OpenAI fails to prove Navier-Stokes, mathematicians claim
OpenAI presented a 166‑page potential solution to the Navier‑Stokes problem in early September, claiming it followed the approach of Spanish mathematicians Diego Córdoba and Luis Martínez‑Zoroa. Experts disagree. Gonzalo Cao‑Labora of EPFL says the proof’s details differ and only surface‑level similarities exist. The proof, written in Lean, is under review by the Clay Mathematics Institute.
Key points
- OpenAI released a 166‑page Navier‑Stokes proof that experts say only superficially follows Spanish mathematicians’ approach
- NYU’s Tristan Buckmaster accused OpenAI of using his and Anthropic’s prior work, a claim OpenAI denies
- OpenAI ran 10,000 model instances for 88 hours and 130 billion tokens, but the multi‑agent contribution is under 10 %
Before publication, NYU mathematician Tristan Buckmaster accused OpenAI of building on work he and Anthropic’s Levent Alpöge had done with publicly available models. OpenAI denied the claim, saying it never used Buckmaster’s data. Camillo De Lellis of the Institute for Advanced Study notes that a definitive answer is unlikely because language models are black boxes.
In early October, OpenAI released roughly 400 more results on GitHub, angering mathematicians who felt the collection cheapened the field. The company also deployed about 10,000 model instances for roughly 88 hours and 130 billion tokens to develop the solution. Noam Brown says the multi‑agent setup contributed less than 10 % of the success, with the base model doing most of the work.
The story so far
8 episodes →- OpenAI fails to prove Navier-Stokes, mathematicians claimthis story
Mathematicians say OpenAI doesn't understand its own Navier-Stokes proof
The Decoder · 17 October 2026
Loading the full article…
This text was published by The Decoder and written by Manuel Uth. It is reproduced here with attribution so you can read it in full; the rights remain with the publisher. Read it at the source ↗
The headline, key points and digest above were generated by Digest AI's editorial model from the linked sources. Automated summaries can contain errors: the sources are the record. Spotted a mistake? Tell us. Published by Martin K., who runs Digest AI and handles corrections.
More in Research
All →- Opinion: AI may never fully decode human genome regulation · 1 src
- You could have invented transformers from memory · 1 src
- Anthropic’s Claude Mythos 5 redesigns quantum circuits better than prior research · 2 src
- Danaher and NVIDIA partner to integrate AI into scientific instruments · 3 src
- LitGene transformer integrates text and ontology to improve gene representations · 1 src
Comments
via GitHub Discussions