Animal brain simulation is becoming a new trend in the tech community. An engineer at cryptocurrency exchange Coinbase has gone on to become a day trader on the Stonkfly project.
Animal brain simulation is becoming a new trend in the tech community. An engineer at cryptocurrency exchange Coinbase has gone on to become a day trader on the Stonkfly project.
The open-source project recreates the virtual brain and eyes of a male fruit fly (Drosophila), which is shown standard charts with historical cryptocurrency prices. The fly can buy, sell, or hold the selected currency, and receives a virtual "reward" for profitable transactions, writes Tom's Hardware.
The brain of a virtual fly consists of 166,700 neurons and 25.6 million connections. The virtual fly sees the chart through its left and right eyes at a resolution of 320×180 pixels. It is the visual perception of RGB pixels that is simulated, not direct digital price data.
An increase in the value of the portfolio triggers the release of dopamine as a positive reinforcement signal to 15 brain cells. Losses or trading commissions activate two “disgust” cells. Learning occurs through changes in synapses.
The engineer notes that the simulation lacks mechanisms for pain or emotion. At the same time, the fly demonstrates a more measured approach than many human traders — it cannot use leverage or trade short.
By default, the simulation makes a decision every 500 milliseconds, and market data is updated every 60 seconds. The maximum bet per order is $10 (up to 24 trades per day).
You can watch Stonk auctions live at the link .
The developer emphasizes that this small project only demonstrates the connection between incoming visual signals and changes in synapses. He cautions against drawing conclusions about the insect's real trading abilities, especially against the backdrop of the general growth of the crypto market, where "any buyer can look like a seasoned expert."
Anyone can run their own version of Stonkfly. All you need is a macOS or Linux computer with at least 16GB of RAM, Python 3.11, and a C++ 17 compiler. By default, the simulation uses a virtual balance of $100 and simulated trades, but it is based on real BTC/USDC market data.
The project is available on GitHub .
