Summary
Wes Roth delivers a comprehensive account of what may be the most consequential AI story of the year: OpenAI’s announcement that an internal AI system produced a formally verified proof of forced blowup for the Navier-Stokes equations, one of the seven Millennium Prize Problems from the Clay Mathematics Institute. The video explains the underlying mathematics accessibly—Navier-Stokes describes fluid motion, and the open question is whether solutions can spontaneously develop singularities—and distinguishes between the full Navier-Stokes problem (the $1 million prize) and the Euler equations variant (frictionless fluids, a related special case) at the center of the claimed result.
The computational scale is staggering: roughly 10,000 AI agents, approximately $22 million in compute, 300 billion output tokens, and 5 million inter-agent messages over 88 hours. Models used include GPT-5.6, Claude, and Astra.
The story’s darker dimension involves mathematicians Tristan Buckmaster (NYU) and Levaan Elgougi (Anthropic, acting personally), who had spent a year quietly pursuing the exact same unusual mathematical approach—with almost no other researchers working that angle. After OpenAI’s Sebastian Bubeck contacted Buckmaster following leaked rumors of their work, Buckmaster alleges he was offered a choice between a joint release crediting OpenAI’s model or writing it up himself while ceding AI attribution. The unanswered question: did OpenAI’s model train on or access their private Codex drafting sessions?
📺 Source: Wes Roth · Published September 09, 2026
🏷️ Format: News Analysis
