The Navier Stokes situation is insane

Summary
Presented by Michael B. Paulson (ThePrimeagen) on his channel The PrimeTime, this video examines the controversy and priority dispute surrounding OpenAI’s claimed solution to the Navier–Stokes Millennium Prize problem. Paulson breaks down mathematician Tristan Buckmaster’s four-page statement alleging potential data misuse from OpenAI Codex sessions, OpenAI’s counter-claims, and the public response from OpenAI researcher Sébastien Bubeck.

What is shown

• [00:05] OpenAI’s announcement post stating an internal multi-agent model and GPT-6 Astra resolved the Navier–Stokes Millennium Prize problem.
• [00:13] Summary slide listing the Clay Mathematics Institute’s Seven Millennium Prize Problems.
• [00:35] Posts and replies on X accusing OpenAI of stealing results from mathematicians Tristan Buckmaster and Levent Alpöge.
• [01:26] Excerpts from Tristan Buckmaster’s statement detailing their results on finite-time blowup for porous media, Boussinesq, and 3D incompressible Euler.
• [01:48] Sponsor segment demonstrating Namespace devbox and sandboxing infrastructure for AI agents.
• [03:34] Reading through Buckmaster’s four-page letter documenting his timeline, tool usage (Claude, Codex, GPT-5.6 Sol, Astra, Lean), his communications with OpenAI, and negotiations with Sébastien Bubeck.
• [07:42] Sébastien Bubeck’s public clarification on X alongside screenshots of text messages with Levent Alpöge.
• [10:05] OpenAI chart titled “Performance of GPT-6 Astra and our Internal Model on a curated set of open math problems” across test-time compute.

Claims & numbers

• The presenter states OpenAI utilized around 10,000 concurrent agents running for roughly 88 hours, followed by 17 hours of Lean formalization and verification with GPT-6 Astra.
• The presenter estimates OpenAI’s compute spend on the proof run was likely north of $10 million for a $1 million Millennium Prize bounty.
• Buckmaster’s document claims he and Alpöge obtained LLM-generated blowup results on August 15 and verified them in Lean on August 22, 2026.
• Buckmaster claims he was informed OpenAI’s internal proof for forced blowup in $\mathbb{R}^3$ and $\mathbb{T}^3$ was approximately 100 pages long.
• Bubeck claims in his response that he never requested Alpöge be removed from authorship of his own work and sought to coordinate discovery announcements.

Notable quotes

• [01:06] “Wait a second. Am I... am I being deceived?”
• [07:23] “The reply was, ‘If you don’t want me to be nice, then I don’t have to be nice.’”
• [10:36] “They cannot let Anthropic win. The fact that Anthropic could be winning, that is the worst thing for them.”

Assessment
This is a commentary and review video analyzing a major academic and industry controversy. No technical tools or model outputs are demonstrated live by the presenter; he reviews and reacts to publicly posted statements, screenshots of correspondence, and official OpenAI announcements.

Described by gemini-3.8-flash on 2026-10-18 from the video’s audio and frames.