Autonomy at the edge: why mass is coming back to the fight

For three decades the Western way of war traded mass for exquisiteness: fewer platforms, each more capable, each more irreplaceable. That trade made sense against the adversaries of 1995. Against an industrial peer, it is a liability with wings. When a single airframe costs more than some countries' defense budgets, you cannot afford to lose one — which means you cannot afford to use them the way a real fight demands.

Attritable autonomy inverts the trade. A system cheap enough to lose changes commander psychology before it changes anything else: you can send it into the seam you would never send a crew, keep it on station past prudent, and mass it in numbers that saturate defenses built to kill the exquisite few.

But the demo-day version of this future skips the hard part. The edge is a hostile place for autonomy: GPS is denied, links are jammed, and the cloud is a continent away. Systems that phone home to a data center die in the first hour of a real war. The autonomy that matters runs inference on the aircraft, degrades gracefully, and completes the mission with zero bars of signal.

That is why our diligence in this thesis looks more like an automotive teardown than an AI review. Bill of materials, EW assumptions, swarm coordination under link loss, and the production line that gets unit cost down seventy percent by lot three. The model is the easy part now. The margin is in the manufacturing.

Mass is coming back to the fight one way or another. The only question is whose factories it comes from.