Bitcoin Day Trading with a Fruit Fly Brain Simulation—Coinbase Engineer Releases Open-Source Project

Coinbase software engineer Alex Warmoth has released an open-source project that simulates the neural connectivity of a fruit fly's brain to trade Bitcoin. When the simulated fly generates profit, reward signals are sent to dopamine neurons; neural activity patterns are then read and converted into buy, sell, or hold decisions.

Warmoth announced the project, named "Stonkfly," on the social platform X on the 12th (local time). It is built on the "MaleCNS v1.0" dataset, which maps the brain and ventral nerve cord connections of an adult male fruit fly. The simulation incorporates approximately 166,700 neurons and roughly 25.6 million synaptic connections.

Warmoth previously developed "DOOMFLY," a project using the same dataset to play the first-person shooter game DOOM. Stonkfly applies the technology built for that project to cryptocurrency trading.

Rather than presenting a finished investment strategy, this project is closer to an experiment testing whether a simulation reflecting real neural connectivity can respond to market information and modify its behavior based on trading outcomes.

Reading Candlestick Charts Through a Fly's Eyes

Bitcoin price data is fed into the neural network in a manner similar to how a fruit fly receives external information. The virtual fly views candlestick charts on a 320x180 resolution screen spanning both eyes, with simulated photoreceptor cells receiving raw RGB pixel values rather than price data.