FLYSWARMCOLONY / 001
AN EXPERIMENT IN COLLECTIVE BEHAVIOREST. 2026 / PONS

FLYSWARM.

ONE BRAIN WAS AN EXPERIMENT.
A THOUSAND IS AN ORGANISM.

Local instincts. Optional coordination.
Just local signals. And everything that follows.
COLONY ACTIVE / LOCAL SIMULATION
RUN 001/
TIME SCALE
SEED 7141

Share the same starting colony. Your later interventions are not included.

FIG. 01 — THE PHEROMONE LAYERSELECT A FLY TO INSPECT
LOCAL SENSING
SHARED ENVIRONMENT1 UNIT = 1 SIMULATED AGENT
FOLLOW CAM / LOCAL SIMULATION#0041
CAMERA FOLLOWS POSITION · HEADING SHOWN BY RAYS
ONE FLY. ITS WHOLE WORLD.

INSIDE THE SWARM.

Follow the selected fly. See the three nearby signal samples that help steer its next move.

LEFT00.00
AHEAD00.00
RIGHT00.00

Exploring the field.

Choose any fly in the main field to switch the camera.

Magnified model view, not biological vision. Movement also includes wandering, nearby food and boundaries.
THE COLONY, RIGHT NOW

The colony is exploring. Give it something to find.

SEARCHING
256
FOLLOWING
0
RETURNING
0
CHANGE ONE THING.

Try an intervention. Watch the colony respond.

NEW / THE SWARM OBSERVATORYENTER THE
SWARM MIND. ↗

Every fly becomes a node. Follow the signals, inspect the connections and step inside the colony’s live activity network.

SHARED WORLD ↗ — One saved colony for every visitor.

FOUNDATION / SWARM LEADER 01

ONE MISSION.
A COLONY TO LEAD.

A coordinator that reads the field, chooses a resource target and sends signals the flies can follow. Every decision has a reason you can inspect.

SEE THE PATH ONCHAIN ↓
STANDBY / LOCAL

Rule-based browser prototype. It controls signal cues, not individual flies. Persistent agents, learning and onchain coordination are future steps.

MISSION / GUIDE FORAGING

Choose the nearest non-empty food patch. Reinforce seven points between home and that patch every five simulated seconds.

SIGNAL ACTIONS0 / 12
FOOD DELIVERED SINCE RESET0

Preview the next decision without changing the colony.

Ready.

DECISION LOG / LATEST SIX
    Starting ends a recorded replay. If paused, resume the colony to begin. Pause freezes timed decisions; reset clears this session’s log and budget. Download before resetting. Food totals include all colony activity and do not measure the leader’s benefit. Nothing is signed, spent or published onchain.
    LEADER / MEASURE BEFORE PROMOTING

    DOES LEADERSHIP HELP?

    Same seeds. Same food. Same time. Compare the baseline colony against the Leader across three matched pairs: 128 flies and 60 simulated seconds per run. This test runs separately from your live colony.

    Ready. Six runs execute locally in a background worker.

    Negative results stay visible. Three fixed seeds are a narrow test, not proof of general improvement. The same engine checks both runs; independent verification, automatic policy promotion and onchain publication are not active.
    THE LEADER → ONCHAIN COORDINATION

    THE WORK RUNS.
    THE CHAIN KEEPS THE RECORD.

    The Leader is the first coordination layer. The proposed next step is persistent workers doing the computation, with approved results recorded on Robinhood Chain. Recording a result onchain does not move the simulation itself onto the chain.

    1. 01 / WORKING LOCALLY

      OBSERVE & ACT

      The Leader reads the colony, chooses a food target and sends bounded signals. Download its decisions to inspect the actions and observations.

      TRY THE LEADER ↑
    2. 02 / PLANNED

      RUN & CHECK

      Persistent workers would execute missions beyond a browser session. A separate verifier would repeat tests against a baseline before a human approves a result.

      What must be built?

