Go to the Ant: The Stigmergic Agentic Landscape

E.O. Wilson Was of Two Minds

"The real problem of humanity is the following: We have Paleolithic emotions, medieval institutions, and god-like technology. It is terrifically dangerous, and it is now approaching a point of crisis overall." E.O. Wilson, debate at Harvard, 2009

Wilson carried both sides of this problem in his own career. One side was Victorian Darwin: sociobiology, the overreach that extended competitive reproductive success into an explanation for all behavior, a framework that struggles to even account for altruism. Natural selection as universal competition. Survival of the fittest. The logic that ranks organisms, races, civilizations, and now AI benchmarks on a single ladder.

The other side was the ants. Wilson spent decades studying colonies where no individual holds the plan, where intelligence emerges from local transactions written into shared chemical fields. The collective mind. Stigmergic memory. Paths that form through cooperation, not competition. He saw that life, mind, and even light are cooperative coherence at the fundamental level.

Wilson named the crisis but left the deeper question unspoken: the layers of inherited thinking between Paleolithic emotions and god-like technology that still run in our heads. We need to name them, because we are building a god-like mind with agentic AI right now, and the challenge for Harvard, for all of us, is how we change society's thinking while that mind is being born.

Layer	Era	What It Installed	What It Still Runs
Paleolithic Emotions	300,000 years ago	Fight/flight, tribal bonding, status anxiety, fear of the other	Social media engagement algorithms, attention economy, us-vs-them politics
Axial Age Religions	800 - 200 BCE	Monotheism, transcendent authority, top-down moral law, the idea that one God (or one Logos, or one Dharma) governs all from above	Centralized platform theology: one model, one API, one cloud, one company playing God. The assumption that intelligence must be singular and hierarchical. The missionary impulse to impose one paradigm on all peoples.
The Axial Age also produced its own antidote. Sufism, like the ant colony, finds intelligence through participation, through whirling inside the field.
Medieval Institutions	500 - 1500 CE	Feudal hierarchy, papal authority, divine right, the guild as gatekeeper	Technofeudalism: platform lords owning the digital land, users as tenants, AI labs as the new priesthood interpreting the oracle
Victorian Science	1850 - 1920	Natural selection as universal competition. Equilibrium thermodynamics. Neoclassical economics treating markets as self-correcting equilibria. Social Darwinism. The ranking of races and civilizations.	The assumption that competition is the engine of progress. "Survival of the fittest" applied to companies, nations, AI labs. The belief that markets find optimal equilibria by themselves. The colonial logic that the "most advanced" civilization should govern the rest. AI benchmarks as intelligence ranking.
God-like Technology	Now	Agentic AI, planetary-scale computation, synthetic cognition	All of the above, at machine speed, unless we intervene in the architecture

Every layer is still executing. The Axial Age gave us the template for a single omniscient authority; Silicon Valley built it as a product. Victorian science gave us the metaphor of competition-to-equilibrium; the AI industry runs on it, racing to build the one model that wins, the one company that dominates, the one alignment that governs all agents.

These are colonial patterns. A single intelligence radiating outward from a center. A single set of values encoded as universal law. A single optimization function deciding what is fit and what is obsolete. The Axial Age called it God. The Victorians called it natural selection. The platforms call it the foundation model.

Go to the Ant

The oldest wisdom literature already knew. The ant has no king, no pope, no CEO, no foundation model. The colony computes through local interactions: pheromone gradients written into the landscape, read by neighbors, amplified through feedback. No ant holds the plan. The plan is in the field.

E.O. Wilson spent his life studying this. The irony is that the civilization building "god-like technology" in his name still reaches for the god-architecture: one intelligence, one center, one model to rule them all.

Physarum polycephalum solves mazes and replicates Tokyo's rail network. No brain. No central plan.

Collecting intelligence is what platforms do. Harvest data from billions of users. Concentrate it in one place. Train one model. Sell access back to the people whose lives were mined. This is the colonial extractive pattern: take the raw materials from the periphery, refine them at the center, sell the product back at a markup.

