The Trojan horse in AI-written math: LLM-generated Lean code floods Mathlib
Machine Learning Street Talk · youtube · 2026-10-27
Kevin Hartnett, author of The Proof in the Code, discusses the catch in AI autoformalisation on Machine Learning Street Talk: LLMs can now turn maths papers into Lean code in a fraction of the time a human team would take, but the Mathlib community is finding it's a Trojan horse — a flood of Lean code lacking the precision and economy Mathlib's canonical definitions require.
It's the same problem everyone else faces with the volume of AI-generated everything: fast output whose quality doesn't match community standards, forcing maintainers to pay for the low-quality speed. The clip is from the documentary 'Mathematics Just Changed Forever' featuring Kevin Buzzard, Jeremy Avigad, Leo de Moura, and Paul Lessard.
More from Research
- OptiMer Wins EMNLP2026 Outstanding Paper: Post-Hoc Data Mixing With 15-35x Lower Search Cost — prajdabre · 2026-10-28
- Citrini: Claude designed scorpion antivenom in 64 hours — monkeys are the next bottleneck — citrini · 2026-10-28
- Flow matching ported to the browser via ONNX and JavaScript, runs locally — NoenD_i0 · 2026-10-28
- Rebuilding 'Democracy Does Cause Growth': long-run GDP effect rises to 33.6% with 2024 data — ChenhaoTan · 2026-10-28
- YC-backed Wondersearch claims OpenAI-level retrieval on FiQA with no embeddings, no vector DB — antoine_chaffin · 2026-10-28
- AAAI/ICLR Reviewer: About Half of Reviews Look AI-Generated, Peer Review in Crisis — dsmirza · 2026-10-28