If AI-Enabled Weapons Are So Smart, Why Do They Keep Hitting Schools?

On Feb. 28, a girls’ school in Minab, Iran, was struck three times during school hours. The roof collapsed, killing between 175 and 180 people—most of them girls aged 7 to 12. The U.S. eventually confirmed it had hit the site.

This week, Cy takes Virginia (and you) deep inside Project Maven—the Pentagon’s AI-powered targeting system, how it was built, who built it, and why the school was just the tip of a very bloody iceberg.

What we cover:

* The origin story: Marine intelligence officer Drew Cukor in Afghanistan in 2001, blind on the battlefield with nothing but Microsoft Office and Google Earth, and his two-decade obsession with solving that with technology.

* How Palantir—Peter Thiel’s data company, bootstrapped with CIA venture capital—became the backbone of the most expensive targeting system in U.S. military history. And why they nearly didn’t make it.

* What a kill chain actually is, why compressing it was always the point, and how layering large language models into Maven took targeting from 20 targets a day to 5,000.

* The Millennium Challenge war game of 2002 — a $250 million exercise where a retired Marine general sank the U.S. fleet in 10 minutes.

* How Pete Hegseth gutted the civilian harm teams, fired the JAGs, and eliminated the “roadblocks”—the people whose job was to check whether a school was still a school.

* And the fundamental question underneath all of it: is any of this actually making us safer, or is the product never really the point?

Thanks for listening to Omnishambles! Subscribe for free to receive new posts and support our work.

Mentioned in this episode:

* Project Maven: A Marine Colonel, His Team, and the Dawn of AI Warfare, by Katrina Manson

* AlphaGo (Netflix)

* Millennium Challenge 2002

* Operation Spiderweb (Ukraine, 2024)

This is a public episode. If you would like to discuss this with other subscribers or get access to bonus episodes, visit omnishamblespod.substack.com