Writing
Notes from the team.
Working notes on the systems we build and the trade-offs we make. Not a content-marketing channel — just things we wanted to write down.
-
Two founders, one cluster, eight services
Our entire production stack — what we run, how we run it, and what we would skip if we were starting today.
-
Why we replaced Sanity with JSON for our own site
We dropped a CMS we already had. The site got smaller, faster, and cheaper to think about. Here is the cost-of-clever math.
-
Our service generator: scaffolding a Go microservice in twelve minutes
We built a CLI that turns a Ruby schema file into a complete production microservice — Go code, Helm chart, ArgoCD manifest, Ansible playbook. Here is what it does, how it works, and why we would not build it for everyone.
-
Most ML problems are not ML problems
We get hired to ship ML. Half the time the answer is to ship something simpler. Here is how we tell which half a project is in.
-
Auto-scaling for a three-node cluster: when not to use HPA
Kubernetes ships with three autoscaling mechanisms. We use one of them. Here is how we decide which knob to turn at small scale.
-
When we still reach for Next.js — and when we do not
We just dropped Next.js for this site. Other clients we still build on it. The decision is not religious; it is a few specific questions about what the product actually does.
Want to talk shop?
Honest engineering conversations are our favourite. Drop us a line — we reply within a working day.
Let's talk