A boring, cheap stack that ships a v1 — and the list of things never to build yourself

The frame Your competitive advantage is not your database choice. Every hour on infrastructure is an hour not spent on the thing customers pay for. So the goal for a v1 is the stack with the fewest unknowns: technology you already know, or that is so widely used every problem you hit has a top answer already written. Default stack for a team of two · One full-stack web framework you know well. Not a microservices split — one deployable app. · Managed Postgres. Not a database you run yourself. Postgres because it does relational, JSON, full-text search and simple queues well enough that you won't need a second datastore for a long time. · A platform host that deploys on git push. No Kubernetes, no VPS you patch on weekends. · Error tracking from day one. You cannot fix what you cannot see. Never build these yourself · Authentication — sessions, OAuth, password reset, MFA. Use a library or a service. · Payments and billing — use Stripe or the regional equivalent. Do not touch card data. · Email delivery — use a transactional email API. Deliverability is a full-time job. · Search infra, feature flags, analytics pipelines, a job queue beyond Postgres — buy it or defer it. When to add complexity Add a cache when a query is measurably slow and already indexed. Add a queue when a request does work the user shouldn't wait for. Split a service out when one part has a genuinely different scaling or deploy cadence — not before. Each of these is a good decision at the right time and a tax paid early. Further reading · "Choose Boring Technology" — Dan McKinley on innovation tokens: you get about three, spend them where they matter. · The Twelve-Factor App — the checklist for an app that deploys cleanly. · The Grug Brained Developer — on keeping complexity out, written in the voice of a caveman. Funnier and truer than most architecture books.