Most nights, before I close the laptop, I talk to my business.
Not metaphorically. I open Claude, tell it I want to run my nightly triage, and we work through the day. What slipped, what shipped, who is waiting on me, what I owe other people, and what I am quietly worried about. I talk to it the way I would talk to a sharp chief of staff who happens to remember everything. Tonight it reminded me about a follow-up I had promised someone and let slip for a week. I had forgotten it completely. The system had not.
Then it did the thing that still catches me off guard. It did not just hand me a to-do list. It pulled up the standing worries I had logged as open risks, ranked them by how much damage they could do, and pointed out that one of them was about to collide with something already on my calendar that week. I had never connected those two facts. The system connected them for me.
I built that system for myself over a handful of evenings. I call it GC Brain.
What it actually is
It helps to start with what it is not. It is not a prettier to-do app. A normal task manager stores what you have to do. This stores everything an AI needs to actually help you do it.
The app itself is nearly invisible. There is a database, a small amount of plumbing, and a bare-bones admin screen I almost never open. The real interface is Claude, on my laptop, on my phone, wherever I happen to be working. The trick is that Claude already reaches into my working life through its connectors: email, Slack, calendar, call transcripts, documents. GC Brain does not duplicate any of that. The connectors do the fetching. The brain does the remembering.
What it remembers is organized into a handful of first-class things, and all of them connect to each other. Tasks are the obvious one. But it also tracks the people I work with and the companies behind them, sorted into clients, prospects, and partners. It tracks decisions, including the options I weighed and the reason I landed where I did, so a choice I make today becomes a record I can search a year from now. It tracks commitments, which are the mirror image of the things I am waiting on: what I promised, to whom, and when. It tracks concerns, which are standing worries that each carry a severity, like a personal risk register. It keeps a journal and a research library. None of it sits in isolation. Ask what is connected to a given deal and it can actually answer, because the person, the promises, the open risks, and the source documents are all linked to the same thread.
The piece that changed how I work is context. In most tools, context is a description box you fill in once and never touch again. Here it is a set of versioned snapshots that Claude compiles from my real emails and call transcripts, each one stamped with where it came from and how old it is. So the system can tell me, on its own, that what it knows about a deal is nine days stale, and offer to refresh it from the original sources. I stopped trying to hold my business in my head, because holding it is no longer my job.
One more piece makes the brain feel like it works while I sleep. A routine runs on its own each morning, before I have opened anything. It reaches through the same connectors, reads the previous day across email, chat, and call transcripts, and updates the brain on its own: new tasks, changed commitments, people I now owe, and risks worth watching. By the time I run my triage, the system has already done a first pass and caught me up. What I sit down to do at night is a second look at a brain that is already current.
Why this has been a big deal for me
The real measure of a personal tool is whether you keep using it. I have built and abandoned a lot of systems over the years. This one stuck, because the friction is close to zero and the payoff compounds.
I can capture a task with a single sentence and let the detail accrete later. I can ask what to work on with twenty minutes and low energy and get a genuinely useful answer instead of a guilt-inducing wall of everything. It surfaces the person I have left waiting too long before that turns into a problem. It remembers a promise I made out loud on a call months ago and quietly carries it until it is done. The result is that very little falls through the cracks, and I get to spend my attention on judgment instead of on bookkeeping. For one person, it is the most useful thing I have built in twenty years.
The idea is not mine, and it is not new
None of this is a private invention. The whole industry arrived at the same place at once, and even settled on a name for it this year. People are calling it the "company brain," and the framing is everywhere. Y Combinator went as far as naming it a missing primitive in its recent call for startups. Meta has described rolling an internal AI second brain out to sixty thousand of its people.
The reason it surfaced now is simple. AI agents mostly do not fail because the model is not smart enough. They fail because they have no memory of your world. The model shows up brilliant and amnesiac every single morning. A brain fixes the amnesia, and suddenly the same model is useful in a way it never was before.
It is not even a new ambition. More than a decade ago I worked at a company called Brainspace, in the world of legal discovery, where the entire point was pulling meaning and connections out of enormous piles of documents. We were reaching for a version of this back then with the tools of the time. What is different in 2026 is not the dream. It is that the harness to build it finally exists, and an individual operator can stand one up in a few evenings.
From one to a few is easy. From a few to a company is not.
At Gadoci we run lighter versions of this for our team of six or seven, and getting from one person to a small, high-trust team is not much of a leap. When everyone basically trusts everyone with nearly everything, a shared brain mostly just works.
The enterprise is a different animal, and I do not want to oversell where any of us are with it.
The power of my personal version comes precisely from the fact that it was built for one. It is locked to a single account. Every piece of context lives in one vault that one person is allowed to see. That single assumption is carrying enormous weight, and it is the first thing to break when you walk into a real company. What breaks is governance. Who is allowed to see what. Where the data is permitted to live. What is even allowed to flow into the brain in the first place. In a regulated or publicly traded company, those questions are not footnotes to the project. They are the project.
The wider conversation backs this up. By one count, around 82 percent of organizations are already using AI agents while only 44 percent have security policies for them. And the pattern people keep rediscovering is that the personal second brain which works so well as an individual weapon falls over as a corporate platform, because it has no role-based access control. Access control for an AI is also harder than the human version. A person's permissions can be defined in advance. An agent tends to discover what it needs as it works, at runtime, which is exactly the moment a static set of rules helps least.
What I see companies reaching for is opt-in. Let people decide what they contribute upward, with personal brains, team brains, and a company brain layered on top of one another and membranes in between. That feels closer to right than one undifferentiated pool of everything. Nobody has fully worked out how those layers share what should be shared without leaking what should not. That part is genuinely unsolved. We have a very good answer for one person and a workable one for a small team. The enterprise answer is still being figured out, by us and by everyone else.
Why I could build it at all
One last point, because it is the quiet enabler. I could stand this up in a few evenings largely because Claude brings skills and connectors together more reliably than anything else I have used. The open Model Context Protocol and the growing library of connectors are what turn a brain from a research project into something composable. My system does not integrate with email or Slack directly at all. Claude does the reaching through its connectors, and the brain just remembers what it brings back. The routine that updates everything each morning is nothing more than that same arrangement running on a schedule instead of waiting for me to ask. That division of labor is the whole trick, and it gets better every month.
So here is where I have landed. For one person, a company brain is real, buildable today, and it genuinely changes how you work. For a small team, sensible derivatives of it already work. For the enterprise, the appetite is real and the technology is largely here. What nobody has solved yet is who gets to remember what. That is the question worth working on now.