PlanetScale shipped TIN on September 16, 2026 — a Postgres access-method extension for full-text search with BM25 ranking, proximity queries via TINQL, and MVCC-correct visibility. Vendor benchmarks claim 25× faster than ParadeDB and 541× vs built-in GIN on a 150M-document corpus.
TIN runs inside the Postgres transaction — no Elasticsearch sync pipeline. Lead provides TIN-compatible SQL for CI testing but is intentionally slow (full table scans).
TIN is not self-hostable today; available on PlanetScale Postgres and Neki only. Self-hosted Postgres users remain on ParadeDB, GIN, or external search services.