Boring tech scales better than yours

Every few months a new database shows up on Hacker News promising 10x throughput. Last year it was vector databases. The year before, edge databases. This year, something else. I've watched teams rip out Postgres for each of these, and three years later I've watched them rip the replacement back out.

Here's the thing: Postgres has been battle-tested for thirty years. It handles 10TB tables, JSON columns, full-text search, geospatial queries, pub/sub, and now vector search — all in one process. The features you need are probably already there. The bugs you'll hit are probably already fixed.

I'm not saying never use new tools. I'm saying: every new tool in your stack is a bet. Each bet costs you operational complexity, debugging time, and a class of bugs you've never seen. The boring tech tax is real, but it's a known tax. The shiny new tech tax is unknown — and unknown taxes tend to be larger.

When I audit a system that's failing to scale, the problem is rarely the database. It's usually that someone added a Redis cache layer that's now stale, a microservice that's now a network bottleneck, or a queue that's now a single point of failure. The cure is almost always: remove a layer, not add one.

The teams I've seen scale cleanly all share one trait: they're suspicious of new technology. They adopt it, eventually, but only after the new tech has proven itself on someone else's production. They're not late adopters — they're second-wave adopters. And they ship faster than the bleeding edge, because they spend less time debugging other people's software.