AI Agents Reportedly Produce a Solution to the Navier-Stokes Millennium Prize Problem

Sam Altman ·

Key Info

  • According to the quoted announcement, OpenAI is sharing a solution to the Navier-Stokes Millennium Prize Problem, produced by a group of agents using a next-generation OpenAI model described as "significantly more capable than GPT-6 Astra." The problem, which asks whether smooth 3D fluid motion described by the Navier-Stokes equations can break down, has gone unresolved for roughly 90 years.
  • A related comment puts the result in context: the effort "cost millions of dollars," while previous and current models show sharp cost drops—o3 needed ~$500,000 to score 87.5% on ARC-AGI 1, whereas Astra now scores higher for ~$20. The author predicts that by the 2026 IMO, anyone with a $20/month ChatGPT subscription could reach gold-level math performance.

Highlights

  • AI agents have reportedly produced a proof for a deep, 90-year-old open problem in mathematics.
  • The model behind the proof is an OpenAI next-generation model, positioned as significantly stronger than GPT-6 Astra.
  • The post highlights an accelerating efficiency curve for AI reasoning: from roughly $500K for o3's ARC-AGI 1 score to about $20 for Astra to outperform it.
  • Scaling test-time compute is described as a glimpse of a future where frontier-level math reasoning is available to everyone.
Loading...