All insights
Boring tech, on purpose
Why we still pick Postgres, Rails, and a single Hetzner box for new bets — and the two cases where we don't. The cost of "interesting" is paid in 2 a.m. pages.
We pick boring tech for new bets — Postgres, Rails or Astro/Node, a single beefy Hetzner box, Cloudflare in front. It runs.
The two cases we deviate: when we need GPU inference at the request path (then it's Modal or RunPod), and when the data shape is genuinely event-y (then it's ClickHouse).
The cost of 'interesting' tech is paid in 2 a.m. pages. We'd rather be bored at midnight.
Got a problem worth shipping?
If this resonated and you've got a project queued up, we should talk.
Book a call