Cross-verifying AI math proofs with hundreds of agent calls instead of Lean
DimitrisPapail · x · 2026-09-15
- Dimitris Papail pushes back on Lean formalization demands: he hasn't seen a single case where models like Sol/Astra claimed a proof was correct and it wasn't, with results cross-verified by hundreds of agent calls across all 3 models.
- QuangVDao argues that without Lean, purely computational results are hard to trust; checking large certificates in Lean is already non-trivial in his own project.
- Papail counters that such proofs involve masses of inputs, bounds, and numerical tables that don't compress well, so Lean may not help — e.g., an upper bound proof hinging on distributions and tables satisfying prescribed inequalities.
More from coding & agent
- Dev builds 3D universe explorer with 20 galaxies, 767 planets using Codex GPT-6.1 Sol — OpenAIDevs · 2026-11-02
- ChatGPT, Claude and Perplexity's fetchers don't run JavaScript, leaving client-rendered sites invisible to AI — gaganghotra_ · 2026-11-02
- SGLang boosts multimodal inference >10% by replacing GPU memory bookkeeping with a dict lookup — charles_irl · 2026-11-02
- kivi: hold Fn, talk, and it types polished text anywhere on your Mac — itsOmSarraf_ · 2026-11-02
- A base LLM plus 11 lines of code is all you need for a decision model — helloiamleonie · 2026-11-02
- Fix codebase structure before asking AI agents to write tests, says engineer — arpit_bhayani · 2026-11-02