Google’s Male Fruit Fly Connectome: Largest Brain Map Powers Doom, Mario 64 Experiments

The Fruit Fly Connectome Just Changed How We Study the Brain

Estimated reading time: 9 minutes

Key Takeaways

• In early September 2026, a team from Google Research and the Howard Hughes Medical Institute's Janelia Research Campus published a complete, fully proofread map of every neural connection in the brain and central nervous system of an adult male fruit fly.
• They call it the fruit fly connectome — specifically, MaleCNS v1.0.
• Within days of the data going public, developers had already used it to build a simulated fly brain that plays Doom. Then Super Mario 64. Then Minecraft.
• The dataset includes 166,000+ neurons and approximately 125 million synaptic connections.
• The full dataset is released under a CC-BY license, meaning anyone can use it.

Scientists Mapped Every Single Neuron in a Fruit Fly. Then Someone Made It Play Doom.

In early September 2026, a team from Google Research and the Howard Hughes Medical Institute's Janelia Research Campus dropped something that the neuroscience world had been working toward for a decade. They published a complete, fully proofread map of every neural connection in the brain and central nervous system of an adult male fruit fly.

They call it the fruit fly connectome — specifically, MaleCNS v1.0.

And within days of the data going public, developers had already used it to build a simulated fly brain that plays Doom. Then Super Mario 64. Then Minecraft.

This is not science fiction. This is what happens when a decade of cutting-edge research meets an open-access dataset and a community of developers with curiosity and fast laptops.

Let's break down exactly what was built, why it matters, and what comes next.

---

What Is a Connectome, and Why Is the Fruit Fly a Big Deal?

The Brain's Wiring Diagram

A connectome is essentially a complete wiring diagram of a nervous system. It shows every neuron — every single nerve cell — and every connection between them. Those connections are called synapses, and they are the physical pathways that allow signals to travel from one neuron to another.

Think of it like a city map, except instead of roads and intersections, you're looking at billions of microscopic biological wires.

For decades, building a connectome of any complex organism was considered nearly impossible. The brain is dense, three-dimensional, and extraordinarily fragile. To map it, researchers must slice it into layers thinner than a human hair, photograph every layer under an electron microscope, and then stitch all those two-dimensional images back into a three-dimensional model.

That takes years. And it requires enormous computing power.

Why the Fruit Fly?

The fruit fly — Drosophila melanogaster — is one of the most studied animals in biology. Scientists have used it for over a century to understand genetics, development, and behavior.

Its nervous system is small enough to be mappable with today's technology, but complex enough to be genuinely interesting. The adult male fruit fly has around 139 brain regions, and its central nervous system handles everything from vision to courtship behavior to motor control.

Mapping the full human brain's roughly 86 billion neurons is not currently possible. So researchers are working up the ladder, starting with smaller organisms and learning how to build better tools along the way.

The fruit fly is the perfect training ground.

a connectomics milestone

MaleCNS v1.0 — What the Dataset Actually Contains

The Numbers Are Staggering

The MaleCNS v1.0 dataset is the first finished connectome of an entire adult male Drosophila central nervous system. Not just the brain — the full package.

Here is what it includes:

• 166,000+ neurons mapped in full
• Approximately 125 million synaptic connections between those neurons
• The brain, the optic lobes (which handle visual processing), and the ventral nerve cord (which runs down the body and controls limb movement)
• 262 sex-specific cell types found only in males
• 114 sexually dimorphic cell types — cells that exist in both sexes but are structured differently depending on sex
• The sex-specific and dimorphic cells account for about 4.8% of the central brain

To give that scale some context: the team's previous benchmark — a 2020 map of half the female fly brain — contained only 25,000 neurons and 21 million connections. MaleCNS v1.0 is not an incremental improvement. It is a structural leap.

How They Built It

The reconstruction workflow is a combination of physical precision and AI-assisted analysis.

Here's how the process works:

1. Slicing: The central nervous system is cut into millions of ultra-thin sections — each one just nanometres thick.
2. Electron Microscopy: Every slice is imaged in high resolution.
3. 3D Reconstruction: A system called flood-filling networks (a type of AI) stitches those 2D images into 3D neuron models.
4. Human Proofreading: A system called PATHFINDER flags potential reconstruction errors, which human annotators then review and correct.
5. Visualization: Researchers can explore the finished map using a tool called Neuroglancer, and the full dataset is released under a CC-BY license, meaning anyone can use it.

The combination of AI automation and human quality control is what made a dataset of this size actually usable. Without the AI, the proofreading alone would take lifetimes.

male cns connectome

Why This Matters for Neuroscience and Biology

Comparing Males and Females at the Circuit Level

One of the most significant scientific contributions of MaleCNS v1.0 is that it allows researchers to directly compare male and female fly nervous systems at the level of individual neurons and connections.

Before this dataset, scientists could observe behavioral differences between male and female flies — courtship rituals, aggression patterns, responses to stress — but they could not trace those differences back to specific neural circuits with precision.

Now they can.

The 262 sex-specific cell types and 114 sexually dimorphic cell types identified in the male connectome give researchers a clear list of biological candidates. These are the exact circuits that may drive male-specific behaviors. Studying them could shed light on how sex differences in the nervous system emerge — and potentially how similar differences operate in more complex animals, including humans.

Repairing Damaged Neural Pathways

Google's stated goal for this research goes beyond curiosity. The team has described the project as a step toward understanding how nervous systems "perceive the world, react to stimuli, and how damaged neural pathways might one day be repaired."

That last part is important. Conditions like Parkinson's disease, spinal cord injuries, and certain forms of vision loss all involve damaged or disrupted neural connections. If scientists can study intact, healthy connectomes and understand exactly how those connections function, they build a foundation for developing therapies that restore or reroute broken pathways.

