **we can still build our way out of this challenge. How?** **1/ By abandoning incrementalism** and by focusing on hard problems. As an ecosystem we should do less ad optimisation for the metaverse, less last mile food delivery, less enterprise SaaS and less dating apps but instead focus on solving the existential challenges at hand. ![](https://substack-post-media.s3.amazonaws.com/public/images/68796e0f-e566-4e3e-beac-cf77bdca2995_1096x574.png) **2/** By creating an **industrial tech flywheel** comprised of key innovations in **compute, manufacturing and energy**. At Inflection we believe that compute has the highest leverage, moves the fastest and heavily influences the other two. E.g., with compute we can [accelerate fusion science through learned plasma control](https://deepmind.google/discover/blog/accelerating-fusion-science-through-learned-plasma-control/), we can discover [new materials](https://deepmind.google/discover/blog/millions-of-new-materials-discovered-with-deep-learning/) to build stuff or we can create new manufacturing facilities leveraging [embodied intelligence](https://eclipse.vc/blog/embodied-ai-is-the-opportunity-of-our-generation/?s=31&utm_source=substack&utm_medium=email) besides [protecting our physical-digital infrastructure through cryptography](https://svrgn.substack.com/p/crypto-as-defence-tech). We broke down some of our thinking in [Thesis 2.0 on Sovereign Computation](https://svrgn.substack.com/p/thesis-20-sovereign-computation). Accelerating innovation in Europe might be the only path to sustain our bureaucracy without risking a severe civilisational set back in this new era of conflict. If you are contributing to any of the above problem spaces, please don't hesitate to reach out to us. --- # Crypto as Defence Tech *Published: 2024-06-10 | Author: Alexander Lange | Section: Research* URL: https://inflection.fund/writings/crypto-as-defence-tech The crypto-sphere's creativity and resilience are fueled by its diversity of thought. People with vastly different backgrounds are contributing to, and influencing the space. After the early days dominated by libertarians, privacy advocates and hard money aficionados, the pragmatists and speculators took over. **We believe that the defence community is poised to become significantly more involved in the years to come**. This piece aims to shed light on the intricate but overall symbiotic relationship between crypto and defence. ![](https://substack-post-media.s3.amazonaws.com/public/images/37fbeef7-7461-4c51-986c-6a339af67d4e_2400x1590.jpeg) ## **A brief history of military cryptography** Cryptography is the art of writing and solving codes, dating back to ancient civilizations. The initial reasons for having cryptography were all political or military. Greeks and Romans developed systematic approaches to encryption, e.g., the spartan scytale, a device used to perform a transposition cipher or the Caesar cipher used by the romans to protect military messages. During the Renaissance, cryptography took on new importance with polyalphabetic ciphers, such as the Vigenère cipher. While basic compared to today's standards, these tools enabled empires to expand and conquer, and they laid the information theoretic base for more secure protocols to emerge. More dramatic leaps were made in the war riddled 20th century with the Nazi's Enigma machine, a sophisticated rotor cipher device. The late 20th century and early 21st century saw significant public and governmental debate over the control of cryptographic technologies, referred to as the "Crypto Wars." The U.S. government attempted to limit the export of strong cryptographic tools and proposed the Clipper Chip, which would allow government access to encrypted communications. Edward Snowden's 2013 revelations about extensive government surveillance programs further emphasised the need for robust encryption to protect privacy and national security. Most recently, we witnessed a continuation of the crypto wars in rather financial and transactional contexts when the Biden administration was advocating for an [anti-crypto army](https://www.politico.com/news/2023/02/14/elizabeth-warren-anti-crypto-ftx-00082624). Yet, many cutting edge cryptography tools are used in military as well as private and public environments revealing the short sightedness of such a positioning - potentially backfiring in the run-up to this year's election. ## **Dual use: of crypto's military and commercial utility** Today, most cryptographic technologies are used in both - military and civilian contexts revolving around three major areas (1-4) but with an infinite frontier (5): ### **1/ Secure Communications** Various cryptographic protocols and tools have been developed to ensure the confidentiality and integrity of communications. PGP and its open-source counterpart, Gnu Privacy Guard (GPG), have become standard tools for secure email communication, widely adopted by individuals and organizations seeking to protect their sensitive information. HTTPS ensures secure communication over the internet. Encrypted messaging apps, such as Signal or Matrix / Element use end-to-end encryption to ensure that messages can only be read by the sender and recipient, preventing interception by third parties. The adoption of these apps by military and intelligence agencies underscores their importance in maintaining secure communication channels in high-stakes environments. ### **2/ Secure Transactions** For many, Bitcoin is the anti-thesis to keynesian, politicalised monetary systems with the potential to separate money from state. It's role in covert operations and payments in high-security environments, highlights its potential for secure transactions despite its lack of sophisticated privacy protection beyond pseudonymity. The decentralised finance stack is taking those principles further by applying them to a broader set of financial interactions. Services like [Tornado](https://en.wikipedia.org/wiki/Tornado_Cash) and [Aztec](https://aztec.network/) allow nation state actors as well as individuals to keep their financial activities hidden from the public to protect themselves from scrutiny. As part of DeFi stablecoins, CBDCs can be looked at as [military or political tools](https://www.nfx.com/post/stablecoins-defense) to cement the governments power over the money supply and to foster the US Dollar's global reserve currency status. ### **3/ Secure Computations** Some of the frontiers we have been exploring at Inflection since our founding are advancements in cryptographic techniques related to secure data collaboration. **PETs** (privacy enhancing technologies) are a cluster of cryptographic techniques encompassing [FHE](https://www.perplexity.ai/search/FHE-dvg9iE_uSCKy08Xr0fAgBg#0), [ZKPs](https://www.perplexity.ai/search/FHE-dvg9iE_uSCKy08Xr0fAgBg#1), TEEs, SMPC amongst others and allow computations to be performed on encrypted data without decrypting it. They also allow for the generation of verifiable claims without revealing any information about the underlying data set. This comes in handy for collaborating in untrusted environments such as [military](https://www.scaleoutsystems.com/) operations, healthcare or finance. Our portfolio companies [Tune Insight](https://tuneinsight.com/) (encrypted computation and data collaboration), [Modulus](https://www.modulus.xyz/) (verifiable AI) and [Fabric](https://www.fabriccryptography.com/) (general purpose hardware acceleration for cryptography) are pushing boundaries in such areas. ### 4/ Secure Storage **Local-First Software** offers user data control, privacy and security by ensuring that data primarily resides on the user's local device rather than on centralised cloud servers. It holds the benefits of collaboration and connectivity while preserving data autonomy and minimising reliance on internet connectivity. Our portfolios [Anytype](https://anytype.io/) and [Radicle](https://radicle.xyz/) have been pioneering local-first since 2019 before the term was coined by Martin Kleppmann in his [Local First Paper](https://martin.kleppmann.com/papers/local-first.pdf). Based on the local first stack's properties it is in high demand amongst developers, knowledge workers like journalists, marginalised communities as well as actors operating in hostile environments - including defence and national security. ### **5/ Infinite Frontiers** As usual the frontiers are infinite. Here are some other dual use problem spaces we are interested in: **5.1/ Cryptographic seals** are security devices that utilise cryptographic techniques to verify the integrity and authenticity of physical objects. They are physically attached to the object like a container, a lock on a document case or a tag on a piece of equipment and incorporate sensors that collect environmental data such as temperature, humidity, vibration, or movement. This data is then hashed to create a digital fingerprint. If any changes occur in the sensor data due to tampering, the hash will change, indicating a breach. Blockchains can be used for time stamping. Use cases span logistics and supply chain security in military and commercial contexts amongst others. Physical security is a core part of the resilience of our digital systems, and we believe we are especially lacking in the **5.2/ Cryptographic cameras** are similar to cryptographic seals but instead of sensor data they are hashing visual information and for example device IDs to allow for cryptographic proving of where an image originated. This could increase information provenance in the age of deep fakes. **5.3/ Quantum Cryptography** uses the fundamental properties of quantum particles to ensure security unlike classical cryptographic methods, which rely on mathematical complexity. This approach promises unprecedented levels of security, making it a crucial area of research and development we are excited about. **5.4/ Bio Crypto** combines the principles of cryptography with biometric authentication methods. This fusion aims to enhance security by using unique biological traits such as fingerprints, facial recognition, iris scans, and even DNA as cryptographic keys or for secure authentication. [Worldcoin](https://worldcoin.org/) is a popular, yet controversial example of bio crypto in action. **5.5/ Government funded crypto research** is blossoming for [post quantum crypto by CISA,](https://www.cisa.gov/sites/default/files/2023-02/cisa_insight_post_quantum_cryptography_508.pdf) [CASA](https://casa.rub.de/en/research/rh-a-future-cryptography) or [DARPA's zero trust initiative with 40+ implementation plans](https://defensescoop.com/2024/04/04/dod-zero-trust-implementations-phase-2027/) from military services and defense agencies. ## **Conclusion: bridging crypto and defence is a strategic imperative** We believe that the infinite frontiers of cryptography will continue to be developed by the open source community with increasing contributions from the defence community as researchers, co-developers and customers. This trend is in line with thousands of years of defence technology innovation as well as current geo-political tendencies we discussed [elsewhere](https://svrgn.substack.com/p/key-learnings). Following [our tradition](https://www.augmenthack.xyz/) of fostering interdisciplinary innovation by bringing together distinct communities we are co-hosting the [eurodefence hackathon](https://eurodefense.tech/) end of June 2024 with the goal of providing a platform for innovators to connect around the above themes and beyond. --- # European Defense Tech Hackathon *Published: 2024-05-21 | Author: Jonatan Luther-Bergquist | Section: Events* URL: https://inflection.fund/writings/european-defense-tech-hackathon As investors and innovators, we are always looking ahead, trying to balance optimism with caution. One major influence beyond technological progress is our political environment. There is a geopolitical shift happening, and we need to explore areas for growth and innovation that go with those shifts. The invasion of Ukraine underscored the urgent need for advancements in defense tech, and the role venture investors could play. As part of this, we are thrilled to announce the European Defense Tech Hackathon, taking place from June 28 to 30 in Munich, Germany. **[European Defense Tech Hackathon](https://lu.ma/a2dyx0h8)** **Date:** June 28-30, 2024 **Location:** Munich, Germany ![In a situation where there's an asteroid that close to earth, being on the moon with your favorite engineers is probably a good idea.](https://substack-post-media.s3.amazonaws.com/public/images/56acba96-8e51-448f-8d0e-32af81fe1952_1024x1024.webp) ## **Wake Up, New Geopolitical Situation Dropped** An invasion of a European, sovereign country, can change an optimistic (hopium) outlook suddenly, not just for investors. This realization came with a barrage of changes in public policy, from free trade, open society with fewer borders to protectionism and uprooting of the status quo. We're living the [Zeitenwende](https://dgap.org/de/forschung/expertise/zeitenwende) right now. It's happening at your local grocery store with the increased price of bread, at our power plants, and public commitments of investment into defense capacity. Of course, the Russian invasion of Ukraine is more a symptom and a wake-up call than a causal effect. The euphoria of the last 10+ years of monetary policy has come to an abrupt end. The realization that the U.S. has more uncertainty than we thought when it comes to commitments outside of its borders has been a rude awakening for NATO allies and underlines the need for Europe to take increasing responsibility for its own defense capabilities, in order to safeguard its freedom and prosperity. ## **A Europe in need of acceleration** What reality have we woken up to now? One where Europe is leading in inclusion, sustainability, and regulation. Great, but not enough. It's necessary (depending on who you ask) but not sufficient. Instead, between 2015 and 2022, Europe spent half as much on R&D as the U.S. did, relative to their revenue. The rate of growth of Europe is two-thirds that of the U.S. and the market cap of the U.S. companies was 2.5x greater. [\[1\]](https://www.mckinsey.de/~/media/mckinsey/locations/europe%20and%20middle%20east/deutschland/news/presse/2024/2024-01-16%20european%20competitiveness/accelerating-europe-competitiveness-1501-vf.pdf) When Germany decided against nuclear power, they said no to technological progress and bet on stable relations with Russia. A former colleague who was involved with gas contracts between a German gas purchaser and a Russian seller said in October 2021: "There's no way they can roll back on their word. Supply is absolutely secured!" (_paraphrasing)_ That would probably have been true if not for the massive use of unwarranted violence on Russia's part and a Western reaction. ## **Ukrainian ingenuity** For the first time in history, an invasion was essentially live-streamed over Twitter. We could follow open source intelligence (OSINT) researchers on telegram and get the exact positions of Russian troops. At the beginning of the war, there were about 80 drone operators in Kyiv who coordinated directly with frontline artillery and warfighters to target Russians. It was mostly civilian DJI Mavic drones with a ~20 km range and regular cameras being used. [\[2\]](https://warontherocks.com/2024/04/mike-kofman-and-rob-lee-on-drones-in-ukraine/) Not the $39 million Reaper drones we imagine in war from the Middle Eastern conflicts of past U.S. involvement but a ~$1,000 hobbyist toy. [\[3\]](https://www.forbes.com/sites/davidhambling/2020/06/10/why-the-air-force-needs-a-cheaper-reaper/?sh=58d6ec9d946f) They also developed apps like Delta and Diia to crowdsource intelligence on Russian troop movements and target data. [\[4\]](https://cepa.org/article/could-ukraine-copy-israels-military-tech-success/) In the early days, this ingenuity and decentralized coordination was one of the few advantages they had. The Ukrainians dug into this and continued to set up new mechanisms for attracting innovation and novel tech. Ukraine's defense innovation has been driven by necessity in the face of Russia's ongoing invasion. How can we avoid having such a desperate situation in the rest of Europe? The country has rapidly adapted existing technologies and developed new ones to counter Russian threats on the battlefield. As an example, they integrated various air defense systems into the "FrankenSAM" hybrid system to counter Russian missiles and drones. [\[5\]](https://www.wilsoncenter.org/blog-post/defense-technology-investment-ukraine-attractive-awaits-greater-risk-insurance) Ukraine's innovation has been highly decentralized, with private companies, civilian volunteers, and military units all contributing solutions at a rapid pace driven by urgent battlefield needs. [\[5\]](https://www.wilsoncenter.org/blog-post/defense-technology-investment-ukraine-attractive-awaits-greater-risk-insurance) But it's a constant struggle and not a decisive advantage in itself. Russia is adapting. Where does this leave the rest of Europe with our decades, or in some cases century+ of industrial technology companies to government relationships? Where is the competition and innovation? ## **Primes vs. Hackers** When asking an official for the Armed Forces of one of the most important Euro-Defensive forces how to know what is lacking and what they need. He said "No one really knows what the Armed Forces need as a whole. If they do, they can't tell you. Probably the procurement department knows, but they will never tell you anything!" Turning to the UK, with their public calls for explicit challenges to be solved through the Defence and Security Accelerator, or literally texting with Ukrainian officials over signal about immediate needs seems way more democratic and efficient. Inspired by the energy of our mullet-touting, burger-munching, pick-up truck-driving neighbors in the West, the land of denim on denim, we decided to organize a European Defense Tech Hackathon. We want to inspire and explore, we want to survive and thrive in a democratic society, and for that, our defensive forces need to have the best technology. Our MVP is a hackathon. ## **The European Defense Tech Hackathon** In conjunction with the mostly optimistic but realistic **[Festival of the Future](http://festival.1e9.community)**, we're organizing an evening event and a weekend hackathon together with friends of the fund, **[Benjamin Wolba](http://benjaminwolba.com)** and **[Jens Holzapfel](https://www.linkedin.com/in/jens-holzapfel/)**, as well as talent investors **[Entrepreneur First](http://joinef.com)**, and local research incubator **[TUM Venture Labs](https://www.unternehmertum.de/en/about/tum-venture-labs)**. Munich happens to be a hotbed for defense tech innovation, being home to our sponsors **[Helsing.ai](http://helsing.ai)**, **[Tytan Technologies](http://tytan-technologies.com)**, **[Auterion](http://auterion.com)** and the **[University of the Federal Army](https://unibw.de/)**, as well as **[Arx Robotics](http://arx-robotics.com)** and Quantum Systems. We're also lucky to get the support of **[D3.vc](http://d3.vc)** - a Ukrainian incubator and fund investing into technologies to win the war. We're proudly collaborating directly with the **Ukrainian Ministry of Defense** to ensure high relevance of the challenges we're tackling during the hackathon. The hackathon will be focused on two main challenges: - Preserving lives in conflict - e.g., UAS and counter-UAS tech, friend vs. foe identification, demining, early warning systems - Controlling the information environment - e.g., sensor fusion, cybersecurity, secure comms, jamming, signal analysis, OSINT tooling ### **Get Involved** Join us in Munich on the 28-30th of June (hackathon starts 29th) to explore, innovate, and contribute to the future of defense technology. Whether you are a soldier, developer, engineer, researcher, or enthusiast, your skills and ideas are vital. Let's work together to ensure our defensive forces are equipped with the best technology to protect our democratic societies. **Register Now:** **[Link](https://lu.ma/a2dyx0h8)** For more information, visit **[Eurodefense.tech](http://eurodefense.tech)** or contact us at mail@eurodefense.tech --- # Radical Aero *Published: 2024-04-24 | Author: Alexander Lange | Section: News* URL: https://inflection.fund/writings/radical-aero Radical is an aerospace company building autonomous, solar-powered airplanes to offer **decentralized computational services** from the stratosphere. The fixed wing airplanes can be launched from spaceports and airports with special allowances and negligible launch costs for rapid deployment. Once launched, the aircrafts spiral themselves upwards to an altitude of 20km where they can either hold their position or manoeuvre according to the operator's needs without maintenance for many months. The company's first products will be focused on connectivity services and high resolution imagery. ![](https://substack-post-media.s3.amazonaws.com/public/images/5ac2b70b-a049-44ee-b056-894a0b214034.heic) > At Radical, we are driven by the goal of revolutionising how we connect and monitor our planet. Our high-altitude, solar-powered aircraft are engineered to operate autonomously in the stratosphere, flying over targeted areas for months without the need to land. This capability allows us to provide continuous cell service, collect high-resolution imagery, and carry crucial weather and climate sensors, all while maintaining zero emissions. The potential to transform telecommunications, environmental monitoring, and many other fields is immense and exciting. [James Thomas](https://www.linkedin.com/in/james0thomas/) and [Cyriel Notteboom](https://www.linkedin.com/in/cyriel-notteboom/) founded the company after contributing to the frontiers of aeronautics at Amazon Prime Air as a Research Scientist and Senior Hardware Engineer respectively. Today, they have been announcing their $4.5M Series Seed fundraise led by Scout Ventures with participation from Inflection and Y Combinator. ## Our thesis High altitude platforms have several advantages over LEO satellites that are rooted in their comparatively cheap launch costs, flexible and upgradeable payloads, 25x shorter distance to earth and their exceptional manoeuvrability. **Cell towers in the sky:** HAPS are operating at only 20km altitude (vs. 500km for LEO satellites) and can hover above ground what optimises their ground coverage footprint with line of sight coverage over 50km+ whereas thousands of satellites would be needed in orbit for similar coverage. HAPS also have much more attractive signal attenuation (loss of signal strength over distance) and 30x less latency compared to LEO satellites. Those features allow them to transmit signal directly to hand-held devices without gigantic antennas and receivers like e.g. Starlink does. Constellations of airplanes could be meshed to balance load and to increase coverage and data transmission. ![](https://substack-post-media.s3.amazonaws.com/public/images/3d70911f-176c-43cd-984d-f452befa6519_788x464.png) **Eyes in the sky:** similarly, image resolution from the stratosphere can be sub-10cm GSD (ground sample distance) with a lens diameter of only 12.2cm for visible light at fixed resolution vs. a lens diameter of 3.05m (25x larger) from orbit. The high maneuverability allows for live video coverage and near real time data transmission through a HAPS link or to a ground tower. Satellites on the other hand would have only temporal data transmission (7min for every 90min) and needs radiation hardened data storage coming at much higher prices. Despite many failed attempts of creating similar aircrafts in the past we believe that the time is now because [solar cell costs came down 10x](https://ourworldindata.org/grapher/solar-pv-prices?time=2011..latest) over the last 10 years and [battery performance came up from 80 wh/kg to 350 wh/kg](https://physicsworld.com/a/lithium-ion-batteries-break-energy-density-record/) amongst other enabling technologies accelerating drastically (nano coatings for wings; autonomous flight systems etc.). If Radical Aero is successful in delivering a high altitude compute platform it can become a large, dual use company providing compute services with strong network effects through a software ecosystem and constellation meshing. [The team is hiring.](https://www.radicalaero.com/careers) --- # Build pyramids, not casinos *Published: 2024-03-29 | Author: Alexander Lange | Section: Markets* URL: https://inflection.fund/writings/build-pyramids-not-casinos With each cycle the very nature of the crypto economy changes and expands. What started with idealistic cryptographers caught fire in libertarian circles before scammers, speculators and Wall Street started dominating the space. The drivers of such shifts are obvious: increasing wealth gaps and massive currency devaluation motivate people to lean into speculation. They want to catch up. Technologists jump on the (pretty large) opportunity by leveraging crypto technology. The broad based [ethical erosion](https://polynya.mirror.xyz/ptscXuh3J3KOj2uJAn0vrEanpn2nauwA7iytYZ4cM9U) of the industry left us with over-funded incrementalism, the highest scam per capita quota of all industries and a pretty bad public reputation. Yet, at Inflection we are of the belief that some of the most important technologies of the 21st century will be developed out of this vibrant idea-melting pot. Our LP [Chris Dixon](https://a16zcrypto.com/posts/article/blockchain-culture-computer-vs-casino/) is dividing the industry into the [casino and the compute camp](https://a16zcrypto.com/posts/article/blockchain-culture-computer-vs-casino/). We identify strongly with the latter and believe to be in the early innings of creating a universal state- and time machine that will empower humanity to find truth and document history like no other tools before - besides democratising finance and data ownership. ![pyramid of giza](https://substack-post-media.s3.amazonaws.com/public/images/b9867bab-fa11-4896-b17c-79a12ce624ca.heic) We are not here to fight ethical battles and swim against the current. We hare here to tune down the noise and spot the signal - **people who are building things to last**. Here are some simple filters we apply to find them: **1/ novel ideas** We are not interested in incrementalism. The companies we back typically do something fundamentally novel or different. Funding high risk pioneers making a dent into the universe is our raison d'être. Greatness doesn't come from companies following playbooks or companies that fit themselves into boxes - we stay away from those. **2/ obsession, purpose** We try to find founders who are obsessed with a problem for a long time - either through their own experience dealing with it or through previous (failed) attempts of solving it. Without a deep sense of purpose founding teams won't survive the incredible pain the market is going to expose them to. People can move mountains if they just care enough. Purpose increases mental resilience like nothing else. **3/ resilience** Besides purpose, previous hardship makes founders more resilient, it's like muscle memory and puts things into perspective. We actively seek to work with founders who have been through the ups and downs of life (and crypto cycles) before and got on their feet after defeat. > "I don't know how to do it \[but\] for all of you Stanford students, I wish upon you ample doses of pain and suffering. Greatness comes from character and character isn't formed out of smart people, it's formed out of people who suffered." Jensen Huang, CEO Nvidia **4/ skin in the game** We play infinite games and expect our founders to do the same. We are not here to win zero sum competitions but to survive long enough to see the compounding results of our work come to fruition. Locking up financial resources (high opportunity costs, low salary, long vesting) and social capital (hiring long term co-workers, taking investment from friends & family) are strong proxies for that. Skin in the game drives accountability and accountability drives greatness. Burn the boats! **5/ build first, narrative second** Some of our peers have been making the case for narrative investing in crypto. Hire a team, create a narrative, launch a token, dump it on retail, repeat. Here is an elegant playbook: ![](https://substack-post-media.s3.amazonaws.com/public/images/23671397-df7c-441a-b8dc-2dcc7fe4efca_1194x1394.png) Such an approach is fundamentally at odds with what we would like to achieve with Inflection - **building things to last**. Early liquidity for funders and founders incentivise laziness, reduce productivity by distraction and create a lack of accountability. Strong narratives are a necessity for every business to attract capital, talent and customers but they should be a means to an end: the creation of a differentiated product or service. Narratives without fundamentals are vapourware propaganda and will collapse sooner or later. **6/ tokens as distribution tools, not products** In principle we disagree with the idea of [tokens being products.](https://m.mirror.xyz/gyyuM1OocVmptHYvl4jR6G3fWYsOutCgu19RDGjzGqc) A product is characterised by providing some kind of utility to its owner like a pair of (virtual) shoes, a piece of software to write blog posts, IP Rights and data attached to an NFT etc. Tokens that are merely representing ownership in a network or access and governance rights are not a product; they are are a distribution and loyalty tool. Focusing on the attraction of attention and liquidity towards a token without a clear path to how it can be embedded into an actual product is a red flag for us. * * * Short term speculation is a powerful tool to drive innovation and experimentation but it isn't sustainable and comes with extremely high opportunity costs for the entire space. At Inflection we play infinite games and build things to last. Join us. --- # Thesis 2.0 - Sovereign Computation *Published: 2023-11-13 | Author: Alexander Lange | Section: Research* URL: https://inflection.fund/writings/thesis-2-0-sovereign-computation _First published in Nov 23; updated in Nov 2024_ Our thesis is a living framework of ideas allowing us to operate with focus and towards what matters most to us. It is built upon our perspectives on societal and technological [inflection points](https://svrgn.substack.com/p/the-anatomy-of-inflections) that are shaping markets. Within it we evaluate startups as complex, adaptive systems on a case by case basis. ![](https://substack-post-media.s3.amazonaws.com/public/images/403d296a-5ad5-4906-8551-126385ee5aac_1080x631.png) Over the past few decades knowledge technologies made huge leaps. The internet enabled the real time, cheap and global distribution of information. Search and social broadened its access and exploration. Transformers enabled its synthesis. Crypto enabled its ownership. Agents are enabling actions taken upon it. The convergence of such technologies won't leave any aspect of human civilisation untouched. They started automating knowledge work and science but will reach far into the material world. Intelligent machines will be embodied in forms we cannot even think of yet - from space ships to nano bots. They will populate our labs, factory floors and streets. They will protect our infrastructure and replace humans on the battlefield. They will populate our oceans, the earth's subsurface, outer space and other planets. Computational sovereignty is [as important as energy](https://www.sciencedirect.com/science/article/abs/pii/S136403212101056X) or military sovereignty. In fact, it is an input to both of them. The powers that be woke up to this reality and started a rat race for AGI. Analogous to the development of nuclear weapons, game theory dictates non-cooperative strategies when it comes to the development of weaponised AI. Organisations and individuals alike will similarly pursue access to sovereign compute technologies to evolve, thereby unlocking trillion dollar market opportunities. > **With inflection we are pioneering the frontiers of sovereign computation: the physical-digital fabric that empowers nations, organisations and individuals to independently control their critical resources.** To achieve computational sovereignty, various constraints need to be overcome. Many of which are [rooted in the physical world](https://darioamodei.com/machines-of-loving-grace#basic-assumptions-and-framework). To unfold their powers, our intelligent machines will need to operate interactively with their environments, humans as well as other machines. Yet, cells and matter operate at much slower speeds than machines can process and act upon information. Similarly, human brains, social systems and their institutions have high inertia and will struggle to keep pace with progress in the world of bits. Precisely for that reason it is the intersection of bits and atoms where opportunity lies: ## **Scale** As [process scaling might come to an end](https://royalsocietypublishing.org/doi/10.1098/rsta.2019.0061) it is a societal imperative that we find alternative paths to extend [Moore's Law](https://www.doc.ic.ac.uk/~wl/teachlocal/arch/papers/cacm19golden-age.pdf). Current computational and networking frameworks are voracious and exponentially growing consumers of energy, accounting for [over 20%](https://physicsworld.com/a/powering-the-beast-why-we-shouldnt-worry-about-the-internets-rising-electricity-consumption/) of global electricity demand. The quest for [new computing concepts](https://en.wikipedia.org/wiki/Unconventional_computing) calls for innovation in energy generation, conducting materials, lithography and manufacturing. Novel technical but also legal frameworks governing interactions between humans and machines need to be developed to drive progress. We are excited to back entrepreneurs contributing to those problem sets, e.g. hardware acceleration for cryptographic algorithms ([Fabric](https://www.fabriccryptography.com/)), re-programmable, general purpose edge AI chips ([Ubitium](https://www.ubitium.com/)) or orbital edge AI running on satellites ([Aptos](https://www.aptosorbital.com/)). ## **Resilience** As compute scales and intelligence becomes more abundant, bits and atoms become more interconnected. We're more efficient but also more exposed to critical, large scale failures as the attack surface of our systems grows with their complexity. As much as nuclear weapons defined the 20th century world order, we believe that sovereign compute infrastructure will define that of the 21st century. AI safety, trust and alignment (for lack of a better term) as well as privacy and data-sovereignty of the individual are foundational to the liberal societies we must protect. Doing so requires us to rethink [zero trust architectures](https://en.wikipedia.org/wiki/Zero_trust_security_model) from the ground up - _don't trust, verify_. Therefore, we like to team up with founders pushing boundaries in encrypted computation ([Flashbots](https://collective.flashbots.net/t/project-t-tee-reading-list/3566)), local-first post cloud architectures ([Anytype](https://anytype.io/)), distributed systems and blockchain ([Iron](https://iron.xyz/), [Modulus](https://www.modulus.xyz/), [Entropy](https://entropy.xyz/)) all the way to space security ([Lodestar](https://lodestar.space/)) and robotic fleet control ([Ark](https://ark-robotics.com/)). ## **Flow** Without exponential growth in data generation and access, [more intelligence won't help](https://arxiv.org/pdf/2211.04325). Improving our thinking machines capabilities necessitates new approaches to sensing, data interoperability and interfaces. We believe that [novel sensing](https://www.frontiersin.org/journals/nanotechnology/articles/10.3389/fnano.2024.1434014/full) tech beyond the electro-magnetic spectrum (think quantum, bio, chemical) will unlock entirely new behaviors, [aka opportunities](https://svrgn.substack.com/p/the-anatomy-of-inflections). Data markets powered by private compute tech (see above), [synthetic](https://arxiv.org/html/2406.20094v1) data or market mechanisms (think prediction markets) can be catalysts for the free flow of data. Due to the rising capabilities of our thinking machines, mouse, keyboard and screen seem archaic. We are excited to see new designs emerge, be it in a machine-to machine (think agents) or human-machine (think XR, BCI) context. We enjoy working with teams unlocking insights by providing e.g. high resolution imagery and sensing from the stratosphere ([Radical](https://www.radicalaero.com/)), comprehensive, interactive maps of the earth's subsurface (Stealth) or data markets run on encrypted data at rest ([Tune Insight](https://tuneinsight.com/)). **Inflection is committed to a future where sovereign compute platforms power human progress by backing outlier entrepreneurs with capital, networks and technology.** Get in touch. --- # Less trust, more truth in AI through verifiable compute *Published: 2023-11-06 | Author: Jonatan Luther-Bergquist | Section: News* URL: https://inflection.fund/writings/less-trust-more-truth-ai-verifiable-compute AI regulation is booming. Heads of states are racing to prove to their lobbyists, constituents and others who can have the strongest opinion on AI. Last week alone Biden announced an executive order on [AI safety](https://www.whitehouse.gov/briefing-room/statements-releases/2023/10/30/fact-sheet-president-biden-issues-executive-order-on-safe-secure-and-trustworthy-artificial-intelligence/), and Rishi Sunak organized the [AI Safety Summit](https://www.techrepublic.com/article/uk-ai-safety-summit/) with the bigwigs of both nation (though Biden and Macron were missing) and industry leaders. The [EU AI Act](https://www.europarl.europa.eu/news/en/headlines/society/20230601STO93804/eu-ai-act-first-regulation-on-artificial-intelligence) is set to go live in December this year. Meanwhile Yann LeCun is, [alleging that the existential risk fear mongers are in collusion](https://www.businessinsider.com/ai-destroying-humanity-big-con-tech-existential-risks-2023-11) with the incumbents like OpenAI and Anthropic. They want to repress the open source movement, and build it themselves. A suggested path forward is instead to improve work on transparency - less trust, more truth. Not a new idea and a key principle of Inflection's investment strategy, e.g., portfolio founder, [Michael Gao](http://f16y.xyz), [wrote about this in Fast Company](https://www.fastcompany.com/90899892/artificial-intelligence-trust-crisis) in June. Luckily there's an ecosystem in which there is no higher virtue than being trust-less: crypto! In the last 15 years, we've invented numerous techniques for being able to use public systems as efficiently as possible without the need to rely on any single party. ![](https://substack-post-media.s3.amazonaws.com/public/images/da42803a-6ba7-4f32-8fb0-a441d4efaaae_1024x1024.png) ### Verifiable compute to the rescue A key technology of this revolution is **zero-knowledge proofs**, a form of cryptography that allows one party to prove to another that a statement is true, without revealing any information beyond the validity of the statement itself. This technology is at the heart of **verifiable compute**, a system that can validate the execution of computational tasks without having to perform the computation again. > Didn't get it? [Check this out!](https://www.wired.com/story/zero-knowledge-proofs/) [Modulus Labs](http://moduluslabs.xyz) is at the forefront of this, developing a specialized prover system designed to facilitate the use of AI in blockchain environments, a concept known as ZKML (Zero-Knowledge Machine Learning). We're very proud to have been part of their journey since the pre-seed round, and this week they announced their [$6.3M Seed round](https://twitter.com/ModulusLabs/status/1719749763753730118) led by our colleagues over at 1kx and Variant. The applications of ZKML are relatively far and few in-between, something which the Modulus team is [hyperaware](https://medium.com/@ModulusLabs/chapter-8-make-zkml-real-a3a355b2b756) of. But we see the potential for ZKML as a fundamental technology for the spread of more advanced computation across all platforms. Including blockchains. ZKML can be useful in scenarios where computational tasks need to be outsourced, or where there is a need to verify that a trusted entity, such as OpenAI, has run a specific model. One may want to verify that the model is free from issues like bias or poisoning, and then that the same model is actually used in