https://www.youtube.com/watch?v=fY-pGkrLXg4

You're building a RAG app. Your team says: Postgres for the data, Pinecone for the vector search. You nod — because that's what you always do, one database per job.

Here's the thing nobody tells you up front: Postgres is the database that became a platform. The vector search, the geospatial queries, the time-series rollups, the fuzzy text search — all of it might already be Postgres. And three specific design decisions are what made that possible.

MVCC: Readers Don't Wait for Writers

Two transactions on the same row. One reads. One updates. Same instant.

In a traditional row-lock database, one of them has to wait. In Postgres, for ordinary reads and writes, neither one waits — and the reason is one fact that almost nobody teaches in an intro.

In Postgres, an UPDATE does not change the row. It writes a new row.…