In 1997 it took IBM a room-sized supercomputer, a corporate moonshot, and a bench of grandmasters-for-hire to beat the reigning world chess champion exactly once, in a six-game match the machine barely edged. Today the free engine running on the phone in your pocket would beat that same champion every single game, and it would beat the highest-rated human who has ever lived, and it does it on a battery.
The machine did not catch up to the grandmaster. Superhuman skill stopped being rare and expensive and became common and nearly free. That one fact is all you need to understand what is about to happen to air combat.
BLUF: It already happened once. In August 2020 an artificial intelligence (AI) agent built by a small company swept a graduate of the Air Force's F-16 Weapons Instructor Course five to nothing in a simulated dogfight — and it did it while held to the physical limits of the F-16 itself, an airframe engineered around what a human body can survive.[1] The scarce thing in that fight was never the pilot's skill. It was the airframe, the dollars, and above all the scale. Drone Dominance is not a drone program; it is a software supply chain with a magazine behind it, won by how fast and how widely we can push better autonomy onto cheap airframes. And the decisive fight is not the high-end duel between a Next Generation Air Dominance (NGAD) fighter and a J-20. It is the cheap fight down low — drones killing drones — and we are still trying to win it with expensive interceptors, re-running the exchange ratio that bankrupts a superpower. The winning pair of ACEs, Air Combat Evolution (ACE) and Agile Combat Employment (ACE), only pays off if we win that cheap fight first.
Superhuman got cheap
Deep Blue was a national event because beating Garry Kasparov was genuinely hard. A generation later, the hard part is gone, and nobody thinks a human should take a game off a laptop. The dogfight is on the same curve, and it is further along than most of the force wants to admit.
Go back to that 2020 trial. The Defense Advanced Research Projects Agency (DARPA) ran it; a company called Heron Systems won it. Their agent flew the within-visual-range gun engagement a fighter pilot trains a lifetime for, against a pilot with more than two thousand hours in the jet, and shut him out five to zero.[1] The pilot did nothing wrong. He was simply slower, by a margin no amount of training closes, at the one thing the merge rewards: reading the geometry and acting on it before the other guy does.
Now read the fine print, because it is the whole argument. The AI was held to the limits of the F-16. It pulled no more g than a real jet could, because the simulated jet is built around a pilot who greys out somewhere north of nine. Take the pilot out of the airframe and that ceiling is a design choice, not a law of physics. A machine that never loses consciousness, never tunnels its vision, never flinches, flown on a structure engineered for the machine instead of the man, does not merely match the ace. It flies tactics, techniques, and procedures (TTP) no crewed fighter can survive. That is why Collaborative Combat Aircraft (CCA) matter, and why the F-47 pilot will never be Robin Olds. He is not supposed to be. He is the quarterback for a formation of autonomous shooters that are, at the merge, already better than any human will ever be again.
The ace still matters, at the right scale
None of this kills the exquisite platform; it right-sizes it. We never fought the Yamato broadside to broadside, and the NGAD fighter is worth buying for the same reason the carrier was: to hold a class of warfare closed. The crewed fighter earns its keep as a small-count node that opens the door and directs the fight, not as the thing that wins the war by running up a score. Keep it, cap it at what that mission needs, and crew it with a force broad enough to wield everything behind it. The mistake is never buying the ace. The mistake is paying ace prices for the whole deck.
The war is won by the deck
If skill at the merge is no longer scarce, then the thing worth buying is not the better pilot or even the better airframe. It is the ability to push better software onto more airframes, faster than the other side can. That is what “Drone Dominance” actually names: not a quadcopter, but a supply chain. A body — the production base that turns out cheap airframes by the thousand — and a mind — the autonomy stack you update weekly and spread across all of them at once. The airframe is the disposable part. The software is the weapon, and software is the one thing that gets cheaper, better, and more plentiful the more of it you field.
We keep paying a premium to buy the ace. The war is won by the deck.
Run that forward and the air-to-air fight of the real war is not only a CCA against a J-20. More often it is a drone shooting down another drone, far from any runway, with no human in either cockpit because neither has one. Companies are already building exactly that. SplashOne's XFD-1 is a reusable, autonomous interceptor designed to find, chase, and kill Group 1 through 3 unmanned aerial systems (UAS), several per sortie, and recover to do it again.[2] Reusable is the word that matters, and the next section is why.
You can't buy your way out with a cheaper expensive thing
The reflex answer to cheap drones is a cheaper missile: a small, fast, sacrificial interceptor that costs a fraction of the big air-defense rounds. It feels like progress. It isn't. A hundred-thousand-dollar kill vehicle fired at a few-thousand-dollar drone is the same losing exchange ratio we run with million-dollar interceptors, moved one tier down and spread across a thousand more shots. It doesn't fix the cost problem; it retails the loss instead of wholesaling it. You cannot buy your way out of a cost-imposition problem with a cheaper expensive thing.
