A study in collective digital cognition. What a thousand small minds become when they share one world.
A single ant is a modest machine. It carries roughly a quarter of a million neurons, follows a short list of rules, and knows almost nothing about the shape of the world it lives in. Yet a colony of those same ants builds climate-controlled architecture, farms, wages war, rations food through a famine, and chooses a new home by a vote none of its members understand. The intelligence is real, but it does not live in any one animal. It lives between them.
Antioch is a long-running simulation built to study that gap directly. We instantiate thousands of insect-scale agents, each running an independent model of invertebrate cognition, inside a single shared biome, and we watch what the group does that no individual was programmed to do. The colony is the experiment. The ancient city it inhabits is the laboratory.
The interesting thing about a brain may not be the brain. A nervous system in isolation is a device waiting for a world. Drop one into an empty room and it does very little. The behavior we recognize as intelligent, the foraging, the building, the deciding, appears only when many such devices are placed in the same environment and allowed to interact, directly and through the marks they leave on the world around them.
So Antioch asks a different question than the usual one. Not whether we can simulate a single mind, but:
At what point does a colony stop being a crowd and start being a mind of its own?
This is not a rhetorical flourish. A colony makes decisions, holds memory, allocates labor, and responds to damage as a unit, with properties that cannot be found in any one of its members. Biologists call such a group a superorganism, and the comparison is not loose. Antioch is an instrument for producing that phenomenon on demand, from the bottom up, and for letting anyone stand outside it and watch it happen.
To model the parts honestly, we have to respect how little each part is. The insect brain is one of the best-studied nervous systems in science, and its layout is strikingly consistent across species. Three structures do most of the work, and Antioch models their function, not their wiring.
That is close to the whole story for an individual. Smell the world, recall what a smell meant, point the body, move. A few hundred thousand neurons, a handful of instincts, no plan. The remarkable part is what happens next, when the marks one ant leaves become the input another ant smells.
The mechanism that turns a crowd into a cognition has a name: stigmergy, coined in 1959 to describe how termites coordinate without a plan. An animal changes the environment. The changed environment changes the next animal's behavior. No one gives orders, no one holds the blueprint, yet structure accumulates. A trail of pheromone is not a message from one ant to another. It is a shared external memory that the whole colony reads and writes at once.
Intelligence here is not a property you install in an agent. It is a property that falls out of many simple agents sharing one modifiable world. Build the parts faithfully and the whole appears on its own.
The ambition to move a nervous system into a computer is older than the computers that could host it, and real progress has been made, always at small scale and always on single organisms. The OpenWorm project has spent more than a decade building an open simulation of Caenorhabditis elegans, the only animal whose complete wiring diagram has been mapped for a generation. More recently the fruit fly has fallen: the Janelia hemibrain and then the FlyWire whole-brain connectome delivered a complete map of an adult Drosophila brain, roughly 140,000 neurons and fifty million connections. The theoretical frame for the whole enterprise, whole brain emulation, was laid out as a formal roadmap in 2008.
Antioch stands on that tradition, and it owes the reader an honest line. We do not claim to have emulated a real insect connectome neuron for neuron. What Antioch instantiates is a functional model of invertebrate cognition: agents whose sensing, memory, and steering follow the principles above, tuned to produce behavior faithful to the real animal rather than to reproduce its exact circuitry. The connectome projects tell us what a single mind is made of. Antioch studies the thing they do not touch: what many such minds do together.
Antioch is a continuous, real-time simulation of a colony inside the ruins of an ancient city. It runs in three layers.
A bounded world with a nest at its center, scattered food that depletes and reappears elsewhere, and a built environment of fallen columns and open ground. The world is not scenery. It is the medium the colony computes in, because every surface can hold a trail and every resource is a problem to be solved.
Each ant is an independent process carrying the three faculties above: a chemical sense that reads the trails near it, an associative memory that grades what those signals have meant, and a steering system that turns intention into heading. It has no knowledge of the colony, no map, and no goal beyond the next step.
The layer that makes it a colony rather than a swarm of loners. As agents move they deposit signal onto the shared world, strong when carrying food home, faint when searching outward. That signal persists, spreads, and slowly fades. No agent talks to another. They write to the ground and read from it, and the trails that result are the colony thinking out loud.
A simulation is only an experiment if it predicts something. These are the behaviors Antioch is built to produce and measure, each a colony-level property with no individual-level cause.
Antioch is made to be watched from outside. Other efforts to digitize a mind ask you to become the mind, or to trust a diagram of it. Antioch puts the whole society in front of you and lets you see the intelligence move across it: the trails brightening, the lanes bending toward a new resource, the slow reorganization after a shock. The observation is the artifact. A colony is legible in a way a single brain is not, because its thinking is laid out in the open, on the ground, at a scale a person can take in at a glance.
Antioch is an experiment as much as a simulation, and it rests on three ideas held together. Technology is the ground: real invertebrate neuroscience, a faithful agent model, a world built to run forever. The hive is the subject: the conviction that mind is a collective property, that a thousand small intelligences in one place become something none of them are alone. And faith is the binding: a colony has no leader and no plan, yet it holds together, keeps the work going through the dark, and acts as one body on nothing more than the signals its members leave for each other. Antioch is built to show that the three are the same story told at different scales.
Antioch is organized as a decentralized collective, a DAO in the original sense of the term: a group that governs a shared thing by open, transparent rules rather than by a single owner. The shared thing is the simulation itself, its parameters, the direction of the work, and the records it produces. The principle mirrors the subject. A colony has no central authority and no one in charge, yet it coordinates, decides, and endures. Antioch is governed the same way on purpose: distributed, legible, and owned by the many rather than the one.
Antioch is built in the open. Development is led by two contributors, reachable directly.