Sure, what I mean by that is that LLM makes different kind of mistakes than humans, they usually take the shortest direct route to accomplish their task. You can see that with the Bun Rust rewrite, I don't think any human coder would put as many `unsafe` and `Clone()` and `Arc<Mutex<T>` in their code, so a lot of time, they would just attempt to bypass the borrow checker if they see it get in their way.

Bun is an unrepresentative example of a Rust codebase for this purpose because (1) it was a direct port from a memory-unsafe language, and (2) it has a major C++ dependency (JavaScriptCore) whose objects in memory are deeply entangled with its own in very lifetime-complicated ways, which is something like the worst-case scenario for Rust's model of memory safety encapsulation.

Ah, so it's more that "LLMs just bypass Rust's safety" rather than "LLMs write perfect C"? That's a more fair argument.