• Between late May and late September 2026, Ukraine's 3rd Army Corps ran Operation Vivaldi north of Lyman with drone-dropped strike robots and a deliberate bluff; it claims 176 km² retaken.

• Deception and doctrine decided the outcome, along with a willingness to risk robots before people.

• NATO units keep losing to small Ukrainian drone teams in exercises: REPMUS 2025, and Combined Resolve and Aurora 26 in 2026.

• Humans are coming off the battlefield but stay in the loop; robots coordinating with robots in combat is still ahead.

• Inflection's thesis: European defense's gap is operational know-how, and we want to back the company built by people who have it and can spread it.

Between May and September, Ukraine's 3rd Army Corps took back what it says is 176 km² with an offensive led by drone-dropped ground robots and decided by a bluff. Europe's armies can buy the same robots; what they lack is the know-how to run them, and at Inflection we think that is where the next important European defense company gets built.

After a night train, a tall soldier picks me up in a rental car and drives me deep into the forest: sweat-wicking polo neatly tucked in, sporty sunglasses, cargo pants. The night train would barely have been dark were it not for the blinds, so high is the sun at these latitudes this time of year. Finally on an unmarked road, which after a few hundred meters has a strict boundary of military/non-military grounds, we get out.

We're here for a tech demo, together with a rep from an unnamed development office of the MoD, and a few other military officers preparing their soldiers for a potential russian confrontation. They need to figure out effective ways to protect themselves against new threats: drones, large scale surveillance, cyber-attacks, and how to counter the russian counter-measures. Not only do they need to test, evaluate and procure the new technologies, they also need to train, maintain and figure out TTPs (tactics, techniques and procedures) for operating the technology at scale, effectively. Quite the challenge given the conditions (procurement frameworks) and liberties granted.

Ukraine and the rest

I leave the forest without conclusions. No grand plan has been made, no commitments or clarity on next steps. I'm vaguely optimistic for the soldiers and citizens that at least there are parts of the forces who understand and engage actively in the technical details of the new reality. They are all admittedly on the younger side, not top ranking (and thus less politically driven), and in some cases with very solid technical backgrounds on top of their special forces training. This should be an easy path to field new technologies, the need is obvious and urgent, and everyone understands it.

Unbeknownst to me, meanwhile in Ukraine a new model of operating land offensives is being put to test. Call sign Makar, commander of NC13, the strike UGV unit of the 3rd Assault Brigade in the 3rd Army Corps, is experimenting with new ways of running fully robotic offensives. It doesn’t sound so surprising to hear that it’s possible to airdrop a weaponised UGV with a multi rotor drone (the brigade calls it the “Gryphon” tactic), but the element of deception and surprise was critical to the success. And it was incredibly successful! Russian milbloggers were reporting Ukrainian counterattacks from late May; the corps kept silence until 15 September, then announced it in "seasons". Season one cleared 75 km² of infiltrators and liberated 10 km² (Serednie). Season two took Shandryholove and Derylove and secured Drobysheve. Season three, announced 28 September, took Nove, Ridkodub and Katerynivka for 51 km², bringing the corps' total to 176 km², with claims of about 5,000 Russian casualties across all phases and 250+ prisoners in season three alone. On the 8th of October ISW suggested Ukraine may be re-running the same playbook across the Siverskyi Donets near Kryva Luka. The operation is appropriately called Vivaldi, which of course leaves us to wonder what the fourth season will entail…

The effectiveness of robot-robot-human teaming

We can’t know precisely what the cost of this operation was, but presumably it helped keep many Ukrainian soldiers alive by keeping them out of the battlefield. You can risk a lot of robots before you should be willing to send a human. This is the exact opposite of the Soviet-style warfare of Russia and indeed conservative parts of the Ukrainian military, where the sacrifice of the individual soldier means very little. Instead, what seems to work and what we believe is the only plausible long-term future, for operational reasons and for maintaining public support for the war effort, is to remove humans from the battlefield. In fact, this is close to the mission of our portfolio company, Ark Robotics, “to protect human lives”, born out of the pain and suffering of lives lost to soviet-style tactics. A public report of one of their products, the ARK-1 UGV, being airdropped behind Russian lines carrying anti-tank mines surfaced this summer. To be fair, humans followed in the Vivaldi-operation in order to clear out areas not accessible by robots, and to actually claim land capture, at least for now we think humans are needed in the loop.

