What's the point of open sourcing something you one shot with an LLM? At that point just open source the prompt you used to generate it.

Testing. If you share something you've tested and know works, that's way better than sharing a prompt which will generate untested code which then has to be tested. On top of that it seems wasteful to burn inference compute (and $) repeating the same thing when the previous output would be superior anyway.

That said, I do think it would be awesome if including prompts/history in the repos somehow became a thing. Not only would it help people learn and improve, but it would allow tweaking.

To save time and energy?