The never-ending discussion of LLM-generated code in binary terms is tiring. It’s either all completely useless slop (which is the author’s stance), or it’s infallible and should be maximized at all costs. Yes there is slop, but there is also good or even great code produced by those making use of LLMs. I understand why a maintainer of a large project like Zig may outright reject LLMs for sheer practicality matters, just to stay on top of the wave of slop generated contributions. But it makes very little sense when a consumer of code, such as the author, takes a black and white stance on if LLM generated code is accepted by the language. If anything, it feels like an intellectual crutch. No need to understand technical nuances when you can fallback on dogmatic zeal.
0 comments
—