Outside of tests, LLMs sure are useful for slop and buggy software. Whether they’re useful for production-ready code and higher long term productivity rather than short term gains at the expense of users and maintenance later (look at Windows 11), seems at best unclear from the studies and experiments I saw so far.
As for tests, I personally wouldn’t really want to use LLMs for that either, since just because it’s possible to make it work doesn’t mean it’s the better approach long term. But I guess it seems like one of the lesser risk uses. I feel like a community isn’t owed fixes at the speed only an LLM might manage, hwoever, and I think that’s one of the problems of the debate.
Outside of tests, LLMs sure are useful for slop and buggy software. Whether they’re useful for production-ready code and higher long term productivity rather than short term gains at the expense of users and maintenance later (look at Windows 11), seems at best unclear from the studies and experiments I saw so far.
As for tests, I personally wouldn’t really want to use LLMs for that either, since just because it’s possible to make it work doesn’t mean it’s the better approach long term. But I guess it seems like one of the lesser risk uses. I feel like a community isn’t owed fixes at the speed only an LLM might manage, hwoever, and I think that’s one of the problems of the debate.