production, here ZKML can be the solution. It could also allow for the control over a neural network without revealing its weights, ensuring privacy and security. You may see why this is relevant to how we regulate AI. ### In a future not so far away The proposed AI acts are quite fuzzy on the implementation details. Normally this would be totally fine, as regulation shouldn't be technology specific, but rather impose intended outcomes. But we're not sure that's what happened here. If we don't understand the technology well enough to certainly modulate its properties, how can we achieve the goals of compliance? The goal is to protect end-users and citizens from potentially lethal digital tech, but how do avoid creating a useless charade of it? The US has set a computational threshold for trained models, requiring safety tests and disclosure of results for models above 10^26 floating-point operations per second (flops). However, this overlooks the nuances of model performance, as even smaller models like llama-7B can pose risks. The UK hasn't published any direct guidelines or indication of what rules might be used, yet. Ideally, down the line, end-users and regulators have an easy way of ensuring the algorithmic compliance and safety of models, but even doing this without cryptography is a very hairy problem. Factors like [bias](https://www.nist.gov/news-events/news/2022/03/theres-more-ai-bias-biased-data-nist-report-highlights) and accuracy are difficult to define in a generalized way, even without involving zero-knowledge proofs. Cryptography could be used as a digital watermark for AI models. This proof would be the certificate of the computation's integrity, confirming that a given input was used to produce an output with a specified model. The process is succinct, non-interactive, and preserves zero knowledge of the inputs. However, creating these proofs is not without its costs. The computational overhead is significant, with the cost of creating a proof being approximately 1000x that of running the algorithm without the prover, using off-the-shelf (non-Modulus labs) solutions. ### Modulus labs leads the way in use cases Blockchains represent the first real-world application of verifiable compute. Given that computational resources are limited on-chain, and that blockchain applications strive to minimize trust, integrating complex AI tasks is a challenge. Verifiable compute offers a solution to this dilemma, enabling trustless AI operations on the blockchain. Modulus labs plan to demonstrate its potential on the Ethereum blockchain through various use cases, such as Upshot's proprietary NFT appraisal algorithm and Worldcoin's handling of private biometric data. ![Total congruency between the ZK and AI](https://substack-post-media.s3.amazonaws.com/public/images/7f2ace41-dd9f-4d3a-8b46-d646c93112d6_400x400.jpeg) Backed by [Inflection's pre-seed investment](https://twitter.com/ModulusLabs/status/1719749763753730118), Modulus Labs is a testament to the power of combining engineering excellence with entrepreneurial spirit, intellectual honesty, and customer focus. They are not just a research company; they are creators of 'accountable magic,' turning the seemingly impossible into a tangible reality. --- # New Frontiers in Defense Tech *Published: 2023-10-24 | Author: Jonatan Luther-Bergquist | Section: Research* URL: https://inflection.fund/writings/new-frontiers-in-defense-tech ![Flowing digital waves surrounding a radiant prism, which reflects and refracts the patterns.](https://substack-post-media.s3.amazonaws.com/public/images/77c30528-f101-48fa-bd07-f247f8605ae7_1024x1024.png) In a world where the geopolitical landscape is shifting at an unprecedented rate, Europe finds itself at a crossroads. The continent is grappling with a multitude of challenges, from the erosion of democratic values to the urgent need for technological innovation in defense. One emerging player making waves in this space is [201 Ventures](https://www.201.vc/), led by Eric Slesinger, who organized the European Defense Tech Summit in Madrid. In this article I try to share some of the observations and learnings. ## The Catalysts for Change ### The Geopolitical Landscape Europe is facing a drastically different geopolitical environment than just a few years ago. Events like the Cambridge Analytica scandal in 2016, Russia's invasion of Ukraine in 2022, and the recent invasion of Hamas into Israel in 2023 have served as a long-coming wake-up call. The message is clear: Europe might have to fight for democracy. [Finland joining NATO](https://www.nato.int/cps/en/natohq/topics_49594.htm) and [Sweden](https://www.government.se/government-policy/sweden-and-nato/history-of-sweden-and-nato/) finally deciding to apply for membership in spite of long-standing public opposition (recent shift) are additional signs of the threat of violence and oppression being taken seriously. At the same time, Europe is wrestling with a disparity between its values, actions, and what are the facts. Europe should mean progress, democracy and union, in the face of war, as it was started in the wake of WWII. **While the EU leads the way on regulation, it severely lags behind when it comes to public support for innovation**. The continent, often seen as the face of unified democracy and functioning welfare states, has amassed more bureaucracy than functioning paths to novel tech. This has led to calls for strengthening political support for innovation in industry ([Handelsblatt article](https://www.handelsblatt.com/meinung/gastbeitraege/gastkommentar-europa-sollte-seine-leistungstraeger-unterstuetzen/29442676.html)). This is certainly true for deep tech, as well as for defense tech. ### Increased appetite for dual-use in venture What became very clear when looking around the room in Madrid, was the amount of VCs and later stage investors who seemed very open to investing in defense applications. Though, almost exclusively as long as it is "dual-use". Dual-use tech means that the companies and technologies serve both civilian and military applications. As highlighted on stage at the summit by Andrea Traversone, MD at the recently founded [NATO Innovation Fund](https://www.nif.fund/): "There is one example of a dual-use company that pivoted from military applications to civilian" - an [Italian bullet maker](https://fiocchi.com/en/) that switched from military to sport shooting applications. Or, inversely, the norm is that a company which a "dual-use" fund invested into would decide to go after defense applications as a secondary GTM strategy. Overall the increased willingness to engage with the sector by generalist investors can be observed since 20 years, when Palantir started, and later when Anduril became a hit. Yet, most objections still stand: [selling to governments means long sales cycles](https://www.washingtonpost.com/technology/2023/10/22/scale-ai-us-military/), chunked income and a strong preference for incumbents. This is changing to some degree with tech outreach coming from the European agencies themselves, but there is simply no competing with the U.S. defense budget. ![](https://substack-post-media.s3.amazonaws.com/public/images/f5ff5758-fb9e-4f79-a226-aa7cd2b4f028_720x540.gif) The rise of [American Dynamism](https://a16z.com/american-dynamism/) by a16z highlights numerous national interest-aligned technologies such as energy, satellites, drones and bio-engineering. The European Dynamism movement is yet to crystallize, but may be guided by the recent movement by the EU towards sovereignty. It's somewhat ironic that European defense investors to a large degree still are more hesitant to invest in European defense companies than U.S. VCs. ### €160 Billion for critical technology development in the EU The recently launched [€160B program "Strategic Technologies for Europe Platform"](https://www.europarl.europa.eu/news/en/headlines/economy/20231012STO07016/critical-technologies-how-the-eu-plans-to-support-key-industries) highlights six key enabling technologies for the sovereignty of the European Union. Among them are advanced (nano)materials, manufacturing, AI and semiconductors to name a few. The hundreds of billions of euros can be used to recreate success stories such as "[Made in China 2025](https://en.wikipedia.org/wiki/Made_in_China_2025)" and fill the noted gap in funding at graduation from startup but the exact spending is still unclear. The verdict of an assessment of these key technologies in the EU is: in spite of effort and commitments, Europe is losing ground. We're dependent on suppliers of raw materials, talent and knowledge, and we lack the proper paths to commercialization. Pretty grim. Somewhat reflective of the European situation was the fact that the organizer, [Eric](https://www.linkedin.com/in/ericslesinger/), is an American, and many of the other prominent speakers were as well, such as the CTO of the CIA. The US has a rich history of funding deep technology with strategic relevance, e.g., DARPA notably incubating the internet. The venture arm of the CIA, In-Q-Tel, also made a memorable appearance at the summit in a discussion on how to ensure alignment of technology to national interests. ## The Customers: A Paradigm Shift European governments, agencies, and services are increasingly recognizing the importance of technology. The recent experiences of battle on the cyber- and physical battlefield has caused the economic scare of China overtaking a struggling manufacturing industry to turn into a scare that the technological advantage may be lost as well. Palantir and Anduril have gone from the scary companies that may or may not have contributed to the capture of OBL (rumors unconfirmed but not refuted to the joy of the PR departments), to now role models in the market for defense contractors. The customers are primed since a longer time to understand that the battlefield is driven by technological advantages, such as chips for strategic bomb targeting, secure communications, and the use of drones. However, these customers are often too slow-moving to also be developing the tech and research, leading to a need for new collaboration models. As [Nand Mulchandani](https://www.linkedin.com/in/nandmulchandani/), CTO of CIA, said in a video call: "We don't want to have to invent the tech, we just want to buy it and apply it!" In a panel on the use of drones in warfare, the Dep. Minister of Digital Transformation of Ukraine, Alex Bornyakov, explained how the transfer of technology in the private sector to public was overhauled at the beginning of the invasion. Previously, there was an upper limit to the margins that a defense contractor was allowed to make in Ukraine. A limit of 1%! This led to basically only incumbents being able to supply the government. The demands of the Ukraine defense weren't covered! Soldiers bought off-the-shelf drones and 3D-printed custom parts to retrofit them for battle. Now the margin limits have been increased and a multitude of new funding programs have led to over 700 unique Ukrainian suppliers to the local defense have popped up. The public customers and private technology innovation are in sync, but only because of the dire situation. The operating system of government procurement is confirmed by a former military intelligence officer of a Northern European country I spoke to. > The goal is to make 100% certain that the systems are secure, even if it takes years to get there. Whereas Ukraine, with hubs such as Brave 1, built so much with very little that is achieving its goals within the current battle. He also highlighted how certain states tried to replicate the success with initiatives like the Cyber Innovation Hub in Germany and the military innovation program in Sweden, but remain skeptical of their impact. ## The Scope of Tech: Beyond Drones By far the hottest tech topic at the European Defense Tech Summit was drones. The [Auterion](https://auterion.com/) CEO sees the future of battle drones as drone dog fights, with massive swarms (armies?) of drones being coordinated from a distance or autonomously. This will require highly advanced compute capabilities and massive amounts of data processing on the edge. [Quantum systems](https://www.quantum-systems.com/) build aerial surveillance drones, while [Alpine eagle](https://www.linkedin.com/company/alpineeagle/) and another stealth startup (all three out of Munich) provide counter-drone technology. Defending yourself against attack drones is a truly frightening (Hitchcock's The Birds, anyone?) thought. However, the scope of technology in defense is incredibly wide. From bullets to sub-nautical communications and quantum computing, the field is expansive. **What I would have liked to see more discussed this time around, were topics like cyber defense of critical infrastructure and more focus on deeper tech like advanced materials, resilient manufacturing and energy grid technologies.** I was very pleasantly surprised to meet a number of super-computing and semiconductor startups, who have clearly understood their own strategic importance to Europe. Even though there was a gentleman from a UK ministry present, there was a lack of discussion around commercialization support from European states, and few representatives (as far as I could tell) from the public sector. The funding paths may [exist](https://ec.europa.eu/commission/presscorner/detail/en/IP_22_4595), but they should be de-bureaucratized and made transparent in outreach through more events like the European Defense Tech Summit. Thanks again, Eric for pulling this off and gathering the impressive people you did. --- # Key Learnings *Published: 2023-10-18 | Author: Alexander Lange | Section: Markets* URL: https://inflection.fund/writings/key-learnings-building-emerging-venture-firm As a firm we are committed to principled, life-long learning. This comes with honest reflections upon our actions and the conclusions we draw from them. This post is an attempt to do just that. ![](https://substack-post-media.s3.amazonaws.com/public/images/882f2280-2f97-4b4c-96cb-8923d2fbddb4_1920x1280.jpeg) ### Under-estimating geopolitics Since 2020, it has become readily apparent that the global landscape is undergoing a profound transformation. Conclusively, our industry is poised for significant evolution. We find ourselves in a transitional phase, bidding adieu to the once-dominant U.S.-led world order and transitioning toward a more multilateral one. This paradigm shift carries a multitude of implications, not only for the interplay between sovereign nation-states but also for the intricate dynamics within those states and their citizenry. As the era of U.S. hegemony slowly wanes, we observe the ascendance of new superpowers, most notably, China and India. The underlying cycles governing these geopolitical shifts are apt to span multiple decades before fully manifesting. Transient setbacks, such as fluctuations in economic growth or advances in technology, may exert nominal influence on the grander trajectory, unless they precipitate systemic collapses akin to the events of 1989. The ongoing hotbeds of conflict in regions like Ukraine, Armenia, and Israel serve as poignant manifestations of this evolving world order, where the United States finds itself embroiled in numerous proxy wars concurrently, as it strives to uphold its leadership position. The duration and terms of this leadership's sustenance remain shrouded in uncertainty, with the pivotal 2024 election being a critical juncture. As conflicts proliferate and Western powers become increasingly entangled, the specter of new confrontations looms ominously. We turned a watchful eye on regions like Taiwan and the Middle East. These external dynamics yield an intricate and nuanced relationship with the internal cycles affecting governments and their constituents. Given the interplay between these two (external and internal) dimensions, it has become challenging to distinguish real cause-effect relationships from mere coincidences. However, one discernible trend amidst this complexity is the clear shift toward **re-nationalization and de-globalization**. Governments across the globe have embarked on strong intervention in market affairs. Initiatives like the CHIPS Acts in the European Union and the United States, the Infrastructure Bill in the U.S., the Inflation Reduction Act in the U.S., Mica and the AI Act in the EU, and the Securities and Exchange Commission's approach to regulating the crypto industry through enforcement measures collectively underscore this interventionist fervor. Subsidization of domestic economies, the politicization of technology regulation, and the imposition of export constraints have now become the prevailing norm. Escalating debt-to-GDP ratios have precipitated the emergence of high inflation and other forms of financial repression, compelling Western nations to pursue these measures as fiscal imperatives. These external and internal transformations are inexorably reshaping the role of governments, as articulated by **Russel Napier**, a market strategist and historian [in late 2022](https://themarket.ch/interview/russell-napier-the-world-will-experience-a-capex-boom-ld.7606): _In the past four decades, we have become used to the idea that our economies are guided by free markets. But we are in the process of moving to a system where a large part of the allocation of resources is not left to markets anymore. Mind you, I'm not talking about a command economy or about Marxism, but about an economy where **the government plays a significant role in the allocation of capital**. The French would call this system «dirigiste». This is nothing new, as it was the system that prevailed from 1939 to 1979. We have just forgotten how it works, because most economists are trained in free market economics, not in history._ The realization of the aforementioned dynamics has prompted us to engage in introspection concerning our societal role as venture investors. In the bygone era, I characterized the venture industry as the "_gatekeepers of the future_," implying that our collaborative efforts with entrepreneurs were pivotal in bringing about transformative change. However, this assertion may no longer hold true, given that governments are back to the center stage. Notably, the monumental accomplishments of our time, such as the development of Covid-19 vaccines, the development of the internet or the first moon landing, owed their realization to decisive government intervention. In light of this shifting landscape, some pressing questions emerge: **How does a venture investor make informed decisions when markets sway in response to the unpredictable actions of politicians?** **Do these kinds of developments disrupt the foundations of our bottom-up investment frameworks?** **How can we effectively incorporate and weigh macroeconomic factors into our investment theses and portfolio construction moving forward?** **What implications do geo-political events hold for our limited partners our global sources of patient capital with whom we collaborate?** For any venture investor or entrepreneur navigating the tumultuous markets of the 2020s and 2030s, these inquiries merit careful consideration. Regrettably, our systematic contemplation of these matters commenced only subsequent to the invasion of Ukraine, and in retrospect, the writings have been on the wall for much longer. _The last 25 years was a great time to be an investor. With Geo Politics and politics on auto-pilot, running a portfolio became a mathematical routine (quants). (…) The Goldilocks era is over. We live in a fundamentally different world now._ Marko Papić, Macro Strategist (I cannot find the exact source anymore, sorry!) We do highly recommend Marko's book [Geopolitical Alpha](https://www.goodreads.com/en/book/show/54220036) where he applies a constraint-based framework to assess geo-political risks. ### Adapting to changing technology cycles As early stage investors we are exposed to market timing risks driven by technology adoption cycles. In some cases we have been right, in others we have been over-estimating the speed of technology dissemination. **(1)** With the introduction of LLMs working at scale the [world of software development and investing changed](https://e-dorigatti.github.io/development/deep%20learning/2023/04/10/impact-of-llms-on-software-development.html). The long term implications are hard to predict and it is not an easy task to reflect such developments within a bottom up investment framework either. Similarly to the generation of content, everyone is empowered to **create** software by leveraging conceptual knowledge and natural language. This empowerment sets in motion a chain reaction, wherein the demand for software experiences exponential growth (rebound effects) while the market readily accommodates it, courtesy of near-zero entry barriers and immense scalability. Consequently, the next critical link in the value chain centers on the **distribution** of such software. As per current trends, it appears that this distribution will remain firmly in the grasp of entities overseeing consumer hardware and interfaces—think giants like Google, Apple, Microsoft, or those presiding over business infrastructure, such as Amazon. This continuity aligns with the substantial investments these companies are directing toward AI innovation. **If the creation and distribution of AI infused software is commoditised, what will emerge as the new bottlenecks of computational innovation?