Every autonomy pitch deck says the same thing: our system performs better than the competition's. Almost none of them can prove it under conditions anyone outside the company controls. That gap — between claimed performance and verified performance — is the real bottleneck in defense tech right now, and this week two stories from opposite ends of the procurement pipeline make the case better than any think-piece could.
The first is Valkryn, a Washington-based startup that just came out of stealth with $17 million and a founding team pulled from Applied Intuition, Palantir, and SpaceX. Valkryn isn't building a drone or a sensor. It's building the referee. Co-founder Hugh Zabriskie put the problem plainly: "Test is scarce. It's really hard to do. It's really rare to do." Companies wait months for a shot to demonstrate themselves, and when they finally get it, there's often no rigorous, government-grade way to score what happened. Valkryn's "Battle Lab" is designed to instrument real-world tests and generate scored records that program managers can actually compare apples to apples.
That matters because the Army is simultaneously trying to widen the funnel on its end. The Army's G-TEAD just expanded from issuing a single OTA per Accelerated Capability Event to five or six, explicitly to pull more small and mid-sized vendors into prototyping. Col. Christopher Hill framed it as bringing "more people under the umbrella contractually." Good instinct — but more vendors entering the pipeline only helps if the Army can tell which ones are actually delivering versus which ones are good at demos. G-TEAD's own operational assessments, the 45-day evaluations that follow the ACE, are essentially trying to do in-house what Valkryn is proposing to do as an independent layer.
I'd put the two together like this: procurement reform is finally admitting more startups to the table, which is the right call. But admitting more vendors without a trusted, replicable way to measure them just moves the "integration hell" problem upstream, from fielding to selection. A testing company isn't as flashy as a hypersonic missile or a swarm contract. But if Valkryn's thesis is right, independent, scored test infrastructure is the unglamorous plumbing that determines whether all this new OTA money buys real capability or just more unverified promises.
Watch whether G-TEAD's undisclosed Hawaii ACE winners get paired with third-party test data before their next OTA — that's the tell on whether this problem is actually being solved.
