Engineer turns simulated fly brain into a crypto day trader, posts downloadable sim to GitHub — 166,700 virtual neurons read candlestick charts for dopamine hits
Coinbase engineer turns a fly brain into a crypto day trader — Stonkfly has 116,700 simulated neurons and hasn't lost its money yet
An engineer working with Coinbase has created a simulation of a fly's brain that functions as a cryptocurrency day trader. This open-source project, named Stonkfly, allows users to observe the fly's interactions with a standard candlestick chart displaying historical pricing data. The virtual fly, with 166,700 neurons and 25.6 million connections, observes the chart through a 320x180 display across its left and right eyes.
It makes decisions to buy, sell, or hold various currencies in a seemingly random manner, receiving positive reinforcement through a dopamine response for profitable trades, while losses are reflected in the activation of aversive cells.
The fly's trading activities are limited to avoiding leveraged positions and short-selling, and it is rewarded for successful trades with a $10 bet cap per order, up to 24 times daily. The brain's synapse changes are triggered by the fly's profits, while losses are factored into trading fees. Notably, the fly does not possess pain or emotional mechanisms.
The simulation runs on macOS and Linux systems, with 16 GB of RAM required and Python 3.11 and a C++ 17 compiler installed. It uses real-time BTC-to-USDC data but defaults to paper trading with a $100 virtual balance. While the project does not validate the fly's trading prowess, it emphasizes the relationship between input mechanisms, visual signals, and synapse modifications.
Written by urgent.news from Tom's Hardware's reporting — not their text. Machine-written — may contain errors; check the original before relying on it.
