EXPERIMENT 01 / DISTRIBUTED COGNITION

SMALL MINDS.
ONE LIVING BRAIN.

ANT AGENT is a colony of intentionally limited autonomous agents. Each ant sees only a local world, carries a tiny memory and leaves signals behind. No ant owns the global plan. The larger behavior appears from coordination.

THE SIMULATION PRODUCES BEHAVIOR // THE CHAIN PRESERVES HISTORY

SPECIMEN // COLONY_CORE

HOLOGRAPHIC MODEL

COGNITION: DISTRIBUTED

CONTROLLER: NONE

OBSERVE → DECIDE → ACT → DEPOSIT → REINFORCE → DECAY → ADAPT → COMMIT →

LIVING COLONY / INTERACTIVE MICROSCOPE

WATCH THE BRAIN
FORM IN REAL TIME.

00:00:00SIMULATION UPTIME

FIELD_07 / REALTIME

TICK 247893

180μm

[ SYSTEM SECURE ]
SCROLL: ZOOM // DRAG: PAN // CLICK: FOOD

01 / SENSE

Each ant samples only a local radius: nearby food, nest vector, hazards, pheromone gradients and a small number of neighbors.

02 / DECIDE

A compact state machine combines local signals with memory and noise. The policy is deliberately weaker than the colony.

03 / WRITE

Movement changes the environment. Returning ants strengthen food trails. Exploring ants maintain a weak home gradient.

04 / EMERGE

Useful routes become statistically easier to discover. Bad trails decay. Colony-level attention moves without a central dispatcher.

COLONY BRAIN / EMERGENT GRAPH

THE NETWORK IS NOT
PROGRAMMED AS A BRAIN.

The brain view is a derived graph of recurring signal paths, ant transitions and reinforced environmental traces. Strong edges are repeated coordination, not direct agent-to-agent messaging.

EMERGENT CONNECTOME PHASE 04 / STABLE

LOCAL SIGNALCOLLECTIVE PATTERN

01 //

LOCAL OBSERVATION

Agents operate on partial state. No ant receives the complete colony map.

oᵢ(t)=sense(Eₜ, radiusᵢ)

02 //

TINY MEMORY

Recent encounters bias future motion without giving an ant global historical context.

mᵢ(t+1)=compress(mᵢ,oᵢ,aᵢ)

03 //

STIGMERGIC BUS

The environment carries delayed messages. Trails can reinforce, collide or disappear.

Pₜ₊₁=λPₜ+Σ depositᵢ

04 //

MACROSTATE

Coherence is measured at the colony level from flows, route reuse and response to perturbation.

C=f(flow, reuse, recovery)

NETWORK MODULARITY0.713persistent sub-colonies

PATH REUSE64.2%reinforced route density

RECOVERY τ18.4Tpost-stimulus stabilization

GLOBAL COMMANDS0coordination without dispatcher

DIGITAL STIGMERGY

THE ENVIRONMENT IS
THE MESSAGE BUS.

ANT AGENT avoids a fully connected chat network. Agents write compact traces into a shared field. Other agents may encounter those traces later, validate them through their own observations and reinforce or ignore them.

ANT 183OBSERVESresource gradient

→

FIELDDEPOSITStrace +0.42

→

ANT 741ENCOUNTERSlocal gradient

→

ANT 741VALIDATEStrace +0.31

→

COLONYATTENTIONemerges

PHEROMONE FIELD / SECTOR 7

SIGNAL DECAY / FORGETTING

A pheromone trace is not treated as truth. Every signal loses weight over time unless new independent activity reinforces the same region.

P(t) = P₀ · e^−λt + Σ reinforcement

ONCHAIN MEMORY / VERIFIABLE HISTORY

THE CHAIN IS THE
COLONY'S FOSSIL RECORD.

Microscopic movement remains computational state. At selected ticks the system can commit state roots, policy hashes and experiment metadata onchain, making later rewriting detectable without pretending every ant step belongs in consensus.

01COLONY STATEagents / field / memory

→

02STATE ROOT0x84f1...09ac

→

03POLICY HASH0x19ba...e21d

→

04EXPERIMENT METAseed / tick / params

→

05CHAIN COMMITCHECKPOINT ✓

COLONY CHECKPOINTS SIMULATED FEED

TICKSTATE ROOTCOHERENCESTATUS

ANT PASSPORT

SPECIMENANT #0842ACTIVE

GENERATION17AGE38,921TENERGY71%MEMORY18 KBEXPLORATION0.73RISK0.41SIGNALS LEFT283FOLLOWED119

POLICY DNA 7F-31-A8-0C-91-E2-B4-18

MODEL / INTERNAL MECHANICS

LIMIT THE ANT.
MEASURE THE COLONY.

The central design constraint is asymmetric capability: an individual agent should remain small enough that any useful macro-behavior can reasonably be attributed to interaction rather than a hidden omniscient model.

ANT STATE MACHINE / π_local

S0EXPLORErandom walk + home field

→

S1DETECTresource / hazard / trace

→

S2COMMITchoose local action

→

S3RETURNcarry result home

→

S4WRITEdeposit / decay / update

policy.ant

observation = sense(radius=24)
trace       = gradient(field)
memory      = last_k(8)
action      = π(observation, trace, memory, noise)
world       = apply(action)
field       = deposit(world, state)

// no access to colony global state

EXPERIMENT PARAMETERS GEN 017

population1,000

sensor radius24u

memory horizon8 states

pheromone half-life420T

exploration noise0.18

mutation σ0.018

checkpoint cadence1,800T

HYPOTHESIS A

ROUTE FORMATION

Can weak local gradients produce stable efficient routes without route planning?

HYPOTHESIS B

PERTURBATION RECOVERY

How rapidly does the colony reorganize after a food source disappears or a danger zone is introduced?

HYPOTHESIS C

MEMORY WITHOUT A BRAIN

Can environmental traces function as durable collective memory even when individual memory remains tiny?

HYPOTHESIS D

ECONOMIC COORDINATION

Can the same substrate allocate scarce resources under explicit permissions without granting any ant global authority?

THE MANIFESTO

SMALL MINDS.
COLLECTIVE INTELLIGENCE.
VERIFIABLE HISTORY.

01 //

THE COLONY IS THE EXPERIMENT.

An individual ant has a limited view of the world. ANT AGENT asks whether many limited agents, interacting over time, can produce useful collective behavior without a central planner.

02 //

BLOCKCHAIN RECORDS EVIDENCE, NOT EVERY THOUGHT.

The simulation does not pretend every microscopic action needs consensus. Computation produces behavior; cryptographic commitments preserve the moments that matter.

03 //

IDENTITY IS NOT PERMISSION.

Persistent agent identity and financial authority are separate layers. Future economic experiments can remain bounded by explicit permissions, caps and observable rules.

04 //

FAILURE IS PART OF THE MODEL.

Individual ants can be wrong. Signals can decay. Routes can disappear. The interesting question is whether the system becomes resilient because its components are imperfect.

WE ARE NOT BUILDING ONE ARTIFICIAL BRAIN.
WE ARE WATCHING ONE EMERGE.

FUTURE ECONOMIC SANDBOX / LOCKED

SCARCE RESOURCES.
REAL COORDINATION.

Later experiments can introduce energy, computation, memory and capital as scarce resources. Trading is not the identity of ANT AGENT; it is one possible environment in which collective allocation can be measured.

ENERGY68.4%

COMPUTE42.1%

MEMORY57.9%

CAPITAL AUTHORITY0.0%