Claude has memory. Obsidian gives it a brain.

Claude's memory is weak. That's the whole problem.

Claude's native memory holds onto fragments. Your name. A few preferences. Sometimes the last thing you said. But the moment you start a new session it has no idea what you actually worked on yesterday, what decisions you made last week, or what's currently on your plate.

So you spend the first five minutes of every conversation re-explaining yourself. Then Claude gives you generic answers because it doesn't know your specific situation. Multiply that across every agent you use — Hermes, OpenClaw, Codex, Antigravity, Free Claude Code — and you're losing an hour a day to "context re-loading."

This was the question Richard asked inside the AI Profit Boardroom last week. I answered him on video. This is the cleaned-up read-along of that answer.

The three-piece stack that fixes it forever

Three tools. Each one does one thing. Together they form a continuous loop where every conversation makes the system smarter.

Combine the three and you get a second brain that captures itself, organises itself, and feeds every AI agent you run.

OMI — taking notes on your day automatically

OMI is the listener. Wear it, or set it on your desk. It captures every conversation, every idea, every task you mention out loud — transcribes them, structures them, surfaces tasks, links to other apps. Yesterday I gave a coaching call; OMI captured the whole thing, surfaced 14 action items, and quietly logged them.

The killer feature for this guide: OMI exports to Obsidian. Every memory it takes on you can flow straight into a markdown file in your vault. Set it up once, forget it. From that moment on, every conversation you have in real life becomes content for your AI agents to read.

• What's free: OMI itself + the export-to-Obsidian feature
• What's paid: OMI chat (subscription). You don't need it for this setup — Claude is the brain.

The capture loop. Voice in. Memory accumulating. Intelligence flowing back. Every day you live, the brain grows.

Obsidian — your second brain, in markdown

Obsidian is a free note-taking app. The crucial detail: notes are stored as plain markdown files in a folder on your machine. Not a proprietary database. Not a cloud silo. Just .md files you can read in any text editor for the rest of your life.

Why this matters: every AI agent in 2026 speaks markdown. Claude reads it natively. So do Codex, Hermes, Antigravity, OpenClaw, Free Claude Code, Gemini. Your vault becomes a universal memory layer that every model can use.

The interconnection happens via [[double brackets]] — you link a note to another note, and Obsidian builds a graph view that maps the relationships. It's how the second brain becomes a brain (a network) rather than a stack of files.

Claude — the reader

Two ways to wire Claude to your vault. Pick whichever fits.

• MCP Obsidian (cleanest): Install obsidian-mcp. Claude reads + writes the vault automatically on every conversation. Permanent connection. The pattern Free Claude Code → uses too.
• Open Folder (quickest): In a fresh Claude session, click "Add to context" → "Open folder" → point at your Obsidian vault folder. Claude reads it for that session. Simple, no install. Good for trying it out first.

Once wired, you can ask: "Check through my memories and surface some ideas for what I could automate based on what I do day-to-day." Claude reads the vault, finds patterns, suggests automations rooted in your actual workflow — not generic SaaS advice.

That single shift — from generic answers to answers grounded in your real context — is the 100× multiplier. Every other improvement compounds on top of it.

Claude — the organiser

This is the move most people miss. After Claude reads your vault, ask it to reorganise it. The prompt:

"Based on my Obsidian vault, can you restructure it into a useful second brain using PARA? Create maps of content so the navigation is useful, not just tidy."

Claude will restructure the whole vault. Mine went from "untitled documents and one giant memories.md file" to a clean PARA structure with daily notes, maps of content, an SOP file, and a graph view that actually links together.

PARA — the only second-brain structure you need

PARA is Tiago Forte's structure for second brains. Four buckets. That's it. Claude already knows it — you don't need to teach it the rules.

The four temples. Projects, Areas, Resources, Archive. Pathways of light connecting them. The map of your life and work in four directions.

Claude builds you a Home dashboard at the root, a README that explains the rules (the SOP your future agents can read), and a Maps of Content file for each section. The graph view becomes meaningful. The system becomes navigable.

The flow that compounds — both directions

The system isn't read-only. It's a two-way loop:

• Vault → Claude. Every chat starts with full context. Claude pulls relevant notes, surfaces what's on your plate, references your actual decisions.
• Claude → Vault. Every chat writes back useful insights — notes on the conversation, action items, decisions, new resources. The vault grows.

So the more you use it, the smarter it gets. Every conversation feeds the next. By month 6 your second brain has more useful context on your business than you yourself can remember.

This is what I call the Infinite Context Engine. Same pattern that powers The Goldie Infinite Knowledge Engine → with NotebookLM as a content multiplier on top.

Three beliefs holding you back

The setup is free. The blockers are mental.

30-day roadmap

One piece per week. Each one builds on the last.

The Goldie Second Brain™ — recap

Claude alone is impressive. Claude wired to a vault that captures your life and organises itself is a different category of tool. The setup is free, the maintenance is automatic, and every agent you'll ever use will pull from the same memory.

OMI captures. Obsidian stores. Claude reads + organises. PARA gives the structure. The two-way loop makes it compound. One weekend of setup, then years of leverage.