Rumor: Anthropic's Claude solved Navier-Stokes; but is AI solving open problems just reward hacking?
burny_tech · x · 2026-09-07
Andrew Curran predicts Anthropic has solved a Millennium Prize Problem — specifically Navier-Stokes via Claude, now out for expert review, with an announcement before the IPO (pure speculation, unverified).
liuying04 offers a contrarian take: AI solving famous open math problems is itself a form of reward hacking — the reward is fame, and the hack optimizes for the answer while bypassing the original purpose. Problems like Hilbert's were posed to spawn new methods and connections, not just proofs; a 10-million-line Lean proof may just be benchmark optimization. His fix: turn open problems into meta-problems that ask what new structures and cross-field connections the solution reveals.
Related event: Rumor: Anthropic Has Solved Navier-Stokes with Claude(2 posts)→
More from AGI Musings
- "The world will change unbelievably fast once intelligence itself starts scaling" — Dr_Singularity · 2026-10-23
- Why short-timeliners should expect tens-of-billions AI lab acquisitions, not technical evals — herbiebradley · 2026-10-23
- Frontend dev left holding two senior backend jobs after company bet on Claude Code — ParticularFan1122 · 2026-10-23
- Grady Booch revives Weinberg's 1971 quip: software so buggy one woodpecker could destroy civilization — Grady_Booch · 2026-10-23
- Why Netflix keeps losing to YouTube: UGC platforms professionalize easier than pro platforms go casual — ZeroStateReflex · 2026-10-23
- AI killed the coding moat: indie devs now compete on demand and distribution — sven_ai · 2026-10-23