FLY FUND IThe experiment ↗Markets become signals.
Five- and twenty-minute momentum, volatility and a smoothed trend become neural inputs. Consistent trends stimulate approach; sudden spikes and noisy moves are filtered out.
BIOLOGY MEETS THE MARKET
Fly Fund I
Can a fruit fly outperform the market?
A $1,000,000 fund. Ten assets. One fly circuit making the calls. Watch the signals, follow the decisions, track the returns.
Follow the portfolio ↓Connecting to the experiment
READING MARKET INPUTS
THE MARKET10 ASSETS
SamplingWaiting
5m momentum--
20m momentum--
Trend clarity--
OBSERVINGThe next allocation starts with a signal.
THE NEURAL RESPONSE0 SPIKES
1,045 NEURONS
17,224 CONNECTIONS
17,224 CONNECTIONS
Approach input0.0 Hz
Avoid input0.0 Hz
Neural score0.0
Waiting for the first shared market observationRecorded inputs · neural activity replay
FUND PERFORMANCE
The results, in real time.
Portfolio value$1,000,000.00Starting balance: $1,000,000
Total P&L$0.000.00% since launch
Against equal weight0.00 pp10-asset buy-and-hold benchmark
Capital at work0.0%No trades yet
Fly portfolioEqual-weight basket
SINCE LAUNCHPerformance begins with the first observed prices. No invented history.
The portfolio.
$1,000,000 cash| Asset | Price | Allocation | Value | P&L |
|---|
Recent decisions.
0 TRADESObserving first price changes.
Actual neural readouts and portfolio trades will appear here.
THE QUESTION IS SIMPLE
A brain built to forage.
A market full of choices.
Price movement is the stimulus. Each asset is a different sweet. Can the circuit's approach and avoidance responses produce a useful investment strategy?
The circuit responds.
A spiking model runs 1,045 connected fly neurons. Motor-output activity produces a buy, sell or hold score. The fly visits each asset as its recorded inputs are replayed.
The ledger keeps score.
Confirmed signals allocate up to 2.5% per trade, scaled down as volatility or fund drawdown rises. New buys respect a 12.5% per-asset limit and 60% total exposure ceiling. Each simulated fill includes 0.10% fees and 0.05% slippage. Results are compared with an equal-weight basket.
Model, data & experiment rules +
This is a fly-connectome investment experiment, not a trained financial model or the complete fly brain. The recorded connectivity comes from a MaleCNS locomotor circuit; the market encoding and output decoder are designed for this experiment. Positive outputs do not establish predictive skill.
Ten selected Solana assets: SOL, cbBTC, TSLAx, NVDAx, SPYx, GOOGLx, AAPLx, MSFTx, AMZNx and METAx. These are not a market-cap ranking. xStocks are tokenized exposures; cbBTC is wrapped Bitcoin. Quotes come from Jupiter and represent the tokens, not direct stock-exchange quotes.
The shared ledger starts with $1 million in hypothetical cash. Quotes are sampled at most once per 30 seconds while the feed is being followed. Only fresh, advancing quotes with six observations can trigger trades. Gaps are not backfilled with invented orders. Fees and slippage are estimates; liquidity, market impact, taxes and token corporate actions are not modeled. The chart includes cash and marks positions at the latest available token prices.
The fixed readout is buy-motor Hz minus three times sell-motor Hz. Scores of at least 5 qualify for buying; scores of at most −3 qualify for selling. The strategy requires two matching neural responses on distinct source blocks, with fifteen-minute asset cooldowns and two minutes between buys. A 2.5% fall from an observed position peak stimulates protective avoidance, which can bypass the normal exit delay if neurons produce a sell signal. New buys pause at an 8% portfolio drawdown. These controls do not guarantee a stop price; gaps can prevent trading. No real funds are held or traded.
This is a fresh launch run starting with $1,000,000 in simulated capital, not real deposited funds. Its timestamp and performance begin at launch. The earlier development run is preserved separately in the pre-launch archive.
The spark display replays actual modeled spikes on recorded connections at a slower visual speed. Background anatomical dots are context, not additional simulated neurons.