The Cartographer of Connections
The man who mapped the world's professional relationships is now betting that those relationships are about to become the least interesting thing about a person. LinkedIn, founded by Hoffman in 2002, was the first piece of social infrastructure to take seriously the idea that a person's economic life is a graph — nodes connected by edges of trust, reputation, and capability. The platform now holds over a billion identities. Every recruiter, every founder, every salesperson navigates it the way ancient mariners navigated star charts. And yet Hoffman, in 2025, argues that the graph is not where value will live in the AI era. Value will live in the agents each node commands.
Hoffman's intellectual training pointed him here long before the technology arrived. At Stanford he studied Symbolic Systems — a curriculum invented to fuse philosophy of mind with cognitive science and computation. At Oxford he wrote on Wittgenstein and language games. His earliest companies were not about software but about how humans coordinate meaning at scale. SocialNet, his pre-LinkedIn attempt at online dating, failed because the network was too small to be useful. LinkedIn worked because Hoffman understood Metcalfe's law in his bones: a network's value is the square of its participants, which means the first thousand users are nearly worthless and the millionth user is a giant.
The lesson of that arithmetic is brutal and unsentimental. In network markets, you do not get partial credit. You win the geography or you do not exist. Hoffman articulated this in Blitzscaling as a doctrine of speed over efficiency — burn capital, hire chaotically, ship before you are ready, because the only sin in a winner-take-most market is being late to the position. It is the doctrine that built LinkedIn, Airbnb, OpenAI, and a dozen other companies he funded or counseled.
But networks built on human nodes have a ceiling. Each person can only sustain about a hundred and fifty meaningful connections — the so-called Dunbar number — and even that requires constant maintenance. LinkedIn's genius was to externalize that maintenance into a database. Hoffman now sees a further externalization coming. If each person can field an AI agent capable of doing the cognitive work of a small team, the value of the human-to-human network compresses. Why know a tax expert when you have a tax agent? Why know a designer when you have a design agent? The graph does not disappear, but its center of gravity moves.
This is not a cynical thought. It is a Hoffmanian one. He has spent his life believing that the deepest leverage in the world is the ability to amplify what one person can do, and he has spent it convinced that networks are the right metaphor for that amplification. The question is whether the network of the next decade is a network of humans, a network of agents, or a hybrid in which the line between them stops mattering. Hoffman's bet is on the hybrid, with the slider moving rapidly toward agents. He has not stopped believing in networks. He has stopped believing the nodes have to be people.
It is worth pausing on what that means for the social fabric LinkedIn was supposed to weave. If Hoffman is right, the professional connections that defined the twenty-first century — the warm introductions, the curated graphs, the careful self-marketing — start to look like a transitional technology. They got us from the rolodex to the database to the agent. And now the agent will do the connecting itself. The cartographer turns out to have been mapping a continent that was about to liquefy.
