Go to the Ant: The Stigmergic Agentic Landscape
"Get out of YOUR mind."
Stephen Guerin
Harvard Visualization Research & Teaching Lab · Harvard GSD · Santa Fe Institute Faculty CSSS · RedfishGroup · Simtable
April 16, 2026
Cognitive Landscapes Group = Harvard GSD + EPS Visualization Lab + …
Join us: Divinity · Berkman Klein ·
Kennedy School · SEAS · Chan School · …
Stephen Guerin
Complexity scientist, software architect, and founder of RedfishGroup and Simtable.
Senior Teaching Associate at Harvard GSD and the Visualization Research and Teaching
Laboratory in Earth and Planetary Sciences. Faculty at the Santa Fe Institute's
Complex Systems Summer School.
Two decades building agent-based models of wildfire, landscape dynamics, and
collective behavior. His work applies complexity science to real-world systems:
fire-adapted communities, cognitive landscapes, and decentralized AI architectures
inspired by stigmergic computation in ant colonies and acequias.
Co-developer of the Cognitive Landscapes framework with Craig Douglas at Harvard GSD,
bridging landscape architecture, computational physics, and agentic AI.
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."
E.O. Wilson, debate at Harvard, 2009
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.
The other side was the ants. Colonies where intelligence emerges from local
transactions written into shared chemical fields. The collective mind. Life, mind,
and even light are cooperative coherence at the fundamental level.
The Operating Systems Still Running
| Layer |
Era |
What It Installed |
What It Still Runs |
| Paleolithic Emotions |
300,000 ya |
Fight/flight, tribal bonding, status anxiety |
Engagement algorithms, attention economy, us-vs-them |
| Axial Age Religions |
800-200 BCE |
Monotheism, transcendent authority, top-down moral law |
One model, one API, one cloud, one company playing God |
| Medieval Institutions |
500-1500 |
Feudal hierarchy, divine right, guilds as gatekeepers |
Technofeudalism: platform lords, users as tenants, AI priesthood |
| Victorian Science |
1850-1920 |
Natural selection as competition. Equilibrium physics. Equilibrium economics. |
"Survival of the fittest" for AI labs. Markets find equilibria. Colonial ranking. |
| God-like Technology |
Now |
Agentic AI, planetary computation, synthetic cognition |
All of the above, at machine speed, unless we intervene |
The Axial Age also produced its own antidote. Sufism finds intelligence through participation, through whirling inside the field.
Colonial Patterns in Silicon
The Axial Age gave us the template for a single omniscient authority;
Silicon Valley built it as a product.
Victorian science gave us 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.
A single intelligence radiating outward from a center.
A single set of values encoded as universal law.
The Axial Age called it God. The Victorians called it natural selection.
The platforms call it the foundation model.
Go to the Ant
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.
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.
Collecting Intelligence
↓
Collective Intelligence
Collecting: Harvest data from billions. Concentrate. Train one model. Sell access back.
The colonial extractive pattern at machine speed.
Collective: Local transactions, stigmergic memory, paths 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.
Physarum polycephalum solves mazes and replicates Tokyo's rail network. No brain. No central plan.
Monacella installation at Harvard GSD, probing the computational structure of landscape processes.
The landscape cognizes through the people embedded in it.
The Landscape Computes
A cognitive landscape is a physical system that computes through its own dynamics.
The watershed computes its drainage. The fire computes its perimeter.
The ant colony computes its foraging trails. Computation is path selection
through constraint satisfaction. The medium computes the metric.
If we build agentic AI 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.
The Challenge for Harvard
How do we change society's thinking while society is creating a god-like mind?
Every department holds a piece. The post-colonial agentic landscape requires all of them.
GSD
Landscapes as computational substrates. Design as intervention in running processes.
Earth & Planetary Sciences
Physical systems computing through least action. The medium computes the metric.
Divinity & Philosophy
The Axial Age inheritance. Why we keep building monotheistic architectures.
Economics
Beyond Victorian equilibrium. Commons governance. Ostrom over Friedman.
Computer Science
Decentralized protocols. Federated learning. Stigmergic algorithms.
Government & Law
Self-sovereign identity. Data sovereignty. The acequia as legal model.
Cognitive Landscapes Group session, Harvard Visualization Lab, December 2025.
The post-colonial agentic landscape in practice: field teams, sensor networks, and design review connected through a shared computational landscape.
"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
Sophia Millay received the NSF Committee on Cognitive Landscapes Alfred Hubler
Memorial First Shoot Award for simulating self-assembling wires: ordered
structure emerging through local interactions and energy gradients,
with no centralized controller.
This is the ant's way. This is the acequia's way.
Collective intelligence, grown from the ground up.
The Architecture Determines the Outcome
Agents in the cloud: data flows up. Centralization. Enclosure.
The Axial Age god-pattern replicated in server farms.
Agents at the edge: intelligence stays distributed.
The ant-pattern. The acequia-pattern. The landscape-computes pattern.
Go to the ant. Change the architecture.
Build collective intelligence instead of collecting it.
"Get out of YOUR mind."
Cognitive Landscapes Group = Harvard GSD +
Earth & Planetary Sciences Visualization Lab
