Software engineer Alex Wormuth has developed an open-source project called Stonkfly, which connects a digital simulation of a fruit fly brain to a Bitcoin trading account. The model utilizes a connectome consisting of 166,700 neurons and approximately 25.6 million connections, based on the nervous system map of an adult male fruit fly.
The system processes candlestick charts as images split across the simulated eyes of the fly. Trading decisions are based on visual input, with profitable trades triggering 15 reward neurons, while losses and trading fees activate two neurons associated with negative feedback. The bot is restricted from short selling or borrowing assets.
The project operates using virtual currency, starting with a balance of $100. In live mode, the software limits individual orders to $10 and caps activity at 24 attempts per day. As of Saturday, the public dashboard indicated that no trades had been completed by the system.
Wormuth noted that the experiment does not demonstrate that the simulated brain can learn to trade profitably through synaptic changes. The project aims to explore whether a mapped biological brain can adapt to reward mechanisms that were not part of its original evolutionary development.
