Y Combinator Co-founder Paul Graham Says Arms Testing Depends on Ukraine's Needs
Y Combinator co-founder Paul Graham wrote on October 8 that he recently spoke with someone familiar with defense technology, who judged that the current testing standards for most military hardware depend on whether the Ukrainians want it. The reasoning is that they are engaged in a hot war against a determined opponent and know what works. The post did not name this conversation partner or specify the equipment.
Graham did not frame this statement as an investment memo but rather as a filter: procurement demonstrations and lab metrics give way to whether orders are placed at the front lines.
The Ukrainian government-supported defense technology agency Brave1 has launched "Test in Ukraine" in July 2025. The public announcement invites companies and governments to send unmanned vehicles, drones, anti-drone systems, munitions, maritime unmanned vessels, lasers, electronic warfare, and artificial intelligence platforms for testing, either by participating themselves or by handing them over to Brave1 experts, in exchange for frontline feedback and performance reports. The agency claims that testing can address urgent frontline needs and provide immediate feedback in combat.
This filtering has already appeared in consumer markets. Low Earth orbit satellite terminals were used by Ukraine after ground networks were destroyed; iterations of disposable loitering munitions and modified civilian drones are adjusted weekly rather than on a ten-year defense cycle. Graham's post summarizes these disparate cases into one statement: if you want it, you get it.
Traditional military procurement testing occurs at firing ranges and over years of bidding. Hot wars compress obsolescence, interference, and supply pressures into the same week. Unwanted equipment remains in warehouses, while desired items are requested for increased production. The post did not say that NATO procurement standards would change as a result, only that those who know what works are at the front lines.
This is a demand signal pricing military industry, not a new contract. The sellers are companies that need to send prototypes into testing programs, while the buyers are Ukrainian forces depleting their inventories and the aid countries paying for them. Beneficiaries are drone and electronic warfare suppliers who can adjust the next version based on interference feedback, while those under pressure can only present exhibition prototypes that the front lines do not request. Money flows from aid and defense budgets into specified models, not into roadshows without orders.
Source: Public Information
ABAB AI Insight
The old testing for defense procurement is based on firing ranges and congressional milestones. Lockheed and Raytheon operate on ten-year cycles, with samples first passing test sites before entering budgets. Ukraine has reversed this order: after 2022, civilian drones, Starlink terminals, and loitering munitions were sent to the front lines first, then required to be modified for interference resistance. Brave1 is making this process accessible in July 2025, allowing unmanned vehicles, anti-drone systems, lasers, and electronic warfare to be sent for testing in exchange for performance reports, not just exhibition photos.
Capital follows the demand list. Companies like Anduril, Shield AI, and several defense firms with Y Combinator backgrounds have shifted their selling points from theoretical kill chains to whether they can fill gaps caused by interference. Graham himself is a co-founder of the accelerator, and this post effectively sets the front line's requests as the admission criteria for his network. Aid country funds are directed to specified models, which then pressures production lines to adjust firmware weekly.
Comparable examples include the Soviet Union and Britain directly deploying American tanks into battle during World War II, where ineffective models were discontinued and effective ones were ordered in greater quantities; as well as Israel incorporating combat feedback into the next generation of missiles. The difference now is that the cycle has shifted from batches to software releases. The industry is in a transitional phase: lab metrics still exist, but the authority to explain orders temporarily lies with those engaged in combat.
Essentially, this represents a transfer of pricing power. The pricing authority of weapons has shifted from procurement committees to those who consume them. The mechanism is that hot wars destroy inventory daily, and unwanted hardware does not receive a second budget. Knowing what works is not just an opinion here; it serves as a filter for the next procurement.
ABAB News · Cognitive Laws
Returns in hot wars are faster than firing range evaluations
Wanting something equals acceptance; not wanting it means inventory
Iteration cycles must match consumption cycles for orders to remain