Google Research and its collaborators spent about 10 years building MaleCNS v1.0, a full wiring map of a male fruit fly’s central nervous system. Three days after the dataset was released on Sept. 3, Coinbase software engineer Alex Wormuth had already connected it to Doom for a survival-learning experiment. He later reworked the same connectome into an open-source Bitcoin trading system.
Google and research partners published the connectome
MaleCNS v1.0 is described in the report as a fruit fly “connectome,” a map showing which neurons connect to which other neurons and how many times they connect. In a post on its official blog, Google Research called it “the brain map with the largest number of neurons ever.”
The work was completed by Google, HHMI Janelia Research Campus, the U.K.’s MRC Laboratory of Molecular Biology, and the University of Cambridge. The paper was published in Cell. To build the dataset, the team sliced a male fruit fly into millions of ultrathin sections, scanned them one by one with electron microscopes, reconstructed 3D neurons from 2D slices with Google image-recognition models, and then manually checked the results. The process took about a decade.
The final map contains 166,691 neurons, about 125 million synaptic connections, and more than 11,700 cell types. It also brings the brain and ventral nerve cord into the same map for the first time. The report explains the ventral nerve cord as roughly comparable to the human spinal cord, carrying signals between the brain, six legs, and two pairs of wings.
Cambridge neuroscientist Greg Jefferis said the result lets researchers “trace from the eye all the way to the leg.” The report compares the release with a female fruit fly brain map published in 2024, which had 139,255 neurons but did not include the nerve cord. The new male version has more neurons and gives scientists a full basis for comparing male and female brains in social behaviors such as courtship and aggression.
The dataset has been released through the open-source tool Neuroglancer, making it available for public browsing and download.
A Coinbase engineer quickly connected it to Doom
Not long after the release, the first person to turn the map into a working project was not a neuroscientist but Alex Wormuth, a Coinbase software engineer who described himself as an aspiring neuroengineer. In a Sept. 6 post on X, he wrote: “I'm training a fly brain to play Doom using the full MaleCNS v1.0 fruit fly connectome. Each Doom frame stimulates sensory neurons. Neural activity is mapped to game controls. Damage triggers a stimulus to two PPL101 dopamine cells as reinforcement. Will the fly learn to…”
Wormuth turned that idea into an open-source GitHub project called DOOMFLY. According to the report, each frame from Doom is converted into 3,335 brightness signals and 811 color signals to mimic the stimuli that would reach a fruit fly’s photoreceptors. Those signals are then run through simplified neural dynamics across the retained 166,000 neurons and 25.6 million synaptic connections. A fixed output interface translates activity in selected neurons into turning, moving forward, and firing in the game.
When the fly takes damage, the system injects an artificial signal into two PPL101 dopamine cells as a punishment cue. That setup adjusts the weights of about 4,184 connections, while the rest of the wiring remains unchanged.
The project README is blunt about the current result: it is “training live” and has not shown that it has learned to survive. The latest candidate version did not pass visual, conditioning, and survival tests, and longer training failed to produce a stable learning curve. At this stage, the digital fly still cannot stay alive in the game.
The same map was later adapted for Bitcoin trading
Wormuth later connected the fruit fly brain map to real-money trading, using it to read candlestick charts and place Bitcoin buy and sell orders through a Coinbase account. That project is called Stonkfly, and its source code was also released on GitHub.
The report adds that less than a week after the data became public, engineer Jessica Paquette connected the connectome to Super Mario 64, making Mario repeatedly jump and hit walls based on fruit fly neural activity. Others used it with Beat Saber or inserted it into Minecraft as a new creature.