I made a fruit fly trade crypto
I have been trading crypto for a while. So I tried something I probably should not have.
Google recently released a complete map of a male fruit fly's brain, more than 166,000 neurons and their connections. To build it, researchers sliced the fly's brain and body into incredibly thin sections, imaged millions of those slices, and used AI to reconstruct them into a 3D map of its neural circuitry.
And somehow, my first thought was: what if I made the fruit fly trade crypto?
The build
My friend Krishang Sharma and I spent the weekend building exactly that. We took my actual trading patterns, the setups I look for, when I enter, when I stay out, how I react to different market conditions, and tried to teach it my way of trading.
Then we wired the actual fruit-fly connectome into a trading system and put it on Binance Spot Testnet. It gets live order-book data, watches liquidity and market conditions, makes its own decisions, and sends actual orders through the Testnet: BTC, ETH, SOL, BNB, XRP, DOGE, ADA, AVAX.
Underneath it, I am running the fly's mushroom-body circuit, thousands of real neurons, with dopamine-gated synaptic plasticity driven by realised P&L. Basically, the fly gets rewarded when its trades make money and changes its behaviour based on what just happened.
Watching it trade
Which sounds pretty reasonable, until you watch it trade. It will sit there waiting for liquidity, enter a position, get absolutely wrecked, unwind a huge chunk of the portfolio, then somehow claw its way back.
And instead of deciding it has made enough money and stopping, it keeps going, because I gave it one extremely simple objective: make enough money to stop trading. It never thinks it has enough.
So now the fruit fly is basically stuck in an infinite trading loop. Buy. Sell. Get wrecked. Recover. Make money. Lose money. Repeat. Somewhere inside 166,000 neurons, there is apparently a tiny trader staring at the order book thinking: "Just one more trade." That might be the most realistic part of the entire experiment.
We built the whole thing over the weekend and open-sourced it: fruit-fly trader on GitHub.
