FULL STORY
OpenAI's Navier-Stokes Proof: From Vindication to Doubt
After mathematicians verified OpenAI's disputed Navier-Stokes proof, the math community erupted into debate over AI formalization, with Scientific American arguing the result is overstated.
2026-10-11 ~ 2026-10-14 · 4 episodes · 12 posts
Episode 1 · OpenAI's Navier-Stokes Proof Verified as Correct in Lean (2026-10-11, 3 posts)
Mathematicians including Obhishek Saha have verified that OpenAI's Navier-Stokes proof is correct, with its Lean formalization passing compilation, ending earlier cheating allegations. Meanwhile, a Reddit review of 242 Lean notes in OpenAI's math repo found at least 10 headline results lack formal verification.
- Reader finds 10+ results in OpenAI's math repo whose headline claims fall outside their Lean checks — MysteriousAvocado580 · 2026-10-11
- Mathematicians reviewing OpenAI's Navier-Stokes proof find the Lean formalization sound — burny_tech · 2026-10-11
- Mathematicians confirm OpenAI's Navier-Stokes proof is correct; Lean version compiles — basedjensen · 2026-10-12
Episode 2 · Mathematicians Debate the Value and Risks of AI-Formalized Proofs (2026-10-11, 4 posts)
Mathematicians are split over LLM-generated Lean proofs: critics call them unverifiable "Leanslop" lacking proper checking, while others argue that like poorly written human papers, AI proofs should not be dismissed outright.
- Critic: nobody can verify LLM-generated Lean formalizations, so Slop Lean isn't a solution — _onionesque · 2026-10-11
- Researcher's lesson from an LLM formalization debacle: never use slop, even in scratch notes — _onionesque · 2026-10-11
- Mathematicians debate OpenAI proofs: horribly written, but impossible to ignore — _onionesque · 2026-10-11
- Math community clashes over AI formal proofs: knowing results vs being a mathematician — zetalyrae · 2026-10-12
Episode 3 · Researchers Detail How AI-Generated Lean Proofs Can Go Wrong (2026-10-12, 3 posts)
Mathematician Gregor Kambauer outlined multiple ways AI-generated Lean proofs can fail, from relying on sorry or implicit axioms and formalizing the wrong statement, to being unreadable or exploiting kernel bugs, criticizing OpenAI for unchecked formalization gaps.
- Four ways AI-generated LEAN proofs can go wrong, from formalization drift to kernel bugs — gklambauer · 2026-10-12
- LEAN proofs can be 'correct but alien' or exploit kernel bugs, researcher warns — gklambauer · 2026-10-12
- AI Lean proofs can quietly rely on sorry or bypass the kernel via native_decide and unsafe — gklambauer · 2026-10-13
Episode 4 · Scientific American: OpenAI's Navier-Stokes Proof Overhyped (2026-10-14, 2 posts)
Scientific American re-examined OpenAI's widely touted Navier-Stokes result, clarifying that the proof covers a much narrower claim with limited conditions and does not solve the Millennium-level fluid problem as hyped.
- OpenAI's blockbuster Navier-Stokes proof isn't what we thought it was — scientificamerican · 2026-10-14
- OpenAI's celebrated Navier-Stokes proof isn't what it seemed, Scientific American finds — scientificamerican · 2026-10-14