Introducing wipnote: Local-First Observability for AI-Assisted Development
Every AI coordination tool I looked at stored state in SQLite or JSON. Queryable, but you can’t diff them in a pull request. You can’t open them in a browser and read them. And they don’t express relationships between items.
HTML gives you all three for free. Browsers render it, parsers query it, git tracks every change. And links between HTML files form a graph — traverse them to find related features, prior decisions, things that didn’t work. The web is already a graph database; you just have to use it as one.
wipnote (launched as wipnote) started from that idea. The architecture evolved from there: SQLite joined as a cache layer for fast reads, a Go binary replaced the original Python scripts, and the design settled through daily use rather than upfront planning.