second-brain
Second Brain Structure: PARA Done Right
The PARA folder layout we use across 50+ projects. Projects, Areas, Resources, Archive, adapted for AI-native workflows.
Why Your File System Is a Liability
Most knowledge systems collapse under growth. Folders named by month, project names that mean something today and nothing in six weeks, screenshots dumped in Downloads. The chaos is fine when you are solo, until you add a second brain, a second engineer, or a second client.
The structure below is what we use across 50+ projects at AY Automate. It is git-controlled, AI-queryable, and designed for Claude Code to navigate without a map. Every folder has a contract. Every file has exactly one home.
PARA Refresher: The Four-Category Rule
Tiago Forte's PARA method gives every piece of information one of four buckets. The rule is simple: ask where you will need this next, not where it came from.
P: Projects
Has a start date, an end date, and a goal. Client delivery. Product launch. Content series.
A: Areas
No end date. Ongoing standard to maintain. Sales process. Marketing machine. Operations.
R: Resources
Timeless reference. Agent patterns. Workflow templates. Prompt libraries. Useful always.
A: Archive
Inactive but worth keeping. Completed projects. Old strategies. Deprecated playbooks.
The AY-Adapted PARA
Stock PARA does not account for capture, binary files, or system metadata. We extended it with three prefix folders so Claude Code always knows where to write without asking.
Folder Tree: Interactive Explorer
Click any folder to see its purpose, file types, and real examples from inside the second brain.
PARA Method
Click a folder to explore
A knowledge system with four categories. Every file has a home.
P: Projects
Time-bound, clear end goal
A: Areas
Ongoing responsibilities
R: Resources
Reference, reusable anytime
A: Archive
Done or inactive
Per-Folder Conventions
Conventions eliminate decisions. These rules apply across every folder, every project.
snake_case everywhere
Folder and file names use underscores, not spaces. No spaces in paths, ever. Claude Code handles these without escaping.
Dated captures in Inbox
Files captured to 00_Inbox/ get the date as a suffix: relay-notes-2026-02-22.md. Sorting is trivial. Context is preserved.
README.md at every folder root
Every non-leaf folder has a README.md explaining its purpose, current state, and what belongs inside. Claude reads this first.
No deep nesting
Maximum three levels below the top-level PARA folder. Anything deeper is a smell. It belongs in Resources or should be archived.
Links over copies
Never duplicate a file. If a resource is needed in two places, reference it from one canonical location. 05_Attachments/ holds the source.
Archive on project close
The day a project ends, move its folder to 04_Archive/. Do not rename it. The history stays searchable and git tracks the move.
Sample Contents: The 4 Client Files
Every active client folder under 01_Projects/Clients/Active/ contains exactly four markdown files. No more, no less. Four files, zero ambiguity.
README.md
Overview, contact info, assigned engineers, engagement model, and key links. The single source of truth for who the client is and what we are doing for them.
Communication-Log.md
Every call, email, Slack message, and decision logged chronologically. Nothing falls through the cracks. Searchable history of all interactions.
Technical.md
Stack, architecture, integrations, repos, and deployment details. What is being built, how it connects, and where the code lives.
Progress.md
Sprint logs, deliverables, blockers, and weekly updates. The running record of what shipped, what is stuck, and what is next.
Migration Tips: Move Without Losing Your Mind
Migrating an existing knowledge dump to this structure takes one focused session. Here is the order that works.
Start fresh. Do not try to sort in place. Create the seven top-level folders in a new directory, then move files into them one category at a time.
Dump everything into 00_Inbox/ first. Resist the urge to sort as you go. Get everything into one place, then process the inbox in a single pass.
Sort by asking one question per file: when will I need this next? If it is part of an active project, it goes to 01_Projects/. If it is an ongoing responsibility, 02_Areas/. If it is general reference, 03_Resources/. If it is done, 04_Archive/.
Add a README.md to every non-leaf folder as you go. Two sentences is enough: what belongs here and what the current state is. Claude Code will thank you.
Initialize git at the second-brain root. Every file move is a commit. You get full history, blame, and the ability to revert a bad reorganization.
Run one Claude Code session after the migration: 'Read the README.md in each top-level folder and tell me if anything looks misplaced.' It will find the edge cases you missed.
The payoff
A structured second brain is not about tidiness. It is about giving Claude Code the context it needs without burning tokens on search. When every folder has a contract and every file has a home, your AI agents stop guessing and start shipping.
Want this running in your stack?
AY Automate builds AI automation systems for production teams.