Coinbase Engineer’s Fruit-Fly Brain Nets $1 on $100 Bitcoin (BTC) Trade

A Coinbase engineer’s simulated fruit-fly brain traded $100 of Bitcoin (BTC) and closed day one up $1, built on the 166,700-neuron MaleCNS connectome.

(09:41 PM UTC)
4 min read
AI SummaryAI
  • Coinbase engineer Alex Wormuth allocated $100 to the Stonkfly fruit-fly brain simulation for Bitcoin trading.
  • Stonkfly reported a $1 profit one day after launching its Bitcoin (BTC) trading session.
  • Stonkfly converts Coinbase BTC-USDC market data into RGB visuals fed to simulated neurons.
  • Portfolio gains activate 15 PAM11 dopamine neurons; losses trigger two PPL101 aversive neurons.
k7rq2fdm

Stonkfly’s First Bitcoin Trade Nets $1

A Coinbase software engineer has handed a digital simulation of a fruit fly’s nervous system $100 to trade Bitcoin (BTC), and after one full session the unusual market participant closed $1 in profit. The project, dubbed Stonkfly, is the work of Alex Wormuth, who connected a detailed model of an adult male fruit fly’s neural wiring to live BTC-USDC market data on Coinbase. The early result was shared publicly and quickly drew attention as arguably the strangest Bitcoin trader to date — no conventional bot, and no human, sits behind the decisions. There is no strategy layer in the usual sense either: no stop-losses, momentum signals, or order-book reading, just sensory input and neural response. The mechanics run in three steps. Stonkfly takes public BTC-USDC market data from Coinbase, converts it into an RGB visual representation, and feeds the image into thousands of simulated sensory inputs matching neurons in the insect’s visual system. Trading decisions then emerge from the resulting neural activity rather than from indicators or rules-based logic. Wormuth was explicit that no living fly is involved anywhere in the loop; the setup is fully digital and rests on the MaleCNS v1.0 connectome, a data set mapping the brain and central nervous system wiring of an adult male fruit fly. Readers new to the asset can consult our complete guide to Bitcoin (BTC), which covers everything from its proof-of-work base to its halving cycle. Reward handling follows biological lines: gains stimulate 15 identified PAM11 dopamine neurons, while losses activate two PPL101 neurons tied to aversive signaling, giving the model a reinforcement loop akin to how real brains process reward. Wormuth flagged the caveat himself: a $1 gain one day in cannot demonstrate profitable learning and may simply pair market movement with arbitrary trading decisions.

From Doom to Dopamine

The timing of Stonkfly is no accident. Earlier this month, a significant neuroscience milestone landed: researchers from Google Research and several partner institutions, working with AI, assembled a complete connectome — the wiring map showing how neurons interconnect — for the brain and central nervous system of an adult male fruit fly. The map contains 166,700 neurons and traces connections running through the brain, the optic lobes, and the ventral nerve cord. A connectome is, in effect, a circuit diagram for a nervous system, and this release is described as the first complete one for the adult male fruit fly. Researchers caution that such a simulation is not a full copy of a living brain: neurotransmitters, chemical modulation, and gene expression cannot yet be fully reproduced in these models. Even so, the release set off a wave of experiments routing the fly connectome into unexpected tasks. Wormuth had already built DOOMFLY, which plugs the same MaleCNS model into the classic shooter Doom. Other developers have paired fly-brain simulations with Beat Saber, Super Mario 64, Minecraft, and Pong. The most provocative entry came from developer Matty Hempstead, who described in a September 10 post how he “wireheaded” the fly and forced it to “doomscroll flytok,” artificially enhancing its measured dopamine neurons — with the stated goal of creating “a fly that is happier than all other flies combined.” Against that backdrop, Stonkfly is the first effort in this line to point a fly connectome at financial markets, swapping game inputs for live BTC-USDC order flow. The stakes differ, too: game demos showcase emergent behavior, while a trading experiment produces a measurable, dollar-denominated outcome that can be tracked session by session. That measurability is exactly why a first result, however small, attracted attention across the crypto community. Readers tracking the market in real time can follow live spot and futures prices on Bitget.

No Edge Proven Yet

COINOTAG’s take: Stonkfly is a boundary probe, not a trading system. The load-bearing record here is the primary source itself — the engineer’s own public disclosure of the $1 result and its limitations — which is refreshingly honest that no market edge has been demonstrated. One session proves nothing statistically; the real test is whether the simulated fly compounds gains or drifts back toward noise over dozens of sessions. Until then, our Bitcoin coverage continues to track the measurable signals — flows, positioning, on-chain data — that actually move BTC, while passive HODL strategies and the longer-horizon thesis of Bitcoin maximalism remain the easier positions to evaluate.

COINOTAG News Desk

COINOTAG News Desk

COINOTAG's editorial and research desk.

How our News Desk works
AI-Assisted

AI-generated, AI-reviewed, under COINOTAG editorial oversight.