Prismatic × Sig Nihl // Technical articles
These articles describe how Catharsis as a Service™ was made: the architecture of a bilingual static publication, the way one person directed an AI coding agent (Claude Code) under the supervision of Majordomus, the data model that keeps every public claim dated and graded, the tests that stand between a commit and a release, and what the records say about time.
They follow the same rule as the research notes: numbers come from records (the Git history, the GitHub Actions runs, the release list and the Majordomus ledger), each is linked to its source, and estimates are labelled as estimates. The case study presents the same records as an interactive timeline; these articles explain the engineering behind them.
Engineering 01
The architecture of a bilingual static publication and its build pipeline end to end: Markdown and TOML in, validated HTML and a versioned JSON API out, deployed only after the live URL has been checked.
Engineering 02
How one person directed an AI coding agent under Majordomus supervision: the kinds of instructions, scoped tasks, commit and push hooks, the pipeline, live verification and releases, and a record-based account of who did what.
Engineering 03
How every public claim on the site is recorded as data with a type, an evidence level, a confidence, sources and a review date; how validators and a weekly workflow enforce it; and how pages render grades from the ledger instead of restating them.
Engineering 04
The layers of checks between a commit and a release: an ordered validation script, Python validators that are tested to fail, Playwright on a subpath preview and on production, Rust contract tests, and a command registry whose every entry is executed.
Engineering 05
How several AI workers extended the site at the same time on a single Git working tree: content agents that write only their own files and propose shared data, a main session that merges, a peer session that commits by explicit path, and what the records show went wrong.
Engineering 06
How a site with no server runs simulations and search in the reader's browser: Alpine.js models drawn as SVG, including the Kuramoto model of synchrony, diacritic-insensitive search over an exported index, and per-page social previews rendered once and checked by hash without a browser.
Engineering 07
An honest account of time for a site built in one day by a supervised AI agent: measured elapsed times from Git, the pipeline and the Majordomus ledger; a labelled bottom-up estimate of the same scope for a human team; and the reasons this is not a controlled comparison.