Full Second Brain Setup, Just put the .md in chat

Hi Clief Notes! Attached is a walkthrough for setting up your own second brain.

You can drop it into whatever AI model you use, and it will interview you one question at a time, help you design the system around your actual needs, and generate the complete scaffold. It includes CLAUDE.md, CONTEXT.md, Doctrine.md, AGENTS.md, the File Clerk, and the rest of the associated structure.

It can also guide you through setting it up as an Obsidian vault and connecting Cowork, Codex, or another agent tool, but those parts are optional. You can stop with the generated files or continue through the full setup.

I’ve coined this an Analytical Second Brain, or ASB, because it does more than store notes.

A normal second brain mainly helps you save and find information. An Analytical Second Brain compares new information with what is already known, separates raw sources from accepted conclusions, preserves decisions and their reasoning, shows how ideas and projects connect, and requires your approval before important changes become part of the maintained system.

It does this through a staged and gated process called the File Clerk.

In everyday use, the flow looks like this:

Begin a session

→ open the vault

→ direct the agent to CLAUDE.md

→ agent reads index.md and the applicable CONTEXT.md files

→ use QUERY for existing knowledge

→ place unchanged new material in 00_inbox/

→ directly invoke the File Clerk

→ review Gate 1 proposals

→ approve exact mutations

→ verify them at Gate 2

→ use projects/ for live work

→ run daily enrichment when material exists

→ run a weekly macro review

→ run LINT periodically

→ leave the approved state and traces on disk

The File Clerk does the searching, comparing, organizing, cross-referencing, and drafting. You retain control over what is true, what matters, and what gets added.

I’ve also included universal-file-clerk-handoff-prompt.md. You can drop it into almost any chat or coding session alongside the material you have been working on. It produces a structured handoff that can later be placed into the File Clerk.

At the end of the day, or after a few days, you run those handoffs through the staged File Clerk process to synchronize the brain. It helps prevent the system from becoming overburdened with duplicate, rewritten, reworded, stale, or contradictory context. It can also identify how subjects are becoming connected over time and propose links between them.

This system starts from pain, not from plan, much like

@Jake Van Clief

’s system does, but it goes a few steps further.

Pain usually comes from a misalignment between the current process and the desired outcome or ideal scene. The walkthrough helps you identify the recurring pain, examine the workaround you already invented, define what good would actually look like, and then design the smallest system that moves you from the current scene toward that ideal.

I don’t make content yet—looking at you,

@Matthew Creamer

—but I tested it by having it build an ASB for me as though I were a content creator.

It worked incredibly smoothly, especially considering that the original policy was written from the perspective of business operations and data processing. I only used GPT Instant, not Fable, Sol or Kimi, and it produced an Analytical Second Brain that might genuinely work for scripts, hooks, voice, style, and content projects. I was impressed!

GPT Inst did drift twice, so I had to reorient it to the document. It also asked a lot of questions. Good questions, but still a lot. That is partly the point: the system is trying to understand what you actually need before inventing a folder structure.

Hope it helps. Let me know in the comments what you think or what you run into.

8 comments

Full Second Brain Setup, Just put the .md in chat