Quantico, September 30
In his State of the Force address at Quantico, Secretary Hegseth announced the Autonomous Warfare Command, a four-star functional combatant command for drones, AI, and the command and control systems that direct them. The memo he signed the same day sets October 1, 2027 as the target date and sends the department to Congress for the authorities it needs: its own budget, its own manpower, and new career fields for the people who will run unmanned systems.
Until then, Project Agincourt does the groundwork, led by Defense Innovation Unit director Owen West and Navy SEAL Senior Chief Max Strasiser. It follows the creation three months earlier of a direct reporting portfolio manager for unmanned systems under the Deputy Secretary. The Senate's version of the defense bill proposes its own robotic and autonomous systems command with narrower acquisition authority. The details will move. The direction will not.
The SOCOM precedent
The closest precedent is U.S. Special Operations Command. Congress created it in 1987 after Desert One exposed how badly the services' special operations forces worked together. The decisive feature was its own budget line, which let the command buy equipment that worked across all of its components instead of waiting for each service to agree.
SOCOM's lasting advantage came from owning the things the services could not do alone, and interoperability was near the top of that list. AUTOWARCOM will face the same test. Its components will arrive from every service and from dozens of vendors, and they will have to work as one force from the first day.
Three things to require on day one
If we were writing the first requirements for AUTOWARCOM's software, we would start with three. A common vocabulary for observations, tracks, assets, and tasks that every vendor's system must produce and consume. Tasking that works when the link to higher is gone, because autonomous systems will operate exactly where the spectrum is most contested. And a signed record of every tasking and engagement decision, human or machine, so the command can show what happened and why.
Those are the foundations EdgeLance is built on. Its components exchange typed observations and tasks, its tasking runs on the team's own devices and mesh, and every approval, dispatch, and receipt lands in a record that survives the mission. We think that is what a command for autonomy will need underneath it, whatever the final legislation looks like.
The assignment logic underneath is the same one we tested against counter-swarm scenarios this summer, where a central solver that loses its link loses the fight.
The command will be judged by coordination
Buying drones is now the easy part. The Department has shown it can order them by the hundred thousand. AUTOWARCOM will be measured by whether those drones, robots, and the software directing them fight as one force. That depends far more on shared language and shared tasking than on any single platform.