The Second Brain

# The Second Brain: how to make your notes answer questions instead of collecting dust

Your notes app is a graveyard. You bury a thought there and never see it again — because a
normal notes app *stores* things, and storage is where ideas go to be forgotten. A second brain
does the opposite: it doesn't store notes, it answers questions, because every note links
into a graph the moment you write it. This is the exact setup I use, packaged so you can run it
in about ten minutes on any machine with Claude Code and a notes folder.

The kit is free. On Instagram you can also comment BRAIN and the bot will DM it to you;
everywhere else the download button above is it.

It's three small pieces that work together — no database, no account, no subscription:

- The capture ritual (second-brain-skill-template.md): a Claude Code skill that turns every
learning into ONE atomic, pre-linked note, and lets you query the vault by asking it a question.
- The note format (note-format.md): the single-idea, claim-titled, pre-linked shape that
makes the whole vault *linkable* instead of a pile of long documents nobody re-reads.
- The query guide (query-guide.md): how to ask your vault a question — day one through your
assistant, and later (optionally) through a real knowledge graph.

The trick that makes it compound: every note links into what's already there, so each new note
makes *every future answer* smarter. The graph only ever grows, and it only ever compounds.

1. Make the vault. mkdir -p ~/second-brain/learnings ~/second-brain/maps. Notes live in
learnings/, one idea per file.

2. Install the capture skill. Copy second-brain-skill-template.md to
~/.claude/skills/brain/SKILL.md. Now /brain captures a note in any session, and
/brain query "<question>" answers from your vault.

3. Learn the note format. Read note-format.md once. The only rule you must internalise:
one idea per note, and the title is a claim, not a topic. Everything else follows.

4. Turn the habit on. Add two lines to your CLAUDE.md (they're in the kit's README) so your
assistant captures the non-obvious learning at the end of each task and answers "what did I
learn/decide about X" straight from the linked notes.

5. Write your first note and ask it back. Capture one real thing you learned today, then run
/brain query "what did I learn today?". It answers from the note you just wrote. That loop —
capture once, ask anytime — is the entire product.

Level 1 — ask through your assistant (works day one). Because every note is a single-claim
title with pre-linked links:, your assistant can sweep the titles, follow the [[wikilinks]]
between the relevant notes, and answer from them — citing the exact files. No index, no database.
This alone replaces most of what people use notes-search for.

Level 2 — build the real graph (optional, ~30+ notes). Once your links form real clusters,
index the folder into an actual knowledge graph with community detection, and ask the *graph*
questions: "what's the shortest path between A and B?", "explain this concept from my own notes",
"what cluster is this note part of?" You'll need your own API key for whichever indexer you
choose — set it as an environment variable, never hard-code it. (The kit needs no key for Level 1.)

A notes app makes every note an island you have to go *find*. Here, every note links into what's
already there, so ten notes answer ten questions — but a hundred *linked* notes answer questions
no single note could, because the answer lives in the connections. The longer you run it, the more
your second brain can answer, with zero extra effort. That compounding is the whole reason to
build one.

- [Claude Code](https://docs.claude.com/en/docs/claude-code) — the assistant that runs the
capture ritual and the querying. The only requirement.
- A notes folder — plain markdown, on any machine. No database, no account.
- The kit itself — the capture-ritual skill, the note format, and the query guide, all in one
zip: download it here.

They treat their second brain like a filing cabinet — dump everything in, organise it into
folders, and never look back. Then it becomes exactly the graveyard they were trying to escape.
The fix is the opposite: capture *less* (only the non-obvious learning), make each note a single
claim, and always pre-link. A small vault of linked claims answers questions; a huge vault of
unlinked documents just buries them deeper.

Create ~/second-brain/learnings/, write one real thing you learned today as a single-claim,
pre-linked note, and then ask it straight back with /brain query. The moment your notes answer
a question instead of making you go find them, you'll never want a plain notes app again — and
every note you add from there only makes the answers smarter.