Meanwhile, at NATO's REPMUS exercise off Portugal in September 2025, a Ukrainian-led red team using Magura sea drones (made by UFORCE) won all five scenarios and “sank” an allied frigate, according to a Ukrainian participant. At Combined Resolve this spring, on the U.S. Army's own training range in Hohenfels, Germany, Ukrainian drone teams "easily spotted and defeated" U.S. units, the Wall Street Journal reported in August. In Karlskrona, Sweden this May, we hosted a Naval Defense hackathon with the Swedish and Ukrainian Navy present. During Sweden's Aurora 26 exercise happening in parallel, FPV crews from the Ukrainian 1st Corps (Azov) working with Swedish pilots knocked out 28 of about 32 major pieces of equipment in one drill. Andreas Gustafsson, who commanded the Swedish brigade defending Gotland, said "They are miles ahead of us when it comes to drone warfare...". We're still playing catch-up.

Up until now, it's mostly been drone pilots controlling drones (human-robot teaming). In this case, the claim of robot-robot teaming (or unmanned-unmanned) feels somewhat contrived to be honest, without detracting from the immense achievement. True robot-robot teaming would be them coordinating amongst each other, like how claude spins up agents in ultracode, a UAV would tell a UGV "I'm coming to pick you up here" and then use other drones at its disposal for protection. And this is really the direction we're heading, the future is just not arriving at the speed of PR.

Operational intelligence and tactics innovation as the bedrock of high-tech warfare

In a lab we can make all these scenarios reality, but just like in any robotic application, be it on the battlefield or in a warehouse, humans in the loop are still needed for common sense and exception handling. The ground and aerial robots are incrementally improving all the time, thanks to engineers both in the armed forces and in the companies delivering to them, but those improvements really only compound when the doctrine is turned on its head. Vivaldi's decisive ingredient wasn't technical.. The tech is an enabler, but ingenuity of the individual and understanding for the complete picture, including intelligence on mindset of the opposing forces made it ultimately succeed.

From intercepted radio chatter, we can tell that the Russian soldiers first thought the UGV fire was friendly, as they couldn’t imagine how such a robot could have got so far behind their lines. This was driven by months of deception and information operations, planting false information even at the corps’ own highest command level. So Russia committed reserves to the Shandryholove–Zelena Dolyna axis while the real blow went at Nove.

How can we allow for the wide-spread tactical and operational innovation across NATO forces? The technologies change constantly, or rather, will change very rapidly as we approach adoption, and with the change TTPs need to be as adaptive and imaginative and go through cycles of iteration. Being inventive and prepared for the unknown is the specialty of certain units, but we believe this will be required behavior for everyone in the armed forces. It's a muscle that can be trained, just like any other.

We're investing in an operations security company

We're looking past another DJI replacement co, long-range strike, maritime sensor company or even "industrial base" manufacturing company, we think there's space for an operational security company. It's a novel type of private military contractor/R&D/advisory/defense tech deployco to be built on the operational learnings of the new reality of warfare. These are likely to come out of early adopters of the technology wave, and have very strategic thinkers at the top who understand the ways of the old as well as have a huge vision for the future of deterrence and security, both private and public. European and NATO MoD's will be buyers of tech and services, this will a delicate political operation, and there will be a clear divide between the countries who adapt and adopt the new doctrine and those who are left behind.

We think this is not only critical to our security and democracy, but will also produce at least one very very big mission-driven company. We think the lessons about how a modern armed force operates compounds and can turn into a completely new category of companies. This might be more like Palantir, with a heavy consulting and technology component, or like a multi-national QinetiQ, or more akin to a technology R&D + deployment company like Tiberius plus Ode by Anthropic. We can only see the outline at this point, not the inner workings, but they will emerge as we get closer.

If this resonates with you, please get in touch.