My comment is not about that. The person doing the genning has to rewrite it before they can submit it. They’ll probably fix the problems as well if they put more than a sliver of effort, which would be needed to get past the pr audit.
Note the guidelines specifically say LLM is fine, but the submitter most be able to actually own the code. Basically if they can’t tell, then fine. However they don’t need to expand a great deal of effort rationalizing rejection of obvious first pass codegen fodder, they can cite this policy.
Basically a reaction to various people saying “you don’t need to know how to code or have skills to contribute code” (literally a quote from an email I received about Gemini 3). If all you have is the idea, just do the feature request. If an LLM can handle it, then one of the devs can do it. Going into an LLM knowing what it is and directly iterating on it might be helpful, but treating LLM input with a human middleman is just maddening.
My comment is not about that. The person doing the genning has to rewrite it before they can submit it. They’ll probably fix the problems as well if they put more than a sliver of effort, which would be needed to get past the pr audit.
Note the guidelines specifically say LLM is fine, but the submitter most be able to actually own the code. Basically if they can’t tell, then fine. However they don’t need to expand a great deal of effort rationalizing rejection of obvious first pass codegen fodder, they can cite this policy.
Basically a reaction to various people saying “you don’t need to know how to code or have skills to contribute code” (literally a quote from an email I received about Gemini 3). If all you have is the idea, just do the feature request. If an LLM can handle it, then one of the devs can do it. Going into an LLM knowing what it is and directly iterating on it might be helpful, but treating LLM input with a human middleman is just maddening.