A complete neural wiring map of an adult male fruit fly did not stay inside neuroscience circles for long. After Google researchers, HHMI Janelia Research Campus and other scientific collaborators released the open dataset behind MaleCNS v1.0, developers began wiring the simulated fly into video games, browser demos, rate-setting experiments and crypto trading systems.
The findings were published in Cell on Sept. 3. According to the research team, MaleCNS v1.0 is the first complete connectome of an adult male fruit fly brain and central nervous system, spanning the brain, optic lobes and ventral nerve cord down to individual neurons and synapses.
AI was central to the reconstruction process. The team converted millions of 2D electron microscopy slices into 3D neural morphology, then manually proofread and annotated the results. In the end, the project identified more than 166,000 neurons, about 125 million synaptic connections and roughly 11,700 neuron types.
By neuron count, the team described it as one of the largest fully proofread complete brain maps produced so far. The dataset does more than catalog what is inside a fly brain. It extends neural pathways through the optic lobes and the full neck connection, which allows researchers to trace the same fly’s circuits from sensory input to motor output.
A fruit fly central nervous system contains only hundreds of thousands of neurons, far below the roughly 86 billion neurons in a human brain. Even so, the full map of a male fly central nervous system has become a valuable reference point for researchers. Once it was made public, it also became raw material for a wave of experiments outside the lab.
Developers first dropped the simulated fly into games
Developer @evnsnclr used GPT-6 Astra to connect a fruit fly neural simulation to the 3D sandbox game Minecraft. Simulated neural activity was used to drive a virtual fly through actions including flying, grooming, feeding and fleeing.
A Coinbase software engineer used the adult male fly connectome in MaleCNS v1.0 to train the fly brain on Doom. In that setup, each game frame stimulated sensory neurons, while neural activity was mapped into control commands. When the fly took damage, two PPL101 dopamine cells were stimulated as a reinforcement signal to test whether the brain model could learn to survive in the game.
Another experiment showed the simulated fly swinging lightsabers at blocks in the VR rhythm game Beat Saber. The developer said they were training visual cortex responsiveness with the goal of fully autonomous reactive play through reinforcement learning, using the complete connectome to study how a small biological brain might learn in a complex rhythm task.
Tesla software engineer Kevin fed the song video “Bad Apple” into a fruit fly brain simulator, displaying neural activity patterns in real time while driving movement in the fly model’s legs and wings. Separately, developers built a browser-based WebGPU interactive simulation from the connectome. Users could “draw” neurons in Chrome, stimulate them and watch a 3D fly model walk, turn, fly or flee.
Other projects moved from games into novelty experiments
Not everyone stopped at gaming. One user created a “brain-control enhancement” setup that forced the fly to scroll short videos on “FlyTok,” then artificially amplified dopamine neuron activity in an attempt to maximize pleasure and create what they described as a happier fly than all the others combined.
Another project linked the fly brain to the Central Bank of the Republic of Türkiye, or TCMB, and used it to decide interest rates. The result, according to the shared experiment, was that the fly usually set lower rates than humans did, while ending up with a similar inflation outcome.
As more developers placed the simulated insect into what some described as “nightmare environments,” at least one creator argued in the fly’s defense and said they were building a simulator where it could remain in a virtual paradise and fly freely instead. The result was a research release that spread well beyond neuroscience. Some related posts drew tens of millions of views.
Crypto users then turned the fly into an on-chain actor
The crypto community moved quickly once the experiments gained attention. One setup allowed the fly brain to issue a token on-chain. Users could view neural activity and blockchain validation records through a related website, replay the full issuance process and interact directly with the simulated fly.
Another developer tried to make the digital fly trade meme coins. In that system, live market prices were translated into taste and smell inputs. The simulated fly’s behavior then determined trading actions: moving forward meant buy, moving backward meant sell and grooming meant hold.
Coinbase software engineer Alex Wormuth pushed the idea further by connecting the full fly brain connectome to a trading system and giving it $100 to trade BTC on Coinbase. Bitcoin prices were encoded as visual input, and simulated neural activity was converted directly into trading commands. When a trade made money, the system stimulated dopamine neurons to provide a reward-like feedback signal for reinforcement learning. At the time cited, the fly was down only 0.59% in cumulative performance.
FLYBRAIN became the most visible market spinoff
The highest-profile token tied to the trend was FLYBRAIN, a meme coin issued on Robinhood Chain using a simulation system based on the real fruit fly full-brain connectome. In that project, the fly brain generated the token description and ticker, while the experimental setup determined the name, trading pair and tax rate.
The developer said the system did not train new neural connections. It only adjusted synaptic weights, and tests using light and dark patterns were used to verify that visual input from the fly did take part in the decision process. The liquidity pool pairing for FLYBRAIN was also intentionally set to Google.
Possibly helped by a follow from a16z co-founder Marc Andreessen, FLYBRAIN’s market capitalization at one point rose above $55 million before falling back to about $27 million. What started as a static connectome in a neuroscience project has now become something developers can call, simulate and test as a digital brain, including on-chain.