Mathematician: in 6 months AI went from 'slop' to finishing proofs that take me months
RexDouglass · x · 2026-09-10
Mathematician ObhishekSaha traces AI's leap in his field: in May, the best public models still produced "absolute slop"; the Erdos unit distance problem jolted mathematicians; his first AI-assisted theorem proof (ChatGPT 5.5 Pro, early June) felt like "a competent, indefatigable PhD student"; Sol in July stepped up again. Today he gives Astra a bare proof sketch — sometimes less — and gets a correct, complete (if terribly written) argument within hours, work that would take him weeks or months, plus mostly autonomous formalization. Quote-tweets Noam Brown: OpenAI mathematicians and physicists had their "Lee Sedol moment" watching the model solve open problems they'd struggled with for years, in minutes.
More from Models
- Teacher's Demo: LLMs Answer When Humans Would Ask Clarifying Questions — qtzbra · 2026-09-10
- GPT-6 'Astra' Does 34 Math Steps in Latent Space, 4x More Than Sol — MaartenBaert · 2026-09-10
- NVIDIA details Alpamayo 2 Super, its L4 autonomy model for robotaxis — drmapavone · 2026-09-10
- OpenAI users report usage quotas wiped to zero as weekly reset dates shift by two days — ___Patrice___ · 2026-09-10
- Prediction: V4.1-Flash to score 36-38 on new AA index, agency at 42 — teortaxesTex · 2026-09-10
- Perplexity benchmarks 13 retrieval models: pplx-embed-v1-4b leads two of three categories — perplexity_ai · 2026-09-10