Every engineering team gets a limited number of innovation tokens. Spend them on your core product. For everything else, choose the most boring technology that solves the problem.
What 'Boring' Actually Means
Technology with documented failure modes, handled edge cases, and Stack Overflow answers. PostgreSQL, Redis, Nginx — boring, reliable, performant.
The Hidden Cost of New
Learning curve, monitoring gaps, missing runbooks, 3 AM debugging. I've seen teams adopt Kubernetes for a single-service app that could run on a $20/month VPS.
When to Be Exciting
2M events/hour needs ClickHouse, not PostgreSQL. Real-time ML inference needs specialized infra. Be exciting where it matters, boring everywhere else.
My Decision Filter
Can a senior engineer debug this at 3 AM cold? Can I hire for this? Are there 3+ production case studies at my scale? If no — it needs a very compelling reason.