[int_19h]

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.

[thawawaycold]

that is not what OP was talking about though.

[rorytbyrne]

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

[warkdarrior]

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