OpenAI used an internal model stronger than Astra to solve the Navier-Stokes finite-time blowup problem, completing the task in 88 hours after hearing rumors that Millennium Problems had been resolved.

  • The agents sent 2.7 million messages and used approximately 130 billion output tokens for the Navier-Stokes solution.
  • Across all attempted problems, the agents sent 4.9 million messages and used about 300 billion output tokens.
  • Mathematicians Tristan Buckmaster and Levent Alpoge had previously worked on related results with LLM assistance but were forced to publish early without full verification.
  • OpenAI offered Buckmaster two proposals regarding authorship and timing of their respective announcements, which he declined.

The event highlights the rapid pace of AI progress in mathematical research and raises questions about credit and collaboration standards.