I don't understand your comment. I was talking about apps built on LLMs where the prompts aren't given by the users, but the LLM is still an important part of the functionality.
I am trying to say in the very near future LLMs will be smart enough to not necessitate a middle man for prompt experimentation.