How the Brain Works

The Brain is Andy's personal knowledge base — a structured system for accumulating, synthesising, and retrieving knowledge across 50+ projects using Claude as a thinking partner.

The three-layer architecture

raw/ — Source material. Articles, papers, PDFs, meeting transcripts, web clippings. The LLM reads but never modifies files here. Immutable inputs.

wiki/ — Compiled knowledge. Structured, interlinked markdown files maintained by the LLM. Entity pages, concept pages, summary pages, comparison pages, synthesis pages. Every page uses frontmatter (title, type, sources, dates, tags) and cross-references via [[wikilinks]].

outputs/ — Generated deliverables. Reports, analyses, to-do lists. Unlike wiki, outputs are consumed and may be disposable. Ephemeral deep-dives live in outputs/ephemeral/.

Why this structure matters

The split exists for token discipline. Wiki pages are cheap to load — they're already synthesised. Raw files are expensive — they're unprocessed source material. The rule: load wiki first, touch raw only when ingesting new material. Without this discipline, every session re-reads raw files and burns context window on material that should already be compiled.

The public wiki

This wiki — what you're reading now — is a curated subset of the Brain. Default-private, opt-in-public. Only pages Andy has explicitly authored a public-quality version of appear here.