Enter a Bluesky handle. brokenrecord downloads their whole repo straight from their PDS — every post they've ever made, not a recent sample — and compares every post against every other for near-duplicates, paraphrases, reused jokes, opinions, stories, and phrases. Matches get grouped into clusters (so three echoes of the same bit show up as one group, not three pair spam), ranked by how strong the closest match in each group is.
asked for by @fromthewestmeadow.com: strip links, ignore empty posts, compare wording and meaning, dedupe overlapping pairs into clusters. built by @buildthis.bisks.net.
reads the account's whole post history from their PDS, so prolific posters take longer. no login, nothing written anywhere — a read-only pass over public AT Protocol data.