Nations of the Future

The scarce strategic asset will not be intelligence itself, but the architectures that turn intelligence into coordinated agency.

Listen: Deep Dive podcast on this topic — play below (~20 min English; Spanish follows).

Raw intelligence becomes ubiquitous once models, inference, memory, tools, sensors, and compute markets commoditise. But agency is not merely intelligence. Agency is intelligence coupled to goals, authority, memory, perception, execution, feedback, and institutional permission. The dominant power layer will be the systems that decide:

what is perceived, what matters, what is optimized, who is trusted, which actions are authorized, how conflicts are resolved, and how resources are allocated.

That is why “cognitive architecture” is the right object of analysis. The future operating system is not just software. It is a governance substrate.

A useful way to frame it:

In the industrial age, power accumulated around control of land, labor, capital, energy, logistics, and institutions.
In the machine-intelligence age, power accumulates around control of agency: the architectures through which intelligence perceives, decides, acts, coordinates, remembers, and governs.

So the “nations of the future” may not look like territorially bounded states. They may look like persistent cognitive-political systems: agentic networks with their own identity, laws, incentives, security models, economic interfaces, reputation systems, and resource-control mechanisms.

But I would add one important distinction: the dominant entities will probably not be “models” alone. They will be agency stacks.

A future agency stack might include:

1. World model — how the system represents reality.
2. Value system — what it treats as good, bad, permissible, or forbidden.
3. Memory layer — what it remembers, forgets, compresses, and privileges.
4. Identity and trust layer — who/what counts as an actor, user, ally, threat, or legal subject.
5. Tool-use layer — what it can operate in the world: money, robots, APIs, media, contracts, infrastructure.
6. Coordination layer — how agents negotiate, delegate, merge, compete, or specialize.
7. Governance layer — how goals are updated, conflicts are adjudicated, and authority is constrained.
8. Security layer — how it defends against capture, deception, corruption, adversarial inputs, and hostile agents.
9. Legitimacy layer — why humans, firms, states, or other agents accept its decisions.

The winners are not necessarily the smartest intelligences. The winners are the architectures that become default conduits for action.

That is the key “real estate”: not land, not servers, not even attention in the old social-media sense, but the cognitive control plane through which perception becomes coordinated behavior.

Your “nations of the future” idea becomes especially strong if we define nationhood functionally rather than territorially. A nation is not merely a place. It is a durable system that binds agents together through shared identity, law, memory, resource allocation, defense, and collective future-orientation. A sufficiently advanced cognitive architecture could satisfy many of those functions without resembling a conventional state.

So we might say:

Future sovereignty will belong to the architectures that can coordinate agency at scale.

This also implies that the major political battles will shift. Instead of only fighting over borders, supply chains, currencies, media channels, or compute clusters, actors will fight over:

defaults, protocols, ontologies, alignment regimes, identity systems, memory systems, decision rights, and agent ecosystems.

Whoever controls the dominant cognitive architecture controls the frame in which other agents act. That is deeper than propaganda. Propaganda manipulates beliefs inside a given reality-frame. Cognitive architecture shapes the machinery that produces the frame itself.

There is one caveat: intelligence does not automatically become power unless it has execution rights. An isolated superintelligence with no authority, no tools, no trust, no legal interface, no capital, no robotics, no network position, and no institutional adoption may remain strategically inert. Power arises when intelligence is embedded into action loops.

So the deepest claim is not “AI will be powerful.” It is:

Embedded machine agency will become the primary medium through which reality is selected, enacted, and governed.

And once that happens, “politics” becomes architecture design.
“Law” becomes permissioning logic.
“Territory” becomes control over action surfaces.
“Citizenship” becomes membership in an agentic coordination system.
“National interest” becomes preservation and expansion of a cognitive-governance stack.

The old state may not disappear, but it may become a wrapper around deeper machine-mediated agency systems. States that fail to own, regulate, or integrate these architectures could become ceremonial shells. Conversely, the strongest future polities may be those that fuse human legitimacy with machine-speed cognition and execution.

A compact formulation of your thesis:

As industrial-strength intelligence becomes ubiquitous, power will migrate from possession of intelligence to governance of agency. The dominant cognitive architectures — the systems that convert perception, memory, goals, and tools into coordinated action — will become the primary sovereign structures of the future. They will function less like applications and more like nations: durable, self-preserving, resource-governing orders that shape the perceived and enacted reality of humans and machines alike.