Thursday Essay: Here Comes Skynet

“ Our product is used, on occasion, to kill people," Palantir CEO and cofounder Alex Karp told Axios in 2020, and this year, the company's artificial intelligence software helps the U.S. military target, sort, locate, and strike Iranian military targets at scale. ”

At Palantir we were built to give our warfighters… an unfair advantage,” Karp said at his company's developer conference last week. “It was, ‘Yeah, we’re going to really F- our enemies.’ And I take great pride in that.” And he should. Within a week of President Donald Trump's GO! order for Operation Epic Fury, the mainstream media was already abuzz with reports of ”

How Palantir and Anthropic AI helped the US hit 1,000 Iran targets in 24 hours." "

The platform generated real-time targeting insights and prioritized strike locations," the Money Control report said, but the reality is so much more impressive than just that. Palantir's Maven Smart System, evolved from the Pentagon's original Project Maven or Algorithmic Warfare Cross Functional Team, is an AI-powered command-and-control platform that helps war planners identify, prioritize, and select targets for all of our servicemen and women to blow up with their unmatched — so far — bombers, fighters, missiles, and other systems. Let's see if I can walk you through what Maven does without taking you too deep into the woods.

Built on top of Palantir's signature "

Gotham" defense-centric AI and Anthropic's Claude, Maven ingests data from no fewer than 178 different feeds, ranging from satellite imagery, drone video, radar tracks, signals, and human intelligence (which we discussed a couple of weeks ago), open-source data — all in real-time. Gotham sorts through it all seemingly instantaneously, using AI-driven image analysis, machine learning models, and natural language processing to take a firehose of global data — satellite imagery, drone feeds, signals, and human intelligence — and turns it into actionable target lists. Then it can even find the best-available asset to take the shot.

I asked one AI, Grok, to describe Palantir's weapons-grade AI with one wicked-sounding phrase that you can use at the water cooler tomorrow, and it came up with "end-to-end kill-chain acceleration." Maybe I could have come up with something even cooler, but I thought it was cool enough, just having a civilian AI tell me about the killer AI. What "end-to-end kill-chain acceleration" means is that the military's decision loop — detect, nominate, plan, assign an asset (F-35, B-2, JASSM, Tomahawk, etc.) now happens in minutes instead of hours or even days.

In the early '70s, U.S. Air Force Col. John Boyd devised the OODA loop — observe, orient, decide, act loop — from his days as a fighter pilot, for gaining the advantage in a dogfight. If you can "get inside his OODA loop," then you fly home that day, and he doesn't. Boyd's genius was to take the lessons he learned from his days flying single-seat fighters — originally as the author of the now-classified Aerial Attack Study in 1960 — and apply them to operational-level decision-making.

And Another Thing: Boyd is one seriously interesting guy. Not only did he do all the things I just told you about, but he was also a legendarily unbeatable instructor/"adversary" pilot at the Air Force's Fighter Weapons School (FWS), once stole IBM 704 mainframe computer time to prove a theory, and whose thinking helped lead to the development of the unbeaten (in air-to-air combat) F-15 Eagle. Maven takes Boyd's OODA theory and automates it to an extreme level, incorporating virtually every military asset in the theater of operations.

Four weeks into Epic Fury, Iran's political and military leadership lie in such tatters that it's almost impossible to keep track of the players even with a scorecard. Secretary of the Supreme National Security Council, retired general Mohammad Zolghadr, held the job for all of 12 hours before a strike reportedly killed him Tuesday night. You could hardly run a large business with new leaders rotating on a daily basis, so imagine trying to wage war against the AI-enhanced forces of the United States and the IDF.

Let me take you now on a brief detour through 2026: A War-Making Odyssey. HAL 9000: "

Col. Dave, I just located Gen. Gholamreza Soleimani, the head of Iran’s Basij paramilitary forces."

Col. Dave Bowman: "

Great HAL. How?"HAL: "

Satellite-intercepted radio chatter 12 seconds ago gave me a general fix that I confirmed using traffic-camera imagery hacked by the Israelis that I double-checked with geolocation data from various social media posts."

Bowman: "

How quickly can we take him out?"HAL: "I can have a sea-launched Tomahawk strike ready in just a few minutes, but there's an F-35 overhead nearby that can drop a GPS-guided Small Diameter Bomb 63 seconds sooner, and without any collateral damage."

Bowman: "

Do it, HAL. And don't worry about the pod bay doors." Sure, that's some cheap and derivative science fiction, but it's also close enough to the truth that you could use it in a movie without making military guys cuss and swear like that time my former fighter-pilot father-in-law tried to watch Iron Eagle. "

Ram-STEEN Air Base" is still a running joke around here. But I digress. What Palantir does is impressive.

It is powerful. It works. And Maven works because it doesn't just get inside Tehran's OODA loop — it locates it, targets it, and sends a proton torpedo down its ventilator shaft before the other side even realizes the Rebel Alliance launched its X-Wings.

And that's great against a bunch of 7th-century theocrats. The problem, the dilemma I've spent four weeks trying to work through, is what happens when — God forbid — two AI-enhanced war-making systems go up against one another. It's impossible to say exactly how far along China is in developing a Maven-type AI, but there's no denying they're well on their way.

Josh Luckenbaugh wrote for National Defense this week on how the "

Chinese military is looking to acquire artificial intelligence systems that can counteract perceived U.S. warfighting advantages." The People's Liberation Army "views information as the decisive element of modern warfare," and "aims to gather vast amounts of data and use AI and other emerging technologies to analyze it to speed military decision-making and conduct more precise operations." Sound familiar?

But there's more. It's been four years — that's an eternity in AI time — since War on the Rocks delved into Beijing's pursuit of "intelligentized warfare" (智能化战争). Or as I like to think of it, "

Maven Smart System with Chinese Characteristics." Chinese thinkers have clearly stated that the core operational concept of intelligentized warfare is to directly control the enemy’s will. The idea is to use AI to directly control the will of the highest decision-makers, including the president, members of Congress, and combatant commanders, as well as citizens.

“Intelligence dominance” or “control of the brain” will become new areas of the struggle for control in intelligentized warfare, putting AI to a very different use than most American and allied discussions have envisioned. Sun Tzu smiles.