Tin: full-text search for Postgres

PlanetScale released Tin, a general-availability full-text search extension for Postgres. It supports boolean, phrase, fuzzy, and BM25-ranked queries while keeping correct MVCC transactional visibility, a property bolt-on search engines like Elasticsearch do not share with the source database. The team credits three specific design choices for the speed gain over existing Postgres text search: 48-bit document IDs that keep the index compact, work elision that skips redundant scoring work, and segment merging that keeps the index from fragmenting under continuous writes. For teams currently running a separate search service purely to get ranked full-text queries, a native, transactionally consistent option inside Postgres removes an entire piece of infrastructure and the sync lag that comes with it.