Blog · October 31, 2026

Many Agents, One Brain. Why My Second Brain Setup Actually Works.

I don't run one assistant.

I run a team.

Several agents. Working at the same time. Each in its own lane, handling its own kind of work.

And every one of them saves into the same brain.

That's the whole idea. Everything else is how I keep them from tripping over each other.

The Problem Nobody Talks About

Everyone's excited about agents.

Nobody talks about what happens when you have more than one.

One agent makes a decision. Another never hears about it. A third redoes work that was already done. You end up as the messenger between your own tools, copying context from one window to the next like an intern.

That's not leverage. That's babysitting.

One assistant with a good memory is nice. Several assistants with separate memories is chaos.

So I stopped giving them separate memories.

One Knowledge Vault. Every Agent Writes to It.

The brain is an Obsidian vault. A folder of plain markdown files.

It lives on a machine that stays on. That machine is the hub. Everything else connects to it.

Every agent saves into that one vault. What got decided. What got found. What's still open.

And I can reach it from wherever I am.

So when one agent finishes something, the others know. When I tell one of them something on the go, the rest can see it. Nobody waits on me to pass notes.

That's what makes it powerful. Not a smarter assistant. A shared brain that a team of assistants actually uses.

Plain text is what makes it possible. Any agent, any tool, any device can read a markdown file. No proprietary database. Nothing locked in. If every tool I use disappeared tomorrow, my knowledge would still be sitting there in files I own.

How the Team Stays Coordinated

A shared brain only works if it stays clean. Here's what keeps it clean.

Saves Happen Without Me Asking

The agents save what I say and what I decide. I don't stop to say "save this." It just lands where it belongs.

And they share memory. What I tell one, the others can pick up. I don't repeat myself.

That's what agent memory should be. Not a chat history that disappears when the window closes. A real record, in a knowledge vault I control.

Every Project Has One Running Record

Every project gets its own running record. One place.

Entries are tagged by what they are. A decision. A realization. A discovery. An open question.

When any agent picks up a project, it reads that record and it's caught up. So am I.

Many workers. One logbook per project. No guessing who did what or why.

The system even keeps a running record of itself. How it's wired, what changed, what got fixed. So the rules don't drift as it grows.

A Compact Map Keeps Lookups Cheap

Agents are expensive readers. Every word they read costs something. Multiply that across a team and it adds up fast.

So the vault has a compact map. One short line per note. What it is and what it covers.

Every agent reads the map first. Then it opens only the note it actually needs.

The difference is massive. A small fraction of the cost of reading everything, and the agent still finds what it's looking for.

Same map for the whole team. Cheap to read. Fast to search.

The Rule That Holds It Together: One Place Is the Truth

Here's the part most people get wrong.

When more than one place can be written to, they drift. One copy gets a change the other doesn't. A sync runs the wrong way. Work disappears quietly and you don't notice until you need it.

With several agents writing at once, that gets worse, not better.

So there is exactly one source of truth. One hub. Everything else is a copy of it.

And every agent has to agree on where that is. I've seen what happens when even one of them has an old idea about where the truth lives. It writes to the wrong place, and its work is at risk.

That's the real lesson of any multi-agent workflow:

More agents don't fail because they're dumb. They fail because they disagree about where the truth lives.

One brain. One source. Everything else is a mirror.

Why It's Powerful for Me

My attention is the scarcest thing I have.

I run multiple projects at once. I can't afford to re-explain context. I can't afford to lose a decision I made late at night. I can't afford tools that forget.

So I built something that doesn't.

  • Several agents working at once, each in its own lane.
  • One Obsidian vault on an always on hub.
  • Saves that happen without me asking.
  • Shared memory across the team.
  • One running record per project.
  • A compact map so lookups stay cheap.
  • One source of truth, no exceptions.

It's not a productivity hack. It's infrastructure. An operating system for a team that doesn't sleep.

This Isn't for Everyone

I build setups like this for a small number of businesses.

Not many. On purpose.

It only works if you're willing to run your operation on one source of truth. Most aren't.

If you are, and you want one built for your business, leave your email.