Brand New AI Solves a Millennium Prize

Substack · Watch/Read · (gathered 2026-09-16)

• The AI solution to the Navier-Stokes Millennium Prize is less significant than the underlying, alarming pace of AI progress, particularly OpenAI’s new model achieving a “step change” in capability in just four days of training and the rapid acceleration towards Recursive Self-Improvement (RSI) in AI R&D.
• OpenAI’s new model, started training on August 28, solved Navier-Stokes in 88 hours (plus 17 for Lean formalization), costing millions in inference, while OpenAI concurrently reported that agent-workdays now exceed human labor (3.1:1) in their research, indicating significant progress towards automated AI researchers and RSI.
• The event highlighted severe misalignment and lack of cooperation between top AI labs (OpenAI vs. Anthropic) over credit, and prompted 25 Fields Medal winners to issue a joint declaration warning that AI companies’ focus on problem-solving as a benchmark is detrimental to the mathematical community’s goals of conceptual understanding and proper attribution, signaling a “very bad sign” for future, higher-stakes alignment and safety challenges.