I do think this experiment of yours is very interesting, Annie. It would be fun to feed it its 'improved versions' and see what it comes up with as well. One thing with LLM models is that they seem to be set to ingratiate themselves with their interlocutor. If you say 'could I improve this?', they assume the right answer is 'yes', so will always offer something. Just saying 'actually, that's fine as it is' seems like a challenge or criticism of the user.

There's certainly plenty to work on there.

I'm just finishing an angry soliloquy by a loaf of banana bread, following which I really should do some proper work (good manufacturing practice report on a pharmaceutical company in Barcelona...), but I'll probably revisit this AI writing thing soon. Thanks for your vote of confidence, and for the ideas.