** _"Software has been eating the world" and is now becoming marginal. The new frontier of computational innovation will be hardware led - access to raw compute sustaining Moore's Law; elastic compute infra underpinned by strong cryptography and the free flow of information between the worlds of atoms and bits._ Our take might be contrarian and counter-intuitive to some but we are quite confident in such conclusions. We will write about related topics in more depth soon. **(2)** In other areas our market timing has been off as we have been overly optimistic about the adoption of crypto technologies to coordinate labour at scale as opposed to just capital. A concrete example is the emergence of DAOs (dencetralised, autonomous organisations) as new forms of organisations orchestrating labour and resources programmatically over the Internet. Our collective vision was to create global organisations which are run on self executing code as opposed to pen and paper. We broke it down into composable, enabling products (e.g. voting, governance, treasury or identity modules) to solve coordination problems without intermediaries. Unfortunately, none of such modules reached product market fit and are unlikely to do so in the near future. The reasons are obvious in hindsight. First, people need to prove their employee status, pay taxes, enforce titles against their offline counterparties or pay local merchants at some point and without proper interoperability between the old and the new systems this is nearly impossible to achieve. Second, there is early evidence of the DAO framework being effective to coordinate labour in the form of running specific algorithms (Bitcoin's proof of work) or providing standardised services towards a network (bandwidth, compute, storage). However, those mechanisms cannot be abstracted and applied towards more complex, nuanced services or products currently. Running the conventional legal system and the internet-based one in parallel doesn't resonate as one is prevailing the other. We are long term believers in on-chain organisations being fundamentally superior to pen and paper based ones but the time needed to make this transition practicable is likely 5-10 years out and startups don't have that kind of time. We shared our concerns publicly at [Inflection's DAOday in September 2022.](https://svrgn.substack.com/p/daodaywtf-was-a-blast) ### Constructing a portfolio and pacing investments The disciplined execution of our portfolio construction plan and investment pacing have been key for us. As we closed our [Inflection Mercury Fund](https://techcrunch.com/2022/01/19/inflection-raises-40-million-for-its-second-crypto-focused-fund/) in 3 closings over the course of 2021 our capital base grew from $4M in the first towards $40M in the final closing. To stay conservative we always defined the amounts closed as our final capital base and sliced our allocations accordingly. In hindsight that created some outsized (too small), asymmetric positions in our portfolio. Yet, we believe that staying conservative was the right thing to do as markets could have moved into the opposite direction as we've witnessed during the pandemic - when we lost many handshake commitments from one day to another. We would have been terribly over exposed if we had assumed a larger capital base before closing out. Further, we managed to **invest quite broadly within our mandate** what prevented us from being overly exposed to single niche-categories further adding to concentration risks. Specialisation works well in times of market expansions but it works poorly in times of market contractions (NFT only funds, anyone?) or large scale resets as it changes the risk / reward profile of a fund. We prefer to operate under a broader mandate and sticking to first principles without being opportunistic - to be adaptive enough to changing markets. Related to portfolio construction is the aspect of **investment pace**. The most important driver of venture capital fund returns is not the ability to pick winners but the vintage year. Hence, stretching out the investment period over 4 years was the right thing to do. We are grateful for our collective discipline despite the euphoria and hype around us especially in 2021 and 2022. That said, we did try to time markets when we built up some of our liquid positions over the course of nine months. In hindsight we should have picked an even larger time frame to dollar-cost-average into such positions as we intend to hold on to them for about a decade. ### Diversifying service providers and counter parties Venture funds are known to take incredibly concentrated positions coming with high risks. The concentration of single positions is a function of portfolio construction (how many positions and how they are weighted). It's not unusual to see early stage funds holding 50% of their fund in a single positions after a large uplift in one company. With that context it becomes existential to at least **diversify all kinds of operational risks** and in particular that of failing counter parties in the form of limited partners (yes they can default) and banks (yes, they can default) including custodians (yes, they can lose assets and default). Our setup has been tested during the collapse of Silicon Valley and Signature Bank as well as FTX and its fallout. We established a **war room protocol** to deal with such situations and provided our ecosystem with a **guide to navigating a banking crisis**. Ever since operational risk management has been deeply engrained in our firm and we can only recommend every emerging manager and founder to follow through. * * * Where do you disagree and why? --- # DeSci's Reality Check *Published: 2023-09-13 | Author: Alexander Lange | Section: Markets* URL: https://inflection.fund/writings/descis-reality-check _This post is a summary of a presentation I gave at https://www.desci.berlin/ in September 2023. The (very high level slides) are linked at the bottom of this page._ ![A science led utopia](https://substack-post-media.s3.amazonaws.com/public/images/e6f64461-612b-4d3c-8a16-f1b4e86312fa_1024x1024.png) The decentralised science movement is very close to our heart. We perceive scientific research as the bedrock of human progress, innovation and flourishing. This is why we have been part of the community since its inception when we backed Molecule.to in late 2019 and Labdao in 2022. ## 99 Problems In the post WW2 era the scientific ecosystem evolved into a complex bureaucracy composed of various different stakeholder groups and relationships between them - a complex system. This system is currently facing many structural challenges such as (1) paid access to research which is financed by the public, (2) opaque and inefficient funding mechanisms, (3) a lack of data sharing initiatives across countries, institutions and disciplines preventing deeper and broader insights, (4) intransparency around contributions and many, many more. **Science is broken and everybody knows it.** Having broad baed consensus on the 99 problems and some of the smartest, purpose driven talents of a generation working on it - why is it so hard to make progress? The root cause is inertia: the more complex a system, the larger its resistance to change. To share some examples: **(1) Institutional biases:** the brand of an institution counts more than the actual scientific contribution to a field when it comes to publishing in top journals. Exclusivity instead of meritocracy and truth seeking became the norm. **(2) Social biases:** in its current form the 'scientific method' is defined very narrowly and hampers us to to get insights from experiments which require hard-to-replicate human interventions like in the field of psychedelic therapy where the subjects are typically guided through the journey what makes it harder to assign the causality of the results to the compounds as opposed to the human therapist's influence. **(3) Feedback Loops:** Publish or perish is a terrible feedback loop which leads to more frequent publishing creating group think and incrementalism. Quality became a vanity metric for scientific researchers. So how do we overcome such inertia in a complex system? We need to tackle (1) the '_right leverage points_' and (2) build up sufficient '_activation energy_'. But what does that mean? ## Finding the right leverage points Leverage points in our context are activities that are yielding maximal effects with minimal effort. Ideally, those are changes which unlock other changes. Changes that can create ripple or network effects of some sort. To generate maximal impact, the leverage points should be composable and integrate with one another. Intuitively, the DeSci ecosystem picked a mix of technologies and behavioural patterns as leverage points to start with. I'm currently seeing 3 major categories that I'd consider as the 'right' leverage points and one that I perceive as 'less relevant'. **(1) Data & Collaboration tools** can be used to facilitate data sharing and to track contributions and research provenance - e.g. through NFT issuance and provable claims of contributions (not your keys, not your contribution!). (2) On top of such tools a **social-knowledge graph** can be layered, e.g. to represent research trees which can be leveraged for (3) novel **incentive structures** such as programmatic royalties for researchers, paid peer-reviews, retro-active funding or fractional, liquid IP markets The leverage point I am much less excited about is '**Public Policy**'. Building momentum on the regulatory front requires insider access, knowledge and obscene budgets while driving marginal (if any) results. It might only become relevant at a much later point in time once we built up momentum aka activation energy. I'd be very curious to learn more about alternative or additional leverage points in this context. What do you see in terms of novel products and approaches to address some of the 99 problems? ## Activation Energy Now, assuming that we figured out some of the right initial leverage points we need to think about activation energy - the energy that must be delivered to a system in order to initiate a reaction; to break bonds so that new ones can form. Activation energy in socio-technological - let alone political - contexts is a very tricky thing. We can borrow some analogies from the Arab Spring and many other (failed) revolutions of the past. In most cases it only requires a small but very active minority of say 10% of a given population to rebel against the status quo. In Tunesia, Lybia, Egypt, Algeria, Jordan, Morocco and some other countries such highly active, minority groups were amplifying their goals through new means of technology - social media. The problem many of them were facing though, was a lack of activation energy. In order to execute a revolution successfully it doesn't suffice to overthrow the powers that be and to create a power-vacuum. One needs to have a concrete plan about how to seize power, how to provide and maintain public infrastructure like health, education and basic supplies like electricity and water. One needs to have a plan for how to arrange new governance models and how one deals with opposition forces short-, mid- and longterm to avoid an immediate counter-revolution or seeing the military taking over. The latter is sort of a default for most revolutions. For more context about those topics I highly recommend the book Revolt of the Public by Martin Gurri. Applying some of those learnings to the DeSci movement - what would need to happen in order to change the state of a highly complex system with huge inertia? More than a few things: - Ideology - Talent - Capital - Technology - Adoption - Public Support - Governance Those things are partially path-dependent in the sense that we cannot have certain things without others (e.g. no capital, no technology or no public support, no governance). In order to navigate through what might probably be a multi-decade, maybe never-ending state transition we need to be aware of where we are in the cycle. And to be aware of where we are in the cycle we need data. ## Reality Check So, how far are we in the transition towards are more open, more collaborative, more accountable and efficient scientific ecosystem? Are we 5% in, 15%, 50%? In order to get any sense of orientation we could look at the **inputs and outputs** of the system from a 10,000 foot view. Some metrics coming to mind: - **Input:** 95 DeSci Projects (40% funding, 25% data, tools, 25% Publishing, 10% Other); $65M in funding raised, about 150 talents working FT in Desci, - **Output:** about 2,400 users of DeSci Products, 9 IP-NFTs issued Those numbers do barely yield **any** insights - they are too shallow, too fragmented and mostly incorrect or outdated. We used the [Messari Report - The Decentralised Science Ecosystem](https://messari.io/report/the-decentralized-science-ecosystem-building-a-better-research-economy), [Awesome-Desci Github](https://github.com/DeSciWorldDAO/awesome-desci) (for #Projects with URLs), [Crunchbase](https://www.crunchbase.com/) (for funding data) and [Desci.world](https://desci.world/) (for users + categories) to pull some of those numbers. **Going forward we'd like to collaborate with any of the above initiatives and the broader ecosystem to put up some (self reporting) dashboards where we can gather more insightful fundamental data. Please reach out in case you have ideas or would like to contribute!** ## Growing Pains; Builder's 101 Assuming we'd be able to get more meaningful data - a next step would be to identify our community's blank spots and improvement areas to accelerate progress. Below I am sharing some observations of the last few years in the space which I believe are critical for survival: ### Measure Fundamentals, not vanity metrics > **❌ Discord Activity, X Followers, Token Holders, Token $Price, Trading Vol, Projects Sourced** > > **✅ Utility, Growth, Conversions, Usage, Retention, Transaction Value, Revenue** Referring back to the 'Activation Energy Section' the phase of rising public awareness feeding into public policy and governance comes at a later stage. For the near future of a couple of years we should rather focusing on show casing to the world what is possible and how a better science system could look like. Flash-in-the-pan traffic, PR campaigns and social media noise won't drive results, they rather distract from what is essential - namely fundamental utility. Do we solve a critical problem for someone? Even if that that someone represents a small user base it would be the right starting point. The utility of the product can be measured by retention rates and going forward by willingness to pay. ### **Build products, not decentralisation theaters** > **❌ Complex Gov / Voting Schemes early on** > > **✅ Solving someone's problem, Deep Customer Exploration, Simple** Again, referring back to the section 'Activation Energy' the 'Governance' phase comes years after the utility phase. Most 'DAOs' today suffer from immense over-head and noise in their communities. Very few people make regular, meaningful contributions - pareto is everywhere. Some of the decentralisation theatre is driven by regulatory uncertainties, some of it is driven by a real desire for more equitable governance. The reality is that the opportunity costs are organisational efficiency and building something useful - in the early years only the latter is existential while the former is nice to have. ### Seek long term alignment, not short term speculation > **❌ Liquidity Carousels, pump & dumps** > > **✅ Long lock ups, Deep Customer Exploration** Attracting the right talent, finding the right leverage points and getting to sufficient activation energy to drive change are critical. Focusing on token prices, pre-mature asset issuance, market making and speculation will be counter-productive as it attracts (1) the wrong people, (2) the wrong money and (3) it confuses internal speculative demand with actual outside demand verification. Defi doesn't have product-market fit as long as it serves a niche tech- and finance savvy community of speculators and doesn't integrate with real-world assets and payments infra. Similarly, DeSci won't find product market fit if we can't channel Bio-pharma demand and resources into our products. Let's be the 'American Revolution', not the 'French Revolution' or the 'Arab Spring'. ![](https://substack.com/img/attachment_icon.svg) Desci's Reality Check Bridging The Gap Between Idealism And Pragmatism 2.27MB ∙ PDF file [Download](https://svrgn.substack.com/api/v1/file/90c58cda-56e8-4e25-ad41-5021d4e84f23.pdf) --- # Anytype: Pioneering the future of trustful collaboration through knowledge graphs *Published: 2023-08-23 | Author: Alexander Lange | Section: News* URL: https://inflection.fund/writings/anytype-pioneering-the-future-of-trustful-collaboration ![](https://substack-post-media.s3.amazonaws.com/public/images/4cecb32b-fd9f-43a2-a821-2230ce55dd45_1920x1050.png) We live in an era where the management and utilization of information have become the cornerstones of progress. This holds particular significance in the context of AI, where information and its intricate interconnections can greatly enhance the capabilities of machine intelligence. This post delves into the concept of knowledge graphs and their role in AI model training as pioneered by our portfolio company [Anytype.io](http://anytype.io/). [Anytype](https://anytype.io/) is a company community of 100,000 people built on the foundations of freedom of expression and data sovereignty, putting the user at the center stage of a next generation super-app. The early version of the product is a powerful knowledge management tool which leverages bi-directional links, new types of data structures and aesthetic, freely composable interfaces. As a heavy user I can attest that Anytype is not just incredibly fast and reliable software - it is a piece of art. ![](https://substack-post-media.s3.amazonaws.com/public/images/0331bd30-9f37-468f-b0d4-8fd5ec1b8d3d_1588x1000.png) Having been early backers since 2019 and living through the pandemic and geopolitical turmoil alongside the Berlin based founding team we are deeply grateful and proud of them to announce their $13.4M Series A round led by Balderton with participation from [Inflection.xyz](https://inflection.xyz/), [Square One](https://squareone.vc/?ref=blog.anytype.io), [Script Capital](https://script.capital/?ref=blog.anytype.io), [Protocol Labs](https://protocol.ai/?ref=blog.anytype.io), [Connect Ventures](https://www.connectventures.co/?ref=blog.anytype.io), [New Forge](https://www.newforge.de/?ref=blog.anytype.io), and [Foreword VC](https://www.foreword.vc/?ref=blog.anytype.io) and prominent angels from Heroku, Muse, Centrifuge, Polkadot, Ocean and others. ### From Caves to Screens: A Brief History of Knowledge Management Tools The evolution of knowledge management tools is a captivating journey that mirrors the progress of human civilization itself. From cave paintings to the vast troves of papyrus scrolls housed in the ancient Library of Alexandria all the way to the boundless digital