alter_1: drowning in plumbing
The first attempt at a serious platform. It sank in three weeks, and taught the most important lesson of the whole project.
Late June 2025, the serious rebuild began: alter_1, the “Alter platform.” Real foundations from the very first commit, separate services for sign-in and for routing requests, a web client, automated tests and deployment (a pnpm monorepo with a gateway, an auth service, CI, and Docker, and notably, AI-agent-driven and spec-conscious from commit one, shipping instruction files for coding agents alongside the code).
What it was. All skeleton, no body. I had jumped straight into distributed-systems engineering (the kind of architecture serious products use) without ever deciding what the product actually was.
What went wrong. The commit log tells the story with brutal honesty: it ends with “comprehensive Docker troubleshooting report,” then a commit deleting the troubleshooting report, then silence. Three months of it. I had drowned in infrastructure, fighting the machinery of the platform instead of building the thing the platform was for.
What I decided. The most important decision in the project’s history: stop coding. If jumping to implementation keeps failing, invert the order. Go and learn to write the entire product down first, then build. What came next was the most consequential stretch of the whole project, and there wasn’t a line of code in it.
Next era: The hinge