GPT 6.2 credited with proving an open math conjecture despite poor communication
apples_jimmy · x · 2026-11-20
A mathematician reports ChatGPT (Astra 6.2) not only proved a conjecture he and a collaborator had been chasing, but discovered unexpected new phenomena and posed new conjectures. Jimmy Apples argues 6.2 is highly intelligent and people over-index on its poor communication — a flaw similar to o3, just more pronounced.
More from Models
- Simon Willison visualizes Gemini 4 effort levels with the pelican-on-a-bicycle test — hugobowne · 2026-11-20
- Not every agent task needs a general LLM — specialized small models by task type — ziv_ravid · 2026-11-20
- Start custom LLMs at 30B: runs on a single GPU and works with LoRA — ziv_ravid · 2026-11-20
- Before fine-tuning, distill traces into skills in-context to test if training is worth it — ziv_ravid · 2026-11-20
- Classification problems don't need a general LLM — use classifiers plus text encoders — ziv_ravid · 2026-11-20
- OpenAI policy chief touts research transparency, but METR task-horizon gap is widening — soumitrashukla9 · 2026-11-20