Hi HN. This is a small experiment using GPT atop a recommendation graph I've been developing for a few years. While GPT is broadly aware of books and their topic areas (dependent on its existing corpus), I found it quite fun and challenging to tune the prompts, parse out and normalize output to make a list of meaningful book titles. I reckon it's about as capable as a mediocre but widely diversified librarian? What do you think? Feedback really welcome.