Coinbase engineer wires simulated fly brain to Bitcoin trading

A Coinbase software engineer linked a digital version of a fruit fly's brain to a Bitcoin trading account. The open-source project goes by the name Stonkfly.

The system displays a price chart to the simulated brain, and profitable trades trigger a burst of virtual dopamine. The simulated fly never sees any price numbers directly.

How a Simulated Fly Brain Executes a Trade

Alex Wormuth published the code on GitHub alongside a public dashboard. The model includes 166,700 neurons and roughly 25.6 million connections among them.

These numbers come from a connectome, a published wiring diagram of an adult male fruit fly's nervous system. Neuroscientists spent years constructing it. Wormuth instead aimed it at a candlestick chart.

Stonkfly draws the chart as a small image and splits it across the fly's two virtual eyes. Light-sensitive cells read the raw pixel colors, after which the brain decides on buy, sell, or hold.

Profit activates 15 reward neurons. A loss activates two neurons that register something unpleasant, and trading fees count as losses. The fly cannot borrow money or bet on a price drop, giving it an edge over many AI bots trading stocks.

The Engineer Says the Experiment Proves Nothing

According to the Coinbase engineer, the experiment does not demonstrate everything.

“Synaptic changes do not establish that it learns to trade profitably,” Alex Wormuth said.

The program runs on paper money by default, starting at $100. In live mode, each order is capped at $10 and attempts are limited to 24 per day. The public dashboard still showed zero completed trades as of Saturday.

That caution has a different impact in a falling market. Bitcoin (BTC) was trading near $77,286 on Saturday, up 0.6% on the day but well below its peak.

The real question is not whether the fly makes money. It is whether a mapped brain can rewire itself for a reward that evolution never provided.