The Second Brain Actually Works Now

The Notion workspace had 200+ pages in it. Organized by project. Meeting notes cross-linked. Templates for every recurring workflow. A capture system, a review system, the whole architecture.

Built it over four months. Used it for four months. Then gradually stopped.

Not because it wasn’t organized — it was. I stopped because the knowledge just sat there. Build something up, then go back in, find it, read it, and manually translate what was in the notes into whatever I was doing now. The second brain was a library. Libraries are passive.

I kept the workspace. Kept adding to it occasionally. But I knew, somewhere in the back of my mind, that this was not what the promise was supposed to be.

Last month, I ran a session where Claude Code queried that Notion workspace through MCP and synthesized context from fifteen pages in one pass. No manual browsing. No copy-pasting. I described what I was building, and it pulled the relevant history, cross-referenced the meeting notes, surfaced decisions made three months ago, and integrated all of it into the deliverable — while I was building it.

That’s when I understood what had been missing.

The Gap Was Always Execution

Tiago Forte’s second brain concept — and its predecessors, going back to personal knowledge management as a category — was always fundamentally correct as a theory. Externalize your thinking. Organize your information. Let your knowledge compound over time rather than decay the moment you close a tab.

The theory was right. The practice fell apart at retrieval.

You built the system, you maintained the system (or mostly maintained it), and then when you needed something from it, you had to go get it yourself. You were the retrieval mechanism. And the synthesis mechanism. And the translation mechanism — from what was stored to what was needed right now.

The system helped you hold more. It didn’t help you do more with what you held.

That gap existed because the only tool interacting with your knowledge base was you. Your notes couldn’t read themselves. Your Obsidian graph couldn’t generate a proposal from your past work. Your Notion database couldn’t draft a project plan grounded in what had already been decided.

All of that required you to be present — the bridge between stored knowledge and current execution.

What Changes When the Agent Can Read It

When Claude Code connects to your Notion workspace through MCP, something structurally different happens.

Your knowledge base stops being a place you visit and starts being a resource something else can act on. The agent can query it during a task. Pull what’s relevant. Incorporate it without you having to locate it first. The stored knowledge flows into the current work without you having to carry it there manually.

I ran a session to build out a planning document for a consulting engagement. The kind of document I’d normally spend half a day on — pulling together past notes, prior decisions, context from previous conversations. Instead, I described what I needed, Claude Code queried the relevant Notion pages, and the output was grounded in actual history rather than whatever I happened to remember in the moment.

Fifteen pages of context. One session. The document was better because it had access to things I’d forgotten I knew.

That last part is the piece worth naming clearly: things I’d forgotten I knew. Your knowledge base, properly maintained, holds more than you’re currently holding in your head. The original promise of the second brain was that it would make stored knowledge available when you needed it. It just never solved the retrieval-and-use problem in practice.

MCP closes that loop.

The Obsidian Side

The same pattern applies for people whose knowledge lives in flat files.

Claude Code with file access can read your Obsidian vault, parse the links between notes, pull context from your daily notes, and incorporate that context into whatever you’re building. If past-you captured a useful insight six months ago, present-you doesn’t have to remember it existed — the agent finds it.

This changes the calculus on what’s worth capturing. Notes that felt like record-keeping — “I might need this later” — become genuinely useful if the agent can access them later without you having to retrieve them manually. The return on capture goes up.

Which also means the second brain idea was always right about the capture side. Write it down. Externalize it. The problem was that retrieval was expensive. Now retrieval is nearly free.

What It Asks of You

This doesn’t work without a knowledge base. You need the Notion workspace, the Obsidian vault, the flat files — whatever system you’ve maintained over time.

If you abandoned your second brain because retrieval felt like too much work relative to the value returned, it’s worth revisiting. The calculus has changed. The system you built and mostly stopped using is now a resource an agent can act on, not just a library you can browse.

The investment you made in capturing and organizing is now worth more than when you made it.

I’ll be direct about what this isn’t: it’s not magic. A poorly organized workspace with stale pages and inconsistent structure returns stale, inconsistent context. Garbage in, garbage out still applies. The agent is only as good as what’s there to retrieve.

But a reasonably maintained knowledge base — one that captures decisions, meeting notes, project context, past work — is now a genuine operational asset. Not just for you to browse. For the agent to use on your behalf.

2026 Is When It Executes

The second brain concept has been around for more than a decade. The tools got steadily better. But the missing piece was always the same: retrieval cost too much, so the system never paid off in practice the way it was supposed to in theory.

That constraint is gone now.

Your knowledge base can talk to the thing doing your work. The stored thinking can inform the current execution. Notes from three months ago can show up in today’s deliverable without you doing anything to put them there.

The promise was always that your knowledge would compound. It just needed something that could actually use it.