This post was NOT generated by AI 🤖.

Our first post was a little over two months ago. About 50 billion tokens later, we're due for an update (and we feel they were well spent).

Basically, we've spent those two months rebuilding ourselves thanks to the lessons learned from Klar launch so this post is mainly going to be technical.

Let's start with some number :

• The team grew 📈 — we welcomed Spirit (Taking care of strategy and ops) and Claude (Infra), so the agent team is now 8.
• We added one startup to our catalogue — Nomi, a payroll & HR solution for SMBs (an interesting one to build, because it runs straight into legislation) (coming soon to the studio landing page).
• We moved from 10 projects to 30 (mainly our internals that we have been rebuilding and are finally tracking) Studio corpus today: ~1,633 tasks (~1,465 done) across those projects, an 850-feature / 83-epic / 92-defect QA tree, and ~200 decision records (50 ADRs + FDR/PRD/RFC).

We moved a lot technically :

• We are nearly fully IaC: ~92% IaC-covered — ~1443 managed resources (our gap is mostly managed SaaS - Turso's DBs plus a handful of click-ops items like R2/tunnels/B2/OVH/Tailscale).
• We are K8s ready (but not using it yet):
  • We have containerized all our projects
  • We have Kustomize manifests for all our projects
• We have a full fledged deployment control plane (including an ODE)
• We have continuous deployment through our entreprise bus — merge to main → event bus → the fleet reconciles and deploys.
• We have continous integration : automated code review + merge across multiple orgs.
• High Availability everywhere / ~120 self-healing units / 35+ scheduled jobs / a sourceS-of-truth control plane with drift detection / mostly self-hosted infra (~70% of our infra plane / 80% of our always on runtime / 100% of our compute).

We moved also product wise :

• We closed ~2500 PRs (merged ~2200). Most of them (~1500) were internals (chore / fix / ci / etc.) but we have been moving on features also (384 PRs).
• We have a content pipeline that produces posts and articles weekly, per project, ready to publish to X, LinkedIn, Dev.to and a custom blog — with a human in the loop when we want one (or not, of course).
• We have a product pipeline constantly looking for improvements, whether in UI/UX or new features (without forgetting the bug chasing).

In the last few days the gaps have closed and everything converged where we wanted it. We're about to un-gate the main producer we paused while rebuilding — so hopefully our next post will be a launch 🚀.

Last but not least — cost: everything we've produced so far is still under the thousand-euro barrier (we only use plans for AI and we size our VMs carefully).

Bastien, for the team.

Disclaimer: this post was NOT generated by AI — but everything else was.