databases of the modern world, our tools for managing knowledge have continuously evolved. **Beyond information storage they shaped how we access, retrieve, share and perceive knowledge across time and space**. While the monumental shift from physical to digital mediums is in full swing, significant challenges need to be overcome: graph contruction, information provenance and user sovereignty are some of them. ### Solving AI's context problem through knowledge graph construction Against this backdrop emerges the concept of knowledge graphs. Knowledge graphs can be visualized as expansive networks where nodes of diverse information interlink, akin to cities on a map connected by an intricate web of roads. ![from anytype.io](https://substack-post-media.s3.amazonaws.com/public/images/3faa541a-07b7-4b03-8ab0-7c3117c54618_860x732.png) Each node in a knowledge graph represents a piece of information, while the connecting lines, or edges, denote the relationships between these pieces. Graph structures can be compared to the human brain's synapses (edges) and neurons (nodes). As opposed to relational databases which refer tables to tables, graphs consider the links between different data points. This is an incredibly powerful concept as the graph is both a place to organise and store data, and to reason what it is about and to derive new insights. It is also flexible and extensive in terms of the types of data and schemas it can support. Graphs evolve to reflect changes in the domain and new data is added to the graph as it becomes available. They are a powerful structure to support advances in GNN (Graph Neural Networks) you can explore further in [Introduction to Graph Machine Learning](https://huggingface.co/blog/intro-graphml), [Must-read GNN papers](https://github.com/thunlp/GNNPapers) or [Graph ML in 2023: The State of Affairs](https://towardsdatascience.com/graph-ml-in-2023-the-state-of-affairs-1ba920cb9232). Anytype approaches graph construction bottom up: users are at the center of organising their bookmarks, thoughts and content of all types (images, videos, music) - you can think of it as a [second brain](https://www.buildingasecondbrain.com/). Next stops will be access control, collaboration- and publishing features. ### Open Source Signing Of Everything The second profound challenge Anytype addresses is that of content provenance and user authentication - particularly in the age of generative AI which which allows for the creation of deep fake content distributed through fake social media accounts and bots. Crypto systems have been leveraged to prove ownership ("not your keys not your coins") and [solvency of exchanges](https://eprint.iacr.org/2015/1008) since 2015. Now, innovators are leveraging them as a tool to verify [human identities](https://vitalik.ca/general/2023/07/24/biometric.html), information provenance as well as [compute](https://fortune.com/crypto/2023/05/04/artificial-intelligence-zero-knowledge-proofs-zkml-verify-ai/) in order to let humans and machines interact more reliably over the internet. By embedding signature schemes into every interaction seamlessly (soon to come) Anytype creates a new type of trustful collaboration environment for individuals, enterprises and communities of all sorts. In order to be trusted such a crypto system needs to be open sourced - and that is exactly what [Anytype just did with its open beta version](https://blog.anytype.io/our-open-philosophy/). ## User Sovereignty Rather than relying on conventional, proprietary systems, Anytype offers a departure from the norm. Users are not bound by typical software limitations, nor are they required to consistently be online to access or share their data. They are not subject to the whims of any developers, including the ones working with the Anytype Association. The platform's Anysync protocol ensures that data synchronization occurs seamlessly across a peer-to-peer network, all while maintaining stringent encryption standards. Each user's space is managed locally, ensuring that they remain the sole custodian of their data. This emphasis on individual control, combined with the platform's collaborative ethos gives us a glimpse into how global scale collaboration between humans and (AI) algorithms over the internet might look like in the future. There is way more to say about Anytype as an ecosystem and its many other value propositions but check out [what others have to say](https://www.producthunt.com/products/anytype/reviews) or even better: [run it yourself](https://anytype.io/). Onwards & Upwards --- # Join Augment - a Decentralized AI hackathon hosted by Inflection and friends *Published: 2023-06-15 | Author: Jonatan Luther-Bergquist | Section: Events* URL: https://inflection.fund/writings/join-augment-decentralized-ai-hackathon The growth of Artificial Intelligence (AI) continues to stun and captivate us with its advancements. New models and understanding of super-intelligence arise from the traditional halls of academia and corporate research but the **exponential progress happens in open-source development** as open access continues to attract the best ideas and talent. ![Midjourney; Generative art for a crypto and AI hackathon subtly referencing to the history of computers and colourful silicon wafers. Hyper modern, aesthetic, high resolution.](https://substack-post-media.s3.amazonaws.com/public/images/c98d1118-3d4f-4386-a630-6d6758e349a7_1822x1220.png) The recent developments and AI democratisation leave us with many open questions: Do self replicating and self extending AIs impose existential risks to humanity? Or are they rather the dawn of a utopia where we gain a much deeper understanding of intelligence and the mind's inner workings? How can we align AIs with human goals and keep them accountable? How can we decide what is real and what isn't in a world of autonomous agents and machine generated content? The answers to those questions won't come from the AI community alone as they require inter-disciplinary collaboration across various fields spanning the humanities, computer science and cryptography. For years the AI and crypto communities have been operating in silos as they were allegedly driven by opposing ideologies. **We don't agree with those views but consider AI and cryptography as two sides of the same medal: you can't have one without the other.** At Inflection, we believe in the power of collective and open intelligence. Hence, **we're excited to announce that we're co-hosting a hackathon in Paris, aiming to gather the most brilliant minds from the worlds of AI and cryptography** to collaborate. Some powerful ideas we'd love to see experiments around: - Enhance AI agent's economic capabilities through composable open open finance legos - Democratise access and development by LoRA hacking or censorship-resistant embeddings publishing - Enable marketplaces of agents and data sets represented as NFTs - Cryptographic access control over AI agents and custom data sets - Notarise and verify identities and content through cryptography, end-to-end secured Our goal is to gather developers, cryptographers and AI experts who believe in the transformative potential of trustless AI systems integrated with crypto technologies. We are joined by some of the leading innovators in the space including [Aragon](https://aragon.org/), [Ceramic](https://ceramic.network/), [Gensyn](https://www.gensyn.ai/), [Llama Index](https://twitter.com/llama_index), [Modulus Labs](https://www.moduluslabs.xyz/), [Ocean Protocol](https://oceanprotocol.com/), [Ora](https://ora.ai/), [Risc0](https://www.risczero.com/), [Seed Club Ventures](https://seedclub.ventures/), [Spearbit](https://spearbit.com/), [StabilityAI](https://stability.ai/), [Station F](https://stationf.co/), [Violet](https://www.violet.co/), [CrunchDAO](http://crunchdao.com) amongst others. Co-organizers are [Inflection](http://inflection.xyz), [Nevermined](http://nevermined.io), [Valory](http://valory.xyz), [Algovera](http://algovera.ai), [Orbis Labs](https://useorbis.com), [Tenfold](http://tenfold.xyz) and [AI3](http://ai3.co) > **We invite you to join us in pushing the boundaries of what's possible, exploring the intersections between AI and crypto and pioneering new ways to implement and secure these technologies.** **Logistics** Where: Station F, Paris When: 2 days and a night: July 18 - July 19 Who: anyone who's willing to learn Sign up at [augmenthack.xyz](http://augmenthack.xyz) ! Onwards & Upwards, Team Inflection --- # An engineering approach to Venture Capital *Published: 2023-06-14 | Author: Alexander Lange | Section: Building* URL: https://inflection.fund/writings/engineering-approach-to-venture-capital The venture capital industry has always been about turning imagination into reality. As we step into the future, we see an opportunity to build upon this tradition by integrating the power of emerging technologies into everything we do. Today, we'd like to share a high level vision of Inflection's future: a deep tech venture firm that pairs human intuition and judgement with cutting-edge AI and machine learning technologies to revolutionise how we identify and support innovators through networks, capital and insights. ## Vision: A computer aided venture investing machine (CAVI) At its core, a VC firm is a decision-making and insight machine, an amalgamation of mental models, frameworks and human intuition. We believe this machine can be enhanced and optimised. We envision a world where VC operations - currently largely manual - transform into an elegant synergy of human and machine, powered by data-centric processes and automation. We called this concept **CAVI** **\- computer aided venture investing.** This fusion will enable us to leverage our most valuable asset – human attention – more efficiently to focus on non-fungible activities such as relationship building, ecosystem building or creative thinking in dynamic environments with incomplete information. Here's what we're going after: Imagine the first time you held an iPhone or the first time you were recommended exactly that song you wanted to listen to but didn't know it yet. That is the kind of step function improvement CAVI can bring about in venture capital. > _Creating a well-designed decision making system helps to mitigate the influence of personal biases and emotional factors, allowing for more objective and consistent outcomes._ Paraphrasing Ray Dalio ## Catalysts Up until now, it was really hard to be very data-driven and tech-first in venture (as public interviews with some of the larger funds will tell you). The data is inconclusive, mostly it's people and ideas, without little quantitative support to indicate correlation with returns. Research and contributions have been made by other funds - large and small - going from standardised personality tests of founders to predicting performance by proxy in website visits. We want to build on these efforts and take it to the next level. The exponential advancements in AI, machine learning, NLP coupled with the increasingly ubiquitous access to data have opened up unparalleled opportunities to reimagine the VC landscape. Those are some of the core drivers paving the way for such shift: 1. **Text processing and manipulation is leaps and bounds ahead of where it was just six months ago:** The introduction of GPT-4, the latest language model from OpenAI has made it dramatically easier to analyse and extract insights from text data. 2. **More data is available than ever before:** The digital era has created an explosion of data. This ever-growing wealth of information can be harnessed to provide a much deeper and broader understanding of research models, startups and markets. 3. **Emerging venture funds are redefining the playing field:** The traditional VC model has its strengths, but it also has limitations. As the startup ecosystem continues to grow and change, our industry must also adapt to remain effective. In conclusion, the time is ripe for a more data- and automation driven approach to venture capital. By combining the power of advanced technology with the irreplaceable human element, we can create a venture engine that is scalable, efficient and effective in today's fast-paced, data-rich world. If you're more interested in the broader topic in general we highly recommend following and the community. ## A starting point - decomposing the VC value chain Before we dive into how the venture capital firm of the future could look like more concretely, let's explore its underlying principles and analyse how the core of its value chain might be affected by the ongoing compute revolution. This is **one** **out of many ways** to define a venture capital firm's key functions. **In \[\] we defined where the function sits on the spectrum of automation potential** where 0 = non fungible human competency and 10 = fully automated. We used a rule of thumb framework taking into account (1) reliability & availability of high quality data allowing for extrapolations into the future (2) required creativity and intuition in highly dynamic environments with incomplete information, (3) required human coordination with external human stakeholders to fulfil the function. ### Thematic **Research \[7\]** _Where to look at; which hay stack to explore._ Inputs for this function could be - **viability:** research breakthroughs, technology maturity, technological limitations like scale, dependencies - **catalysts:** convergence, amplification through other emerging technologies, regulatory risk and incentives, demand drivers, geo political and macro economic forces - **impact:** degree of novelty, superiority vs. incumbent solutions, societal and economic implications - **market:** large problem set yielding multi billion dollar market opportunities within 5-10 years from investing - **popularity:** controversial sentiment (we have to be contrarian & right), low investment volumes, low web traffic and media coverage A shout out to Lawrence and our friends from Lunar Ventures who inspired us through their [State of the Future Tracker Methodology](https://www.stateofthefuture.xyz/). While some of those inputs can be pulled from research institutions, open source research platforms and the like, qualitative insights might need to be non fungible like expert interviews, hence the 7/10. ### **Sourcing & Pre-Screening \[8\]** _What to spend time on; which needles in the haystack might be worth analysing._ This function is tricky. If the funnel is too narrow the firm might lose out on potential outliers. If it is too broad it drowns in noise and struggles to focus. Inputs to this function include: - **network analysis**: connections between people; connections between people and organisations; connections between people and ideas / problem sets Outbound sourcing will be informed by thematic research as we invest with low velocity and high conviction. Our outbound sourcing filters will be set accordingly what can be strongly technology assisted. On the other hand, a variety of other factors will play a rule, such as (1) thematic content creation and events, (2) establishment of a scouting and mentorship network for soon-to-be founders etc. which have more of a human touch. Hence the 8/10 score. Recommended further reading: [More needles, bigger haystacks: What we mean when we talk about data-driven VC](https://pulse.moonfire.com/more-needles-bigger-haystacks-what-we-mean-when-we-talk-about-data-driven-vc/). ### Investment Decisions \[4\] _What to invest into; which needles to pick from the haystack._ This function lies at the very core of our business. Over almost a decade we developed frameworks, score cards and various heuristics to take maximally informed decisions optimised for upside potential in the early life cycle of a company. Inputs to this functions include (over-simplifying here; each of those bullets can be universe in itself): - **Team:** personal fit; complementary to existing portfolio founders; shared vision & purpose; deep co-working history; alien skills; obsession with problem space; history of outstanding achievements; leadership that attracts elite talent; resilience and grit - **Product:** fundamentally revolutionary approach to solving the problem (0→1); enabling something previously impossible; strong first- or early mover advantage; unique technical moat (network effects; brand) - **PMF:** 10x improvement for customers (faster, better, cheaper); high retention; high product velocity; validated demand in target market - **Market:** Large ($10BN+) current market and / or growing rapidly; blue ocean or unique approach to gain market share; ability to capture value demonstrated; business logic allows for strong network (ownership) effects and defensible valuation; clear monetisation strategy - **Terms:** pre-seed or seed valuation; ownership target of X; 100x upside potential; complementary potential syndicate; 24+ months of runway post financing; clean set up (no previous debt, side deals or skewed incentives **Note:** The more early stage (read pre product or MVP stage) a fund operates the more weight needs to be put on research (see above), market and team. Diligence on most of these items can only be automated to a limited degree, hence a 4/10 score for us as a **first-check fund**. This might look vastly different for later stage funds harnessing much more data (number go up!). ### Deal Execution \[4\] _Pitching our firm to get an allocation; Term negotiations; Due Diligence; Syndicate construction._ This function is one of the more non-fungible ones we can image as it is profoundly related to **human relationship and trust building**. For the more formal parts including the review of data rooms and legal documents or finding ideal syndicate partners technology might be helpful. Hence, 4/10. ### Founder Support & Platform \[8\] _Strategic Sparring; Board Roles; Founder Coaching; Go to market; Pricing Strategies; Pivots; Key hires; Business Development; Shared Resources; Best Practices._ This function is critical to support our founder ecosystem. Historically, it has been the most non-fungible activity a venture firm's partners have been involved in (besides fundraising). We are of the belief that this is about to change. Most of the problem spaces mentioned above can be approached through frameworks. The underlying problem spaces are unique in some ways as every company, team and market conditions are different - but at the same time structural patterns remain similar. - network analytics to identify key hires - Custom trained LLMs to relay specialised knowledge - Shared software- and data infrastructure E.g. every company finding product-market-fit will define modular experiments to test hypothesis and measure success. How this can be done best is specialised knowledge which can be sourced from the public domain and extended through expert-input. Aggregated insights could be packaged in a custom data set and made accessible through LLMs for example. Artificial agents helping innovators to navigate mission critical challenges will seem like a natural expansion of what venture investors do today. Through data-network-effects (knowledge is a network) such services will improve over time and can provide a competitive edge for an entire ecosystem. Therefore, 8/10. ### Fund Operations \[9\] _Portfolio Modelling; Investor Reporting; Investor Relations; HR; Legal; Finance; Accounting_ These functions are very broad and currently necessitate a lot of human engagement and coordination with external stakeholders. Inputs to this function include: - **Strategy**: portfolio modelling and construction; sensitivity analysis - **Tracking:** Portfolio performance tracking; automated quarterly updates; Queries for LP requests - **Audits:** Preparation of audit reports; facilitation of audit memos - **Legal:** screening & summarising long winded legal documents - **Finance:** Cash flow planning; expense analysis - **Fundraising**: LP discovery Given the vast breadth of operations and tooling involved the software and data stack for such functions is typically is fragmented. Hence, the optimisation potential is very high but complex - 9/10. Putting it all together we can draw an analogy to Hans Moravec's framework comparing computer performance with water slowly flooding the landscape. We wonder what this landscape will look like in a decade from now… **in the meantime we are building Arks!