This page requires JavaScript to display.

OpenAI's Millennium Prize math claim

AI & tech · Heat 2.7 / 10 · 4 shows · 13m airtime

On September 7, OpenAI announced it had solved a so-called Millennium Prize problem. The claim widened by October 7 to hundreds of unsolved problems, and on October 13 the company named Navier-Stokes, one of the last six unsolved. It now says a model solved that problem in 88 hours and later announced results for over 350 problems. Terence Tao says there is no one at OpenAI he can call to explain the roughly 300 dumped proofs, and the discussion has turned to whether mathematics' volunteer refereeing can absorb them. The results remain unverified by outside mathematicians.

How 4 podcasts covered it

  1. The Journal., 2026-09-14 (0:05 to 0:53)
    Previews OpenAI's Millennium Prize claim
    Opens the episode on OpenAI's claim to have solved a Millennium Prize math problem, framing it as the latest in a rapid run of AI breakthroughs.
    New angle: OpenAI said it solved a so-called Millennium Prize math problem.
  2. TBPN, 2026-10-07 (10:50 to 13:44)
    Extends the math claim to hundreds of problems
    Extends the math claim from one Millennium Prize problem to hundreds of unsolved problems, and brings in Chollet's RLVR theory to explain why math and code move while other domains plateau.
    New angle: OpenAI says it solved hundreds of unsolved math problems, not just one Millennium Prize problem.
  3. The Daily, 2026-10-13 (18:28 to 21:08)
    Walks through the Millennium Problems and the claim
    Strogatz walks through the Millennium Problems and the September rumor that an AI company was on the verge of cracking one, then reacts to OpenAI's claim that it solved Navier-Stokes - a separate event from the May Erdős result.
    New angle: OpenAI announced it solved the Navier-Stokes Millennium Problem, one of the last six unsolved.
  4. Science Friday, 2026-10-28 (0:06 to 1:17)
    Recaps OpenAI's Millennium Prize claims
    Opens the episode with OpenAI's claims that a model solved Navier-Stokes in 88 hours and later produced results for over 350 open problems, framing the debate over how well they were solved before turning to the Terence Tao interview.
    New angle: OpenAI claimed a model solved the Navier-Stokes Millennium Prize problem in 88 hours and later announced results for over 350 unsolved problems.
  5. Science Friday, 2026-10-28 (1:17 to 5:17)
    Adds the 300-proof backlog and secrecy
    Tao's opening segment frames the AI math claims as a disruption to how mathematicians work, and adds a concrete consequence: a backlog of hundreds of AI-generated proofs with no one at OpenAI to explain them, junior researchers whose projects were made obsolete, and growing secrecy about research directions.
    New angle: Tao says there is no one at OpenAI he can call to explain the roughly 300 dumped proofs.
  6. Science Friday, 2026-10-28 (10:23 to 12:13)
    Argues math's gift economy clashes with AI
    Tao frames the dumped proofs as artifacts with the markings of achievement but no substance, and argues mathematics' volunteer gift economy - peer refereeing, unspoken etiquette - is incompatible with the optimizing approach of AI companies.
    New angle: Mathematics' volunteer gift economy, including unpaid peer refereeing, is incompatible with AI companies' optimizing approach.
Unpacking...