False Collatz Proof via a Lean Bug: Talk Probes Proof Assistants as LLM-Found Bug Targets
marc_lelarge · x · 2026-10-16
A Paris Automath seminar talk by Yannick Forster, "On the Origin of False Proofs by Means of Bugs in Proof Assistants." Proof assistants like Lean are fast becoming accepted as ultimate guardians of correctness — the recent forced Euler blowup result and OpenAI's announced Navier-Stokes solution were both verified in Lean with code mostly written by LLMs. Yet an experiment led by Tristan Stérin showed LLMs are effective at finding proof-assistant bugs: Ramana Kumar published a Lean "proof" of the Collatz Conjecture exploiting a bug, and Daniel Selsam at OpenAI subsequently found a flurry of bugs in Lean and Rocq. The talk examines the tension between proof assistants as guardians for LLM-coded proofs and as targets of LLM-guided bug search.
More from AGI Musings
- Robots enter the tarmac at Japan's busiest airport for a roughly three-year live test — Aiden_Tech_Ai · 2026-10-16
- AI researcher reflects on losing her scientific identity as industry culture erodes autonomy — mariyaivasileva · 2026-10-16
- Aella draws the line on AI art: outsource labor, not creative choices — ZeroStateReflex · 2026-10-16
- AI Is Forcing Everyone Into Management—Until It Eats Middle Management Too — MonkZer0 · 2026-10-16
- Shogun 2 remaster review-bombed over AI-upscaled graphics, sparking dev debate — draginol · 2026-10-16
- Pedro Domingos: Everyone must have their own unique model, anything else is dangerous — pmddomingos · 2026-10-16