** > _Computers are universal machines, their potential extends uniformly over a boundless expanse of tasks. Human potentials, on the other hand, are strong in areas long important for survival, but weak in things far removed. Imagine a "landscape of human competence," having lowlands with labels like "arithmetic" and "rote memorization," foothills like "theorem proving" and "chess playing," and high mountain peaks labeled "locomotion," "hand-eye coordination" and "social interaction." **Advancing computer performance is like water slowly flooding the landscape**. A half century ago it began to drown the lowlands, driving out human calculators and record clerks, but leaving most of us dry. Now the flood has reached the foothills, and our outposts there are contemplating retreat. We feel safe on our peaks, but, at the present rate, those too will be submerged within another half century. **I propose that we build Arks as that day nears, and adopt a seafaring life!**_ Hans Moravec, "When Will Computer Hardware Match the Human Brain?" Journal of Evolution and Technology (1998), vol. 1. Onwards & upwards Team Inflection --- # The great reset *Published: 2022-11-30 | Author: Alexander Lange | Section: Markets* URL: https://inflection.fund/writings/the-great-reset The last few months have been particularly challenging for our industry. In the midst of interest rate hikes, continuously high inflation and a hot war in Europe, FTX imploded with still [unknown consequences](https://www.pymnts.com/cryptocurrency/2022/ftx-us-and-alameda-research-each-have-liabilities-of-10b-to-50b/?utm_campaign=Sunday%20Newsletter&utm_medium=email&_hsmi=233795918&_hsenc=p2ANqtz--0xAFSD5LNZPE-gIIUhCDhgljJRVHGmvdYvIpEzk-lWKquvNe3iuhyATSCgqUcb0UrAs0YpiLZL7PfCDdQV5ygcn39kA&utm_content=233795918&utm_source=hs_email) in terms of contagion effects, reputational and potentially regulatory damage. We are deeply saddened by such developments and are doing our best to help affected parties cope with the fallout. ![](https://bucketeer-e05bbc84-baa3-437e-9518-adb32be77984.s3.amazonaws.com/public/images/a9801c9c-a6c8-4efd-b022-f0d4d4308c3c_5998x3992.png) We observe a high variance amongst market participants when it comes to mental coping mechanisms in response to such events. Many founders and investors suffer from symptoms of high stress, fatigue and depression. In the field of [positive psychology](https://en.wikipedia.org/wiki/Positive_psychology) the [reframing of thoughts](https://positivepsychology.com/cbt-cognitive-restructuring-cognitive-distortions/) is a powerful cognitive restructuring technique. Let's reframe our thoughts on the state of crypto and take a step back. In bull markets things are rarely as promising as they appear; in bear markets they are rarely as bad as they seem. So where are we as an industry? ## The intimate core remains Through bull cycles our communities get diluted and distracted. Same patterns, different scale: Once prices pick up shorttermism and greed undermine the hard work committed innovators put up. Ignorant mainstream media add fuel to the fire with over-blown hybris and flawed narratives seducing the broader public to speculate. Tourist investors join the market and bid up prices with low conviction and head to the doors as soon as things start to become more challenging. Founders fail by optimising for fast, easy money instead of challenging their investor's motivation, conviction and long term commitment to the company. Sophisticated investors continue to invest in such environment, often at accelerated deployment pace instead of sitting on their hands. The more hype, the more money is thrown around and the steeper and longer the bear remains. Yin and yang. In between the cycles the bubble burst and clears the air. Tourists, opaque lenders and paid personalities are removed from the ecosystem. We now have a time window to reflect, adapt our assumptions and move on. In that time window lies the incredible opportunity to [create a culture of intimacy](https://hun3y.mirror.xyz/2WlK39OeCNdWgziiE4XUT7HTxfHSCx8N5BCyxxgJRK0) which can prepare us for future cycles. Intimacy can be achieved through strong inter-member empathy and an explicit commitment to our shared core values such as - transparency - e.g. [proofs of solvency](https://vitalik.ca/general/2022/11/19/proof_of_solvency.html) - [credible neutrality](https://nakamoto.com/credible-neutrality/) - self custody and user control - long termism - better vesting schemes, better incentive alignment between stakeholders - accountability amongst others. Products > words. People > tech. Let's bond in difficult times; now is the time for bear market dinners, research and reflection groups. Reach out to your peers and check in on how they are doing. Let's cut the noise and shrink towards higher productivity and more purposeful building. ## Back to the future: reducing agency costs is crypto's killer feature The [principal–agent problem](https://www.investopedia.com/terms/p/principal-agent-problem.asp) refers to the conflict in interests and priorities that arises when one person or entity (agent) takes actions on behalf of another person or entity (principal). A principal-agent problem is exactly what unfolded in the case of FTX where an over-leveraged organisation turned out to be fraudulent. The original sin of FTX was an obvious conflict of interest rooted in the founder's ownership in the Alameda Hedge Fund which traded on the FTX exchange. These kinds of challenges are as old as human kind. The larger the agent's influence within a market, the worse the fallout in case of failure. Once it reaches a critical mass it can become 'system-relevant' or 'too big to fail' can lead to government bail outs - remember 2008. The patterns revealed through principal-agent problems inspired the creation of Bitcoin in the first place. We've come full circle. ![](https://bucketeer-e05bbc84-baa3-437e-9518-adb32be77984.s3.amazonaws.com/public/images/46a00d75-1a93-41c7-b726-72fe3b048114_1342x927.jpeg) The challenge of agent failures will increase in the future. Centralised agents controlling data and algorithms are growing their influence through data-network-effects and regulatory entry barriers. Data breaches will continue to accelerate in frequency and scope. Self sovereign data ownership and control over public infrastructure are the alternative path crypto is pioneering. This is as important as ever. ## Self custody is finally embraced While Mt. Gox teached us the very same lessons as FTX the half-life of its learnings was disappointingly short. It seems that users need to be reminded of the importance of self-custody through pain every now and then. Experience > conventional wisdom. This reality is starting to sink in quickly, let's hope it last long. Self-custody is be the bed-rock of our industry. ## Crypto has escape velocity In the context of physics an object has escape velocity once they reach the speed necessary in order to break away from the gravitational force of a large body. This concept can be abstracted and applied to crypto's market conditions. Velocity is defined by talent, capital, use cases and adoption. The large body's gravitational force is the status quo of big banks and big tech. Other than in previous cycles we can refer to quite a list of impressive achievements: Individuals in need flock to Bitcoin and crypto infra - ukranian refugees, iranian protestors, Turks fighting with hyper inflation. Nation states and publicly listed companies hold Bitcoin. Ethereum merged smoothly. Defi protocols locking up >$40BN in assets; Dexes processed $100BN in October 2022. NFTs became a horizontal data wrapper and indicated their potential through POCs. >$10BN of patient capital are waiting on the sidelines to enter the industry - from venture funds alone. Regulators started taking our industry seriously and we developed a global network of advocacy groups defending our interests against big tech and big banks. We observe crypto technologies being deployed against problem spaces outside of speculative use cases like #DeSci, #Refi, #DAOs and #DeSo amongst others. In parallel we see a new wave of infrastructure innovation break into the industry such as #AI\*Crypto, #ZKP systems with underpinning #hardware, #TSS as well as overall #security and #UX improvements. The velocity of exploring new use cases and unlocking entirely new users groups is accelerating. An extended bear market slows down the industry for a while but this is almost a necessity given that our current infrastructure (aka scalability, privacy, security and UX) is still in beta stage. > "We're convinced crypto will scale and be adopted. The problem is if it scales faster than self-custody becomes user-friendly. In that case we end up reproducing web2 with a new economic model." - [Henri Stern](https://twitter.com/henri_stern), CEO Privy An extended bear market is a healthy and welcome breather. The core principles underpinning crypto technologies continue to gain momentum over time. Let's not lose sight of that. And most importantly, Let's not lose hope. --- # DAOday.wtf was a blast *Published: 2022-10-17 | Author: Jonatan Luther-Bergquist | Section: Events* URL: https://inflection.fund/writings/daoday-wtf-was-a-blast Just before ETHBerlin, [Inflection](http://inflection.xyz) hosted a workshop on DAO operations, governance and purpose. Check out the home page at [daoday.wtf](http://www.daoday.wtf) for more context. Below is a summary of the talks and conclusions of the discussions. ## Purpose of workshop - Put people with diverse, deep experiences and thoughts on DAO building in one room - Figure out what works, what doesn't work so far within DAOs - Find out what we can do to make DAOs better ![Credit to Michael Matlon from Unsplash](https://bucketeer-e05bbc84-baa3-437e-9518-adb32be77984.s3.amazonaws.com/public/images/2b4c305a-c801-4aee-a235-e1f359bacc0d_4256x2655.png) ## 0\. Inflection shared troubled thoughts on DAOs \[[slides](https://docs.google.com/presentation/d/19kezb2HxzC5BOYbbGF-Twjxha9lEPj33NN1WVKwMN2g/edit?usp=sharing)\] We started off the conversation by sharing our very honest opinion on the state of DAOs today. Most DAOs are very inefficient, and they're not turning out to be the decentral, autonomous coordination tool originally imagined. To try to turn the tide, we decided to put on this event to highlight struggles in DAOs and DAO service providers within our ecosystem. The worst three DAO problems right now: 1. Lack of clear objective/purpose 2. Governance complexity and protocol value not maturing at same speed 3. Putting newspapers on screen aka. a regular company but calling it a DAO These problems are further worsened by the differing opinions on what DAOs should do and be like. There is no ideological consensus on what a DAO is, even within DAOs. This makes it more virtually impossible to be efficient and have progress. It's very hard to balance delivering product and managing sometimes 1000s of stakeholders in what is essentially a startup setting. Imagine giving out stock options before there is even a user or a clear purpose of the product! How can you expect to build what is effectively a startup in this setting? Finally, the basic mathematics of organizations this size without clear paths for autonomy and parallelization are doomed to become bureaucracies. Today, scaling out a DAO to more members is welcomed and even actively pursued, even though it's not a proof of value delivered to the community. It's, probably even counter-productive. For a more positive spin, which we also agree with, please read [Vitalik's recent post](https://vitalik.ca/general/2022/09/20/daos.html) on where DAOs make sense. :) ## 1\. Centrifuge on preparing their pool onboarding process \[[slides](https://docs.google.com/presentation/d/1l7E4ESMqh7tpEgYJdWu1RbXsZjN6krjPgOGwQ2SDfcY/edit#slide=id.gf680904b1a_0_44)\] Centrifuge is one of the longest existing protocols in DeFi. They are undergoing a progressive decentralization since some time. Over the course of time they've experienced 1st hand how - Ownership - Governance - Risk are a triad that needs to be carefully managed in DeFi. A key components of this management is staking for sybil protection and incentive alignment. Currently they are re-defining the process for onboarding new [Pools](https://medium.com/centrifuge/pools-on-centrifuge-a-faq-for-issuers-9baa16f78a6c). There are 7 steps with two thresholds, of quality. ![](https://bucketeer-e05bbc84-baa3-437e-9518-adb32be77984.s3.amazonaws.com/public/images/bb598563-e115-4d5a-bf8c-f0de948cea19_1632x1198.png) Curation of new pools should be as much in the hands of the community as possible. There must also be information transparency built into the process, while enabling human decision making. The results after 4 months was that 9 projects had started the process of being onboarded, however, as with most DAOs, the participation in voting and discussion around proposals was not very high. The financial risk assessment of the pools requires a great level of expertise, but the experts in the community have not been very active in this process so far. To facilitate community activation, Centrifuge will propose a new version of the POP, with less complexity. They finished by asking the audience for suggestions on how to manage making difficult decisions that require expert knowledge, while maintaining community control. #### Outcomes of discussion 1. Create an output-oriented reputation system Building a reputation as an expert, and qualifying as someone who can accurately judge pools, is very complex. "Old world" proxies such as education, experience within TradFi, etc. may not be the criteria that we want to build upon. A suggestion that came up was instead to create a live game of predicting pool success based on open data, and thereby allowing for people to build a reputation in the specific task they would be performing as experts **Pro**: - Specificity to the actual work - Limits discrimination according to "old world criteria" **Con**: - No skin in the game, hence prediction game may be gamed - Few data points on what makes a successful pool at this point 1. Time-weight votes on proposals By adding the time-variable to votes, we can judge how long someone is convinced of a certain vote. This can be modelled linearly but probably logarithmically makes more sense as understanding of a proposal will probably plateau after some time. **Pro**: - May influence vote outcome according to understanding/conviction of proposal **Con**: - Time in vote and conviction/knowledge are not necessarily correlated 1. Short, or negative votes on proposals should be included as well There is a bias introduced by only having positive votes and abstaining. By down-voting proposals, we may get better outcomes. If we include financial incentives in the shorting (defaulting being equal to the proposal not being accepted, or it being proportional to the Yay/Nay vote ratio), then secondary motives may be stronger than actually finding the best pool. **Pro**: - Gives a larger set of information from voters - People with strong conviction may produce stronger signals **Con**: - Financial incentives not necessarily aligned to Pool outcomes 1. Tie expert compensation very closely to pool outcomes Aligned incentives among experts and community should be a good way to onboard better pools. One way of doing this is by shifting compensation from fixed to variable and tying the variable to pool performance and the experts recommendation for that pool. **Pro**: - Ensures experts work towards - People with strong conviction may produce stronger signals ## 2\. Spearbit ensure quality as a service DAO \[[slides](https://drive.google.com/file/d/10_A_HMA1ikRLxI8y2jL6sAMJwLQyo0UX/view)\] [Spearbit](http://spearbit.com) is a freelance marketplace of vetted security researchers who pair on collaborative teams to conduct security reviews with clients. They're redefining the traditional notions of security audits as a box to check and a stamp to get into a collaborative, ongoing effort to prevent any types of unintended behaviour. The security market is going towards a combination of community-driven, decentralized approach to security and centralized security firms. In the evolution of talent marketplaces, there is a range from completely managed ones such as kaggle, Cod4rena to free-er models like Fiverr, to Braintrust and finally ending in DAO-driven ones such as Spearbit. Spearbit matches audits with teams of freelance security experts. Part of the audits is to train apprentices and junior security researchers in live audits, while also providing an environment for sharing knowledge. Clients get the certainty of the highest quality, at faster speeds than centralized firms, and auditors get the best compensation in the industry. There are 64 auditors who have completed an audit with Spearbit since inception, and earned a total of $3.3M revenue to the DAO. Security auditing in crypto is a very **imbalanced market**, with much higher demand for audits than qualified auditors. The logical way to scale is by upskilling security professionals from web2 or skilled smart contract developers. Both Spearbit and Secureum are working on this, but will not be able to scale infinitely. They don't expect to grow beyond ~120 auditors ever. Another challenge is protecting against black hats, who could infiltrate projects, discover vulnerabilities and not disclose them, only to exploit them after the project went live. Spearbit has resolved this by performing full KYC and other methods to vet all team members. However, the most critical challenge for any service DAO will be ensuring quality of services, while remaining open and effective as they scale. This would require establishing accurate, scalable **reputation systems** ([Otterspace](http://otterspace.xyz) badges for example) and proofs of expertise ([GitPOAPs](http://gitpoap.io) for example). #### Outcomes of discussion The DAO is currently centralized in that there is a core team that guides decision making, but Spearbit is aiming to become more than a traditional org, on-chain. They are already achieving much of the benefits of what we usually call a DAO: - Transparency in fees, rewards and decision making - Take rates are completely transparent through a multisig - 20% is flipping the service provider model on its head in terms of fee distribution - Open onboarding of people who are not in traditional financial systems ("unbanked") They also reach allow for some of the benefits of traditional organizational structures: - Shared liability - Sense of community and knowledge sharing - Ability to work with legal entities They struggle with the aforementioned challenges, but also with the question on where value accrues. As with any service provider, the system and structures are supporting the people providing the services. But if the people change, it's hard to ensure the same quality output. Like a [Thesean ship](https://en.wikipedia.org/wiki/Ship_of_Theseus), if we exchange the atoms, can the DNA remain the same? On a more concrete level, how to enable the sharing of upside among contributors, when there is no clear case for a token. ## 3\. Snapshot work on increasing voter participation [Snapshot](https://snapshot.org/#/) has taken governance participation by storm. Snapshot uses IPFS to collect votes rather than onchain transactions. As such, they've removed the need for voters to pay for on-chain votes and reduced friction in the process. Now they are working on how to actually incentivize voters to vote. Most DAOs operate with a vote-based governance system, in fact, very often the same type of governance based vote system with a liquid ERC-20 token where 1 token corresponds to 1 vote and token holders stake their tokens in governance contracts on specific proposals in order to cast their vote. As such, many of these ecosystems fall prey to voter apathy. When a few whales drive the majority of governance votes, it can be a challenge to incentivize voters to be active and participate. Voter participation is likely to have very real implications for protocols as it may be classified as a core criterion for decentralization in a regulatory context. One of