The second brain isn't a brain. It's a folder of text files.
That's the part nobody selling you a vault, a graph view or a linking philosophy wants to lead with. Strip away the plugins, the themes, the twelve-video setup tutorials, and what's left is markdown on a disk. Files. Links. Names. Everything else is a viewer.
I nearly missed this. This week I caught myself speccing out a personal knowledge system, my fourth attempt at one if I'm being honest, and the pull was familiar. New structure. New tool. A clean slate where the thinking would finally get organised. It felt like progress.
It wasn't. It was avoidance wearing a productivity costume.
Here's the tell: every knowledge system I've built has been more fun to build than to use. The building is design work, and design work is where I'm comfortable. The using is the hard part, the shipping part, the part where the notes have to turn into something. A new system doesn't fix that. It postpones it, beautifully.
So I asked a simpler question. What do I actually need from a "brain"?
Three things. My notes have to be mine, readable in twenty years, not trapped inside someone's subscription. My tools have to reach them, and in the AI era that matters more than ever, because a model can read a plain text file natively with no API, no sync layer, no permissions dance. And I have to be able to find things fast.
A folder of markdown files does all three. That's it. That's the system.
If you want a pretty way to browse it later, fine, point one of the popular apps at the folder. They're viewers, and good ones. But the value was never in the app. The value is that your knowledge lives as plain files you own, in a structure any tool can read, including the tools that don't exist yet.
The knowledge worker's version of buying running shoes instead of running is building the vault instead of writing the notes. I've bought a lot of shoes.
So the setup took me an afternoon. One folder, a handful of conventions, an index file at the top. Then I closed it and went back to the actual work, which was the point all along.
What I actually run
For what it's worth, here's mine. Two layers, and neither of them is a second brain.
Notion is the archive. Projects, decisions, budgets, the long tail of things worth keeping. It's a filing cabinet, not a thinking tool, and I stopped asking it to be one. Nothing gets organised for the sake of being organised. It gets filed so it can be found.
Claude is the retrieval layer. It's synced to the archive, so instead of maintaining a web of links and tags to resurface my own notes, I ask. Where did we land on that decision. What did the March review say. The linking work, the part these systems demand hours of, the model does on the fly. The archive stays dumb and durable. The intelligence sits on top, and it's replaceable.
That's the shift the second-brain crowd hasn't priced in yet. When retrieval is a conversation, the elaborate structure was solving a problem that no longer exists. Store it plain. Ask for it later.
Make the folder. Skip the system. Write the notes.