Duplicate detection without an API

Ctrl+Shift+H, an editable title field, the guidelines in a tooltip, and duplicate detection, across Chrome, Edge, Brave and Firefox from one Manifest V3 codebase.

Duplicate detection has no clean solution because HN has no submission-check endpoint. So the extension queries HN’s own domain search and surfaces the highest-scored existing submission for that URL. The awkward half is normalization: the same article arrives with trailing slashes, www, UTM parameters and AMP variants, and unless those collapse first the search reports no duplicate for a story on the front page right now. That is where the false negatives live and it is never finished.

Title optimization is deterministic rather than a model call. HN’s guidelines are a small set of mechanical rules, strip the site name, drop clickbait prefixes, leave a year suffix alone, and rules a person can predict beat a better-on-average rewrite for something you are accountable for.

Version 1.4. The guidelines tooltip converts better than moderation after the fact.