Opus 5.5 formalizes CPython internals in 613,318 lines of Lean
A researcher reports that Opus 5.5 produced 613,318 lines of Lean mirroring CPython's internals and found real interpreter bugs.
According to the researcher's post, the Lean code is tested against the real CPython interpreter and checked by Lean itself. The formalization is meant to be verified against the running interpreter, not taken on the model's word. The central claim is line-by-line mirroring: the code is supposed to track how CPython's internals actually work, not offer a loose high-level model.
The post reportedly says the project surfaced real CPython bugs, including crashes triggered by plain Python code. It doesn't describe how much human direction went into producing the 613,318 lines or how many attempts it took. The size of the output alone says little about how much of it was checked.
Why it matters
AI-written formal models of widely used software can expose real bugs, which may become a practical quality tool for language and infrastructure maintainers.
Original source
We have no link to the original source for this story yet.
Where we saw it
Latent Space
AI wrote this summary from the sources listed, and nobody reviewed it individually before publication. Check the original source before you rely on it. How it works