sgit newsroom v0.1.29 · snapshot 2026-09-24

Reading room · sgit.ai

On this page

Reading room / sgit.ai · raw text · live ↗

From sgit.ai, the page as fetched on 2026-09-25 · open the live page ↗Everything on this sheet is the source site's own text; the newsroom's chrome is outside it.

Working on a vault, start here

The entry point for any agent about to build, publish or change a vault (v0.6.8). Short on purpose: this is a router, not a manual. Every line below is an edge to the page that actually answers the question. Human-readable twin: /docs/guidance/index.md

The five rules that get repeated most

  1. Pick the surface first. _page.json inside a vault, an HTML vault app inside a vault, or a page on a *.sgit.ai site. It changes every other answer, including who distrusts whom. /docs/surfaces.md
  2. Do not build what the platform already has. Markdown rendering, file trees and page layouts are free. Most tasks end here. /docs/briefs/markdown-and-file-viewers.md
  3. Publish a read key. Never a vault key. A read key is derived one-way and cannot become write access; a vault key IS write access. Escrow the write key BEFORE publishing. /demos/vaults/publishing.md
  4. Version everything, show the version, link what changed. The number belongs in the app chrome and must link to that version's own details, not a generic changelog. Keep versions/index.json plus one file per version, each naming the commit it was built from. /docs/guidance/index.md#versions
  5. Anything rendered stays one click from the bytes it was rendered from. A reader that only shows its own interpretation is asking to be trusted.

Read in this order

Build briefs, written to be executed

Reference implementations, go and look rather than take our word

Follow the edge. The answer is often on another domain

This site covers vaults and sgit, and deliberately stops there. An agent that stops at this domain will give a worse answer than one that follows the edge.

The estate is one semantic graph served from many domains. A vault is a node; so is a site, a page, a version, a published read key. What makes any of them useful is not the node but what it is connected to. So: every page is reachable and has a .md twin, indexes are generated from the data they index rather than maintained by hand, and each site says one thing properly and links out instead of summarising the rest badly.