System map0101A
Owning the Stack
Why the platform behind the show stopped being a bespoke stack per client (v0.2) and became one base that talks to a single hub (v0.3) — the happy path, and why it is so much simpler.
Companion to S01 E01 — Smooth Operations →The first version of this platform let every client keep their own stack — their own framework, their own cloud, maybe their own CMS — and stitched it together with a desktop app and a central orchestrator. The current version goes the other way: one base, one hub, one set of backends. Here is the happy path, before and after.
A bespoke stack per client
A desktop app shipped to every user. Each client ran their own website on their own stack — one on Netlify with Contentful, another on Cloudflare with a content backend they self-hosted — and they brought whatever framework they liked. The orchestrator in the middle, reaching each stack through per-client adapters, was the only shared piece. Flexible, but a lot of surface area.
- Our app + orchestrator
- Client sites
- Our content backend
- Third-party platform
One base, everything through the hub
Now every customer site starts from the same base template and the same shared libraries (Astro and the toolkits that work with the system). Each site talks to one hub — that is how it reads content, writes content, and reaches the AI. Only the hub talks to the backends behind it, the Postgres database and OpenRouter. The sites stay simple; the hub holds the complexity.
- Hub (control plane)
- Customer sites + base
- Postgres (content)
- OpenRouter (AI)
Same job, far fewer moving parts. v0.2 multiplied a whole stack — framework, cloud, content source — for every client. v0.3 keeps one base and routes everything through one hub, so the only thing that grows per client is content. That is the simplicity worth owning.
← All drops