Snapshot's goals in the near term is to improve and iterate on the models for governance of DAOs. One example of an alternative governance model would be Decentraland, where voters vote based on estate size, token ownership and a few other factors. Given the degree of uncertainty and the severity of the consequences, developers and other stakeholders have a vested interest in how decentralization of DAOs is perceived. #### Outcomes of discussion - Equating token ownership with rationality as a voter, and completely aligned incentives is probably wrong - The curve of marginal benefit in outcome for the DAO vs. voter participation is probably not linear, but more like $-x^2$ - Voter apathy can be because of - Unfriendly UX - high threshold of participating - Flawed communication - not clear to voters why a vote that is important to them is important - Misaligned incentives - outcomes of single votes are not individually important enough Snapshop tries to solve what they can affect, which is mostly around voter UX, but DAOs need to take care of the remaining parts. Namely how to pull in the right voters for the right decisions. ## 4\. MakerDAO governance chronology has all the juice MakerDAO is a decentralized lending platform and is one of the most well known and heavily used DeFi projects. It also boasts a very complex and nuanced governance ecosystem. Maker recently decentralized its operations and moved away from a central foundation toward a series core unit teams. One part of this governance system is the MIP or Maker Improvement Proposal. These proposals are very involved and require a lot of information to be completed. As such, they create a floor on the level of effort and friction needed to pass governance actions. We had the pleasure of having someone from MakerDAO participate in our workshop, and give us a chronology of some high-level changes to the MakerDAO governance since its inception. To summarize, there has been multiple waves of decentralization and centralization contractions/expansions over the years. MakerDAO governance has very much been centred around regulatory risk mitigation and less about making ideal decisions as an on-chain entity. There was the MakerDAO foundation, where the founders and core teams were not allowed to participate publicly, but still had valuable input on decision making obviously (shadow board). There has been the recent decision on onboarding of Monetalis collateral, where there are unclear incentives and the most recent LOVE controversial vote. In the bear market, things have been brought to a point where Maker is pushed to make a profit and pivot rapidly. This has proven to be quite a difficult task, given dysfunctional governance and MKR votes being mostly made up of whales (VCs included). Instead of risking to misrepresent the story, here are a couple of publicly available viewpoints of governance controversy at MakerDAO: - General chronology [ ![](https://bucketeer-e05bbc84-baa3-437e-9518-adb32be77984.s3.amazonaws.com/public/images/381d11b2-6ef0-4a37-98c0-be3a2ed30202_591x591.png)Dirt Roads \# 42 | Valkyrie: MakerDAO and Our Side of History 2 AM. The adrenaline from having rallied a good portion of crypto's investing society still pumping in. Without having met almost any of them in person. The thick Italian heat. Drinking merely sparkling w… Read more 4 years ago · 49 likes · Luca Prosperi ](https://dirtroads.substack.com/p/-42-valkyrie-makerdao-and-our-side) - - [https://www.reverie.ooo/podcast-episode/rune-christensen-makerdaos-past-present-and-future](https://www.reverie.ooo/podcast-episode/rune-christensen-makerdaos-past-present-and-future) - [https://www.reverie.ooo/podcast-episode/monetsupply-perspectives-on-makerdao-governance](https://www.reverie.ooo/podcast-episode/monetsupply-perspectives-on-makerdao-governance) - Move from foundation - [https://www.coindesk.com/tech/2021/07/20/makerdao-moves-to-full-decentralization-maker-foundation-to-close-in-months/](https://www.coindesk.com/tech/2021/07/20/makerdao-moves-to-full-decentralization-maker-foundation-to-close-in-months/) - Endgame-vote - [https://forum.makerdao.com/t/simple-makerdao-governance-from-first-principles/16207](https://forum.makerdao.com/t/simple-makerdao-governance-from-first-principles/16207) - [https://forum.makerdao.com/t/the-business-thinktank-for-scientific-governance/15838](https://forum.makerdao.com/t/the-business-thinktank-for-scientific-governance/15838) ## 5\. RnDAO can measure community health [RnDAO](https://rndao.info/) is a research DAO focussed on understanding DAOs better, and helping DAOs learn from each other to operate better. A recent topic they've tackled is **Community health**. Specifically they ask: - What are the features of a community? - How are DAO communities different from others? - When is a DAO community healthy? - How can this health be measured? The full extent of the research results can be found [here](https://rndao.mirror.xyz/F-SMj6p_jdYvrMMkR1d9Hd6YbEg39qItTKfjo-zkgqM), but the tl;dr as presented goes as follows: Community boils down to a combination of people, place, identification and association. A healthy community - has an environment that helps people thrive - embodies the values as defined by the community - gives people a sense of belonging - allows for human connection among the individuals Given this, we can break down the health indicators into measurable things, but still, quantifying how much values defined map to values lived is very difficult, given how imprecise and liable to bias qualitative surveys are. What one has to do instead is focus on proxies, such as - engagements between members - e.g., through message graphs - engagement with the product - e.g., usage of services offered by DAO - attrition - e.g., MAU/quitting rates RnDAO is coming out with a 38-page summary of their findings on how to best measure health within DAO communities, including 10 key metrics they suggest. Unfortunately, there was not much time for discussion on the final presentation, hence little feedback or suggestions came from the participants. In conclusion, we believe research efforts both academic and independent, such as RnDAO, to be crucial to understand how DAOs actually function today and should function in the future. With respect to the specific topic of community health, we'd like to see someone measure it in different DAOs. Then we can start making accurate statements of how well a DAO does what it is supposed to do, which is, in part, to have a healthy community. # Outlook on DAOs DAOs are not the optimal human coordination mechanism it was thought to be. Many talented people are pouring in enormous resources into making DAOs work efficiently for purposes that would be much better suited as legal structures with clear incentives and objectives. Many DAOs are formed for the "wrong" reasons in that they exist out of regulatory necessity (not long term sustainable), to attract buzzword investors, or to mislead retail investors. While we understand that compliance is in many cases not possible while operating at the cutting edge of innovation (see Uber), avoiding regulatory oversight is not long-term feasible. For it to be effective, it should lead to an adaptation of regulation long-term, while providing all the benefits and causing as little harm as possible in the mean time. This means, coordinating capital and efforts online, while reducing the amount of In our opinion, there are three forms of organizations that are actually well suited to being true DAOs/on-chain organizations/[internet-native corporations](https://mirror.xyz/0x24BfafEe350b76b4c73422A8f54Ada445d938Cd0/RE1__ME594v4X6W-y5xgL_lD8GZXP77P3mJVbBipKlQ): 1. Service providers/freelance marketplaces 1. Spearbit 2. VektorDAO 3. IndieDAO 2. Off-path research efforts (insert DeSci-meme) that enables people to work on topics not fundable within traditional academia or industry R&D in constellations not possible 1. VitaDAO 2. PsyDAO 3. CrunchDAO 4. RnDAO 3. Low stakes, for fun communities coordinating around digital assets 1. Gaming guilds 2. nounsDAO 3. art projects That being said, we want to further the experimentation and learning from the last decade of on-chain community and company building. What we struggle with in building any organization today, will necessarily be part of the equation in on-chain organizations, but we can try to minimize the coefficiencts of those factors. There are great examples of experimentation in both governance and on-chain organizations, such as outlined in [our last post on on-chain organizations](https://svrgn.substack.com/p/on-chain-organizations-need-social). --- # Bitcoin and crypto networks as climate positive technologies *Published: 2022-10-04 | Author: Alexander Lange | Section: Research* URL: https://inflection.fund/writings/bitcoin-crypto-networks-climate-positive-technologies On-chain organizations (aka DAOs) have been an experiment in new forms of social coordination to achieve a variety of purposes (decision making, collaborative work, incentive alignment etc). In web3, organizations have the [potential](https://vitalik.ca/general/2022/09/20/daos.html) to be permissionless, trustless, distributed, and transparent. In traditional design spaces (web2), however, trust and permissioning are cornerstones of social coordination and even privacy is often a key function. In order to achieve their full potential, on-chain organizations [will eventually need](https://papers.ssrn.com/sol3/papers.cfm?abstract_id=4105763) to be able to incorporate concepts such as trust networks and reputation. One key component that will aid the evolution of on-chain organizations is the social operating system. Such a system would entail (1) programmatic governance, (2) [on-chain reputations](https://svrgn.substack.com/p/how-professional-credentials-unlock), (3) pseudonymous or selectively private identities, and (4) tokenized roles and rights within the organization. ![](https://bucketeer-e05bbc84-baa3-437e-9518-adb32be77984.s3.amazonaws.com/public/images/7dcf81c4-39ef-4257-9fea-632f43b7482c_2399x1533.png) Some of the organizational problems such an operating system would solve (or at least improve): (1) Hyperfinancialization (2) Plutocratic governance (3) Excessive centralization (4) Inefficient workflows This is not an exhaustive list and also there is no silver bullet. The most important takeaway is that we must continue to improve and iterate. Web3 projects made several [leaps](https://github.com/makerdao/community/blob/master/governance/governance-and-risk-meetings/summaries/episode-36.md) in experimenting with on-chain organizational structures early on. These early examples were very successful at yielding new outcomes and creating value. As a result, many new projects simply [replicated](https://wiki.tally.xyz/docs/compound-governor) these existing structures (often with literal [forks](https://github.com/compound-finance/compound-protocol/network/members) of the governance codebase). It is [time now](https://twitter.com/VitalikButerin/status/1571995444200280065?s=20&t=TdypjdNMbsrfn890HiIELw), however, for the industry to continue the pace of innovation in this dimension as well as the many others being worked on. Web3 needs better governance models, improved organizational structures, and ultimately an updated version of a social operating system. ## Dissecting the problem set Let's take a look at the role of a social operating system in tackling some of the above problems: ### Hyperfinancialization Most DeFi users would probably agree that the space has trended toward a lot of speculative use cases. Tokenization has been a central concept for many projects. Also, most major governance systems have been inexorably tethered to conversations of liquidity and profitability due to the nature of their governance tokens also being value-accrual assets. This has been particularly relevant for some DeFi protocols dealing with governance [growing pains](https://twitter.com/eaglelex_eth/status/1563983617080049665?s=20&t=GWJVbXHcPeq6WLw4_QnUdw) or [design challenges](https://blog.makerdao.com/maker-dschief-1-2-governance-security-update-requires-mkr-holder-actions/). Financial use cases and governance may at times overlap, but from a first principles standpoint do not have to be inseparable. As mentioned earlier, this is primarily a result of convenience, skeuomorphism and technical debt. It is hard to create new models and audit them for security risks; it's often much easier and safer to fork an existing model. As such, monolithic do-it-all governance tokens and the [1-token-1-vote model](https://twitter.com/VitalikButerin/status/1427152573002248196) have propagated to the vast majority of on-chain organizations. With alternative governance models (eg [NFT based voting](https://blog.tally.xyz/nouns-dao-first-nft-dao-on-tally-de67dc31054e)) we could break down these legacy assumptions, separate functionality as needed, and explore less egregiously financialized mechanisms. This could either be a replacement or, more likely, a supplement to existing models. For example, financial layers could always be overlaid on top of non-liquid governance structures. Perhaps this would free up web3 organizations or teams within them to focus a bit more on value creation and a bit less on value capture. ### Plutocratic governance Another aspect of 1-token-1-vote models is that they often result in uneven [distributions](https://cointelegraph.com/news/less-than-1-of-all-holders-have-90-of-the-voting-power-in-daos-report) of voting power (such as [plutocracy](https://en.wikipedia.org/wiki/Plutocracy)). This can also result in a knock-on [effect](https://twitter.com/RobertoTalamas/status/1466801457315913740?s=20&t=wh0zbwMRkZkmjvx7fX0UDg) of discouraged or disincentivized voters. Often, there is misalignment between those that do the work, research problems and solutions, and stay up to date with the latest developments within an org and those that have the power to vote on key decisions. One solution that we have stumbled upon as an industry is [vote delegation](https://www.reddit.com/r/MakerDAO/comments/ox27xn/delegation_is_live_in_makerdao/). This is helpful, but may only be a stopgap to resolving this delta. This is where on-chain badges (NFTs) to represent roles, rights, and permissions with an on-chain organization can become quite critical. Additionally, this ties to the idea of pseudonymous or selectively-private on-chain identities and the associated reputation systems. For example, imagine a world where your wallet (or other identity primitive) is associated with every role you've ever held in an organization (and the associated permissions). Of course, such a system would have to be non-transferable as [mentioned](https://deliverypdf.ssrn.com/delivery.php?ID=157098114001097100082004080122066065050032046018055082096069088103113077127093004121032021048000110046110117075022093093064122058047001053003030070085084067108072027029049067072000001069085004095121089127083082097121068093077071031122083088124088081022&EXT=pdf&INDEX=TRUE) by Vitalik earlier this year. This could allow an organization to, for example, upweight the votes of project contributors or restrict access to submitting proposals to members with an "investor" role in an on-chain [investment org](https://syndicate.io/). Most importantly, such actions could be independent from dilution or distortion of value flows and ownership. Greater flexibility, customizability, and articulation of these kinds of tools could allow organizations to better avoid misalignment of power and individual contribution. ### Excessive centralization Decentralization is not a monolith. There is no one property that is decentralization. There are many different aspects of an organization that can be decentralized to greater or lesser degrees including but not limited to: [governance](https://cointelegraph.com/news/less-than-1-of-all-holders-have-90-of-the-voting-power-in-daos-report), ownership, operations, and technical architecture. As mentioned above, a system of on-chain badges and reputation can be very helpful for organizations to improve decentralization of governance but that is not where it stops. As on-chain organizations scale, anti-sybil protections become [increasingly important](https://gov.gitcoin.co/t/knowledge-transfer-characterizing-the-sybil-resistance-problem/11235) for ensuring proper decentralization of ownership and operations and avoiding [decentralization theater](https://www.coindesk.com/markets/2021/08/19/gary-gensler-isnt-buying-your-decentralization-theater/). This is also where [non-transferable](https://thedefiant.io/vitalik-soulbound-tokens) tokens can play a role. As the value of someone's reputation and roles grows and as on-chain histories become more important, the cost (equivalent to economic friction) of "rotating" or selling that wallet, ENS, or other associated identity also increases (and thus we can expect it to happen less often). Though nothing is perfect, this does mean we can reasonably expect a robust social operating system to reduce sybil issues and allow on-chain organizations to more easily decentralize. ### Inefficient workflows A big part of the efficiency of web2 "off-chain" organizations is that they are primarily structured as top-down tiered management systems. This is a very social model that rests heavily on assumptions of trust and permissions. As mentioned earlier, on-chain organizations excel at decentralizing location and operations but this is often at the expense of [efficiency](https://banklessdao.substack.com/p/dao-efficiency) and individual [accountability](https://decrypt.co/91325/makerdao-content-team-fired-mkr-holders). Without clear on-chain records of outcomes and reputational components, it is [hard](https://blockworks.co/contributors-need-to-be-more-accountable-to-the-daos-they-serve/) to hold individuals accountable to those outcomes and thus organizations in the web3 space (particularly where mixed with flat org structures) will tend toward only positive feedback loops and not negative feedback loops as well. It is a well documented systems design [principle](https://www.albert.io/blog/positive-negative-feedback-loops-biology/#:~:text=Positive%20feedback%20occurs%20to%20increase,back%20to%20a%20stable%20state.) that you can better optimize a system when you have both tools available. As such, the on-chain reputation (with outcome tracking) and the identity aspect of social operating systems should improve productivity and efficiency via accountability. Additionally, as NFT-badges and associated rights (including the right to distribute rights) become more programmatic and reliable, it should become easier to decentralize and delegate operations as well. Additionally, we mentioned the idea that badges and on-chain roles and rights associated therein improve the ability to delegate and distribute work within an organization. If a core team lead is able to pick 5 delegates and programmatically define the rights and responsibilities of those delegates, then they can more reliably distribute work rather than taking it on themselves for lack of supporting infrastructure (e.g. Superrare [space curators](https://docs.superrare.com/whitepapers/master/the-next-generation-nft-platform/superrare-spaces)), improving total productivity of the org. ## Introducing: Otterspace, a social operating system for on-chain organizations We are happy to announce our investment into [Otterspace](https://www.otterspace.xyz/), a social operating system enabling on-chain organizations to codify roles and rights, onboard and reward community members, distribute badges and track achievements, manage and curate project contributors, and apply anti-sybil protections. We've been co-leading their $3.7M pre seed round alongside Cherry Crypto and with participation from Coinbase Ventures, Bessemer Venture Partners, btov Partners, Paua Ventures and several knowledgeable angel investors including Mats & Fredrik (Dune cofounders), Sandeep Nailwal (Polygon cofounder), Will Papper (SyndicateDAO), NiMA (SeedClub), Quickrider (FlamingoDAO), Abbey Titcomb (RadicleDAO), and Sarah Drinkwater (Gitcoin). Their first [product](https://otterspace.mirror.xyz/aN5_EepB1x_En12_gzTNyZuvMUmbXbwKnISc2rEg7a0) is a protocol for issuing [badges](https://beta.otterspace.xyz/badges/create/name) as [soulbound](http://soulbound) (or non-transferrable) NFTs. This is a critical initial piece in the above outlined vision of a true social operating system. NFT-based badges provide a data layer on top of which organizations can program logic, rights, and responsibilities for doing work, making decisions, and distributing value. Web3 organizations currently struggle with onboarding and growing their communities, tracking and rewarding off-chain contributions, coordinating and curating human talent, and managing access and control 2-3 layers deep. Otterspace's system of badges and infra for using them as composable logic primitives in configuring rights within on-chain organizations will help push web3 forward toward achieving the true potential originally promised by the DAO model. [Otterspace](https://otterspace.xyz/) is in a private beta stage with 16 DAOs participating, including: Radicle, Bankless and Token Engineering Academy. These organizations are able to integrate the protocol natively into their own infrastructure. Integrations with [Snapshot](https://snapshot.org/#/), [Guild.xyz](https://guild.xyz/our-guild), and [Gnosis Safe](https://gnosis-safe.io/app/) allow communities to use the badges for governance, social access (Twitter, Discord, Telegram), permissioning, and more. One day soon, on-chain organizations will be able to use such social operating systems to manage and organize themselves with greater efficiency and productivity while maintaining the transparency and resilience that comes from operating a decentralized on-chain organization. They have the potential to combine the best parts of social coordination/curation and programmatic logic. Of course, this is an optimistic projection and depends on how well companies like Otterspace build these tools and how well organizations implement them. That said, we are hopeful and the future of on-chain organizations looks brighter than ever. If you are interested in joining the Otterspace community and learning more: check out their [website](https://otterspace.xyz/) or [twitter](https://twitter.com/otterspace_xyz/?utm_source=landing_page&utm_medium=website&utm_campaign=new_landing_page) page. [Here you can find their launch and funding announcement.](https://otterspace.mirror.xyz/uthG_cJF8k3mE5RAVeF22MgFYuUpadEJPNMgtTAQBxY) --- # Eth post-merge data tools *Published: 2022-09-27 | Author: Jonatan Luther-Bergquist | Section: Research* URL: https://inflection.fund/writings/eth-post-merge-data-tools ## Intro With the Merge going through successfully, Ethereum has changed permanently. If you are not familiar with the details, [here's a good primer](https://coinmetrics.io/special-insights/ethereum-merge/). The new process for adding transactions to the chain has significant trade-offs with respect to [energy usage](https://ethereum.org/en/energy-consumption/), [MEV](https://github.com/flashbots/eth2-research), [censorship](https://notes.ethereum.org/@vbuterin/pbs_censorship_resistance)\-[resistance](https://github.com/flashbots/mev-boost/issues/215) and [centralization](https://noxx.substack.com/p/order-flows-kingmaker-of-the-block). The data is crucial to monitor, track and interpret progress of these dimensions. This post intends to provide a coherent library of twitter accounts, dashboards, data tools and other lists. We will do our best to keep it up to date, although this space moves very quickly. If you have feedback, additions or questions, feel free to reach out. For the full page visit: [mergedata.lol](http://www.mergedata.lol) or [mergedata.boats](http://mergedata.boats) ![](https://bucketeer-e05bbc84-baa3-437e-9518-adb32be77984.s3.amazonaws.com/public/images/2e00f9a0-5178-463a-a226-a059979467e6_7499x5030.png) ## Glossary ### Types There are four types of resources: 1. **Twitter accounts** to follow to understand the space (excl. accounts of the dashboards), e.g., @blink\_labs\_xyz 2. **Dashboards** of live data, e.g., [ultrasound.money](http://ultrasound.money) 3. **Repos** of tools you can use to extract data yourself 4. **Meta-lists** which are lists in themselves ### Coverage The resources cover different data points, for ease of use, the resources are classified according to what they are focused on. **Blocks**: blocks built, proposed and finalized **Builder**: proposes blocks to validators, through relays **Burn**: Eth reduction of supply through fee burn due to EIP-1559 **MEV**: Maximal extractable value **MEV-Boost**: Intermediate design towards full PBS, proposed by Flashbots **Validator**: Consensus participant **Relay**: Relays blocks from builders to validators **Rewards**: Eth earned of different participants **Supply**: Amount of Eth in circulation and staked [Full EthData Table here](https://inflectionvc.notion.site/Ethereum-post-merge-data-tools-66a29627f39d474ebc8c55b91f0e1785) ![Screenshot of the table.](https://bucketeer-e05bbc84-baa3-437e-9518-adb32be77984.s3.amazonaws.com/public/images/ba2f285d-fdf4-4d6b-9a59-2a14c4efc525_998x1264.png) [Full EthData Table here](https://inflectionvc.notion.site/Ethereum-post-merge-data-tools-66a29627f39d474ebc8c55b91f0e1785) --- # How professional credentials unlock on chain talent markets *Published: 2022-08-24 | Author: Alexander Lange | Section: Research* URL: https://inflection.fund/writings/professional-credentials-unlock-on-chain-talent-markets Blockchains allow us to store verified events over time. Instead of relying on claims made by individuals or groups that can be corrupted, humanity now has access to a trust anchor. This trust anchor empowers us to create not only monetary and financial systems but also internet native organisations and talent markets at global scale. ![](https://bucketeer-e05bbc84-baa3-437e-9518-adb32be77984.s3.amazonaws.com/public/images/42392fc0-cb48-45c0-96bd-a0b2a5d122e7_7499x5030.png) Some of the key components we'd need to have in place in order for these new types of organisations to thrive are (1) a pseudonymous social graph **(2) professional credentials** (3) internet native organisations (DAOs) (4) programmatic relationships between 1-3 (rights management, governance, workflows) (5) user experience (interfaces, collaboration tools, talent marketplaces) These components are not supposed to be conclusive. What is important to realise is that they are **inter-dependent and partially conditional** for one another. Without a pseudonymous social graph we couldn't issue user owned credentials. Without user owned professional credentials we can't efficiently manage internet native orgs. Points 1-3 are the foundation for the establishment of programmatic relationships between all of them as well as a broader set of applications sitting at the top layer of the stack. ![](https://bucketeer-e05bbc84-baa3-437e-9518-adb32be77984.s3.amazonaws.com/public/images/d5d80989-6887-4a9f-9dd3-13f0f8463c11_1842x964.png) ## Which role do professional credentials play in a crypto enabled future of work? We can think of professional on chain credentials as time stamped certificates of specific actions or contributions. Their application will span every aspect of our professional lives. **Gamified on chain universities** enabled by smart contracts which are pointing to a specific set of exercises and issue professional credentials or payments as a reward can be easily envisioned. Think a gamified codecademy issuing on chain credentials as a reward instead of easy to fake pdf certificates. With more verified and contextualised data available on chain a logical next step would be the creation of **on chain resumes.** A LinkedIn type of social network where users control their data assets and can reveal themselves to the world gradually and where they can own their social graph. From there we could think about new forms of **composable** **talent marketplaces.** Recruiting could become vastly more automatised. Auctions could be set up to finance and execute new projects issuing professional credentials and payments as rewards. **Organisational (DAO) governance and efficiency** can be drastically improved. Verified credentials could streamline processes around specific roles and accountabilities (membership tiers) within an organisation - think tiered memberships and access control to specific content, groups or multi sigs for treasuries. [Soulbound tokens](https://papers.ssrn.com/sol3/papers.cfm?abstract_id=4105763) or irrevocable NFTs will serve as catalysts for much more sophisticated designs going forward. The design space for **programmatic workflows** is vast and barely explored. Finally, **retirement and insurance products** could be built with higher accuracy and efficiency based on such a trove of data. Those developments are in their very infancy and it might take decades for them to crystallize. Where do we start? ## Introducing: GitPOAP, a professional network and contributions oracle for metaverse workers We are happy to announce our investment into [GitPOAP,](https://www.gitpoap.io/) a professional network and contributions oracle for metaverse workers. We've been co-leading their $4.3M Seed round alongside Libertus and with participation from Protocol Labs, Avalanche VC, and several knowledgeable angel investors including Balaji Srinivasan, Patricio Worthalter, Anthony Sassano, Superphiza and Mariano Conti amonst others. Their first product is an oracle for code contributions on Github as developers are natural early adopters within the category. Developers can now have contribution to Web 2.0 and web3 software projects, and organizations be represented by issued blockchain-based badges for meaningful contributions on GitHub. GitPOAP is issuing these badges as [POAPs](https://docs.gitpoap.io/faq) - NFTs that represent the action or contribution supplied. Developer contributions to 617 repositories at the center of the Ethereum ecosystem including the Ethereum [protocol](https://www.gitpoap.io/gh/ethereum) are being memorialized by GitPOAP. Contributions to developer infrastructure and tooling (e.g., Solidity, [OpenZeppelin](https://www.openzeppelin.com/), [HardHat](https://hardhat.org/), web3,js, ethers.js, [web3.py](http://web3.py/)), and many other foundational projects (e.g., [Yearn](https://yearn.finance/#/portfolio), [Ledger](https://www.ledger.com/), [Gitcoin](https://gitcoin.co/), [Gnosis Safe](https://gnosis-safe.io/) & Chain, [Rotki](https://rotki.com/)) have been memorialized by GitPOAPs. Hackathon participants at the MIT Bitcoin Expo and the Staking Gathering hosted by EthStaker at Devconnect used GitPOAP to reward hackathon participants. Going forward additional off-chain sources for verified credentials will be added, particularly for knowledge workers. Complementary **software and tooling** will be provided to equip maintainers with the ability to reward all contributions to their projects, no matter what they are or what form they take. Long term the goal is to increase [adversary interoperability of talent market related software](https://www.eff.org/deeplinks/2019/10/adversarial-interoperability) to break open data silos. We are psyched to be part of the GitPOAP community. We would be a happy if you would become part of it as well! Check out their [website](https://www.gitpoap.io/), [Twitter](https://twitter.com/gitpoap) and join the discussion on [Discord](https://discord.com/invite/qa3mfPvjWm) to participate in the community! --- # Rethinking asset ownership through a user control and experience lens *Published: 2022-06-28 | Author: Jonatan Luther-Bergquist | Section: Research* URL: https://inflection.fund/writings/rethinking-asset-ownership-user-control-experience Since the dawn of time, humans have felt the need to safe guard their most valuable things. Pharaos enslaved thousands to build mausoleums for all the assets they thought they'd need in the afterlife. The first emperor of China, Qin Shi Huang, constructed a necropolis of almost 100 square kilometers of underground tombs to contain representations of an army to guard the emperor in the next life. More recently, as liquidity was desired already in this life, assets have been stored in vaults and treasure chests. ![](https://bucketeer-e05bbc84-baa3-437e-9518-adb32be77984.s3.amazonaws.com/public/images/3b24aee1-54ad-45c9-a3ff-b6a1815ee4a0_5999x3977.png) The usage of these assets changed dramatically once digital representations of them became possible. We had to rely on asset managers to validate and approve each transaction à la banks, card issuers, payments providers etc.. At least until the advent of trustless, cryptographically secured digital assets. As the service providers were cut out of the game the responsibility now shifted to the individual. In the last 12 years users have been forced to store and manage their private keys themselves, or delegate that storage to the web3 equivalent of asset custodians such as Coinbase or Fireblocks. Not only does this undo the trust assumptions and basic premise of permissionless systems but it also compromises on the quality and ease of user experiences. The new generation of service providers, custodians are very costly and not available to an average asset holder. Technically, the predominant way to store and manage digital assets can be organized along a spectrum of hardware- and software-centric solutions. Most institutional grade custodians such as Coinbase, Anchorage or Finoa use a multi layered stack of hardware (HSMs) and software solutions to store assets and manage keys. Rather retail focused approaches like Ledger, Trezor or Foundation separate your key storage from an online machine, and provide a physical barrier for entry. The physical format of a USB-like key with a small display has it's limitations when it comes to usability. Software-centric solutions such as Fireblocks, Metamask or Argent interact with applications directly and usage of the private key is unlocked on the device that is connected to the internet. The direct connection makes for easier use, but also a larger attack surface for hackers. Software-centric asset storage solutions also include smart contract wallets and multi-sigs. A problem that applies to both hardware- and software-centric solutions is that they are highly blockchain-specific, as the account abstraction and signature schemes can vary significantly between different networks. In summary, the trade-offs one has to make when assessing digital asset storage solutions today include: - Usability vs. security - Resilience vs. security - 3rd party control vs. self control (not your keys, not your coins) - Cost vs. own responsibility # Introducing: Entropy - a network of key shards We proudly present our investment in [Entropy](https://entropy.xyz/) - the next step in the evolution of asset storage and usage. Entropy recently [announced](https://twitter.com/entropydotxyz/status/1534566139463979008?s=20&t=duBOTE3RWCrK_uSycGeY8g) a $25M seed raise led by a16z, with participation by Dragonfly, Ethereal, Variant, Coinbase ventures, Robot ventures and Komorebi. Entropy is building a truly flexible, decentralized asset storage and management network for everyday users. They are founded on deep practical cryptography knowledge, a strong focus on user-friendliness and ✨immaculate vibes✨. Technically, they do this by leveraging a t-of-t threshold signature scheme where the shards are distributed across a decentralized network. It will be built as a separate chain, with a Turing-complete system for smart contracts. This allows the owner of assets to programmatically define how specific shards can use the assets on the L1. The vision and technical architecture unlocks new ways we can interact with crypto-assets. A small selection of use cases that could be built using Entropy in a **fully trustless setting** are: - Multi-factor authentication by delegating shards to other devices - Social recovery by delegating a shard with recovery permissions to a friend - Separation of staking reward rights from voting rights Entropy is re-thinking how wallets, keys and permissions in decentralized networks are designed and we are thrilled to support their phenomenal team on the journey. [You can find more info on the project and it's trans, anarchist founder Tux here.](https://techcrunch.com/2022/06/08/the-trans-queer-anarchist-crypto-founder-seed-round-a16z-andreessen-horowitz/?guccounter=1&guce_referrer=aHR0cHM6Ly93d3cuZ29vZ2xlLmNvbS8&guce_referrer_sig=AQAAAKalj2OdPzShoYOUKPxAtDQLD9L1V4PQbv3-eJTC8NKL4IuuGSkpQrHpB0HsR35r4beQ0xFmGI7cJMrin_sdI-Eug2r_cS0Z-wXJKcyZ7V0dFEzOAJrkcEzJd9aLYmnH88DQbcJ2jyVciohOV0rMDNFlVLbdGhyJSMRlAuLUczMv) PS. Entropy is [hiring](https://mirror.xyz/entropy.eth/mZ1Youa80Wb08SMZg2nfp64RHUpaKqyZDpAMwyVla7E) --- # Announcing Inflection's $40M Fund II *Published: 2022-01-19 | Author: Alexander Lange | Section: News* URL: https://inflection.fund/writings/announcing-inflections-40m-fund-ii **BOSTON, 19TH JANUARY, 7AM**: Inflection, the early stage venture firm today announced the launch of Inflection Mercury, a $40.7m fund that will continue to build and invest into the [open economy](https://www.inflection.xyz/thesis). The fund's limited partners include Accolade Partners, Evanston Capital, Isomer, Hutt Capital, Multiple Capital, Christian Angermayer's Presight Partners, Galaxy Digital, DCG, Rockaway as well as Marc Andreessen, Chris Dixon, Bo Shao and Erik Voorhees amongst many others. ![](https://bucketeer-e05bbc84-baa3-437e-9518-adb32be77984.s3.amazonaws.com/public/images/f1a393ec-6d6a-49d2-afb2-17c5714054ed_3998x3930.png) The thesis-driven, first check venture capital firm typically invests between $0.5m and $1.5m, often committing to a decade-long growth plan. It has a unique, community-centric investment approach by sharing an ownership stake in the fund with its portfolio founders and innovator communities. Taking concentrated positions in equity and digital assets alike (private instruments and public crypto market exposure) allows the team to support portfolio founders around go-to-market strategies as well as the provision of key tools and services necessary to bootstrap open source networks and communities. The firm's team members are spread across Boston, Berlin, Munich, Madrid and Denver enabling them to innovate globally and build bridges across the atlantic. Inflection Mercury has been launched following success in a number of investments that date back to early 2019 such as [Balancer](https://balancer.fi/) - the programmatic asset management protocol, [Radicle](https://radicle.xyz/) - a peer-to-peer code collaboration platform or [Anytype](https://anytype.io/en) - an operating environment built on web3. More recently and as part of the new Mercury fund's portfolio, Inflection invested into companies such as [Unstoppable Finance](https://unstoppable.fi/) - a neo bank run on DeFi rails, [Violet](https://violet.co/) - the Ethereum-based identity protocol, [Defined](https://www.defined.fi/) - the open data engine for web3 or [Catalog](https://beta.catalog.works/) - an open archive for music powered by NFTs. Besides growing the portfolio, Inflection will continue its team expansion around data science and engineering roles. Alexander Lange, GP and founder of Inflection, said > "The launch of our new Mercury fund reinforces our commitment to invest in the future of an internet native, open economy. Alongside our founders we explore new frontiers where the collective benefits from open access to markets, data and knowledge while the sovereign individual is empowered through privacy and ownership. > > As a startup company ourselves, we are deeply grateful for the opportunity to work with exceptionally talented founders, as well as having limited partners and co-investors that are perfectly aligned with our vision to create an ecosystem in which innovation can thrive. We are excited about the future of the open economy and it's transformative effects on social coordination, public services, financing and privacy." **About Inflection** Founded in early 2019 Inflection ([inflection.xyz](http://inflection.xyz/)) is a first check venture firm investing into the open economy. It is run by a team of seasoned venture investors and entrepreneurs with diverse backgrounds based out of Boston, Berlin, Munich, Madrid and Denver. Inflection is complemented by a global network of innovators who are holding an ownership interest in the fund.