An engineer has published a downloadable simulation that pairs a virtual male fruit fly brain with cryptocurrency trading, releasing the project on GitHub. The simulation presents the fly with a standard candlestick chart built from historical pricing and lets it choose to buy, sell, or hold each currency it is shown. Currencies appear in round-robin order rather than by any selection process. Profitable trades are rewarded, and the project's author has warned that it demonstrates far less than it appears to.
The reward system runs on dopamine and aversion signals rather than money. When the portfolio value rises, a positive reinforcement signal reaches 15 cells; when it falls, two aversive cells activate. Trading fees are treated as losses under the same mechanism. The author states that no pain or emotional processes are involved. The fly also cannot use leveraged positions or shorts, restrictions the write-up describes as better judgment than most human traders display.
The simulated brain contains 166,700 neurons and 25.6 million connections. It receives market information visually rather than as numeric data: the fly sees the chart as a 320x180 display split across its left and right eyes, with an overlapping center region, and its simulated photoreceptor cells are fed raw RGB pixel values instead of pricing figures. By default the fly thinks and acts every 500 milliseconds, while market data refreshes every 60 seconds. It can wager up to $10 on a single order, up to 24 times per day.
The author is explicit about the limits of the work, saying the small project proves only that a connection exists between the input mechanisms, the visual signals, and changes in synapses. He cautions against treating those changes as evidence of real trading skill. A general rise in cryptocurrency prices, he notes, can make any buyer look skilled. The write-up adds that some observers will likely read more into the experiment anyway.
For a US audience, the project arrives as retail crypto trading remains widely accessible through consumer apps, and as automated and algorithmic trading tools draw ongoing attention from regulators and investors. The simulation imposes its own guardrails that many retail platforms do not, barring leverage and short positions and capping exposure per order. Its trading fees are modeled as losses, mirroring the real cost structure retail traders face.
The project is open source and available for download, and it was covered by Tom's Hardware. Bruno Ferreira, a contributing writer for the publication with decades of experience in PC hardware and a background as a developer, wrote the coverage.
More software news from TechManNews.
