OpenAI's unreleased model cracked hundreds of unsolved math problems, ~3 hours each
haider1 · x · 2026-10-12
- The poster marvels that OpenAI's unreleased model produced hundreds of results on long-standing unsolved math problems, averaging 3 hours of thinking per result.
- The chart shows zero AI-solved Erdős problems in all of 2025, 28 by June 2026, then a jump to 372 problem families (719 manuscripts, 42% computer-checked) after OpenAI's Oct 6 release — and the model only started training on Aug 28.
Related event: Unreleased OpenAI Model Solves Majority of Math Problems(2 posts)→
More from AGI Musings
- Sarcastic jab at naive "don't harm people" style AI alignment goes around — deanwball · 2026-10-12
- Santa Fe's David Krakauer framing: vibe coders are capable, not competent — so are LLMs — MarcJSchmidt · 2026-10-12
- "Results Without Understanding" Is How LLMs Work—and Science Itself Is Changing — fkasummer · 2026-10-12
- Open-Source Dev giffmana: AI Extinction Talk Is 'Complete Nonsense Sci-Fi Fan Art' — giffmana · 2026-10-12
- Programming languages were just scaffolding for thinking — for LLMs and humans alike — MarcJSchmidt · 2026-10-12
- dioscuri publishes 'Bad AI Consciousness Takes Bingo' rebutting common LLM consciousness arguments — dioscuri · 2026-10-12