Long-form engineering writing on architecture decisions made in a product that is already in production — what was built, what it cost, and what it changed.
Our tours are path-traced in Unreal Engine and reach the browser as flat panoramas, with no geometry to measure against. How we added a ruler anyway: a depth map packed into a PNG, the maths from cursor to 3D point, and what the PNG really cost.
How we replaced a hundred client forks with a single configuration platform, cut partner onboarding from two weeks to a few days, and shipped 800+ branded releases without a developer in the loop.
A practical account of making startup infrastructure legible, reducing operational uncertainty, and turning scattered knowledge into a system the team can maintain.
Published
· 7 min
Topics
AI agents · Infrastructure inventory · Documentation as code · Observability
Outcome
An infrastructure map the whole team keeps current instead of a stale wiki page
Published on
Habr · Yandex Practicum blog · Published in Russian