I found lots of good value in their publication as well.

Especially for teachers, who I believe (most at least) have no clue about prompt engineering and how to talk to an LLM.

IMO ’Prompt engineering’ is an implication that the LLM:s are really immature technology. There is no intrinsic value in prompt engineering - it’s ok to wait a bit until LLM:s get a proper product shell you don’t need to walk on eggshells over. I would not promote LLM:s as production ready offerings until this aspect becomes better.

Using an LLM is like having a therapy session - where you the user are the therapist. Humans should not need to learn en masse become AI therapists, that’s a the inverse of what should happen :D

You might be influenced by the perception that LLMs are autonomous.

They're just tools. The outcome depends heavily on the user skills. That's why prompt engineering is a thing.

Heck, even Google results varies depending on the searcher skills, and we're not calling the tool immature...

As with regards to therapy, I have had the opposite experience as you described.

I agree, most don't even know they can tell it how to behave.