Science did something genuinely historic on September 3, 2026. Google Research, working alongside the Howard Hughes Medical Institute’s Janelia Research Campus, published the first complete map of an adult male fruit fly’s entire brain and central nervous system — 166,700 neurons and 125 million synaptic connections, fully traced and made available for free download. The result of nearly a decade of work and an amount of computational effort equivalent to roughly 44 years of human labor, it’s called the MaleCNS v1.0 connectome and it’s being described as the most significant neuroscience dataset ever released to the public.
The internet’s response? Teach it to play Doom within 72 hours. Then Beat Saber. Then Super Mario 64. Then Minecraft. Then, apparently out of guilt, build it a peaceful little garden paradise to retire to. This is where we are as a civilization and honestly it’s hard to argue with any of it.
What Google Actually Did — The Science Behind the Chaos
Before getting into the Doom part — which is coming, don’t worry — the actual scientific achievement here deserves to be understood because it’s genuinely remarkable. A connectome is essentially a complete wiring diagram of a brain: every neuron, every synapse, every connection between them, mapped in full detail. For context, science has only managed to fully map the connectomes of a few organisms before this. A tiny roundworm called C. elegans — with 302 neurons total — had its connectome completed back in 1986 and has been a foundational research tool ever since. A fruit fly brain with 166,700 neurons is a categorically different scale of achievement.
The way Google and Janelia’s team did it was painstaking. They physically sliced a male Drosophila melanogaster fruit fly’s central nervous system into millions of ultra-thin sections — the nervous system was cut into 66 major sections across the process — and scanned each section with high-powered electron microscopes to capture the structure at nanometer resolution. They then used AI algorithms, combined with years of manual verification by human researchers, to trace every micro-pathway through three-dimensional space and assemble the complete picture.
The resulting dataset was released publicly and made available for download. Google said this represents “the largest brain map by number of neurons to date” and that it “promises to be a foundational resource for neuroscience for years to come.” The goal is to give scientists an open-access blueprint for understanding how animal brains process sensory input, control movement, manage learning, and recover from damage — with implications that could eventually extend to understanding human neurological conditions.
It’s also worth noting the distinction from the previous female fly connectome that Google and Flywire (a Princeton University collaboration) had mapped earlier. The number of neurons differs between the two — 160,000-plus for the female versus 139,255 in the Flywire female dataset — and the male brain shows meaningful differences in how certain circuits are structured, particularly around behaviors like aggression and courtship. Those differences are, as you will discover shortly, directly relevant to how the fly brain performs in a first-person shooter.
Doom, Obviously — The First Gaming Experiment
Coinbase engineer Alex Wormuth announced on September 6 that he was “training a fly brain to play Doom using the full MaleCNS v1.0 fruit fly connectome.” He set up a public livestream so anyone could watch the digital fly’s Doom sessions in real time.
The technical setup is genuinely clever. By mapping each frame of the game directly into the fly’s virtual sensory neurons, Wormuth configured the resulting neural activity to trigger movement controls. Whenever the digital fly took damage in the game, the simulation sent an electrical pulse to two specific dopamine-releasing cells — the PPL101 neurons — to act as a biological reinforcement learning loop, rewarding survival and punishing mistakes. The idea is to replicate, in a rough simulation, the kind of reward-and-punishment feedback loop that governs real animal learning behavior.
The fly played thousands of rounds. It still wasn’t doing particularly well as of the livestream’s early sessions, which is both scientifically interesting and deeply relatable to anyone who’s ever watched a new player pick up Doom for the first time. The connectome is a snapshot of a brain’s wiring, not a trained entity — it’s the hardware without the software, in a sense. Getting it to develop any competency at a task through reinforcement learning is the actual experiment.
Beat Saber, Courtship Songs, and the Arousal Sentence of the Year
Developer domi on Bluesky took the fly brain in a different direction entirely: Beat Saber, the rhythm game where you slash colored tiles with lightsaber-style controllers. The implementation involved a little animated fly model waving virtual lightsabers around to the music and slicing through incoming tiles — which, based on video of the sessions, it was handling with a reasonable degree of coordination.
The technical approach domi used for the shooting-adjacent controls is where things get genuinely fascinating and also where we arrive at what is probably the best single sentence in gaming science journalism this week. The fly’s brain was used to determine when to turn, and its “courtship song” — essentially the neural activity associated with the fly’s mating call behavior — was mapped to when to shoot. Domi’s description of the result: “The more aroused the fly is the harder it tries to shoot.”
This is alarming, correct, scientifically coherent, and also somehow extremely funny. The courtship song in male Drosophila is a well-studied behavioral circuit — male fruit flies produce a specific wing vibration pattern to attract mates, and the neural pathways that drive it are among the better-understood circuits in the fly brain. Using the arousal level of those circuits as a proxy for shooting aggression in a video game is the kind of creative leap that probably only happens when you release a complete brain map to the entire internet simultaneously.
