hackerhub is an index of what Hacker News keeps recommending: repositories, papers, videos, books, RFCs, courses, datasets, films, albums, patents, podcast episodes and startup postmortems. A mention is not an endorsement, so every entry shows how many different people linked it, not just how often it appeared.
It currently covers 1,116,238 objects across 18,871,064 links in 43,051,020 comments by 25,610 ranked picks from 2006-10 to 2026-08. Browse a category from the bar above, open Stats for the long view, or search for a title or a domain.
How it was built
- Archive. Start from the public Hacker News dump: every story, comment, poll and job since 2006.
- Link extraction. Every URL in a story's target field and in comment bodies is pulled out, after decoding the HTML entities the archive uses (from 2013 on, a slash is stored as
/). That single fix roughly doubles the number of recovered links. - Object identity. Links are normalised into objects — YouTube video id, arXiv id, ISBN/ASIN, Wikipedia title,
owner/repo, RFC number, patent number, host plus path — so the same work under different URLs collapses into one row. - Aggregation. Each object gets its submission count, summed points, summed comments and the number of distinct commenters who linked it.
- Filtering. Objects need at least one submission or two different recommenders to appear, and per category the top 1,200 by points, comments and recommenders are kept.
How ranking works
- Points — sum of the score of every submission whose URL resolves to the object.
- Comments — sum of those threads' comment counts.
- Recommenders — distinct people who posted a link to the object in a comment. One person can only contribute one, so a single enthusiastic thread cannot dominate.
Sorting by points surfaces the biggest single discussions; sorting by recommenders surfaces the things the community repeats. They are different lists on purpose.
Coverage
What this is not
- Not sentiment analysis. A link in a critical comment counts like a link in praise.
- Not complete for prose mentions. Books, films and podcasts are usually named without a URL, so they are structurally undercounted. Extracting them needs language models over the comment text.
- Not perfect identity. Mirrors, shorteners and re-uploads can still appear as separate rows, and a few journal news pages sit next to real papers.
- Not live. The index reflects the archive up to 2026-08; newer items are not included yet.
Download the data
The aggregated index is published as JSON — one file per category, the same payloads the site renders:
- /data/meta.json — category totals
- /data/paper.json — papers (2,662 entries)
- /data/repo.json — repositories
- …the same path pattern for every category above
Raw source: open-index/hacker-news (ODC-By), generated 2026-11-03.
Privacy
No accounts, no analytics, no advertising. The only thing stored in your browser is your light/dark preference. Full details in the privacy policy.
Credits
Archive by the Hacker News community, mirrored by open-index under ODC-By. Cover images come from Open Library, repository cards from GitHub, thumbnails from YouTube. Comment text belongs to its authors and every entry links back to the original thread.
Inspired by Andrej Karpathy's HackerTube and by Hacker Books. hackerhub is an independent project on hackerhub.app and is not affiliated with Y Combinator, Hacker News or any linked site. Questions: hello@hackerhub.app.
FAQ
- What is hackerhub?
- hackerhub is a ranked index of the repositories, papers, videos, books, RFCs, courses, datasets, films, albums and patents that Hacker News keeps recommending. Every entry links to the original source and to the Hacker News thread that surfaced it.
- How are the rankings calculated?
- Each object receives three numbers: points (sum of the score of every submission whose URL resolves to it), comments (sum of those threads) and recommenders (distinct commenters who posted a link to it). The default ranking is by points; recommender counts are the most resistant to a single viral post.
- Where does the data come from?
- The public Hacker News archive published as open-index/hacker-news on Hugging Face (ODC-By licence): 49,324,111 items, 6,236,741 stories and 43,051,020 comments from 2006-10 to 2026-08.
- Is a mention the same as a recommendation?
- No. The index counts links, not sentiment: an object linked in a criticism counts the same as one linked in praise. Recommender counts mitigate this — one enthusiastic commenter contributes one link — but they do not remove it.
- Why do some categories have fewer entries?
- Because they are named rather than linked. Books, films and podcasts are usually mentioned in prose without a URL, so a link-based index sees far fewer of them than actually get discussed.
- Is hackerhub affiliated with Hacker News?
- No. hackerhub is an independent project. It is not affiliated with Y Combinator, Hacker News or any of the sites it links to.