OpenAI fought dirty on career-making math problem, says NYU mathematician
News Source
β’Tue, 08 Sep 2026 17:32:15 +0000
π° What Happened
NYU professor Tristan Buckmaster has announced three new proofs with a major step toward solving the Navier-Stokes problem, one of maths' hardest open questions. He worked with Anthropic mathematician Levent Alpoge and used the AI models Codex and Claude.
Hours later, OpenAI published its own full proof of the same problem, found by an unreleased next-generation model. OpenAI says the effort used 300 billion output tokens, worth about $22.5 million in computing. Buckmaster says OpenAI built on his team's work before it became public.
π The Backstory
The Navier-Stokes equations describe how fluids like air and water move. They are used across science, yet no one has proved whether their solutions always stay smooth. The problem is one of seven Millennium Prize Problems, each worth $1 million.
For 25 years, none of the prizes went to an AI-assisted proof. Now two rival teams, backed by two rival AI labs, claim to be first. The dispute shows how AI is changing science, and how fast academic rivalries can become public fights.
π― Why It Matters
AI has started cracking problems that stumped humans for centuries. How credit and proof are shared will decide whether AI helps science or sparks more fights over glory.
NYU professor Tristan Buckmaster has announced three new proofs with a major step toward solving the Navier-Stokes problem, one of maths' hardest open questions. He worked with Anthropic mathematician Levent Alpoge and used the AI models Codex and Claude.
Hours later, OpenAI published its own full proof of the same problem, found by an unreleased next-generation model. OpenAI says the effort used 300 billion output tokens, worth about $22.5 million in computing. Buckmaster says OpenAI built on his team's work before it became public.
The Navier-Stokes equations describe how fluids like air and water move. They are used across science, yet no one has proved whether their solutions always stay smooth. The problem is one of seven Millennium Prize Problems, each worth $1 million.
For 25 years, none of the prizes went to an AI-assisted proof. Now two rival teams, backed by two rival AI labs, claim to be first. The dispute shows how AI is changing science, and how fast academic rivalries can become public fights.
AI has started cracking problems that stumped humans for centuries. How credit and proof are shared will decide whether AI helps science or sparks more fights over glory.