alt Hacker NewsBad code is only really bad if it needs to be maintained.
If your AI reliably generates working code from a detailed prompt, the prompt is now the source that needs to be maintained. There is no important reason to even look at the generated code
Replies
AlexandrB • today at 2:24 AM
I think this might be plausible in the future, but it needs a lot more tooling. For starters you need to be able to run the prompt through the exact same model so you can reproduce a "build".
➕ show 1 reply
Well.. except the AI models are nondeterministic. If you ask an AI the same prompt 20 times, you'll get 20 different answers. Some of them might work, some probably won't. It usually takes a human to tell which are which and fix problems & refactor. If you keep the prompt, you can't manually modify the generated code afterwards (since it'll be regenerated). Even if you get the AI to write all the code correctly, there's no guarantee it'll do the same thing next time.