Deterrence is an industrial problem, not a software demo
Every year the demos get better. Slicker autonomy stacks, cleaner sensor fusion, prettier common operating pictures. And every year the same uncomfortable arithmetic sits underneath them: in every serious wargame of a great-power conflict, the munitions run out in weeks, and the industrial base needs years to replace them.
That gap — weeks of magazine depth against years of replenishment time — is the single most important number in defense technology. It is a number software alone cannot move. A brilliant kill chain with an empty magazine is a very expensive way to watch.
This is not an argument against software. Decision advantage is real, and it is worth investing behind. It is an argument about what deterrence actually is: capacity an adversary can count. Adversaries do not measure our TED talks. They measure casting capacity, energetics production, shipyard throughput, and how fast a qualified part can move from a printer to a wing.
An adversary does not count your demos. It counts your magazines, your shipyards, and your production lines.
So when we diligence a company, we ask an old-fashioned question early: what does unit one thousand cost, and when can you build it? Founders who treat production as the product — who design for automotive cost curves, second sources, and surge capacity from the first sketch — clear that bar. Founders who treat manufacturing as a later problem usually discover that 'later' is where defense companies go to die.
The arsenal of the free world will not be rebuilt by the primes alone, and it will not be rebuilt by pitch decks. It will be rebuilt by founders who find the factory floor as interesting as the model weights. Those are the founders we want to meet.
