Hacker Books is a bookshelf of the books people mention in Hacker News comments, from 2007 to 2026. A mention isn’t an endorsement, so the site also tells you how many people actually recommended each book. It currently covers 15,766 books across 281,363 mentions in 225,861 comments by 57,773 different commenters.
Browse Top all-time, see what’s Rising, follow a topic from the sidebar, or open a book to read the comments that mention it.
How it was built:
- HN dump. Start from the public archive of every Hacker News item and comment.
- Seed extraction. Regexes and light NLP find phrases that look like book references (quoted titles, “by Author”, Amazon/Goodreads/Open Library links, ISBNs).
- Open Library resolution. Candidates are matched to Open Library works, which gives a stable ID, authors, first-publish year, covers and subjects.
- Corpus-wide mention counting. Every resolved title is then searched for across the whole corpus, so books are counted even when nobody linked them.
- LLM disambiguation. A small language model is used only for ambiguous titles (“Dune”, “Influence”, “Range”…) to decide whether a comment is really about the book.
Scores weigh the number of distinct commenters more than raw mentions, so one enthusiastic person can’t dominate. “Rising” compares recent mentions with a book’s history.
Recommended vs. mentioned
Counting mentions alone makes a book that people merely discuss (or warn about) look like a favourite. So every mention was also labelled by Claude Sonnet 5.5 as recommend, neutral, critical or not about the book. “Recommended by” counts the different people who recommended a book at least once; “mentioned by” counts everyone. Mentions that aren’t about the book at all (a film, a game, a same-named work) are dropped. The Most recommended page ranks by recommenders. The labels were checked against Claude Opus 5.5 on 400 mentions and agreed 85% of the time, so expect some noise. Simpler approaches (keyword rules, sentiment scoring, a TF-IDF classifier) were tried first and were less accurate.
Privacy
No accounts and no tracking. “For You” is computed entirely in your browser from the books you open, and is stored in localStorage on your device. If you subscribe to update emails, your address is stored by this site (in a Cloudflare database) and used only for those emails.
No comments or reviews, maybe you want to be first?