Ukraine Military History Institute
What a drone assault actually looked like in 2023
The popular picture of a drone assault is wrong, and Ryan O'Leary, who was part of Vadym Sukharevskyi's 2023 experiment with Chosen Company, corrects it from the inside.
Most people imagine drones hitting a position, then infantry walking in. That is not what happened. The drones were not the preparation. They were present throughout, attached to every element of the assault.
One DJI drone followed every four men. One followed every vehicle. Three vehicles on a mission meant three drones overhead. The moment a team dismounted, its drone transferred to following the team on foot. On top of that were FPVs and further DJIs dropping grenades.
Targets were planned position by position, jointly with the drone teams, before anyone moved. Calls for fire were made through FPVs. In O'Leary's account, the whole operation was designed drone-first rather than drone-supported.
His summary of how it looked from outside is the line worth keeping: it appeared that the assaulting troops did the work, and they did shoot a great many people, but preparing the ground was entirely drones.
The uncomfortable claim: this was better three years ago
O'Leary's real point is not nostalgia. It is that capability was lost.
His assessment is that drone assaults as run in 2023 were more advanced than what is commonly seen now. That is a striking thing to say about a war where every other capability has moved forward continuously.
What he wants is unglamorous and specific: put Sukharevskyi in a room with Code 9.2 or whoever currently owns the problem, have him explain exactly how it was done, adapt it to current drones, and run it again. He describes the gap as something that has been badly lacking.
This is a recurring shape in this war. A unit solves something, the people who solved it move or are lost, and the method does not propagate. The technology advanced. The doctrine did not travel.
Why offence is more possible now than it was
Rob Lee takes the frame wider, and pushes back gently on treating "drone assault" as a distinct thing at all. It is offensive operations that incorporate drones. The question is what drones enable that was not possible before.
His judgement is that the prospects for offensive operations are better now than a year ago, and the reason is not the assault itself. It is the strike campaign.
The ability to isolate a section of the front has improved substantially. Previously you could concentrate force on one direction and sometimes succeed. Now, if you can cut an objective's logistics, make resupply hard and make reinforcement hard, you can achieve considerably more with the same assault force.
Isolation, not firepower, is what changed.
The breach nobody exploited
Lee's example is the most instructive part of the conversation, and it is a story about failure at the level above the fighting.
Northeast of Huliaipole, Skala Battalion achieved a genuine breach. They used ATVs, light vehicles and BMPs, pushed through, reached Ternove, and got into firefights with Russian FPV teams in their own positions. That is a real penetration of a drone-saturated defence, which is supposed to be close to impossible.
Then nothing happened.
There was no one to exploit it. Lee is explicit that the opportunity was real: had nearby brigades been pushed through behind the breaching force, told simply to keep driving and create chaos in the Russian rear, it was absolutely possible. Russian Telegram channels at the time were, in his description, extremely alarmed. Ukraine genuinely had a chance at a localised collapse of that section of front.
The breach was achieved by the unit. The exploitation required a decision one level up, and it was not made.
A caution about the numbers
Lee adds a methodological warning that applies to almost every claim of advance in this war, from either side.
Units inflate breach distances, and the grey zone makes it easy. A unit reporting a twenty-kilometre breach may have crossed twenty kilometres of ground the Russians were not meaningfully holding. That is not the same as breaking twenty kilometres of defence.
He is careful to say the Skala breach was real regardless. The lesson is to ask what was actually held before accepting what was supposedly taken.
What to watch
- Whether the 2023 method gets deliberately transferred. O'Leary's request is concrete and cheap: one debrief between the person who ran it and the people running it now.
- Whether breaches get exploited. The constraint is no longer achieving a penetration. It is having a formation ready and authorised to pour through one.
- Isolation before assault. Lee's argument is that the strike campaign is what made offence viable again, which means the interesting question before any attack is what was cut, not what was massed.