How the Second Brain Thinks

The Three Layers

The LLM Wiki pattern has three layers, each with a distinct role. Understanding why the layers are separate is as important as understanding what they contain.

• Raw sources: Your curated collection of documents — articles, papers, transcripts, notes, links. This layer is immutable. The LLM reads from it but never modifies it. It is your source of truth: if a wiki page is wrong, you can always go back to the raw source to check.
• The wiki: A directory of LLM-generated markdown files. Summaries, entity pages, concept pages, comparisons, and syntheses. The LLM owns this layer entirely. It creates pages, updates them when new sources arrive, maintains cross-references, and keeps everything consistent. You read it. The LLM writes it.
• The schema: A document called CLAUDE.md that tells the LLM how the wiki is structured, what the conventions are, and what workflows to follow when ingesting sources, answering questions, or maintaining the wiki. This is the key configuration file — it is what makes the LLM a disciplined wiki maintainer rather than a generic chatbot.

You Never Write the Wiki Yourself

This is the part that surprises people most: you are not supposed to write or edit the wiki pages. The LLM writes and maintains all of them. You are in charge of two things only: sourcing (what goes into the raw layer) and exploration (what questions you ask). Everything else — the summarizing, cross-referencing, filing, bookkeeping — is the LLM's job.

In practice, you can think of it as a collaboration: you are the editor-in-chief, deciding what matters and what questions are worth asking. The LLM is the researcher, writer, and librarian who executes on that direction and keeps everything organized.

How Knowledge Compounds

The self-improving loop has two entry points:

• Every source you add triggers the ingest operation: the LLM reads the source, extracts key information, updates relevant wiki pages, adds cross-references, and logs the event. A single new article might touch ten different wiki pages. The connections are made once and stay there.
• Every valuable query you make can generate a new wiki page. If you ask a synthesis question and get a useful answer, that answer gets filed back into the wiki. Your explorations compound in the knowledge base just like ingested sources do.

Over time, the wiki grows richer with every interaction. The more you use it, the better it gets — not because you are writing more, but because each interaction adds to the compiled knowledge layer.

The Obsidian Analogy

Karpathy describes the workflow this way: "In practice, I have the LLM agent open on one side and Obsidian open on the other. The LLM makes edits based on our conversation, and I browse the results in real time — following links, checking the graph view, reading the updated pages. Obsidian is the IDE; the LLM is the programmer; the wiki is the codebase."

You do not need Obsidian to use this pattern — any markdown file viewer works. But the analogy is useful: you are not a wiki author. You are a product manager for a knowledge base that the LLM actively maintains.

What Makes This Different

Most personal knowledge management systems require significant ongoing effort: tagging, linking, reviewing, writing. They improve only as fast as you can manually add structure. The LLM Wiki pattern flips this: structure is added automatically as a byproduct of your normal workflow — reading sources, asking questions. The system improves whether or not you are trying to improve it. The maintenance is invisible because the LLM handles it behind the scenes.