OpenAI's biggest math breakthrough is getting ugly...

Summary
Fireship presents an episode of The Code Report covering the bitter controversy surrounding OpenAI's claimed breakthrough on the Navier–Stokes Millennium Prize problem. The host explains the mathematical background of the equations and details the clash between independent researchers Tristan Buckmaster and Levent Alpöge and OpenAI's leadership over alleged front-running, credit attribution, and AI model data usage.

What is shown

• [00:00] Intro segment with a humorous meme-coin mock trading UI followed by OpenAI's research post announcement and title sequence dated September 11, 2026.
• [00:49] Historical background on Claude-Louis Navier, Sir George Gabriel Stokes, fluid dynamic equations, and the Clay Mathematics Institute's Millennium Prize problem.
• [01:57] Introduction of NYU professor Tristan Buckmaster and Anthropic mathematician Levent Alpöge, their collaboration using LLMs (Claude Code, Codex), and their August 15, 2026 paper on 3D Euler equations.
• [02:47] OpenAI's announcement post showing GPT-6 Astra benchmark curves and detailing the deployment of 10,000 concurrent agents to derive a Navier–Stokes blowup solution.
• [03:08] Breakdown of the September 3–8 dispute, reviewing Buckmaster’s 4-page statement alleging coercive settlement offers by OpenAI researcher Sébastien Bubeck, alongside counterstatements from Bubeck and Sam Altman.
• [04:26] Public releases of preprints on September 8, 2026, by both sides and OpenAI’s formal response asserting no private Codex session data was accessed.
• [05:05] Analysis of Terence Tao's public commentary warning of the chilling effects of AI front-running on open mathematical collaboration, followed by an ad for Railway.

Claims & numbers

• The presenter states the Clay Mathematics Institute attached a $1 million prize to the Navier–Stokes problem in 2000 (26 years prior).
• The presenter states Buckmaster won the Clay Research Award in 2019 for work on Navier–Stokes and Euler equations.
• The presenter states Buckmaster and Alpöge obtained Euler blowup results on August 15, 2026, after collaborating with AI tools.
• The presenter states OpenAI launched an evaluation across Millennium Prize problems on September 1, 2026, using approximately 10,000 concurrent agents and an estimated $20 million of compute.
• The presenter notes Buckmaster and Alpöge published their results and a 4-page account of negotiations on September 8, 2026, followed hours later by OpenAI's blog post.
• The presenter cites reports from September 10, 2026, that OpenAI is claiming progress toward a second Millennium Prize problem.

Notable quotes

• [04:09] "The reply was, 'Why would you ruin your career?'"
• [04:12] "'If you don't want me to be nice, then I don't have to be nice.'"
• [05:11] "Prematurely solving the problem by purely AI-powered methods... can contaminate this process to the point where it actually becomes a net negative for the progress of mathematics as a whole."

Assessment
This is a tech news roundup and commentary video rather than a primary demonstration or official product launch. The video aggregates screenshots of published preprints, blog posts, and public statements from X/Mastodon, contextualizing them with fast-paced visual humor and memes.

Described by gemini-3.8-flash on 2026-10-07 from the video's audio and frames.