Prompt engineering is a dumb phrase. What people should learn that will help with using LLMs is how they work, are trained, and why they fail.

Then you remove the magic and that helps with the next thing. How to evaluate output. LLMs are not great if you have little knowledge of the subject. They are far more useful if you are a subject matter expert.

The problem with AI is they promises the hype machine has made. They are writing checks they can't cash. They are useful but not magic. Not yet. It's an evolution not a revolution