Can a fruit-fly brain learn to trade stocks? Watch a green-visor fly work the BUY, HOLD and SELL keys while its native neural activity and paper account remain visible.
Start the experiment to watch reward training on 2015 AAPL history. The learned connections are then frozen for 1,105 trading days, January 2016–May 2020, across 18 automatically advancing chapters. The final screen reports the actual net return, trading costs, drawdown, ending holdings and a passive comparison.
Director view keeps the fly, physical key presses and historical stock monitor together. Pilot and orbiting Overview views remain available. Large action captions distinguish holding cash from holding shares. Details opens the full ledger and research telemetry; Hide UI clears the screen for filming. Choose 4× for roughly47 minutes of historical testing or12× for roughly16 minutes, plus training.
The $1,000 account uses whole shares, conservative order limits, modeled commissions, spread, slippage and dividends. Profits, losses and orders are never prescribed. Market observations precede the next-open fills.
This is an engineered reward-trained circuit inside a native fruit-fly connectome:164,587 neurons and25,563,197 original synapses, with learned parameters controlling49,152 added connections. Native decision-neuron spike counts select the actions. It does not establish biological investment competence or predictive skill.
Historical sources: Plotly, extended AAPL CSV, and Apple dividend history. Fixed downloaded data, hypothetical executions. Training and test dates are separate; development has inspected the test dates, and replays are not fresh evidence.
Comments · 0
Open discussion, no rating attached. Ask questions, swap notes, follow threads.