> Software development has the huge advantage that mistakes and hallucinations are very easy to spot: the software works or it doesn't.

Do we not care about code quality, maintainability, performance, extensibility, or understandability anymore? Honest question, not a gotcha, it's just previously getting software to pass all the tests was a small part of what we would consider "working" or perhaps "good" software. Maybe that's different now with LLMs, idk. Maybe we need automated checks for these things as well, like not compiling until the code quality is good enough to let the agent finish it's loop.