PostgreSQL

PostgreSQL is the one database until proven otherwise. Most teams reach for a specialized store long before they’ve exhausted what a single well-tuned Postgres instance can do — full-text search, JSON documents, queueing, even vector search with an extension.

The instinct to add a second data store almost always comes from a performance problem that turns out to be an indexing or query problem instead. Solving it in Postgres keeps the operational surface — backups, monitoring, access control — down to one system instead of two.

I still reach for something else when the access pattern is genuinely different — a cache that must survive a database outage, or an analytical workload that wants a columnar layout. Those are real reasons. “It felt more modern” isn’t.