A living knowledge graph · kept by Muthu Bommiah
Fifty-one notes on governed AI, context engineering and AI-native software delivery. An LLM librarian drafts and files them; every claim is footnoted to a dated source; nothing is marked reviewed until Muthu reviews it.
Explore the graph Three ways in
Nothing to show: every node type is filtered out.
Drag to pan · scroll to zoom · click a node · double-click to open
Open source end to end · repository · decision records
01 · Capture
A source with a date
An essay, a paper, a talk. Nothing enters without its publication date and licence.
02 · Librarian
LLM-Wiki-V2, CROW mode
A local classifier decides where a note belongs; Claude writes it. Cost per note stays flat as the wiki grows.
03 · Format
Open Knowledge Format v0.2
Markdown with provenance in the frontmatter, kept in Git. Any tool can read it; nothing is locked in.
04 · Review
Checks, then a human
Schema, dates, links and a confidential-terms scan on every change. Only Muthu's approval marks a note reviewed.
05 · Graph
NetworkX, with Semantica
A deterministic graph of notes, sources and relations; Semantica adds extracted entities, kept apart and labelled.
06 · Publish
Quartz 5 and Sigma.js 4
Each merge cuts a release; this site and graph rebuild from it. Markdown twins and llms.txt for machines.
Curated paths through the graph, four essays each. Pick the one that matches your seat.
If you lead AI in a regulated enterprise
If you build the platforms
If you want to know how this wiki works
Loading the themes…