The World Model Problem Nobody Is Talking About

Jack Dorsey and Roelof Botha published something worth reading slowly today.

Their argument: hierarchy isn't a management philosophy. It's an information routing protocol. From Roman legions to the modern corporation, 2,000 years of organizational design has been a series of workarounds for the same constraint - humans can only manage 3 to 8 people directly, so you stack layers, and layers slow information down.

From Hierarchy to Intelligence

AI breaks that constraint for the first time. You replace the hierarchy with a world model. You replace the manager with an intelligence layer that composes capabilities into solutions at the moment of customer need. No product manager required. No status meeting. The system recognizes the moment and acts.

It's a compelling architecture. And one line buried in the middle of it is the most important thing in the piece:

"The capability of the system is only as good as the quality of the customer signal feeding it."

Block has a structural advantage most companies won't have. They see both sides of millions of transactions - buyer through Cash App, seller through Square. Behavioral data. Hard to fake. Compounding in real time. They call it the most honest signal in the world, and they're right. People lie on surveys. They don't lie with their money.

But here's what the piece doesn't say out loud.

Most companies trying to build an intelligence layer won't have Block's transaction moat. They'll have dashboards nobody queries. PDFs that were never properly parsed. Years of institutional knowledge that lives in someone's head or an inbox. Data that technically exists but cannot be reached by the systems trying to reason over it.

This is the constraint that predates hierarchy and survives it. The Romans needed centurions because information couldn't travel any other way. Companies today need managers because their knowledge is unstructured, siloed, and inaccessible to the systems that could act on it. You don't fix that by removing the org chart. You fix it by fixing the data.

The intelligence layer Dorsey describes is brilliant in theory. In practice it is only as intelligent as what it can see. And most organizations are asking their intelligence layers to reason in the dark.

91% of AI deployments fail not because the model wasn't capable enough. They fail because the intelligence the model needed was trapped somewhere it couldn't reach. The seeing problem, not the thinking problem.

The companies that will actually build what Block is describing - a company organized as intelligence rather than hierarchy - are the ones that solve the data infrastructure problem first. Liberate what's locked. Structure it with provenance intact. Make it continuously available to the systems trying to act on it. Build the world model before you try to replace the org chart.

That's the layer most people are skipping. And it's the only reason the rest works.

It's what we're building at Inflectiv.