>>ScanTradeCompoundRepeat
FLYWHEEL
A small botA bigger edge<<
  1. SCAN
  2. TRADE
  3. COMPOUND
  4. REPEAT
Flywheel mark: a fruit fly at the centre of a four-blade wheel, drawn in dots

Real time
opportunities
across markets
_

A fly brain, loose in the trenches.

A full simulation of an adult fruit-fly brain, built from the FlyWire connectome: 139,255 neurons joined by 2.5 million connections. Live prices reach it as taste and smell. If it walks forward, it buys. If it backs away, it sells. If it stops to groom, it holds.

Neurons simulated
139,255
Spikes per second
Brain seed
0x1F1E

the brain

Every dot is a neuron.

Each point sits where that neuron sits in the real fly's head. When one fires, it flashes. Neurons collect input, fire at threshold and leak back to rest, following the model in Shiu et al. (Nature, 2024).

FlyWire connectome · 139,255 neuronsdemo feed
drag to orbit
Current decision
HOLD
confidence0%
forward · buy0
backward · sell0
groom · hold0
P&L +0.0%trades 0
Smelling
no scent yet
the coin the fly is smelling shows up here
Regions firing
sugar neurons · buybitter neurons · sellolfactory · one per tracked coinKenyon cells · learninggrooming · hold

the trenches

The fly walks toward what it smells.

A top-down view of the floor. Each coin is a scent selected from active markets, and its size follows market activity. If the brain votes buy while the fly stands on a coin, that coin lights up.

waiting for the braindemo feed
coin sized by activitydim · emerging coinsring · coin heldpulse · boughtpulse · sold

the loop

The flywheel.

Fees fund the fly. Its wins fund the buyback. The buyback shrinks the float, the show draws volume, and volume makes more fees.

fees in$0.00bought back$0.00
  1. Fees feed the fly

    Every $FLY trade pays a fee into the fly's wallet. That's the only money it ever trades with.

  2. Bags set the appetite

    Buy size grows with the wallet, and a fuller wallet lowers the bar for buying.

  3. Activity sets the menu

    The busiest eligible markets become scents, and stronger activity makes the fly sniff them more often.

  4. Wins buy back $FLY

    Half of every realized gain goes into $FLY, and the fly never sells it.

  5. The show pulls the trenches

    A live brain, open bags and a public trade log are worth watching. Watchers trade, and volume turns back into fees.

  6. ↺ back to 01

how it works

A real brain's wiring, plugged into the trenches.

A connectome maps every neuron in a brain and every link between them. FlyWire published one for an adult fruit fly in 2024. It's the wiring only, with nothing learned. $FLY runs that wiring and gives it a market to react to.

1

Sensory input

Sugar ← rising price
When the coin it's smelling climbs over the hour, its sugar neurons get driven, scaled to how much that coin normally moves. GOOGL adds a slow background mood.
Bitter ← falling price
When the coin falls, the bitter neurons fire instead.
Scent ← active coins
Each tracked coin gets its own olfactory channel; market activity sets how often it's sniffed. Coins that paid off smell stronger later, and ones that hurt taste bitter for hours.
2

The brain

The simulated connectome
FlyWire v783 wiring, 139,255 neurons and 2.5 million connections, run as leaky integrate-and-fire neurons after Shiu et al. (2024). Deterministic from a published seed and reproducible from public data.
No alpha, on purpose
The wiring never saw a price chart and wasn't tuned to trade. The seed, the input mapping and the trading rules are public.
3

Motor readout

Forward = buy
When the descending neurons for walking forward win, the fly buys what it's smelling, up to eight coins. A fuller treasury lowers the bar.
Backward = sell
When backing away wins, it sells. Losers taste more bitter the further they fall, which is what makes it let go.
Groom = hold
When it would rather clean itself, it does nothing. Which neurons mean what comes from the connectome's own cell-type labels.

sensory neurons >> interneurons >> descending neurons

why GOOGL

The map was traced by Google's machines.

$FLY pairs with the GOOGL stock token because Google Research made this brain map possible. Biologists imaged the tissue; Google's connectomics team turned tens of millions of microscope images into a wiring diagram.

2018

A whole brain, sliced and imaged

At HHMI Janelia, Zheng and colleagues cut an adult fly brain into roughly 7,000 sections and imaged each one: about 21 million images and 100 terabytes, fine enough to see individual synapses.

2019

Google's networks trace every neuron

Hand-tracing one neuron takes days, and there are about 140,000. Viren Jain's team at Google Research ran flood-filling networks, which grow each neuron's shape voxel by voxel, over the whole volume to produce the first automated reconstruction of a complete adult fly brain.

2020–2024

FlyWire proofreads it

Automated tracing makes mistakes, like splitting one neuron in two or merging neighbours. FlyWire, led by the Seung and Murthy labs at Princeton, put the segmentation online, and hundreds of scientists and volunteers corrected it neuron by neuron, adding synapses and cell types.

2024

The connectome, and a model of it

The finished map, 139,255 neurons and about 50 million synapses, appeared in Nature in October 2024. Grouped by neuron pair, that's the 2.5 million connections simulated here. Shiu and colleagues published a whole-brain model in the same issue, and $FLY runs that model on that map.

Zheng et al., Cell 2018 · Januszewski et al., Nature Methods 2018 · Li et al., bioRxiv 2019 · Dorkenwald et al., Nature 2024 · Shiu et al., Nature 2024