AI Knowledge Center
A personal, self-organizing knowledge base for the Claude tech stack — built to be one
connected brain, not a filing cabinet. I land a doc, article, or blog from the CLI; it gets distilled
into an atomic note, wired to everything it relates to, and placed on a single evolving map. The
goal: master the Claude/Anthropic stack, pass the CCA-F exam, and help small teams adopt agentic AI.
The point isn't to store notes — it's to connect them. Every new idea is force-fit into the bigger
picture so I can see how the pieces fit and how to use them in real scenarios.
Two views
| View | What it is | Where |
|---|
| 🧠 Concepts | Everything I'm learning about the Claude stack in one place: the mental model of every Claude Code primitive, a comparison matrix, a decision guide, primitives, and a self-organizing concept graph wired by typed connections. | docs/concepts/ + knowledge/ |
| 🎓 CCA-F Exam Prep | About the exam, an interactive practice quiz, and the six production scenarios it's built on. | docs/concepts/exam.html + exam-prep/ |
How the Concepts graph works
Three mechanisms turn a pile of notes into a brain:
- Typed connections — notes link with labeled relationships (
builds-on, enables, part-of,
used-with, alternative-to, contrasts-with, used-in), each with a one-line reason. This is the
wiring.
- The spine — every note declares a
spine_layer so it slots into the
Big Picture: Foundations → Claude API → Agent SDK → Claude Code primitives → Agent patterns → Products & consulting. Nothing floats free.
- The big picture — an interactive graph
you explore, plus a narrative synthesis note that explains how it all
fits. Both grow as notes are added.
Notes live as plain markdown in knowledge/notes/; the graph is derived from
their frontmatter by scripts/build-index.py. Structure emerges — no topic
folders, no 101/201/301 levels.
Landing new knowledge
/land <url | file path | pasted text>
The /land skill distills the source into an atomic note (in my own
words, Claude-specific, with a "why it matters" and a "how to apply" section), connects the dots
(2–5 typed links + a place on the spine), updates the big-picture map, regenerates the graph, and teaches
where the new idea fits. A validation hook blocks any note that's
malformed or connects to nothing — the anti-silo guard.
Conventions live in CLAUDE.md. After editing notes by hand, run
python3 scripts/build-index.py to refresh the graph.
Layout
knowledge/
README.md # Home Note (dashboard / entry)
notes/<slug>.md # flat pool of atomic notes (+ images alongside)
maps/big-picture.md # the spine + narrative synthesis
exam-prep/ # markdown companions for the Exam view
docs/ # the two-view GitHub Pages site
scripts/build-index.py # regenerates docs/assets/knowledge.js (the graph data)
.claude/skills/land/ # the ingestion + dot-connecting skill
.claude/hooks/ # note validation / anti-silo guard
LEARNINGS.md # dated log of hard-won lessons (newest at top)
Publishing (one-time)
GitHub Pages serves from /docs: Settings → Pages → Deploy from a branch → main + /docs. The
site rebuilds on every push to main.
