Intelligence at every node
Centralization and its corollary, bureaucracy, were a way to deal with scarce intelligence: get the smartest people in a room, come up with rules, and employ people to apply those rules at scale.
So bureaucracy is a scale technology. Rules are compressed intelligence that allow a few experts to make decisions for the many. The cost is that you have to compromise at the mean. It’s a natural descendant of the industrial revolution in spirit. It’s also a child of the military order and modern nation states. Napoleon built the modern French administration like he built his Grande Armée.
Every aspect of society saw its centralized version win out over the others, in both public and private sectors. An insurer can’t have an army of actuaries study every individual, so it prices risk from aggregated data. A city can’t deeply evaluate every building proposal, so it creates building codes and permitting procedures.
AI lifts the scarcity constraint
What happens when intelligence is no longer scarce? When we no longer need to centralize experts in one room to come up with uniform rules?
A big chunk of current AI applications makes bureaucracy cheaper, faster, leaner. Insurance companies are using AI to process claims. Law firms are automating paperwork… you get the idea. DOGE even applied the same logic to government bureaucracy.
Yet I don’t think this is the best use of abundant intelligence and this is the worst application of AI we’ll ever have. So here’s me trying to understand what we could do instead when everyone has a super-intelligence in their pocket!
1/ From rules to judgment
When intelligence becomes cheap enough, you can replace the rule with judgment.
Let’s look at construction permitting. Today, a city has to encode its understanding of safety, zoning, environmental impact, traffic, fire risk and neighbourhood concerns into thousands of rules. A developer submits a project; bureaucrats check whether it complies. That encoding has a cost of its own, and it kicks in before you ever get to compromise at the mean. To write a rule, you first have to make the case legible: turn a building into a permit application, a person into a set of checkboxes.
Now imagine that every participant has an intelligent agent that understands the underlying objective. The architect’s agent can reason about the building’s design, the city’s agent can evaluate its safety and impact, and neighbours’ agents can review the proposal and flag genuine concerns. Instead of asking, “Does this project satisfy every rule?” the system can ask the question we actually care about: is this a good and safe building?
The same applies to insurance, healthcare, education, law. Making the machine faster is only a tiny portion of the potential upside. We don’t need to compromise at the mean or encode every possible situation into a rule. Unlimited judgment and agency can finally make intelligence live at the edge.
2/ From centralized learning to networked learning
A centre learns once, a network learns everywhere. In a bureaucracy, something learned at the edge has to travel inward, get processed and eventually come back out as a revised rule.
The process is slow, and I’m not sure much “learning” makes the trip. Central planning fails against markets because no central planner can ever possess all the information distributed throughout an economy. AI makes it possible to put intelligence at the edge.
Imagine a network of thousands of intelligent nodes. When one node encounters a new fraud pattern, discovers a better underwriting strategy, or learns that a particular construction technique creates a problem, that learning can immediately become available to every other node. The network doesn’t have to wait for headquarters to update the playbook.
A network of intelligent nodes can be evolutionary. A thousand independent agents can try a thousand different approaches to the same problem. The failures stay local but the successful strategies spread, without a central designer (or bureaucrat or tyrant). The immune system does something similar: distributed detection, local response, continuous adaptation.
Let’s look at education for a second. Our systems are built to homogenise the experience for every student. I’m from France, and in my mind it comes from the Égalité component of our motto Liberté, Égalité, Fraternité. Regardless of where one grows up, they’re a child of the nation and deserve the same curriculum. So it’s designed in the capital and applied everywhere, and every student is tested the same way at the end. Now imagine “evolutionary schools,” where the curriculum is set at the level of a classroom or an individual student, and teachers read the room and adapt it in real time. And teachers could build on the momentum gained: they’d be empowered with real decisions at the edge. Which brings me to accountability.
3/ From procedure to accountability
I get why we built it this way. Centralization brought industrial-grade education to places that could never have found enough great teachers otherwise. But in the age of abundant intelligence, we can do better than a ministry deciding what every classroom does on a given Tuesday.
Centralization also scaled standards and evaluations systems. It was a risk-optimized way to deliver education, since the optimization was at the curriculum (or protocol) level. The catch is that liberating teachers means giving more accountability. It’s a more risky approach and our notion of equality should be distanced from that of uniformity.
Artificial intelligence will allow us to have unique experiences and we should celebrate that. In some strange way, AI is what brings humanity back to the edges of the network.
The first wave of AI will make existing institutions faster, cheaper and leaner. I think the second wave can make many of them unnecessary altogether.
The opportunity is to build networks where intelligence lives at every node: institutions that learn continuously, adapt to individuals, experiment without requiring every decision to pass through a centre.
I’m excited to see them spur and allow more growth, joy and humanity in society.
Vic