Don't invent infrastructure

Distributed systems, storage engines, and compute primitives are hard—really hard. Most teams that invent them accidentally invent outages, staffing cliffs, and multi-year maintenance taxes.

Prefer open systems that have already paid the complexity tax: battle-tested brokers, query engines, orchestrators, and runtimes with real operators behind them. Your job is to compose and productize them for your domain—not to rediscover CAP the hard way.

There is an exception. When the problem is large enough and differentiating enough, inventing infrastructure can be the right call. That exception has a name of its own.