LessWrong article questions the reliability of Lean4 formalized proofs
Start here
On October 12, 2026, LessWrong user Milo Moses published an article critically examining the trustworthiness of the formal proof tool Lean4. He noted that Lean4 is based on a self-developed dependent type theory rather than ZFC, and there is currently no public proof that it is consistent relative to ZFC plus arbitrarily large cardinal axioms; he also said he is 90% confident that the current Lean4 on GitHub contains a True=False proof that the community would regard as a bug. The article also mentioned that Lean FRO discovered seven additional kernel bugs after soundness bug #14576.
Generated by AI from reports · Updated 18 hr ago
Progress timeline
- LessWrong对 Lean4 的批判性审视
LessWrong 用户 Milo Moses 发表文章，批判性审视形式化证明工具 Lean4 的可信度。他指出 Lean4 基于自研的依赖类型论而非 ZFC，目前没有公开证明其相对于 ZFC 加任意大基数公理是一致的，并称自己 90% 确信当前 GitHub 上的 Lean4 存在一个会被社区视为 bug 的 True=False 证明。文章提到 Lean FRO 在 soundness bug #14576 后又发现七个内核 bug，Leo de Moura 称 AI 很擅长利用内核的 soundness bug，嵌套归纳类型研究尚处早期、可能带来一致但行为失控的隐患。作者还批评 Anthropic 在费马大定理形式化中运行 comparator 时关闭了隔离不可信证明代码的沙箱，并称让 GPT-6 Astra 寻找 Lean4 到集合论的嵌入，花费约 25 小时和数千美元仍未成功。
Related discussion
Attention trend
New sources per day
- Oct 12New1 sources (1 outlets,0 accounts)