Coinbase engineer Alex Wormuth has wired up a digital model of a fruit fly brain to a crypto trading simulator. The setup is based on a scientific connectome of the fly, mapping out about 139,000 neurons and 50 million synaptic connections. You can check out the details and the code in the open GitHub repo.

How the experiment works

Market events are translated into stimuli the fly brain model can process: buys are interpreted as "sugar," while sells are seen as "bitterness." The resulting neural activity is then converted into buy or sell commands inside the trading simulator.

No AI agent, no preset strategies

There's no separate AI agent here, and no pre-programmed trading strategy. Every trade is driven purely by how the simulated fly brain reacts to the market signals it gets.

Dev’s previous experiments

Wormuth has tried this approach before—he previously got a digital fruit fly brain to play Doom using similar methods.