Short posts
Wild eval from Vals found that AI models often targeted civilians in drone imagery, including the school in Iran that the U.S. mistakenly bombed
https://www.vals.ai/blogs/autonomous-warfare

Love this from Stuart A. Thompson https://www.nytimes.com/interactive/2026/11/06/business/polymarket-kalshi-prediction-markets-midterm-elections-accuracy.html

I had AI chatbots make voter guides for an address in every county. Half of the time, they did not name a congressional district. When they did, they were wrong 22 percent of the time.
Turns out AI chatbots are citing websites made with AI.





