[zucker42]

Automated software writing from a codebase and a requirement description, with the model asking clarifying questions where necessary, would be an immensely valuable application. We're quite far away from that, but GPT-4 is already rather helpful for some coding tasks.

[gloosx]

In this specific use case though it sucks the most since one will be answering more questions than code needed in sheer volume. Because the code is already straightmost clarification to the initial question. We will never be closer than looking up what you don't understand to solve your problem in the first place and it will always take more time than directly solving it.

I had a good example recently and I'll quote it again:

People seriously using the "Please write an expression for me which adds 2 and 2" prompt in order to get the "2+2" expression they need – advocating that they got it with magical efficiency