My Second Brain Setup

How I built a personal knowledge system using Obsidian and Claude Code — and what makes it actually stick.

I’ve tried a lot of note-taking systems. Most of them died within a month. The current setup has been running for over a year and I still use it every day. Here’s what makes it different.

Why a second brain at all

The problem isn’t remembering facts. It’s staying oriented — knowing where a project is, what decision was made and why, what idea you had three weeks ago that connects to the thing you’re building now. Without a system, that all lives in your head under pressure, and it leaks.

A second brain isn’t about capturing everything. It’s about capturing the right things so your head stays clear.

The vault structure

The vault lives at ~/second-brain/t3hK. Everything in Obsidian, structured like this:

t3hK/
├── 00_Dashboard/
│   ├── Home.md          # command centre
│   └── Dev Sessions.md  # rolling coding log
├── 01_Projects/
│   └── project-name/
│       ├── README.md
│       ├── Tasks.md
│       ├── Decisions.md
│       └── Plans/
├── 02_Areas/            # ongoing, no end date
├── 03_Ideas/
│   ├── Business/
│   ├── App/
│   ├── Creative/
│   └── Wild/
├── 04_Knowledge/        # personal wiki
└── 05_Journal/          # daily notes, weekly reviews

Projects have end dates. Areas don’t. Ideas older than a few weeks without action go to Archive. Simple rules, but they make the difference between a system that stays usable and one that becomes a junk drawer.

Claude Code integration

This is the part that actually changed how I work.

Claude Code reads and writes to the vault. Every dev session starts with /dev-session — it loads the current project context, reads the README and Tasks, checks for blockers, and orients itself. No re-explaining the project from scratch every time.

Custom skills as slash commands: /dev-plan for thinking through a new feature, /checkpoint to save mid-session state, /closeday to wrap up and capture what happened. The memory system means Claude builds up a picture of the project over time.

The hooks run automatically. When a session ends, a stop hook fires and saves key decisions and context back to the vault. It’s not perfect, but it means I almost never lose an insight to the void between sessions.

# ~/.claude/hooks/stop.sh — runs at end of every session
#!/bin/bash
# Prompt Claude to write a session summary to Dev Sessions.md
# The summary includes: what was built, decisions made, what's next
claude --print "Write a session summary to $SECOND_BRAIN/t3hK/00_Dashboard/Dev\ Sessions.md"

What makes it stick

Three things:

Low friction capture. If it’s hard to write something down, you won’t. The vault structure makes it obvious where everything goes. /new-idea creates a file in the right folder instantly.

Regular review. Weekly reviews (/weekly-review) are the maintenance pass. Things get promoted, archived, or actioned. Without this the system quietly decays.

Claude as co-author. Having AI that reads and writes to the vault changes the relationship with it. The vault isn’t just storage — it’s a shared context that makes every conversation better.

The goal is a system where thinking happens faster because the scaffolding is already in place.


Starter vault structure and setup scripts available to download if you want to adapt this for your own workflow.

Download the vault starter

back