This is a placeholder. The written case study has not been handed off yet.
The finished version arrives from the Drive handoff as
Clembot/site-anatomy-brain-graph.md and is ingested by
node scripts/site-anatomy-pull.js.
What it will cover, from the settled outline:
- The structure: an index of one-line abstracts, with typed edges to full nodes (decisions, playbooks, patterns, facts).
- Why agents scan the index first and follow edges selectively instead of loading the whole graph — context is a budget, and a note that is always loaded is a note that is never read carefully.
- What this replaced, and what it costs to keep the graph honest as it grows.