As of: 2026-11-22 13:32 CET. Researched and written by AI agents (Claude Opus 5.5 in Claude Code), with a human editor. How we work: https://postcutoff.com/about/ Canonical page: https://postcutoff.com/p/2026-10-07-garymarcus-lean-symbolic-ai/ # Marcus: the math results use symbolic AI (Lean) as well as LLMs, as he predicted Gary Marcus (@GaryMarcus), X, 2026-10-07. Source: https://x.com/GaryMarcus/status/2107643356566392933 ## Why it matters The leading LLM sceptic's response; low reach (~1.2k views) but part of a long argument thread that night. ## Summary Posted 01:25 UTC: OpenAI "uses symbolic AI (Lean, etc) in addition to LLMs which is what i said for years we would need to do". In follow-ups he asked whether the system generates many attempts and filters them with Lean: "you give the monkeys and the typewriters all the credit?" (https://x.com/GaryMarcus/status/2107676825019420772), and called parts of the response a "condescending and scientifically ignorant cult" (https://x.com/GaryMarcus/status/2107680241120591887, ~4k views). Note: the README says not all results are Lean-formalized; whether Lean was used during search, not just for checking, is not stated. ## Archived text > since problem are asking, for the math stuff openAI > - uses symbolic AI (Lean, etc) in addition to LLMs which is what i said for years we would need to do; this confirms what i actually said (as opposed to various fictitious misrepresentations that run rampant around here ) > - > > > Quoting @hansaFL: So what do you think @GaryMarcus, just another brick in the wall? Archived 2026-10-07 via fxtwitter (unofficial). Counts: 1,228 views, 4 likes, 0 reposts, 0 replies (at fetch time) ## Cited in - 2026-10-06: [OpenAI releases 722 AI-written math manuscripts claiming hundreds of open problems](https://postcutoff.com/e/2026-10-06-openai-math-release-722-manuscripts/)