pa2: PARA + AI
The first real product: a filing system that interviewed you, sorted your goals, and summarised your projects.
Two months after shelving chotu, I came back with the organising principle, PARA, Tiago Forte’s method (from Building a Second Brain) for sorting everything in your life into four buckets: Projects, Areas, Resources, Archives. In about ten days I built pa2, and nearly everything Alter does well today exists here in embryo.
What it was. A knowledge system with AI in the loop from day one. It on-boarded you through a conversation. You talked about your goals, and it filed them into the right buckets for you. It wrote summaries of your Projects and Areas so you didn’t have to re-read everything. And the AI work happened in the background while you kept going (a React front-end, a Python API, a background job queue for the AI calls, more than one AI provider behind a single interface, and a vector database for meaning-based search: three pieces that are the direct ancestors of Alter’s recognition workers, its multi-provider layer, and its semantic search today).
What went wrong. Two things. It was split across two languages, which made a solo project twice the work (Python backend, TypeScript front-end). And the framing was subtly off: it was a filing cabinet that happened to be smart, when what I actually wanted was closer to a thinking partner. It organised my stuff; it didn’t help me think.
What I decided. Collapse to one language (TypeScript), and reframe the product away from “filing system” toward something lighter: a smart notepad and a task board.
Next era: CogniNote