Hacker News Reader

We don't know how to implement a "deterministic, repeatable process" that can look at a bug in a repo and implement a fix end-to-end.

that is not what OP was talking about though.

LLMs are literally stochastic, so the point is the same no matter what the example application is.

Humans are literally stochastic, so the point is the same no matter what the example application is.