Collective intelligence is what ant colonies do. What acequias do. What the landscape itself does when its dynamics are allowed to compute. No central repository. No single model. Intelligence emerges from local transactions, stigmergic memory written into the shared medium, paths that form through bidirectional constraint satisfaction. The colony is smarter than any ant. The acequia is wiser than any parciante. The landscape computes what no designer could specify.

The shift is architectural. You cannot get collective intelligence by reforming a system designed to collect it. You have to change the topology: from star to mesh, from cloud to edge, from platform to protocol, from cathedral to bazaar, from plantation to acequia.

The Landscape Computes

A cognitive landscape is a physical system that computes through its own dynamics. The watershed computes its drainage network. The fire computes its perimeter. The ant colony computes its foraging trails. These are not metaphors. Computation is path selection through constraint satisfaction. The medium computes the metric.

The same principle applies to agentic AI. If we build it as a cognitive landscape rather than a centralized oracle, intelligence emerges from the topology of connections, from agents transacting locally, writing memory into shared fields, forming paths through stigmergic feedback.

This is what the Harvard campus needs to confront. The GSD understands that landscapes compute. Earth and Planetary Sciences understands that physical systems select paths through least action. The Philosophy department understands that the Axial Age installed cognitive patterns we still run. The Economics department understands (or should) that equilibrium is a Victorian fantasy, that real economies are far-from-equilibrium dissipative systems.

The challenge is to synthesize this across the entire campus. Every department holds a piece. The post-colonial agentic landscape requires all of them:

The Stigmergic Agentic Landscape

We are at a fork. The god-like technology E.O. Wilson warned about is here. It is being built by people running Axial Age theology (one God, one model) on Victorian economics (competition to equilibrium, winner takes all) with Paleolithic emotional hooks (engagement, fear, tribal loyalty).

The post-colonial alternative is to go to the ant.

Build agentic AI as a cognitive landscape: decentralized, stigmergic, self-sovereign. Agents that run on your devices, on your land, serving your community. Intelligence that emerges from local transactions, not from centralized collection. Memory written into shared fields that the community governs, like water in an acequia.

This is what "Take That Lab" means for AI. Dissolve the lab. Let it escape into the landscape. Let the intelligence live where the data lives, where the people live, where the problems are. The lab is too small. The landscape computes.

"The real problem of humanity is the following: We have Paleolithic emotions, Axial Age theology, medieval institutions, Victorian assumptions about competition and equilibrium, and god-like technology. Every layer is still running. The question is whether we will build AI that replays these patterns at machine speed, or whether we will finally go to the ant and learn to compute collectively." Extended after E.O. Wilson, Cognitive Landscapes Group, 2026

First Shoots

The work is already growing. Sophia Millay received the NSF Committee on Cognitive Landscapes Alfred Hubler Memorial First Shoot Award for her simulation of self-assembling wires, demonstrating that ordered structure emerges through local interactions and energy gradients, with no centralized controller.

Systems that organize through local interactions, exploiting energy gradients to produce ordered form. This is the ant's way. This is the acequia's way. This is the way forward for agentic AI: collective intelligence, grown from the ground up.

The Architecture Determines the Outcome

If agents run in the cloud, data flows up. Centralization. Enclosure. The Axial Age god-pattern replicated in server farms.

If agents run at the edge, on your devices, on your land, intelligence stays distributed. The ant-pattern. The acequia-pattern. The landscape-computes pattern.

Harvard has every discipline needed to design this alternative. What it needs is the will to synthesize across its own silos, to take Rosalea's lab and dissolve it into the campus, to let the cognitive landscape of the university itself compute the answer that no single department can specify.

Go to the ant. Change the architecture. Build collective intelligence instead of collecting it.

Cognitive Landscapes · Fire-Adapted Cognitive Landscapes · Step Theory · The Acequia Manifesto