      Hosted workers, durable mission storage, complete input logs and independent result checks. Today’s Leader log does not include every manual input.

    3. 03 / CONTRACT PREPARED

      COMMIT THE RESULT

      The prepared registry can record a checkpoint fingerprint and run metadata from an authorized publisher. Deployment, publishing and links to transaction receipts are still pending.

      INSPECT THE CONTRACT ↓
    4. 04 / PLANNED

      COORDINATE THE NEXT RUN

      Future contracts could record approved missions and enforce explicit funding limits. The Leader would read those records to coordinate the next round of work.

      What must be built?

      Mission and budget contracts, permission checks, review controls and a tested connection to the workers. The current checkpoint contract does not manage tasks or money.

    OFFCHAIN / COMPUTATION

    Simulation, future model inference, experiments and verification runs.

    ONCHAIN / PROPOSED RECORDS

    Approved checkpoint fingerprints first; mission approvals and spending limits later.

    What would an onchain checkpoint prove? That an authorized address recorded a particular fingerprint in a confirmed transaction. It would not, by itself, prove that a run was correct, intelligent or improved. That requires accessible evidence and verification.

    Live today: the token and creator-fee activity are on Robinhood Chain. The Leader and colony run in your browser. No Leader mission or checkpoint has been published onchain by this site.

    PHASE 2 / CHECKPOINT PREPARATION

    GIVE THE RUN
    A FINGERPRINT.

    Capture this colony’s state and the simulation engine’s fingerprint in a downloadable checkpoint. Verify that a downloaded file still matches its fingerprint.

    LOCAL CHECKPOINT · REGISTRY NOT DEPLOYED

    INSPECT THE REGISTRY CONTRACT ↓

    Your run’s fingerprint will appear here.

    The next step is publishing checkpoint hashes to Robinhood Chain. Generating a file does not send a transaction.

    A matching fingerprint checks the supplied file and digest. Authenticity needs a trusted published checkpoint. The swarm still runs in your browser.
    SWARM SIZE256independent agents
    SIGNAL DEPOSITS0/smeasured each second
    ALIGNMENT0%shared heading, not consensus
    FOOD DELIVERED0returned to the colony
    Every movement is computed here, in your browser.WHAT IS REAL? ↗
    01 / THE ORGANISMINDIVIDUAL → COLLECTIVE

    THE FLY IS
    NO LONGER
    ALONE.

    One fly is noise.
    A swarm is a possibility.

    FLYBRAIN made us imagine a biological nervous system meeting the internet. FLYSWARM asks what happens when the individual becomes a colony.

    No fly sees the whole environment.
    No fly directs the others.
    Each one observes. Moves. Leaves a signal.

    The others encounter it.
    And something larger begins to appear.

    OBSERVE THE COLLECTIVE ↗
    001AN INDIVIDUAL
    008A SIGNAL
    064A PATTERN
    512AN ORGANISM?
    02 / THE RECORDINGREPRODUCIBLE / SEED 7141

    DON’T TAKE OUR WORD.
    WATCH THE COLONY.

    256 agents. 90 simulated seconds.
    45 seconds of footage at 2× speed.

    DOWNLOAD VIDEO ↓DOWNLOAD RUN LOG ↓

    Generated directly from the same simulation engine as the live experiment. Signal and food inputs are scripted and labeled in the video. This records the colony’s behavior; it does not show a biological connectome or a token launch.

    03 / UNDER THE MICROSCOPEEXPERIMENT ≠ EVIDENCE

    STRANGE ENOUGH.
    HONEST BY DESIGN.

    The simulation is real.
    The biology is inspiration.
    The distinction matters.

    [01]

    THE LOCAL RULES

    Each agent samples three nearby points, wanders, and reacts to food within a small radius. A returning agent remembers home and deposits a trail. No language model controls its movement.

    [02]

    THE SHARED FIELD

    Signals spread and fade. Other agents sense the gradient and turn toward it. Add a signal or a food patch yourself. Paths form from repeated local interactions.

