The format works
Warhacker wrapped June 19 in San Diego, and the short version is that Defense Unicorns ran an exceptional event. No keynotes, no booth theater, no badge-scan lead capture. Around 400 people from government, industry, and academia spent two full days building software against real tactical problems, then briefed out what they made. The room rewards working code over slideware, which is exactly the incentive structure defense software needs and almost never gets.
We wrote about what we planned to demo going in: full mission lifecycle on live hardware, from Zarf-delivered mission pack through local inference, mesh sync, ATAK publish, and cryptographic burn. The fleet ran on the folding table all week. Nothing makes the case for consumer hardware at the edge like a MacBook, a stack of iPhones, and a watch doing the whole job in front of people who can poke at it.
We demoed to the people who matter
The best part of a build-first event is who walks up to the table. We demoed to the Defense Unicorns team and to senior military leaders who came to see what was real. These are people who have watched a decade of AI demos. They do not ask about the model. They ask about the approval workflow, the classification boundary, what happens when the mesh partitions, and who can prove the data is gone after the burn. Those are the right questions, and running the answer live beats describing it.
The feedback was direct and it was good. Some of it confirmed bets we had already made: local-first inference, evidence coupled to every AI output, burn as an architecture rather than a feature. Some of it exposed gaps we had not prioritized, and that is the feedback worth traveling for.
The feedback is already shipping
We took the hard questions home and turned them into capabilities. Several are already landing in the OS, and they are the kind of features that only surface when operators and program folks push on a live system instead of reviewing a deck. That loop, demo to feedback to shipped capability in days, is the entire advantage of being small and building fast. A program of record schedules that loop in quarters. We run it in days.
This is also the argument for hackathons as a serious acquisition signal, not a culture perk. Two days of building alongside the people who will operate and accredit the software surfaces more real requirements than a year of RFI responses. Defense Unicorns proved the delivery layer works in UDS Fleet. Warhacker proved the feedback layer works too.
We will be at the next one
Warhacker earned a permanent spot on our calendar, and we will show up to future hackathons in this space for the same reason: they force focus. A demo that has to survive strangers breaking it for two days is a forcing function no internal deadline replicates. If you run an event where builders ship against operator problems, we want an invite.
And if you saw the demo in San Diego and want the longer version, same answer: reach out. It runs on a single laptop. We will bring the fleet.