Many Brains, One Web

I've been keeping a second brain for a long time. Before Roam made it trendy. Before Obsidian existed. A pile of notes, heavily linked, living only on my machine, and, for most of that time, I thought that was fine. Notes are private things.

I've changed my mind about that.

Here's what shifted: the notes themselves were never the point. What has value is the curation. The way I decided what was worth writing down and what wasn't. The connections I drew between things that look unrelated until they aren't. That's mine in a way a Wikipedia article isn't. It reflects how I actually think. And if that's true for me, it's true for everyone who does this.

See Karpathy Second Brain for the full framing, which I keep coming back to.

So: a second brain that no one else can read is just a notebook. Useful, but private in a way that wastes the curation. A second brain that everyone can read indiscriminately is just a wiki. The signal-to-noise problem degrades it fast. The interesting space is the middle. Curated views of curated brains, shared with intention, composable across people.

That's the bet BrainShare is making.

Why this matters right now

AI agents can already summarize anything on the public web. That's a solved problem, more or less. What they're bad at, still, genuinely bad at, is accessing curated, authored, opinionated knowledge with provenance attached. A published subgraph of someone's second brain is not a webpage. It's a reasoning artifact. You know who wrote it, when they updated it, what they linked it to. The graph structure itself is information.

If we build a protocol where second brains can be shared in slices and called via a stable URL, both humans and agents get something the public web can't offer: structured knowledge with an author, with relationships, with a curation fingerprint. That's worth building. See How BrainShare Works (Behind the Scenes) for how the mechanics actually work.

The mechanic: cross-vault wikilinks

Here's the part I find genuinely interesting. Every note in a BrainShare vault gets a ULID, a stable, non-renameable identifier. You can rename the note, move it, restructure your vault. The ID stays.

If I publish a slice of my vault, and you publish a slice of yours, and I write <span class="wikilink-private" title="not in this published slice">Your Note</span> in one of my notes, that link resolves. Your note is served from your vault. Mine from mine. The link survives because it's pointing at an identity, not a file path.

That's the foundation for a federated knowledge web. Not a platform. Not a feed. Just: stable IDs plus opt-in subscription. You subscribe to someone's published subgraph. You follow their wikilinks. Some of those links point back into my subgraph, some into someone else's. The graph expands as the people composing it expand.

I've seen this fail before

Personal blogs, roughly 2005 to 2015, had a version of this dream. Everyone publishes their thinking, the web links between them, a commons of ideas emerges. It mostly worked, and then it didn't. The noise floor climbed. Aggregators optimized for engagement over signal. The dream got quiet.

I think about that failure a lot. BrainShare's answer to it is two things: subgraph-level publishing is finer-grained than post-level publishing (you share a curated slice, not a stream), and federation here is opt-in subscription, not a global feed. You choose whose brain you follow. There's no algorithm promoting content. The graph only grows where people actively link to each other.

I think that pushes toward signal. Tbh, I'm not certain. We'll find out.

An honest invite

If you keep a second brain, in Obsidian, in Roam, in a pile of Markdown files, I hope you'll consider publishing a slice of it. Not everything. Just the part you'd share with someone you trust.

Build it, publish it, send me the URL. I'll read it. Maybe we link to each other's notes. Maybe that's how this starts.

Install BrainShare in 60 Seconds if you want to get started today.

This might take years to matter. The blogosphere took a decade to build before it collapsed. A federated knowledge web built on second brains might take longer. I'm okay with that. The alternative is everyone's curation staying locked in a vault on their laptop, and I think that's a waste.

Back to the beginning: A letter for my fellow viewers