Appendix · About 5 min
Going further
What my own second brain does on top of this course, for when you want to automate.
This page is optional. Everything here needs a little scripting, or help from someone who does.
Automatic capture
Claude Code can run hooks: small programs that start on their own at set moments. Mine capture each conversation when it ends, extract what deserves keeping, and file it into the wiki afterwards. I no longer add things by hand: the brain fills itself as I work.
A context injected at startup
Another hook runs when a session opens. It no longer loads any notes, just the date, the health of the system and the commands to search. Claude searches only when the question requires it.
A word index
A small script builds a table: each important word → the names of the notes where it appears (in the title, a heading, a link). Claude asks it for names, not text, so finding a note costs almost nothing.
Monitoring
An automation that breaks silently looks exactly like one that works. Mine has failed twice for months without my noticing. Since then, every session starts with a health check: stale index, notes without themes, database stopped. If you automate, monitor at the same time, not later.
Typed graphs and graph databases
You can name relationships in note headers (works_at: [[Acme]]) to get a graph where every line has a type, or move to a graph database such as Neo4j. Tools also exist that generate a graph automatically from a set of documents. None of this is needed to start: a well-linked folder already does 90% of the job.