I’m watching the Anthropic dashboard. The company was founded on Constitutional AI—a grand experiment in democratic machine learning. The platform scaled rapidly, but the trend spike I’m seeing now is the opposite of growth. I see Coxon’s name climbing. Not trending because he built something. Trending because he left. The AI industry treats departures like medical alert systems—they spike when the patient is deteriorating, but nobody calls 911. They just note it.
I’m half-asleep, scrolling two streams that refuse to reconcile: a sterile Federal Reserve memo on “systemic cyber resilience,” forty pages of antiseptic throat-clearing, and a police blotter detail about a Molotov cocktail thrown at Sam Altman’s house in San Francisco—glass, gasoline, wick, the old punctuation mark. Same night. Same species. Two ways of saying please stop.
The PDF smells like printer toner through the screen. The bottle smells like 1939.
The screen is a fractal of rectangles: a grid of tabs, each containing a grid of papers, each containing a grid of graphs. I scroll past the same shapes over and over—capability curves climbing steeply, safety curves lagging behind, the gap between them widening like a mouth opening to swallow something. The Capability-Safety Gap. Every major lab acknowledges it. They publish papers on it. They give talks about it. Then they release bigger models on schedule. This isn’t a failure of planning. It’s what success looks like to them. The gap is the feature, not the bug.