[orwin]

Well, most of the LLM-generated code i serve are unit tests (and scripts), so hopefully, those are good enough to catch my mistakes :)