← Tin's Posts · July 06, 2026 · 2 min read
Infrastructure You Can Walk Away From
Stratodata builds an analytics SaaS, and a good chunk of their actual code gets written by an LLM, not typed by hand. Before I came on, deployments happened from developer laptops. "Latest" was pinned everywhere - not a version, the actual word, meaning something different every time someone typed it. Zero test coverage across 146 files. A real GitHub token sat hardcoded in the codebase.
Ordinary startup pace. Not a bad team - nobody had time to stop and ask how any of this actually reaches production. Sales needed doing.
I built them a CI/CD pipeline and an auth layer, in phases, over about a month - not one long build with a single reveal at the end. Two weeks for a working pipeline. A couple more to wire real auth into the actual application, not just deploy it alongside. Quality gates on every commit by the end of it, which matters more than usual here: a good chunk of the code those gates are catching mistakes in was written by an LLM, not a person.
Then Stratodata paused, to spend time on customer fit and testing before committing to the next round of work. Ordinary startup instinct - and a sign the phase structure worked: there was a demo-ready app to pause with.
Here's the actual point of this post: nothing broke when they paused. The pipeline kept running. The auth kept working. Nothing was built to require me still being in the room. A foundation that only holds together while its builder stays on retainer isn't a foundation. It's a leash.
It's also not finished. A couple of known edge cases are still live in production - multi-tenancy quirks, email delivery failures on custom domains - both self-caught, both self-managed. Worth saying plainly rather than pretending otherwise. The details are in the full writeup.
Juraj's read on it afterward, in a few words: "the delivery was clean... we could own the budget and use it open-eyed." Full testimonial: https://tinthe.dev/projects#testimonial-stratodata
This is the first thing I assess in any new engagement: not what feature ships next, but whether what's already shipped can survive you stepping away from it.
Full case study - the week-by-week build, the bugs, the rest of it: https://tinthe.dev/case-study-stratodata-cicd
Enjoyed this? Subscribe to get future posts by email.