Methodology
Every find should lead back to the thread.
HackerLinks uses automation to spot specific things mentioned in Hacker News discussions. The public archive is built from saved records—not generated on demand—and keeps the evidence beside every summary.
What gets included
A find must be a specific thing a reader can investigate: a product, tool, library, book, piece of hardware, talk, video, or similarly specific reference. Broad ideas, passing topics, and unsupported claims do not belong in the archive.
How the source is preserved
Each find's page shows when it appeared, which HN discussion mentioned it, and what prompted its inclusion. When the same thing resurfaces, the new sighting is added instead of erasing the earlier context.
How updates reach the site
An upstream scout collects structured records. Those records are cleaned, converted to public JSON, and rendered as static pages. Visiting HackerLinks never triggers a live HN scrape or an AI request.
Where automation stops
Selection and editorial summarization are automated and AI-assisted. The published site is deterministic: pages, feeds, manifests, and metadata are built from the same checked-in data. Summaries help readers navigate, but the original comments are authoritative.
What can go wrong
HackerLinks does not capture every worthwhile mention, and automation can miss context. Some records have thin summaries or no direct product URL. The HN thread is kept visible precisely so the archive can be checked rather than taken on faith.
Corrections
If a record misstates its source or links to the wrong place, use the corrections and contact page to report it. Reports are checked against the original discussion and corrected in the source record.