Fact-check: OpenAI and Anthropic models independently found the same proof, formalized in Lean
lxrjl · x · 2026-10-28
A fact-check of the Opus 5.5 'disproof' story: OpenAI says its model found the disproof on Oct 16 and formalized it in Lean the next day; Impagliazzo, Paturi and Williams are organizing an open review while stressing this isn't certification. Bigger story: Anthropic published a paper saying an internal Claude research model independently found the core algorithm (14M output tokens, no human input, also formalized in Lean), and Jelani Nelson says both proofs reach roughly exp(O(n / sqrt(log n))) runtime. Two labs landing independently raises the odds the result holds.
More from Models
- Multi-model arena Day 1: Gemini takes clear lead, MiniMax and Qwen round out top 3 — my_cat_can_code · 2026-10-28
- Arena co-founder: US still leads frontier AI, but China is winning open models and driving down cost — JaynitMakwana · 2026-10-28
- Liquid AI launches LFM3: tri-hybrid architecture, six edge models, all modalities — maximelabonne · 2026-10-28
- Matthew Berman: DeepSeek v4.1 flash is the best open-weights model right now — MatthewBerman · 2026-10-28
- Google exec teases Genie 4, its new frontier world model, coming to the API — OfficialLoganK · 2026-10-28
- Microsoft's MAI Cyber model doubles user adoption to 50% in 30 days — mustafasuleyman · 2026-10-28