DEV TOOLS Signal 52
Coinbase engineer open-sources simulated fly brain crypto trader with 166,700 neurons on GitHub
↗
A Coinbase engineer released Stonkfly, an open-source simulation of a male fruit fly brain with 166,700 neurons that processes candlestick chart images through virtual photoreceptors and trades cryptocurrency using dopamine-like reward signals.
The project provides a working example of connecting a large-scale biological neural simulation to real-time financial data and reinforcement feedback, something developers can now run and modify locally. However, the author explicitly cautions that the experiment proves nothing about actual trading ability, especially given a general rise in crypto prices that makes any buyer look skilled.
Written by elseif from the cluster below · every claim links back to a sourceThe three things worth knowing
Stonkfly simulates 166,700 neurons and 25.6 million connections from a male fruit fly brain, processing 320x180 candlestick chart pixels through virtual photoreceptor cells rather than parsed price data.
Profitable trades trigger dopamine signals to 15 cells while losses activate 2 aversive cells, with the fly deciding to buy, sell, or hold every 500 ms on currencies shown in round-robin fashion.
Running the simulation requires macOS or Linux, 16 GB of RAM, Python 3.11, and a C++ 17 compiler, defaulting to paper trades with a $100 virtual balance using real BTC-to-USDC data.
THE CLUSTER