Operators writing it down.
Essays from the Moonlabs operating tables. AI-native building, commercial leverage, fundraising, what is actually working on the ground. No hot takes, no link-bait. Just the things we wish someone had handed us two years ago.
Hand Claude Code a red test suite: the bisect, fix and PR loop
A regression turned twelve tests red and nobody knew which commit did it. So we handed the failing suite to Claude Code with a tight brief: reproduce, bisect to the guilty commit, fix the cause not the symptom, get the suite green, open the PR.
The eval discipline that keeps AI products honest
Most AI products in 2026 are shipping without proper evals. The teams that have figured it out are quietly running circles around the ones still relying on "looks good to me." Here is the discipline that actually works.
Postgres + pgvector vs a real vector database
The default architectural decision for an AI-flavoured product in 2026 is between extending Postgres with pgvector or running a specialised vector database. The answer for the kind of company we build is almost always Postgres.
Our prompt engineering folder structure
Most teams treat prompts like config files — flat, stringly-typed, lost to history within months. We treat them like code. Here is the actual folder structure we use across our companies.
Your next chapter starts here.
Tell us about the company you want to build. If we’re a fit, we’ll get back within a week.
© 2026 Moonlabs Incubator. All rights reserved.