A complete, fully proofread connectome is not a treatment. But it is the kind of reference dataset that makes treatments possible to design.

A Template for Scaling Up

Google is already working beyond the fruit fly. The team has confirmed it is collaborating with researchers on connectomes for fish and mice. Each step up the complexity ladder requires better tools, more computing power, and refined AI reconstruction methods.

The MaleCNS v1.0 project is essentially a proof of concept. It demonstrates that a whole-CNS connectome is buildable and usable. That demonstration paves the way for tackling progressively larger nervous systems.

complete fruit fly connectome

connectome pursuing fly brain rewired

Stop Doing Everything Manually

Your business runs on decisions, actions, and follow-ups. Most of those happen the same way every time — which means they can be automated. At Flowgrammer, we help businesses remove the repetitive manual work from their operations so their teams can focus on what actually moves the needle. Whether it's lead management, client onboarding, internal notifications, or data routing, we build the systems that run while you sleep.

👉 See how automation can scale your business (automation can scale your business)

The Doom Experiment — A Simulated Fly Brain Plays a Video Game

Who Built It and How

Software engineer Alex Wormuth saw the MaleCNS v1.0 dataset and built something unusual: a simulation that uses the fly brain connectome as the decision-making engine inside a game of Doom.

Here's how the system works:

• Each frame of the game is fed into the simulation as a visual input
• That input stimulates sensory neurons in the simulated connectome
• The resulting neural activity patterns are translated into game control signals — move, turn, shoot
• When the in-game character takes damage, that event sends a signal to two specific neurons called PPL101 dopamine cells, functioning as a basic reinforcement signal

The project is currently being livestreamed online at the simulated fly brain navigate the classic FPS in real time, where anyone can watch the simulated fly brain navigate the classic FPS in real time.

What This Is and What It Is Not

It is important to be precise here. The simulated fly brain is not "understanding" Doom. It is not experiencing the game the way a player does. It is a control experiment — a way of testing whether the connectome data can be used to drive behavior in a simulated environment.

That said, the fact that it works at all is meaningful. It confirms that:

• The neural connection data in MaleCNS v1.0 is structurally usable as a functional model
• Reinforcement-style signals can be mapped onto the connectome's reward circuits
• The dataset is accessible enough for a developer working independently to build a working prototype quickly

Experts in the field have been clear that these game simulations are early-stage demos, not evidence of machine cognition. But as proof-of-concept work, they carry real weight.

make it run doom

Super Mario 64, Minecraft, and What Developers Are Building

The Mario 64 Project

Developer Jessica Paquette built a separate experiment using the same connectome data, this time targeting Super Mario 64.

In footage she posted publicly, Mario can be seen moving through the game — jumping, bumping into walls, exploring the environment in an unpolished but functional loop. The project's code is open source and posted on GitHub at https://github.com/ornata/fly.

In the repo description, Paquette noted that the code was "100% vibe coded with GPT Astra" — meaning she used an AI coding assistant to rapidly prototype the entire system.

This matters for a specific reason: it shows that the MaleCNS v1.0 dataset is not locked away behind academic paywalls or specialized software. A developer with access to modern AI tools and the connectome data can build a working simulation in a very short period of time. The barrier to entry for experimenting with connectomics just dropped significantly.

The Minecraft Simulation

A third developer took the connectome in yet another direction and ran the full MaleCNS v1.0 dataset inside Minecraft. In this version, the simulated neural activity from the connectome drives movement behavior within the game environment.

While less polished than the other two projects, the Minecraft experiment adds to the emerging pattern: this dataset works as a functional tool for prototyping neural-driven behavior simulations. Three independent developers, three different platforms, all working from the same open dataset within days of its release.

That kind of rapid adoption is a strong signal about the quality and accessibility of the data.

Connectomics as a Field — Where This Sits in the Bigger Picture

From Worms to Flies to Mice

Connectomics — the scientific discipline of mapping neural connections — has been building toward moments like this for decades.

The first completed connectome ever mapped was that of the C. elegans roundworm in 1986, a tiny organism with just 302 neurons. It took years of work and produced a landmark paper that is still cited regularly today.

Since then, the field has been pushing toward larger, more complex systems. A partial connectome of the fly larva was completed in 2023. The female half-brain map came out in 2020. Each step has added new tools, better AI reconstruction methods, and a deeper understanding of what connectome data can actually tell us.

MaleCNS v1.0 is the largest and most complete connectome ever published. It is not the endpoint — it is the current frontier.

AI as an Enabling Technology

One of the most important takeaways from the MaleCNS v1.0 project is how central AI has become to making connectomics work at scale.

The flood-filling networks that reconstruct neurons from electron microscope images, the PATHFINDER system that flags errors for human review, the Neuroglancer visualization tool — none of this existed in its current form ten years ago.

The AI didn't replace the scientists. It made the scale of what they were attempting actually achievable within human lifetimes. That relationship between AI assistance and human expertise is likely to be the template for every future connectomics project, including the fish and mouse maps that Google is already pursuing.

An Open Dataset Built for Reuse

The CC-BY license on MaleCNS v1.0 is not a footnote. It is a design decision. By releasing the dataset under an open license, the researchers have made it possible for anyone — academic labs, independent developers, students, and engineers — to build on this work without institutional gatekeeping.

The Doom, Mario, and Minecraft experiments happened because the data was available and usable. Future medical research, AI training experiments, and comparative biology studies will happen for the same reason.

Open data creates compounding value. The researchers who built this map likely could not have predicted all the uses it would be put to within a week of release. That is exactly the point.

FAQ