OpenAI Eyes Second Million-Dollar Math Bounty While Crypto Watches AI Eat Whiteboards 📐

OpenAI has told the New York Times it has made "substantial progress" on a second Millennium Prize Problem, days after claiming a Lean-verified result on the Navier-Stokes equations. The Clay Mathematics Institute has backed each of the seven Millennium problems with a $1 million prize since 2000, and OpenAI has said it is not pursuing the payout, treating the puzzles as a public benchmark for its models.

The company has not identified which problem is in play. Speculation on social media has centered on the Hodge Conjecture, a 1950 puzzle asking whether certain algebraic shortcuts can always count the "holes" in complex higher-dimensional geometric shapes. OpenAI has not confirmed that guess, nor has it confirmed that the model behind the work is a rumored "Aeon" version built for long-horizon tasks and agentic workflows; the only public reference is a September 3, 2026 X post from user @synthwavedd pointing to new slugs in a Statsig feature flag in Codex Desktop.

The prior Navier-Stokes announcement, dated September 8, said an unreleased internal model produced a proof—checked step-by-step in the Lean proof assistant—showing the fluid dynamics equations can "blow up" into physically impossible infinite speeds. OpenAI said roughly 10,000 coordinating AI agents worked for about 88 hours to complete it. NYU mathematician Tristan Buckmaster subsequently released a four-page statement saying he and Anthropic researcher Levent Alpöge had independently pursued a related proof and finished by August 22, before OpenAI's Sébastien Bubeck contacted him and asked him to either let OpenAI publish the following day or write it up alone, since Alpöge works at a rival lab. Buckmaster quoted Bubeck as asking, "Why would you ruin your career?" when he declined.

The same week, Anthropic said its Claude model formalized a 358-year-old proof of Fermat's Last Theorem in 11 days, and mathematician Terence Tao cautioned that AI labs racing each other are burning through math's supply of hard, useful problems faster than new ones can be found.