Over the Mojave Desert, an Anduril-built YFQ-44A released a live AIM-120 at a digital target after an operator assigned the engagement [E1][E3]. The shot completed a deliberate, phased progression and marked the service’s first long-range air-to-air missile launch from a jet-powered autonomous combat drone [E1][E4]. An uncrewed jet had moved from carrying a missile to employing one inside the combat-air system the service intends to field [E1][E2]. The threshold lies in the coupling of an uncrewed aircraft, a live weapon and a program designed to buy them at scale [E1][E2][E4].
After departing Edwards Air Force Base, Anduril’s aircraft first carried an inert weapon, then integrated the datalink needed to receive the target track, then fired the live missile over Mojave airspace [E1][E3]. Lattice software ingested the track, while a human operator tasked the YFQ-44A to engage [E3]. The aircraft executed the instruction and released the AIM-120 as designed [E1][E3]. Sequence matters: carriage, connectivity and fire were tested as linked parts of the same weapons chain [E1][E3].
June’s contract move turned that chain into a procurement structure [E2]. The Air Force advanced General Atomics’ FQ-42 and Anduril’s FQ-44 into engineering and manufacturing development and production, while buying autonomy as a separable software layer [E2]. Six vendors sit in the baseline autonomy pool: Anduril, General Atomics, Lockheed Martin, Northrop Grumman, RTX Collins Aerospace and Shield AI [E2]. Keeping the flight vehicle and control software on different competitive tracks lets the service recompete algorithms without reopening the whole aircraft program, an inference built into its “software sold separately” design [E2].
A force built this way aims to buy combat mass at a lower unit cost than a crewed-fighter fleet [E2][E3]. The Air Force plans more than 150 combat-capable CCA by decade’s end and about 1,000 over time, making the aircraft a source of magazine depth across the force [E2]. A reported target near one-third of an F-35A’s roughly $83 million flyaway cost implies a unit goal below $30 million [E3]. In a Pacific fight, cheaper uncrewed missile carriers could extend the force’s standoff reach and absorb losses that would be harder to replace in crewed aircraft, which is the strategic logic behind the scale [E2][E3].
Controlled conditions carry the strongest counter-case. The target existed in software; no live maneuvering aircraft, electronic-warfare stress or degraded communications tested the system, so the event says little about performance against an adaptive enemy [E1][E3]. A human retained release authority, and the aircraft did not choose its target or cross into autonomous weapons employment [E1][E3]. The one-third-F-35A figure remains a reported target with no disclosed contract price behind it, while split hardware and software contracts carry the familiar risk of buying airframes first and fixing integration later [E2][E3].
Washington is pairing that limited test with an industrial plan built for repetition [E1][E2]. More than 150 aircraft, a thousand-unit horizon and a six-company software contest turn one missile shot into a template for generating combat power [E2]. Success will depend on whether the autonomy stack works under jamming, whether costs survive production and whether human control remains clear at operational speed [E1][E2][E3]. One missile left the rail in Mojave; the Air Force is planning the force by the thousand [E1][E2][E4].