Distillary

Brain Structure

3 min read

Brain structure

The brain is one Obsidian vault. Everything lives inside it — processed sources, cross-source bridges, your personal annotations. The folder structure reflects what the content IS, not when it was created or how it was processed.

Why one vault

The alternative is separate vaults per source. But then:

  • Wikilinks can’t cross vault boundaries — entities in one vault can’t reference claims in another
  • Bridge concepts can’t exist — there’s nowhere to put a note that belongs to both sources
  • Your annotations would be scattered across vaults
  • The graph view would show fragments, not the full picture

One vault means one graph. Everything connected. Bridges grow as sources accumulate. Your annotations link to claims from any source.

The folder layout

brain/
  sources/                     Processed sources
    lean-startup/
      _source.md               Source metadata
      claims/
        atoms/                 Layer 0: individual claims
        structure/             Layer 1: argument groups
        clusters/              Layer 2: chapter themes
        [root note]            Layer 3: thesis
      entities/
        concepts/              Technical terms
        people/                Authors, figures
        companies/             Organizations
        works/                 Books, products
    mom-test/
      ...

  shared/                      Cross-source content
    concepts/                  Bridge entities
    analytics/                 Comparison essays, entity mappings

  personal/                    Your voice
    annotations/               Reactions to specific claims
    notes/                     Freeform thinking
    questions/                 Things to explore

Why sources/ not books/

Distillary processes any knowledge source — books, videos, podcasts, articles, papers. The folder is called sources/ because that’s what it contains. Each source gets a slug derived from author and title: ries-lean-startup, yc-validate-ideas-2024, paulgraham-do-things-2023.

_source.md — metadata

Every source folder has a metadata file with the source’s title, author, type, publication year, URL (if applicable), and ingestion date. This enables filtering by source type (“show me all video sources”), temporal ordering (“what changed between 2011 and 2024?”), and attribution.

---
title: The Lean Startup
author: Eric Ries
type: book
published: 2011
source_slug: ries-lean-startup
ingested: 2026-04-07
---

Why shared/

Shared content doesn’t belong to any single source — it’s the synthesis layer. Bridge concepts unify ideas across sources. Analytics files (comparison essays, entity mappings, graph statistics) describe relationships between sources, not content of any one source.

Keeping shared content separate means re-processing a source doesn’t touch the cross-source synthesis. You can re-run the pipeline on The Lean Startup without losing the bridge concepts that connect it to The Mom Test.

Why personal/

Your annotations, notes, and questions live in a separate folder for two reasons:

First, re-processing a source regenerates its claims/ and entities/ folders. If your annotations lived alongside the generated claims, they’d be at risk of being overwritten. The personal/ folder is safe from the pipeline.

Second, when you publish or share your brain, personal/ is your unique contribution. Two people who process the same book get the same claims and entities. But their annotations — what they agree with, what surprises them, what they want to explore — are different. That’s the value of sharing brains: not the raw distillation, but the perspective.

Your annotations appear in the graph

Because annotations use [[wikilinks]] to reference claims and entities, they show up as connected nodes in the graph view. Your voice is visible alongside the source material, not hidden in a separate system.

Why this structure works for agents

An agent navigating a published brain can predict URLs from the folder structure:

  • /sources/{slug}/claims/clusters/ → main ideas of a source
  • /sources/{slug}/entities/concepts/{name} → entity page with backlinks
  • /shared/concepts/{name} → bridge concept spanning sources
  • /shared/analytics/comparison → cross-source synthesis

No crawling needed. The structure IS the API.

Graph View

Backlinks