One team · 2 to 19 people
Everyone talks to everyone, and it works because everyone is in the room. Twelve people are 66 pairs. Nobody waits, because there is nobody to wait for.
Any two people who need to know what the other is
doing are a pair to keep aligned, and every person you
add makes a pair with everyone already there. The
count grows as n(n−1)/2: twelve people is 66 pairs. The way most companies
cope is hierarchy, and hierarchy is where the waiting
comes from. A world model is the other way.
Hierarchy is the tool every growing company reaches for, and it does its job: each layer cuts the pair count sharply. Each layer also adds a hop between the people who hold the context and the people doing the work. Watch the ring as the company grows.
Everyone talks to everyone, and it works because everyone is in the room. Twelve people are 66 pairs. Nobody waits, because there is nobody to wait for.
The room stops working. The org splits into teams of about eight, each with a lead. 190 pairs become 60. The price is the first hop: context now lives with the lead, and the work waits on the lead to pass it down.
A manager can hold about five groups. Past that, a layer forms above the teams. 990 pairs become 154, and there are now two hops between the person who set the direction and the person doing the work. This is where status meetings are born.
The departments outgrow a span too, and a third layer forms. 24,090 pairs become 815, which is why hierarchy wins. Three hops each way, and a question from the work has to be relayed through three people before it reaches someone who can answer it.
Past a hundred people, the hierarchy is no longer a convenience. It is the only path information has. Send a message and it travels lead to lead, retold at every hop, interrupting most of the people it reaches for nothing. Ask a question and the answer has to come back the same way, with every lead on the path relaying it.
A world model changes the shape. Every person connects to the model, so the pairs are the headcount, not its square. A message is one handoff, and the model already knows who it concerns. A question is answered from what the model holds, and whatever it has to fetch, it keeps. Press a button and both rings start at once: the model has usually answered while the hierarchy is still relaying.
Scale the team, switch the structure, send a message or ask a question. The pairs chart underneath plots all three structures against headcount on a log scale.
Metacosm is the hub at the centre of the second ring: one live model of what the business does, its capabilities, against where it is going. It holds three states, what you designed, what actually happens, and where you are heading, so the transition is visible rather than reported.
Teams read it to see what the strategy asks of them. Agents read it to route work and act in line with where you are taking the organisation. Agent access is read-only by design: people set the strategy and make the decisions. Strategy becomes action the moment it is set, with no hops in between.
Block's essay From Hierarchy to Intelligence describes a company that runs on a world model: how the company understands itself and its priorities, held where every person and every agent can read it, doing the work that middle management used to do by relaying. Its capabilities, what the business does, are the units the model is built from.
Block built theirs at a scale most companies will never reach. Metacosm makes the same model available to scale-ups, at the point in their growth where the hops start to cost them.
Metacosm is being built with a small number of Australian scale-ups making the move from AI-assisted to AI-native. Design partners shape the roadmap, get senior attention from both practitioners, and pay early-partner terms.