Seems like this works best for small pure functions in a strongly typed language.

Is pure core / imperative shell more LLM friendly?