twins.sgit.ai / documents / site-architecture
Site architecture
Summary
The page-by-page specification for this site, the house pattern to copy, and the character it should have: a reference site for a primitive, kept small and canonical rather than broad, where the applications live elsewhere and each one here is three paragraphs and a link out. It also names the one check worth automating, which became this site's release gate.
Key concepts
- Small and canonical — one owner, many referrers
- The name-watch — grep the site for new senses of the word, per build
- The house pattern — validate, tag, publish
Key ideas
- The discipline page is the most distinctive page available, because nothing else in the literature says either we have evidence and it exists, or we do not.
- Each application page is three paragraphs and a link to its owner site.
- Grep the estate for new uses of the word per build: the collision happened once and will happen again.
- A name-watch is this site's equivalent of the licence audit.
On this site
Implemented directly: the page set, the network boundaries, and the name-watch that now fails the build in admin/build/validate.js.
Read the document
📄 Original document · v0.33.62 · 24 August 2026 · rendered from the raw markdown (the source of truth)