Budget documents show what $53.6 billion for autonomous warfare would buy by category, not by system, and the bill meant to carry it has stalled in the Senate.
Read the full essayA presidential order to rewrite Directive 3000.09 is five weeks overdue. The AI that shaped this year’s targeting sits in software the directive may not cover.
The Pentagon fielded its Shahed copy in seven months and used it against Iran. How many it has, and what they hit, remain undisclosed.
The Navy’s new carrier CCA request leans on the MQ-25’s control station and deck procedures, which no public account shows tested with an aircraft at sea.
The Air Force ordered its drone wingmen into production before choosing their software. Readiness should be judged by autonomy builds and test evidence, not airframes.
Uncrewed ships have shown they can navigate oceans. The Navy is buying thirty before showing they can run for weeks without anyone aboard to fix them.
The Navy ran an urgent mine-laying requirement as a research project. Before it buys sixteen more Orcas, it should hold the reviews it skipped.
Central Command called the July 12 Corsair attack a first. The public record shows fast fielding, and little about autonomy, defenses or who commands the boats.
Brussels promised a shield against Russian drones. What exists is national border networks, NATO fighters still doing the shooting, and EU projects that became fundable two weeks ago.
Ukraine’s cheap interceptors and the Army’s Merops buy look like a bargain per round. Cost per kill depends on radar, crews and weather.
Radars, jammers and interceptors are sold one at a time. Tower 22 and the homeland incursions show the failures sit between them.
The Drone Dominance buys finally set rules for motors and batteries, but only at assembly. Chinese cells and magnets remain permitted until August 2027.
The startup behind the Phantom robot cites $24 million in Pentagon contracts and trials in Ukraine. Published accounts and federal data differ on both.
Maven’s user counts show demand, not accuracy. Program offices should label every AI claim by its evidence and contract for revalidation after each update.
Three essays on the gap between a successful trial and a capability people can rely on: what a test shows, what the field adds, and what a buyer should ask.
Read the reportThe Army says it has enough attack drones for every squad but not enough trained pilots, a gap that purchase totals and delivery leaderboards were never built to show.
Read the essayA 95-second excerpt from the Army’s film of its humanoid robot competition finals at Marine Corps Base Quantico.
A 65-second company film from Skylark Labs: its humanoid walks across a field, aims a pistol at targets and is taught tasks by demonstration.
What has been demonstrated? Under what conditions? And what would it take to field it? Our editorial approach