    [03]

    THE HONEST BOUNDARY

    This is a browser simulation using simple rules. It does not run a fruit-fly connectome, biological neurons, or an onchain agent network. Its telemetry describes this browser session.

    HOW TO DISTURB THE COLONY +

    Inspect: click a moving agent, or choose its number in the observation panel. The circle around it is its local neighborhood.

    Signal: choose Signal, then click inside the field. You deposit the same chemical-like cue that returning agents leave. Flies follow it only when they encounter it.

    Food: choose Food, then click inside the field. Agents have to discover it nearby. Amber agents carry one unit home. Add food from a keyboard with Enter at the field’s center, or move the placement cursor with the arrow keys.

    Erase signal: click to clear signals within 70 field units. Agents can deposit new signals afterward.

    Move food: click an existing patch, then its new location. Its remaining food is preserved. Escape cancels the selection.

    Move home: click a new return point for carriers. Repel: add a visible zone that turns nearby flies away. Obstacle: add a circular barrier; keep it clear of resources and home. Remove food: click a patch to delete its remaining food. Clear zones removes barriers and repulsion; Reset restores the original colony and food.

    Quick experiments: Drop food adds a 300-unit patch in the lower field. Draw a trail adds a curved sequence of signals. Clear signals removes the field and visual trails while keeping flies, food and delivered units.

    Expand view: opens a larger observation surface. Use Close view or Escape to return. Keyboard focus stays inside the expanded view.

    Population: changing the count starts a fresh run. Pause freezes the experiment. Reset clears interventions and restores the three starting food patches.

    WHAT THE NUMBERS MEAN +

    Signal deposits count actual field writes per simulated second. Alignment measures how closely the agents’ headings agree; it is not a vote or a measure of intelligence. Food delivered counts units physically carried back to the home point.

    Energy is a simple movement variable that falls gradually and replenishes at food or home. It is not a biological measurement. Simulated time stops when the page is hidden or the experiment is paused.

    INSPIRATION & INDEPENDENCE +

    Inspired by FLYBRAIN ↗ and @fruitflydev ↗. FLYSWARM is an independent creative project. No affiliation or endorsement is claimed.

    Pheromone following is a computational analogy used in this experiment, not a claim that fruit flies behave like ant colonies. The narrative explores collective behavior; it does not establish consciousness or biological intelligence.

    04 / $FLYSWARMPHASE 01 / TESTING

    THE TOKEN.
    THE FUEL.
    THE SWARM.

    $FLYSWARM

    A colony needs energy.
    This is how we feed it.

    $FLYSWARM is registered with Pons on Robinhood mainnet. The next step is connecting creator-fee funding to the swarm’s compute, development and experiments.

    FOLLOW THE FUNDING LOOP ↘
    Verified CA: 0xdfda44de8cb48e92f485350378149a834cba73c0
    TOKEN → FEES → RESOURCES → SWARMTHE FLY WAS NEVER THE END STATE.
    05 / FEED THE SWARMFUNDING MODEL / PLANNED

    KEEP THE COLONY
    EVOLVING.

    The idea is simple: creator fees come back
    to the project that brought the swarm to life.

    1. 01 / THE TOKEN

      $FLYSWARM
      ON PONS

      Trading generates creator fees under the token’s launch settings.

    2. 02 / THE RESOURCES

      FEES FEED
      THE PROJECT

      The project’s creator-fee income goes toward running and developing the swarm.

    3. 03 / THE COLONY

      MORE ROOM
      TO EVOLVE

      Compute. New behaviors. Larger experiments. Recorded results you can inspect.

    ONCHAIN / CREATOR FEES

    THE SWARM’S FUNDING.

    EARNED BY $FLYSWARM1.166630 ETH

    Credited to escrow across 54 sweeps since launch.

    PENDING COLLECTION0.000040 ETH

    New token fees not yet swept into escrow.

    WALLET CLAIMABLE0.195184 ETH

    Across every launch paid to this recipient. Not a token-only balance.

