Updated
Updated · The Atlantic · Sep 15
OpenAI Solves Navier-Stokes in 4 Days, Triggering Math Revolt by 25 Fields Medalists
Updated
Updated · The Atlantic · Sep 15

OpenAI Solves Navier-Stokes in 4 Days, Triggering Math Revolt by 25 Fields Medalists

3 articles · Updated · The Atlantic · Sep 15

Summary

  • OpenAI said a swarm of 10,000 AI agents solved the Navier-Stokes problem in under four days, cracking one of mathematics’ seven Millennium Prize challenges after decades of failed human efforts.
  • The result jolted mathematicians because the 166-page proof is described as nearly incomprehensible, raising fears that AI can now produce landmark results faster than humans can absorb or build on them.
  • Tristan Buckmaster of NYU alleged OpenAI scooped work he and Anthropic researcher Levent Alpöge had advanced with AI tools, though OpenAI said it pursued Navier-Stokes after seeing promising partial results and denied using his inputs.
  • On Friday, 25 Fields Medal winners warned that AI companies’ race to solve headline-grabbing problems is misaligned with mathematics, which values not just answers but methods that generate reusable ideas and human understanding.
  • The dispute has widened into an existential debate over how to reward mathematical work as AI accelerates—from basic arithmetic to elite proofs in a few years—forcing the field to rethink credit, communication and verification.

Insights

Could OpenAI's historic Navier-Stokes breakthrough be plagiarized AI slop that threatens mathematical research?
If an AI solves a century-old problem but no human can understand the proof, did it actually advance science?

The $10M Navier-Stokes Proof: How OpenAI’s AI Swarm Sparked a Mathematical and Academic Uprising

Overview

In September 2026, OpenAI rapidly mobilized a swarm of 10,000 AI agents to solve the Navier-Stokes problem after hearing rumors that competitors were close to major mathematical breakthroughs. This effort, sparked by news of progress from mathematicians Buckmaster and Alpöge, led to a public announcement just hours after Buckmaster shared his own results. The move triggered accusations of academic misconduct and concerns about AI models being trained on researchers' unpublished work. As AI-generated proofs became cheap and fast to produce, academic journals faced a flood of questionable submissions, while traditional mathematics departments struggled with shrinking funding as resources shifted toward AI-driven research.

...