MIT professor: OpenAI and Anthropic's AI proofs contain a radically novel, untraceable step
ChrSzegedy · x · 2026-11-15
MIT professor Ryan Williams says a group of researchers now understands most components of OpenAI's and Anthropic's AI-generated mathematical proofs and can trace them to prior work — but each solution contains one step that seems radically novel: understandable, yet with unclear origin and meaning. He also notes the proofs' writing quality was quite bad, and that the OpenAI proof is now fairly well understood.
More from Models
- Ex-Googler rants: Gemini licensing spans 17+ confusing SKUs across Workspace/GCP — majidmanzarpour · 2026-11-15
- Arohan: Opus 5.5 feels great in every interaction, best since Opus 4.5 — _arohan_ · 2026-11-15
- DeepMind researcher Arohan: Opus 5.5 is a really well done model — _arohan_ · 2026-11-15
- 4-year-old Claude account banned, user restarts from fresh account ahead of IPO — huangyun_122 · 2026-11-15
- If OpenAI's next model can't beat two-month-old Opus 5.5, it may be losing the race — OddConsideration7290 · 2026-11-15
- GergelyOrosz: Every AI Service Is Unreliable, One Subscription Isn't Enough — AccBalanced · 2026-11-15