each row is an app, named by the identifier its records carry (app.bsky is Bluesky, sh.tangled is Tangled). a taller row has written more records overall. a brighter cell means that app was busier in that slice of the archive. ▸ opens an app to show its record types.
+columns are archive position: sealed segments in order, ≈3.2M events each. the dates are when stream witnessed the events, not when they happened — the bootstrap band replayed years of history inside six days from 2026-07-28, so it is dense; the live band is about two minutes per segment. account, identity and sync markers carry no collection and are not counted. "other" is every namespace below the ones shown.
+the archive is stored in chunks of about 3 million records; the columns are those chunks in order, oldest on the left. the dates are when the archive received the records, not when people wrote them: in its first six days (from 2026-07-28) it replayed the network's whole history, so those columns are packed — history, replayed. after that each chunk is a few minutes of live traffic — as it happened. account and identity changes belong to no app and aren't counted. everything else bundles the apps not shown.
-pick an app and every record it has ever written is searched right here, in your browser, from the public archive. Bluesky itself is too big for that and is shown from summaries only.
rows(seq, time_us, did, kind, op, collection, rkey) — queried in your browser with duckdb, straight from the public archive. app.bsky is aggregates-only by policy.