The thesis
Every company is an information-processing machine that pretends to be something else. It pretends to be a product company, a services firm, a bank. But strip away the industry and what remains is a system for moving context from the people who have it to the people who need it. Org charts, reporting lines, weekly standups, quarterly reviews: all of it exists because no single human can hold the full picture of what a 2,000-person organization knows, wants, and is doing right now.
This has been true for so long that we stopped questioning it. When a sales team in Munich notices a shift in buyer sentiment, that signal travels through a regional manager, a VP, a chief revenue officer, and eventually reaches the CEO as a line in a quarterly report. The information is weeks old by the time it arrives. The context has been compressed beyond recognition. The people who relayed it spent hours in meetings whose only purpose was to move data from one level of the tree to the next.
In March 2026, Jack Dorsey and Roelof Botha published an essay called "From Hierarchy to Intelligence," describing Block's move toward what they called a company world model. The argument was simple: hierarchy exists because humans have limited bandwidth for context. If a machine can hold the full context of the organization, the hierarchy becomes optional. Dorsey was describing one company's experience, but the insight generalizes to every organization that runs on layered management, which is to say, nearly all of them.
What middle management actually does
To understand why this matters, you have to understand what middle management actually does. The popular narrative frames managers as either leaders or overhead, but the truth is more mechanical than that. The core function of middle management is information routing. Managers exist to relay context between levels of the organization, to maintain alignment across teams that cannot see each other's work, to pre-compute decisions so that senior leadership doesn't have to process every signal from the ground.
This is not a criticism. For most of the history of the corporation, information routing was genuinely hard. A manager who could sit in three meetings in a morning and synthesize what they heard into a coherent recommendation for their VP was performing an essential cognitive function. The problem is not that managers are bad at this. The problem is that information routing is now a solved problem for machines, and humans doing it at scale introduces latency, compression loss, and misalignment that compounds with every layer of the hierarchy.
Consider the actual workflow of a typical middle manager in a 2,000-person company. They spend roughly 60% of their time in meetings whose purpose is to transfer context. They spend another 20% writing summaries, status updates, and alignment documents. The remaining 20%, the part that requires actual judgment, relationship-building, and creative problem-solving, is the part that justifies their role. The other 80% is coordination overhead.
The shift
What changes when a machine can hold a continuously updated model of the entire company? Not a chatbot that answers questions about HR policy. Not a dashboard that visualizes KPIs. A genuine model: a live, structured, temporal representation of how the organization actually operates, updated continuously from every system the company already uses.
The first thing that changes is that information flows directly to where it's needed. If the model knows what the Munich sales team is seeing, and the model knows what the product roadmap says, and the model knows what the CFO's budget constraints are, then the signal doesn't need to travel up a chain of command. The reasoning can happen at the level of the model itself. The humans who need to act on the conclusion can be notified directly, with full context, without waiting for six layers of meetings.
The second thing that changes is that alignment becomes continuous. In a traditional hierarchy, alignment is maintained through periodic rituals: quarterly planning, monthly reviews, weekly standups. Between these rituals, teams drift. The model doesn't drift. It knows the current state of every project, every commitment, every dependency. When a decision in one part of the organization creates a conflict with a commitment in another, the model sees it immediately.
The third thing that changes is that people become dramatically more effective, and the company captures that effectiveness as higher margins and faster growth. When information flows directly and alignment is continuous, the coordination work that consumes most of a knowledge worker's day simply disappears. A copilot helps an individual do their existing job faster. A coordination layer removes the job that shouldn't exist in the first place. And the gains compound: every layer of coordination overhead you eliminate makes the organization faster, a faster organization generates better data for the model, which makes the model more accurate, which eliminates more overhead.
What we build
Worldmodel is the live intelligence layer. It ingests data from every system the company uses (Slack, email, CRM, ERP, project management, HR systems, financial tools) and maintains a continuously updated model of the organization. Not a data warehouse. Not a knowledge graph. A temporal, structured, reasoning-capable model that understands not just what happened, but what is happening and what is about to happen.
On top of the world model, we build and deploy autonomous AI employees that complete real work. Not copilots, not chatbots. An SDR that runs outbound pipelines. A product manager that turns customer signals into prioritized work. A content specialist that maintains the company's presence. Each worker operates with the full context of the world model, which means each worker acts as if they had read every email, attended every meeting, and understood every strategic priority.
The humans stay at the edge, where they've always been most valuable. The decisions that require judgment, empathy, relationship-building, and ethical reasoning all remain human. What disappears is the coordination layer. The meetings that exist to move information. The status updates that exist to maintain alignment. The management overhead that exists because humans can't hold the full context of a 2,000-person organization in their heads.
Why now
The Dorsey essay is not a curiosity. It is a signal. When the founder of two public companies publishes a detailed account of replacing hierarchy with an intelligence layer, and when Sequoia's managing partner co-signs it, the window is open. Every ComEx in Europe is reading that essay and asking what it would look like for their company. Most of them will wait. Some of them will try to build it themselves. A few will look for the company that has already built the platform.
The first companies to build around a world model will compound advantages that the others cannot catch. An organization that operates on a world model for twelve months has a model that is twelve months smarter. The workers that run on top of it are twelve months more capable. The institutional knowledge that the model holds is twelve months deeper. This is not a feature advantage. It is a structural moat.
The question
If you're reading this and you're running a company, the question we'd ask you is the one Dorsey asked his own team: what does your company understand that is genuinely hard to understand, and is that understanding getting deeper every day? If the answer is that your company's understanding is trapped in the heads of individuals, compressed through layers of management, and slowly degrading as people leave and context is lost, then you're describing the problem we solve. Worldmodel is how you answer yes.