Super Mario 64, Minecraft, and the Full Gaming Roster
The experimentation didn’t stop at Doom and Beat Saber. The MaleCNS v1.0 dataset rapidly became what Tom’s Hardware described as irrefutable proof of Arthur C. Clarke’s Third Law: “Any sufficiently advanced new technology will immediately be forced to play Doom.” And then several other games as well.
Jessica Paquette, who described herself as having “pivoted to fly engineering in 2026,” built a framework using the 166,700-neuron model to pilot Mario through Super Mario 64. Video she posted showed Mario jumping and repeatedly bumping into walls as the simulated insect brain attempted to parse 3D platforming space. Anyone who has spent significant time with Super Mario 64 knows that even human brains with significantly more neurons find some of that camera control challenging, so this result seems defensible.
Developer Evan Sinclair Smith went even further, running the full MaleCNS v1.0 connectome — all 166,700 neurons and more than 25 million connections — inside Minecraft. In a viral demonstration, activity inside the simulated network controlled the movement of a fly within Minecraft. Minecraft events including food, light, nearby creatures, and attacks were converted into inputs for the neural model, with activity within selected groups of neurons then helping choose and adjust actions programmed by Smith, such as flying, grooming, feeding, or escaping. Smith noted the project was built with assistance from GPT-6 Astra.
The Internet Builds the Fly Brain a Heaven
Here’s where the story takes an unexpectedly wholesome turn. After watching multiple developers subject the simulated fly brain to virtual combat, danger, and the general chaos of video games for several days straight, X user Macroblock decided enough was enough.
The developer began building “fruit fly heaven” — an open environment filled with grass, trees, and fruit where the simulated fly can simply move around and feed rather than being repeatedly shot at, slashed through rhythm tiles, or driven into walls by Mario’s physics engine. The contrast with the doom-laden gaming experiments is stark and genuinely touching in a way that’s difficult to fully explain without sounding unhinged.
It’s worth being precise about what all of these experiments actually are, because “consciousness” language started circulating online and got ahead of the science fairly quickly. Developer Evan Sinclair Smith cautioned that the connectome is a snapshot of a dead brain, not a living entity; neither project should be confused with putting a real fly in a game. The MaleCNS v1.0 dataset is a structural map — it shows how neurons were connected in a specific fly at a specific moment. Running a simulation based on that map is not the same as running a fly’s consciousness, which is a meaningfully different thing even if the philosophical edges of that distinction get blurry quickly.
That said, the fact that developers have been able to use the biological neural architecture of an actual animal brain to produce recognizable game-relevant behaviors — turning, shooting, slashing, platform navigation — in different game contexts is not nothing. It’s a remarkable demonstration of both how much Google’s dataset captures and how quickly the open-source developer community can do interesting things with scientific data when it becomes freely available.
Why the Science Actually Matters Beyond the Games
Underneath all the Doom runs and lightsaber-swinging, the MaleCNS v1.0 connectome is a serious scientific tool. Researchers created it by slicing the nervous system into millions of extremely thin sections, imaging them with an electron microscope, and then using computers and AI to assemble the complete three-dimensional picture of every neural pathway. The result gives neuroscientists something they’ve never had before at this scale: a complete reference architecture for how a relatively complex animal brain processes information and generates behavior.
Fruit flies are one of the most important model organisms in biology precisely because their genetics and neural structures are well understood and relatively accessible. Drosophila melanogaster has contributed to research on genetics, development, neurology, and behavior for over a century. Having a complete connectome for a male adult fly means researchers can now trace specific behaviors — aggression, learning, courtship, memory formation, responses to damage — back to specific neural circuits, study how those circuits change under different conditions, and potentially identify equivalent structures in more complex brains including human ones.
The sex-specific differences between the male MaleCNS connectome and the earlier female fly datasets are particularly interesting for this kind of research, since many behavioral differences between male and female flies that scientists have studied for decades can now potentially be traced to specific circuit-level differences in the wiring. That’s the kind of insight that could take years of traditional research to establish and becomes accessible relatively quickly with a complete connectome as reference.
Google’s broader vision for the dataset, as described in its announcement, is as a “foundational resource for neuroscience for years to come.” Whether the researchers imagined that foundation being used to train a simulated fly to navigate the corridors of Doom using dopamine reinforcement learning within a week of release is, generously, unclear. But that’s what happened, and it’s hard to argue the result isn’t at least educational.
The MaleCNS v1.0 dataset remains publicly available for download. The fruit fly brain’s Doom sessions continue to be livestreamed. And somewhere in a digital garden filled with virtual grass and fruit, a simulated neural network based on the mapped architecture of an actual fly brain is peacefully exploring an environment specifically built for its comfort, which is either science or art or both. For more of this week’s most surreal gaming and tech crossover stories, check out everything happening with Rockstar’s GTA 6 security battle against drones and hackers and the full breakdown of the Sony digital ownership lawsuit — because apparently September 2026 is just that kind of month.
