Tamir BaharI still don't understand why "it's now far cheaper to write code" leads everyone to "so it doesn't matter if the code is an unreadable mess" and not "so we can make it really neat and clear and well-made"
Matthew Martin@tmr232 My go-to for code quality is a strict build script. The ones that work well for humans don't necessarily work well for LLMs. For example, an LLM will change the tests to make the tests pass. So now I need a tool to fail the build if the tests changed without my approval. With humans, that is dealt with by social pressure & social consequences.
@mistersql yeah, I'm starting to think what checkboxes I might want to check before even starting to review code. But it's trickier when there are tens of developers.
Currently thinking that the first thing to review should be the tests. Can't trust LLM code without them (it makes different mistakes than humans), and can't trust them to write decent tests (mocks! Mocks everywhere!)
But I need to sharpen my test-reviewing skills for that!