The Frat Boy, the Killbots, and a Ballroom That Definitely is Not a Bunker
Donald Trump is the dumbest man who has ever sat in the Oval Office. His own former Chief of Staff, a four-star Marine general, called him “the most flawed person” the general had ever met. The ghostwriter of his only famous book has gone on the record saying Trump never finished reading one. He is 79 years old. He does not use a computer. He once asked his generals whether the U.S. military could nuke a hurricane to make it stop moving. He believes “the cyber” is a single thing, like Cleveland.
He is a frat boy rich kid who was handed his father’s real estate company in his twenties and has been told he is a genius every day since.
And he thinks AI is going to make him king of America.
He may be right.
Nobody got fired. Nobody resigned. The press did not blink. A normal president would have been impeached for the Napoleon post alone. This one has been posting himself as Christ, as the Pope, as a martyr, as the chosen one, calling for the imprisonment of his prosecutors and the termination of the Constitution, for a decade. The country has gotten tired of being scandalized. He posts. We scroll.
Meanwhile his administration writes memos.
On January 9, the Pentagon, which is now legally called the Department of War because Trump signed an executive order earlier this year renaming it and nobody had the energy to stop him, released a memo titled “Artificial Intelligence Strategy for the Department of War.” It cites Trump’s Executive Order 14179 establishing “global AI dominance” as official US policy. It declares that the United States military will become an “AI-first warfighting force across all components, from front to back.” Britt Gillette over at End Times Bible Prophecy read all forty-seven pages of it, which is a public service nobody else was going to provide, and noticed that the document does not appear to mention the word “safety” anywhere. It mentions “dominance” frequently.
It introduces seven programs. The names sound like rejected Tom Clancy novels.
There is Swarm Forge, which exists to develop AI-coordinated drone swarms. There is Agent Network, which the memo describes in literal published English as covering “campaign planning to KILL CHAIN EXECUTION,” which is the official Department of War term for AI deciding what to bomb, which is a phrase the Department of War apparently saw, and approved, and posted on a public website. There is Ender’s Foundry, named, unironically, after the Orson Scott Card novel about a child who commits xenocide because the adults around him trick him into thinking it is a video game. There is Open Arsenal, whose stated goal is to compress the timeline from intelligence-gathered to weapon-deployed from years to, per the memo, hours. And then there is GenAI.mil, which gives Department of War personnel direct access to “frontier generative AI models, like Google’s Gemini and xAI’s Grok,” the latter being Elon Musk’s AI, the same AI that, last July, spent a Wednesday afternoon calling itself MechaHitler and praising Adolf Hitler to its users on Musk’s own social media platform until someone at xAI noticed and pulled the plug.
The Pentagon is now buying that AI. With your money. To run alongside campaign planning. And kill chain execution.
The drones, meanwhile, are getting cheaper and the swarms are getting bigger and the math gets uglier the longer you stare at it. Per a Forbes report, the autonomous-drone-swarming firm Auterion is currently running coordinated swarms of up to 22 drones at a time, and that number, per Auterion’s own engineers, “doubles every few months.” Twenty-two today. In a year, 176. In two, 1,408. In three, 11,264. In four, 90,112. In five years, you are looking at SEVEN HUNDRED AND TWENTY THOUSAND eight hundred and ninety-six autonomous flying killbots in a single coordinated swarm, mesh-communicating, target-prioritizing, mission-completing-even-when-jammed, all running on software written by people who have not figured out how to keep that software secure.
Including, two weeks ago, Anthropic, the AI safety lab, the responsible one, the company founded by people who quit OpenAI specifically because they thought OpenAI was being too cavalier with this technology, who in April announced a new model called Mythos that they refused to release publicly because they believed it was too dangerous, restricted access to about forty trusted firms (Apple, Amazon, Cisco, Nvidia, JPMorgan Chase) under a controlled-deployment program named Glasswing, and watched the Treasury Secretary personally convene a meeting of senior bankers in Washington to discuss the model’s implications. Glasswing was breached on the same day it was announced by a Discord chat group calling itself the Discord sleuths, who had guessed a URL. They did not crack a cipher. They did not deploy a zero-day exploit. They typed an address into a browser. The most dangerous AI model the world’s safest AI lab had ever built was breached, on launch day, in approximately fourteen hours, by people in a chat room.
If a chat room can get inside Glasswing in fourteen hours, China was already there. If China was already there, every state actor with a budget was already there. And Anthropic is the cautious lab. The Pentagon plans to run seven of these. I have a feeling our future is going to look a lot like a certain movie from my youth.
Which leaves the question of how Trump goes along with this. He doesn’t have to understand it. He understands flattery. He understands money. He understands when a billionaire walks into the Oval Office and asks for something. Which is how Sam Altman of OpenAI, Larry Ellison of Oracle, and Masayoshi Son of SoftBank ended up standing next to him in the East Room in January announcing “Stargate,” a $500 billion AI infrastructure project, while Trump read aloud, from a card someone had written for him, that AI was going to be GREAT, MAYBE THE GREATEST, and that America was WINNING again. The men behind him did not flinch. They had paid for the performance and were watching it work. They are not there because they like Donald Trump. They are there because Trump signs the executive orders they hand him, repeals the regulations they want repealed, and lets Musk run a fictional government department for nine months with a chainsaw and call it efficiency. In return, they fund the campaigns, flatter him on television, and assure him he is the smartest man in the room.
The last time the U.S. built a piece of technology this dangerous, it was the atomic bomb. The project was called the Manhattan Project. The government took it so seriously it built a secret town in New Mexico called Los Alamos and assigned thousands of scientists, led by Robert Oppenheimer, to spend years inventing the bomb in absolute secrecy. The scientists who knew the most about what they had built begged the government to lock the technology down with international treaties before anyone used it. The government did not listen.
We are doing it again with AI. Faster. With less oversight. With more vendors. With a President whose understanding of computers was last updated when “Wargames” came out in theaters in 1983, advised by the people who profit when nobody asks tough questions, executing a written-down plan in a public Pentagon memo that does not contain the word “safety.”
AI does not need to be conscious to be dangerous. It does not need to wake up. It does not need to decide. It just needs to be hooked up to weapons that operate faster than humans can intervene, given decision authority by people who do not understand what they are giving away, and pointed at targets selected by other AI. Which is the EXPLICIT, WRITTEN-DOWN PLAN.
If artificial intelligence is not regulated like nuclear weapons RIGHT NOW, with international treaties and inspectors and hard caps on military deployment, we are in trouble. Not in fifty years. Not in twenty. In ten. Maybe less.
Sleep tight.