OpenAI says 10,000 agents solved Navier–Stokes; mathematicians dispute the credit

reported by 3 outlets· 5 articles · confidence: medium · first seen 2026-09-09 03:10 UTC

What this means for you

Nothing to build on today: no model release, no API, no published proof. The part with teeth is provenance — OpenAI says it did not access Buckmaster's Codex work but cannot rule out that his use of its products improved the model. If your work runs through a frontier lab's tools, you have no way to know whether it did.

On 8 September OpenAI claimed its AI agents had solved Navier–Stokes, the equations that describe fluid flow, one of the Millennium Prize Problems, which carry a $1m reward. The Guardian reports the run used 10,000 agents at an estimated cost of $15m. Mathematicians reacted with alarm; 25 Fields medallists signed an open letter warning of a "severe misalignment" between AI companies and the field. The Guardian's editorial says critics accuse OpenAI of under-crediting mathematicians close to a solution. Tristan Buckmaster says work he did with OpenAI's Codex was seen by its team; OpenAI denies accessing it directly but could not rule out that his use helped improve the model.

Key facts

  • ·OpenAI claimed on 8 September 2026 that its AI agents had solved the Navier–Stokes problem, one of the Millennium Prize Problems. source
  • ·The Millennium Prize Problem carried a $1m reward and, according to the Guardian, had defied mathematicians for decades. source
  • ·The reported run used 10,000 agents at an estimated cost of $15m. source
  • ·Twenty-five Fields medallists signed an open letter warning of a "severe misalignment" between AI companies and mathematics. source
  • ·The mathematician Tristan Buckmaster says work he did on Navier–Stokes using OpenAI's Codex was seen by the OpenAI team. source
  • ·OpenAI denied directly accessing that material but could not rule out that data from his use of its products "helped improve our model". source

What the sources say

  • MIT Technology Review AI — Treats the episode as a case study in what mathematical practice becomes once models can attempt it.
  • The Verge AI — Covers how the announcement landed among academic mathematicians.
  • The Guardian AI — Gives the scale of the run and the reaction it produced in the field.
  • The Guardian AI — Interview with a pure mathematician about the open letter and what the field does next.
  • The Guardian AI — Editorial arguing the attribution dispute, and whether the work stands alone, matter more than the claim.

Sources

The original reporting. Follow these — they did the work.

Related stories

← the wire