Second Brain for AI: Make Your Notes Actually Useful

Updated January 2026 | 6 min read

You've spent 200 hours building a second brain. Organized notes. Tagged pages. Linked ideas. Beautiful wiki structure.

Your AI has never read any of it.

Every time you ask ChatGPT to help with a project, you're copy-pasting information from your notes into chat. Your AI assistant sits in one app. Your knowledge base sits in another. Completely disconnected.

That's not a second brain. That's a filing cabinet you manually transcribe.

The Knowledge Silo Problem

Most note-taking systems create knowledge silos.

Notion stores your notes in a proprietary database. Your AI can't access it.

Roam Research uses an outliner format. Great for human thinking. Unreadable to most AI tools.

Apple Notes syncs to iCloud. Locked to the Apple ecosystem. No AI integration.

Even if you export notes to PDF or text files, those formats don't load into conversations automatically. You have to manually upload them. They expire when the chat ends. Next conversation, you start over.

A real second brain for AI needs three things:

1. Plain text files your AI can read
2. Automatic loading at conversation start
3. Persistent context that doesn't expire

Only one setup delivers all three: Obsidian + Claude Code.

How Obsidian Turns Notes Into AI Context

Obsidian isn't a database. It's a file browser that renders markdown.

Every note you write is a .md file stored on your computer. No proprietary format. No cloud lock-in. Just plain text files in a folder.

Claude Code reads local files. That's the bridge.

When you start a Claude Code conversation, it loads a CLAUDE.md file from your Obsidian vault. That file tells Claude:

• Who you are
• What you're working on
• Where to find more context (paths to other notes)

Your notes stop being reference material. They become persistent AI memory.

The Difference Between Notes for Humans and Context for AI

When you write notes for yourself, you use shorthand. Incomplete sentences. Assumed context.

"Follow up with Sarah re: proposal."

You know who Sarah is. You remember the proposal. The note is a trigger, not complete information.

AI needs complete information.

"Sarah Johnson (sarah@example.com) — Valley Welding client. Sent website redesign proposal on 2026-01-15. Follow up on 2026-01-22 if no response. Proposal includes: new homepage, services page, contact form. Quoted $4,500. Decision maker: Sarah (owner). Timeline: 6 weeks from approval."

Same reminder. But now your AI knows:

• Who Sarah is (full name, email, company)
• What the proposal includes (specific deliverables)
• When to follow up (exact date)
• What you quoted (price and timeline)

A second brain for AI requires rewriting your notes with that level of specificity.

Context Files vs. Regular Notes

Not every note needs to be AI context.

Your daily journal? Human-only. Your random ideas? Human-only. Your book highlights? Human-only.

Context files are different. They're structured information your AI loads automatically.

In your Obsidian vault, separate them with naming conventions:

• _context.md — Domain-specific AI context
• CLAUDE.md — Master AI instructions
• Everything else — Regular notes for human use

When you mention certain keywords in conversation, Claude Code loads the relevant _context.md file. It doesn't load your journal entries or random notes. Only structured context.

Building Your Second Brain for AI

Step 1: Audit Your Existing Notes

Open your current note-taking app. Ask: "What information do I re-explain to AI constantly?"

• Client names and details
• Project specs and deadlines
• Pricing and service packages
• Brand voice guidelines
• Tech stack and tools

Pull those notes into a new document. That's your seed content.

Step 2: Migrate to Obsidian

Download Obsidian (free). Create a new vault. Name it something functional: "Business" or "AIContext" or "Work."

Copy your seed content into Obsidian notes. One note per domain.

If you have four clients, create four notes: CryoBodyWorks.md, LimitlessChiro.md, ValleyWelding.md, KealeChiro.md.

If you have three service types, create three notes: ContentWriting.md, SEO.md, WebDesign.md.

Don't worry about organization yet. Just get the content into plain markdown files.

Step 3: Create Your CLAUDE.md File

This is the master file. It tells Claude Code what to load and when.

Basic structure:

## WHO
Victor Romo. I run Scale With Search (SEO education). Four retainer clients.

## WHAT
I write content briefs, articles, and topical maps for chiropractic and welding clients.

## VOICE
Direct. No fluff. Use contractions. Vary sentence length.

## CONTEXT ROUTING
| Keyword | Load File |
| "Cryo" | /Clients/CryoBodyWorks.md |
| "Limitless" | /Clients/LimitlessChiro.md |
| "Valley" | /Clients/ValleyWelding.md |
| "Ke'ale" | /Clients/KealeChiro.md |

When you type "help me write a blog post for Cryo," Claude Code sees "Cryo" in your message, checks the routing table, loads CryoBodyWorks.md automatically.

No copy-pasting. No re-explaining. It just knows.

Step 4: Test the System

Open Claude Code. Type: "What clients do I work with?"

If Claude lists your four clients by name, your second brain works.

If it says "I don't have that information," check your CLAUDE.md file. Make sure it's in the vault root. Verify the routing table uses correct file paths.

What This Unlocks

Client-aware content. Ask for a blog post and get output that matches brand voice, uses correct terminology, includes relevant keywords. Without briefing the AI first.

Project-aware code. Request a website feature and get code that uses your actual tech stack. Not generic suggestions. Your tools.

Context-aware communication. Draft emails that reference real projects, actual deadlines, correct pricing. Because your AI knows your business.

Memory across conversations. Start a project Monday. Pick it up Thursday. Your AI remembers what you discussed because the context file didn't change.

Why This Is Better Than Uploading Documents

ChatGPT lets you upload PDFs and text files. Claude lets you attach documents to conversations.

That's not persistent memory. That's temporary context.

Uploaded documents expire when the conversation ends. Next chat, you re-upload. If you forget, the AI doesn't have the information.

File-based context loads automatically every conversation. You don't decide what to include. The routing table decides based on your prompt.

It's the difference between carrying groceries by hand vs. installing a pantry.

Maintenance: 10 Minutes Per Week

Your second brain for AI doesn't require daily upkeep.

When something changes (new client, updated pricing, completed project), update the relevant context file. That's it.

The file updates. Next conversation, Claude sees the new information. No syncing. No refresh. Just updated context.

Most people spend 10 minutes per week updating context files. Land a client? Add them. Finish a project? Archive it. Change your service packages? Update the file.

Compare that to re-explaining your business 5-10 times per day in ChatGPT.