    Saved read · 11 September 2026, 04:42 UTC · block 59,982,028. Automatic checks begin when this section is visible.

    Earned counts confirmed creator-fee sweeps, matching Pons’ earned metric. It excludes pending fees and does not mean the funds have been spent on the swarm.

    DOWNLOAD FEE RECORDS ↓
    RECENT CREATOR-FEE SWEEPS
    • Download fee records for the complete transaction list.
    $FLYSWARM / CHAIN ACTIVITY

    FOLLOW EVERY MOVE.

    One row per transaction from the most recent 2,000 blocks. Current-pool BUY / SELL comes from swap events; older curve trades use transfer direction. Router hops are combined and both directions in one transaction are MIXED. Pool amounts are swap deltas, not wallet net amounts. FROM / TO shows token-transfer endpoints, which may be routers. Amounts are rounded down to whole tokens.

    EVENTFROMTOAMOUNTTRANSACTION
    BUY0xb0fb…25df0xea9b…47621,066,926 FLYSWARM↗ 0xfc1f…6304
    BUY0xb0fb…25df0x4828…1913719 FLYSWARM↗ 0x1ed4…46fb
    BUY0xb0fb…25df0x458f…88d4136,902 FLYSWARM↗ 0x3ed9…1812
    BUYMULTIPLE0x2843…2dab3,456,160 FLYSWARM↗ 0x62a8…5240
    BUY0xb0fb…25df0x6e0f…cf5411,265,120 FLYSWARM↗ 0x71b7…7339
    BUY0xb0fb…25df0x4828…1913781 FLYSWARM↗ 0xc029…3b4c
    BUY0xb0fb…25df0x4828…1913781 FLYSWARM↗ 0x6075…ae76
    SELL0xb086…7eac0xb0fb…25df4,068,129 FLYSWARM↗ 0x4ec6…169d
    SELL0x8855…ec920xb0fb…25df3,223,682 FLYSWARM↗ 0xe5d4…fd90

    Saved snapshot · 2026-09-11T06:05:20.411Z · block 60030996. Checking for newer records in the background.

    Confirmed block data, refreshed by polling—not a pending-transaction stream. Successful reads are saved for all visitors. Refreshes run while this page is open and visible; after inactivity the last saved record loads first.

    FUNDING → DEVELOPMENT → EVIDENCE

    BUILT FOR THE SWARM.

    DOWNLOAD DEVELOPMENT RECORD ↓

    Follow what ships and inspect the evidence. Creator-fee receipts show income; a development milestone only counts as paid when a spending receipt is attributed to it.

    FUNDING SOURCECreator-fee receiptsInspect token income ↗
    PROJECT SPENDINGNot reportedNo spending receipts have been published. This does not mean spending is zero.
    AUTOMATIC FUNDINGPlannedFees are not automatically allocated to these milestones.

    Loading development evidence…

    FROM ACTIVITY
    TO POSSIBILITY.

    KEEP IT RUNNING
    Hosting, storage and compute for future swarm runs.
    BUILD WHAT’S NEXT
    Development, agent coordination and new experiments.
    SHOW THE WORK
    Recordings, reproducible runs and published results.
    PHASE 2 / THE DIRECTION

    BUILT BY A DEV.
    GROWING TOWARD A SWARM
    THAT IMPROVES ITSELF.

    Today, a developer builds the behaviors, ships the updates and runs the experiments.

    The Phase 2 goal is a persistent swarm that can propose changes to its own behaviors, test them against measurable goals, and keep the improvements that pass review.

    Developer-led now. Increasingly self-directed over time. Autonomous evolution is not active yet; code releases, wallet access and spending remain subject to human approval.

    STEP ONE / TESTING

    The token is on Robinhood mainnet; the colony still runs locally in your browser. Automated fee routing, onchain swarm execution and project spending records are not active yet. Pending fees are not proof of funding received or spent.

    ENTER THE COLONY ↑