Essays 路 Knowledge graph 路 LinkedIn
I write essays on governed enterprise AI and keep a working reference system for the argument they make: AI can write into a system of record only if every claim carries its source, every change goes through review, and machine output is kept apart from what a person stood behind.
flowchart LR
KF["knowledge-foundation<br/>content 路 rules 路 graph 路 releases"] -- "release assets" --> SITE["knowledge-foundation-site<br/>Quartz 5 路 Sigma.js 4"]
SITE --> WIKI(["wiki.muthub.org"])
MU["muthu-v0-portfolio-website<br/>Next.js"] --> ORG(["muthub.org"])
KF -. "counts, hourly" .-> MU
WIKI -- "canonical essays" --> ORG
| Repository | What it is | Live |
|---|---|---|
| knowledge-foundation | A governed knowledge base: an LLM librarian drafts, a human approves, every claim has a dated source. Open Knowledge Format bundle, a deterministic graph plus a Semantica semantic layer (entity resolution, ontology-typed relations), 16 decision records | releases |
| knowledge-foundation-site | The graph you can walk: a Sigma.js WebGL explorer as the home page, Quartz notes with provenance, strict CSP | wiki.muthub.org |
| muthu-v0-portfolio-website | Essays, initiatives and playbooks, readable by people and agents, with integrity and confidentiality gates in CI | muthub.org |
How the work is done: every change is a pull request with required checks; no bot commits to main; decisions are
recorded as ADRs; each repository has a docs/plan to rebuild it from scratch.
Start here: Six runs, one canary 路 Architecture that runs 路 the graph
Views are my own and do not represent my employer.