Coinbase Engineer Uses Simulated Fly Brain to Trade Bitcoin

Alex Wormuth, a software engineer at Coinbase, connected a simulated fruit fly brain to a Bitcoin trading account. He released the open-source project, Stonkfly, and posted a public dashboard and code repository.

Stonkfly uses a published wiring map of an adult male fruit fly and simulates about 166,700 neurons and 25.6 million synaptic connections. The program converts a candlestick price chart into a small image, splits that image across two simulated eyes, and feeds raw pixel color data to light-sensing cells in the model.

After processing the image the model chooses one action: buy, sell or hold. When a trade finishes in profit the code fires 15 neurons linked to reward; when it loses the code fires two neurons linked to unpleasant signals. Trading fees count as losses. The simulated brain cannot borrow funds or place bets on falling prices.

Stonkfly runs in a paper-money default mode that starts with $100. A live mode limits each order to $10 and caps trading attempts at 24 per day. At the time of posting the public dashboard showed zero completed trades on Saturday.

Wormuth wrote, “Synaptic changes do not establish that it learns to trade profitably.” He described the project as an experiment meant to test whether a mapped brain will change its wiring in response to rewards tied to tasks that did not exist during the species’ evolution. Bitcoin traded near $77,286 on Saturday.