Have been curious what it could look like (and whether it might be an interesting new type of “post” people make) if readers could see the human prompts and pivots and steering of the LLM inline within the final polished AI output.