Pick boring tech first

Why this matters

The thing nobody tells you at 0 users is that every novel piece of infrastructure you adopt is a tax you pay every week for the rest of the company's life. Postgres, nginx, Redis, S3 — these are 'boring' because every problem you'll hit has been hit by ten thousand teams before you, with a StackOverflow answer waiting. CockroachDB, ScyllaDB, Temporal, FoundationDB — each of these has a real reason to exist, but most of those reasons start above 1M users. Picking boring tech at Day 0 means you spend your innovation budget on the product, not on debugging a database your future self has never used.

Demo

Dan McKinley's 'choose boring technology' essay names the rule: a startup should have an 'innovation token' budget of about three. Spend one on a novel database, one on a novel language, and one on a novel deployment model — you've spent them all on infrastructure and have nothing left for product. Most successful B2B SaaS companies at the 100k-user mark are running Postgres + Redis + a monolith in $LANGUAGE_THEIR_FOUNDERS_KNEW. The boring choice is not 'cheap' or 'lazy' — it's deliberately preserving your innovation budget.

Try it yourself

1. List every piece of tech in your current (or planned) stack. Mark each as 'boring' (10+ years in production at hundreds of companies) or 'novel'. Count the novel ones.
2. Read McKinley's 'Choose Boring Technology' essay end to end. It's a slide deck — 15 minutes.
3. For each novel choice on your list, write down a one-sentence answer to 'why couldn't I do this with Postgres?' If you can't answer, swap it for Postgres.
4. Look at the StackShare / public stack of one company you admire (Notion, Linear, Vercel). Note how few exotic pieces they have, and how many of those are recent additions, not Day 0 choices.

References

• Dan McKinley — Choose Boring Technology
• Bryan Cantrill — corporate open source anti-patterns
• Mikhail Bakhterev — why Postgres for everything