I read Prompt Engineering for LLMs.
This O'Reilly book organizes prompt engineering in a systematic way. Rather than collecting ad-hoc tricks, it explains the principles behind why a given prompt works.
Through topics such as tokenization, context length, few-shot prompting, and output evaluation, it turns prompting from "experience and intuition" into something you can design. It gives you a foothold to verify and improve your prompts on your own.
As someone who builds with LLMs every day, I found it practical: it put into words and organized the things I had done by feel.
