Zorto should orchestrate the site and the data behind it.

Zorto started because I wanted to run my static websites on my own static site generator. I wrote up the first version in I’ve switched to Zorto!: fast Rust SSG like Zola with executable code blocks like Quarto. But what I’ve realized is more important is the architecture of “code” in three layers for effective use of AI for static websites and more:

  • content: Markdown; prose, data notes; primarily human-driven
  • config: config.toml; routes, templates, styles, sources, queries, dashboards, knobs; AI & human collaboration
  • code: HTML, CSS, JavaScript, chart glue, .db plumbing, browser weirdness; primarily AI-driven

The next move is build-time data orchestration: zorto build runs SQL, leaves receipts, renders query results into pages, emits a .db, and serves that .db as a static asset for dashboards. Plotly JS, SQLite WASM, DuckDB-Wasm, and friends make browser dashboards boringly possible. Fine. Let machines write more code. Do not make humans edit generated mess to change words, tune config.toml, or inspect the data pipeline. Zorto should be good UX for the whole stack: humans write words, humans and agents tune config.toml, SQL leaves receipts, machines deal with generated code and data, and the site stays static. content above config above code.