Emergent
Superintelligence.

I. A precedent

Intelligence has already scaled beyond the individual.

No single person designed civilization. Our most consequential achievements depend on the ability to learn from others, coordinate effort, preserve discoveries, and improve what previous generations left behind. Human intelligence is not just a property of individual brains; it is also expressed through relationships, institutions, and accumulated culture.

Michael Tomasello's work on shared intentionality helps explain how people coordinate attention and purposes. Research on cumulative cultural evolution, including Joseph Henrich's, examines how knowledge survives and grows across people and generations. These are accounts of human social cognition, not blueprints for machine intelligence—but they suggest that interaction can be a powerful dimension of scale.

II. The artificial frontier

Connecting intelligences is not the same as making them think together.

We have become extraordinarily good at increasing the capabilities of individual models. We are also learning to distribute their work among tools, agents, and specialized systems. But interoperability is only a beginning. Two agents may exchange perfectly formed messages while assigning different meanings to the same terms, acting on conflicting assumptions, or pursuing incompatible goals.

The deeper question is what has to survive the boundary between intelligent systems: not just data, but meaning, intent, authority, evidence, uncertainty, and memory. The difference between agents that are connected and agents that can reason collectively is an engineering frontier of its own.

III. From connected to collective

Coordination changes the problems a system can solve.

Collective cognition requires more than a shared communication channel. Participants need ways to interpret one another's claims, establish which context is authoritative, test disagreements, integrate partial results, and preserve what the group has learned. Coordination can produce better answers; it can also amplify common errors or create confident consensus around a mistake.

This is why the transition from connected to collective deserves its own scientific and systems research. Can we measure not only individual agent performance but also the quality of shared understanding? Can heterogeneous participants disagree productively? Can a group remain coherent as its membership, goals, and evidence change?

IV. The emergent possibility

What would a society of humans and agents become?

Imagine more than a collection of agents executing a workflow. Imagine persistent populations of humans and artificial participants that build shared knowledge, develop conventions, revise beliefs, distribute judgment, and organize around discoveries that none could achieve alone.

Such systems raise new questions about institutions and legitimacy as much as algorithms: who decides, who remembers, who is accountable, and how do participants distinguish knowledge from merely repeated claims? Human–agent societies could extend collective problem-solving, but society does not automatically imply superior intelligence. Nor does emergence imply superintelligence.

To make that stronger claim, we would need evidence of capabilities that exceed those of individual participants and existing engineered ensembles, alongside ways to evaluate reliability, transfer, and safety. This is a direction of inquiry—not a result we can yet claim.

Individuals solve problems. Collectives change the class of problems we can solve.