Zuckerberg Has 70,000 Employees. He Still Needed a Second Brain.
Last week, Futurism ran a story with this headline: “Mark Zuckerberg Secretly Training an AI Agent to Do CEO Job.”
The tech press covered it the way tech press covers everything Zuckerberg does: part fascination, part speculation, part “what does this mean for the stock.”
Nobody asked the more interesting question.
What he actually built
Two systems. One is a personal AI agent that retrieves information Zuckerberg would normally have to route through multiple organizational layers to get. Fast access to interdepartmental data. Decision support without the organizational overhead.
The second is called Second Brain. Built on Anthropic’s Claude. Described internally as an “AI chief of staff.” It organizes tasks, surfaces insights, streamlines retrieval of institutional knowledge. A Meta employee built it. Zuckerberg adopted it.
Read that again: the CEO of a company with 70,000 employees and more engineering talent than most countries needed someone to build him a personal coordination layer to do his job well.
This is not a story about Zuckerberg
The coverage framed this as a Zuckerberg story. It’s not.
It’s a story about what happens when information complexity outgrows organizational structure. When the number of decisions, signals, and open loops a person is responsible for exceeds what a human nervous system can hold — and hiring more people doesn’t fix it because those people generate more signals, more decisions, more open loops.
Zuckerberg doesn’t have a headcount problem. He has a coordination problem.
This is the same problem a startup founder with three employees has. The same problem a VP of Engineering managing four teams has. The same problem anyone has who is responsible for more threads than working memory can hold.
The scale changes. The problem doesn’t.
The failure mode that doesn’t respect your budget
There’s a persistent assumption in knowledge work: cognitive overload is a resource problem. If you had more time, more people, more tools — you’d be fine.
The HBR study published in February tracked 200 employees at a tech company over eight months. Workers using AI worked faster, took on more scope, and extended hours. Not because they were asked to. Because the AI gave them more leverage on execution while nothing changed about the coordination overhead.
More capacity → more work taken on → more open loops → more cognitive load.
Zuckerberg has resources most people can’t imagine. He still hit the same ceiling. Not because he lacked tools. Because no tool was doing the coordination work. The loops were staying open. The context wasn’t held. He was still the bottleneck in his own system.
That’s not a money problem. That’s a layer that doesn’t exist yet.
Why “AI chief of staff” is precisely the right name
The people inside Meta didn’t call it a productivity tool. They called it an AI chief of staff.
Chief of staff is a specific role. It’s not a researcher. Not a scheduler. Not an assistant. A chief of staff holds context across the entire operating surface — tracks what the executive is responsible for, surfaces what needs attention, translates intent into action, and protects the principal’s focus from organizational noise.
The reason that role exists in high-stakes organizations is that the cognitive load of operating at scale requires someone (or something) responsible for maintaining the coherent picture. Without it, the person at the center becomes the coordination layer by default — tracking, routing, deciding what to handle, what to delegate, what to defer.
That’s what Zuckerberg was doing before the bot. That’s what most knowledge workers are doing right now.
The signal buried in the noise
The headline was: Zuckerberg is building an AI to do his CEO job.
That’s not what happened. What happened is: Zuckerberg couldn’t do his CEO job well without a layer above his tools that held context, reduced friction, and gave him back decision-making bandwidth.
That layer is the product.
Not an assistant that responds to prompts. Not a summarizer. Not an agent that does one thing. A system that maintains a coherent model of your work, your priorities, and your open commitments — and acts on that model so you don’t have to.
When the wealthiest person building AI infrastructure decides the most valuable thing he can deploy for himself is a personal coordination layer, that’s not a curiosity. That’s a market signal.
The problem isn’t niche. The infrastructure isn’t built yet. And “AI chief of staff” is the right frame — not because it sounds impressive, but because it’s accurate.
Eliran Keren — Founder of Deeplica, building the coordination layer for humans who’d rather direct than operate.
