I realize that by using chatGPT (and similar), there are certain situations in which they perform really well (e.g. code suggestion) and others where they give bland and generic responses. Therefore I would like to become better at using them.

I've seen a lot around "Prompt Engineering" but so far I haven't seen anything that is significantly better than OpenAI's own documentation about GPT's best practices. Is there actually valid content being produced in this regard?

Furthermore, aside from Prompt Engineering, I would be interested to understand how LLMs work. My goal is to be able to evaluate critically what I can use them for, and what are their limitations (i.e. since the quality of their output changes depending on the area of expertise). If it helps, I'm happy to start studying Machine Learning, but my goal –at least atm– is not to build models from scratch. Is there any resource that is oriented towards a general understanding rather than in-depth application of LLMs?

TLDR: I feel I have a lack of understanding about how LLMs work, and I would like to make up for it in order to employ them better in my work.

Thanks