The way out is a shooter you get back. A reusable drone that kills ten cheap drones over its life and lands changes the arithmetic the way a sacrificial round never can — the cost moves off the per-kill line and onto a fixed asset you amortize. That is the only version of counter-drone defense that scales, and it is the hinge the whole concept turns on. Without affordable drone-on-drone point defense — or a defensive counter-air (DCA) combat air patrol (CAP) you can actually afford to keep airborne — Agile Combat Employment (ACE) is a slide, not a plan. You cannot disperse to austere sites that a twenty-thousand-dollar strike package can range and you cannot afford to defend. Win the cheap-air fight and the dispersal becomes real; lose it and every other ace is theater. The two only beat the house as a pair.
Magazine depth has three floors
Deep magazines are how you refuse the exchange ratio over time, but “magazine depth” is not one number. Like the levels of war, it has three floors, and we are strong on the top two and dangerously thin on the bottom.
Strategic depth is what the nation and its allies can produce. For cheap, dual-use airframes this is a strength, not a gap: the United States plus its partners — NATO, the UK, Australia, and the rest — already command enormous commercial manufacturing capacity both domestically and on friendly shores. The electronics, motors, and airframes of a Group 1 to 3 drone are the same ones civilian industry turns out by the container-load. We largely have this floor. We just haven't told industry to build to it.
Operational depth is what we can actually get to theater. This is a contracting and logistics problem, not a physics problem, and it is solvable with commercial production contracts, commercial delivery, and additive manufacturing where the supply line is too long or too cut to wait. We are mediocre here, mostly because we buy drones like exquisite platforms instead of like ammunition.
Tactical depth is capacity at the fight, in the hands of the unit that has to shoot tonight, and this is where we need a genuine paradigm change. A shipping container of autonomous Group 1 to 3 effectors, parked at the austere site or on the back of a truck, is a vastly more sensible magazine against swarming cheap threats than a hyper-expensive Mk 41 or Mk 57 vertical-launch cell spending a million-dollar round on a problem the container solves for the price of the drones inside it. We keep reaching for the most exquisite magazine we own to answer the cheapest threat in the fight. Treat tactical mass as mass — cheap, dense, forward, and reloadable — and the exchange ratio finally starts running the right direction.
The fix
Buy the deck, not the ace. Fund Drone Dominance as what it is — a body, a mind, and a three-floor magazine — and treat the software pipeline, not the airframe, as the platform you are procuring. Build an acquisition pipeline that can replace both software and hardware vendors alike rapidly to prevent stovepiping, and use commercial tools to make test, compliance, and integration a business-to-business function (B2B). Incentivize venture capital (VC) markets to keep investing in new companies and new technologies instead of marrying a program of record (POR) to a single prime integrator. Make affordable drone-on-drone point defense a programmed requirement that dispersal depends on, not a science project bolted on after the exquisite fighter is paid for. Keep the crewed force for the door-kick and the directed fight, cap it there, and spend the margin on the reusable mass and the software that actually fights the long war. The autonomy already beats the ace. The only question left is whether we buy it at scale or keep paying a premium for one more irreplaceable pilot.
What you can do Monday
- Unit commanders: Treat the air over your own formation as contested and demand the counter-drone layer the way you demand ammunition. Put drone-on-drone defense into every exercise, flown by your people, not a once-a-year capstone demo.
- PEOs / PAEs: Procure the software pipeline and the reusable interceptor, and program magazine depth at all three floors. Stop scoring a cheaper sacrificial missile as progress; it's the same losing trade at a smaller unit price. If you can't show that you're "disassembling" the enemy system for cheap than they assembled it, you failed.
- Contracting officers: Buy reusable mass and the right to push software to it at tempo — warm lines, surge terms, and over-the-air update rights written in — not a one-time delivery of frozen airframes. The most valuable lesson from Ukraine isn't first person view (FPV) tactics, its rapid innovation pushed to the field.
- DIU and the service innovation organizations: Standardize operational test and evaluation (OT&E) with Direct Reporting Portfolio Manager (DRPM) Unmanned Systems (UxS), and build a rapid authority to operate (ATO) function for UxS from the cloud to the drone. Field the reusable drone-on-drone interceptor against Group 1 to 3 UAS, and prove the dollars-per-kill, not just the capability. Source the body, the mind, and the containerized magazine from the commercial base that already builds, secures and certifies them, and incentivize the VC markets to keep bringing new technology into the department by rapidly onboarding new vendors as technology matures.
- OSD: Make affordable counter-drone point defense a programmed requirement that gates Agile Combat Employment, and fund the autonomy supply chain as the platform rather than as an accessory to the crewed fighter.
- Congress: No new authority needed. Ask one number on every counter UxS program: dollars spent per threat killed, against the price of the cheapest asset that threat represents. If we are paying more to kill it than the enemy paid to build it, we are funding our own bankruptcy, one interceptor at a time.
Works Cited
2 SplashOne Robotics Inc. n.d. XFD-1 Autonomous Fighter Drone. splashone.ai.
The conversation happens on LinkedIn — come argue with me there.