tl;dr: The author, inspired by a college job shelving library books, built a free searchable database of ~6,500 award-winning and shortlisted English-language non-fiction books, using Claude and GPT to gather data and enable semantic search for serendipitous discovery. He argues we're in a golden age of non-fiction (peaking roughly 1980s-2000s) enabled by post-war changes like jet travel, broader access to elite institutions, and cataloguing standards like MARC. The post also includes data visualizations of publishers, prizes, and book cover colors, and endorses Pangram as a surprisingly effective AI-writing detector.