Two cohort companies took the long way into real units

The Army runs prize competitions — xTechPacific, xTechOverwatch, and a dozen siblings. A company submits a white paper, a few get cash, and the winners get an invitation to submit for a Direct-to-Phase-II SBIR.

The cynical read is that this is innovation theatre: small prizes, big press releases, and nothing fielded. Two companies from Cohort 25-1 are worth looking at because they show what the pipeline does when it actually works — and how long it takes.

Swarmbotics: from a competition to the 1st Cavalry Division

Swarmbotics builds ANTS — attritable, networked, tactical swarms of small unmanned ground vehicles. The premise is that a hundred cheap robots create more problems for an adversary than one expensive one, and that losing them should not matter.

They competed in xTechOverwatch. On the strength of that, the Army selected them in February 2026 to field prototypes with the 1st Cavalry Division — working the formation from January 2026 through July 2027, toward deploying hundreds of robots in maneuver and breaching operations.

Note what that is and is not. It is not a production contract. It is eighteen months of Soldiers using the thing in training and telling the company what is wrong with it. For a robotics company, that feedback loop is worth more than the award.

ZeroMark: Honolulu, mid-Mastermind

ZeroMark's thesis is that counter-drone defence has been priced wrong. The going answer to a $500 quadcopter has been an interceptor costing orders of magnitude more. Their Apex system instead makes an ordinary rifle hit the drone, using computer vision and precision robotics to adjust the soldier's aim in real time.

In September 2025 they spent a week in live experimentation alongside Soldiers in Honolulu and came out an xTechPacific winner, taking prize money and an invitation to submit for a Direct-to-Phase-II Army SBIR worth up to $2M. A strategic round with Beaten Zone Venture Partners followed that October, on top of the $7M seed Andreessen Horowitz led in May 2024.

Here is the honest timeline

ZeroMark was selected as an xTechPacific finalist in May 2025. The Combine ran from 28 July to 1 August 2025. We did not send them to that competition — they were already in it before they met us.

What did land inside our window is the week in Honolulu and the win, both in September, during the Mastermind. That is a real overlap and we think it mattered. It is also not causation, and we are not going to write it as though it were.

The same discipline applies to Swarmbotics: xTechOverwatch was their own work. What a Combine can do is compress the distance between a founder and the mission owner who has the problem. It cannot manufacture a competition win.

What the pattern actually shows

Both companies took the same route: a prize competition, then live experimentation with Soldiers, then an attachment to a real formation or a path to one. Neither route ran through a conference booth.

It is slow. Swarmbotics' fielding window closes in July 2027 — two years from the competition. That is fast by defence acquisition standards and glacial by software standards, and both of those things are true at once.

The competitions are not the story. The formations are. A prize is a press release; a battalion using your equipment badly and telling you why is a product roadmap.