What it is
A /Brain is not a giant archive. It is a deliberate context layer.
It gives an agent the few files that matter most at session start: who you are and how you like to work, what matters right now, and what should survive across projects and sessions.
The goal is simple: stop restarting from zero every time you open a new AI session.
Why it works
Most AI workflows fail at the start. You waste time re-explaining your priorities, your voice, and the decisions that already exist.
A strong /Brain fixes that by making context:
- local: you control the files
- inspectable: you can see what the agent is reading
- small: startup context stays readable
- durable: important knowledge survives beyond one thread or one tool
This is different from dumping everything into a retrieval system. A /Brain starts from a curated context slice, then uses deeper files or search only when needed.
What belongs in it
A good /Brain separates current context from durable memory:
me/WORKING_WITH_ME.mdexplains how to work with youcontext/now.mdcaptures what matters right nowMEMORY.mdstores durable truths worth carrying forwardframeworks/anddecisions/hold deeper reference material
Short-term capture belongs in places like _inbox/, voice-memos/, or logs/. Those inputs feed the system, but they are not the same thing as memory.
Common mistakes
- Treating the
/Brainlike a vault for everything - Letting startup files grow until they stop being readable
- Mixing short-term noise with durable memory
- Writing generic identity files that say nothing operational
Next steps
- Read Getting Started to create a minimal first version
- Read Obsidian as the Foundation to understand why Obsidian is the best home for a
/Brain - Read Core Files to understand what each startup file should do
- Read Maintenance to keep the system useful over time