Imagine creating a cute 3D character and later discovering that somebody has turned it into the body of a neural network reconstructed from a fruit fly.
“I never expected it to be used like this” is an entirely reasonable reaction.
A 3D avatar is normally made for games, streaming, VR, animation, or desktop characters. The jump from there to
“let us give a fly nervous system a body”
is enormous.
And this is not merely an internet joke. In 2026, researchers released a detailed connectome of the adult male fruit fly central nervous system, including the brain and ventral nerve cord. A connectome is, in simple terms, an enormous wiring diagram showing which neurons connect to which other neurons.[1][2]
Once that data became available, people could turn the wiring into computational neural models and connect those models to virtual bodies and games.
For some reason, one of those bodies became a beautiful anime-style girl.
It may never appear in a neuroscience textbook, but as a piece of 2026 internet history, it feels completely correct.
1. It already feels good when somebody uses your work; this use jumped into another universe
Creators often discover that people use their work in ways they never planned.
Innovation research has a word for existing products acquiring new functions: exaptation. A Management Science study of more than 3,000 IKEA hacks found that nearly half involved functionally novel uses rather than small modifications of the original function.[3]
The unusual part here is the distance of the jump.
This is not “I changed the outfit.”
It is not “I used the avatar in a stream.”
It is “I gave your avatar to a simulated fruit-fly nervous system as its body.”
For the artist, the work has not merely been consumed. It has entered a completely different field.
The work has acquired a second life.
That second life just happens to be “fly body,” which makes the story much better.
2. Nobody is plugging a USB cable into a living fly brain
This is the most important distinction.
When people say “a fruit-fly brain is controlling an avatar,” it can sound as if a living fly is restrained in a laboratory, electrodes are inserted into its brain, and its real-time neural activity moves the character.
That is not what this particular wave of experiments is about.
The 2026 MaleCNS resource is structural data reconstructed from an actual adult male Drosophila central nervous system using electron microscopy. The published dataset contains 166,691 fully proofread and annotated neurons and 11,691 types.[1][2]
Very roughly, the pipeline is:
real nervous tissue → fixation and staining → electron-microscope imaging → AI and human reconstruction → digital wiring diagram.
The biological animal is not still alive inside the computer.
It is more accurate to say that researchers and developers are trying to make the wiring read from a real nervous system run again as a computational model.
3. So is this a “digitized fly”?
As science-fiction shorthand, that is not terrible.
A biological nervous system is mapped, represented digitally, simulated, and connected to a virtual body. Visually, it looks like digitization.
Scientifically, however, one giant warning label is required:
a connectome is not the whole brain.
A connectome primarily records structural connections. A real nervous system also depends on changing states and mechanisms such as synaptic efficacy, excitation and inhibition, neurotransmitters and receptors, ion channels, neuromodulators, plasticity, moment-to-moment electrical activity, glial effects, hormones, metabolism, and sensory feedback from the body.
A complete road map does not tell you which traffic lights are red right now, where the traffic jam is, or where every driver intends to go.
The wiring diagram is extremely important, but it does not automatically preserve everything that a particular brain was doing at a particular moment.
4. The disturbing part is how much you can get from the wiring anyway
Does that mean a connectome is just a gigantic anatomical atlas?
No.
A 2024 Nature study built a simplified leaky integrate-and-fire (LIF) model from the Drosophila whole-brain connectome and neurotransmitter information.[4]
LIF is nowhere near a molecularly complete neuron. It is a compact model in which inputs accumulate, the unit fires when a threshold is crossed, and then it resets.
Yet stimulating sugar- or water-sensing neurons in the model predicted neural activity involved in feeding, and some of those predictions were validated in biological experiments.[4]
So both statements can be true:
“this is not a full reconstruction of a real brain”
and
“the wiring already contains enough structure to predict meaningful functions.”
That is the fascinating part.
We are far from copying a mind, yet a simplified model of the wiring can already reveal pieces of sensorimotor behavior.
5. Why can’t the beautiful avatar simply learn to walk after twelve hours?
The viral experiment reported more than twelve hours of training with little sign that the avatar had adapted, and later reported that it could still barely move forward.
It is tempting to joke:
“The fly studied for twelve hours and still cannot walk.”
But that interpretation is too strong.
A fruit-fly nervous system evolved together with a fly body: fly legs, joints, muscles, sensory organs, body mass, friction, wings, and geometry.
Now imagine suddenly telling that system:
“Your body is a human-shaped anime girl now. The joints, lengths, masses, and control relationships are different. Good luck.”
That is not a normal learning problem. It is a catastrophic embodiment mismatch.
There is another uncertainty: the public posts do not establish exactly what “training” changes in the implementation. It may involve parameters inside the neural model, a readout layer, a controller that maps neural outputs to joints, reinforcement learning outside the connectome, or some combination.
Therefore the careful conclusion is simply that this implementation has not yet found an effective way to adapt connectome-derived signals to the new body.
If the fly model had begun walking naturally as a beautiful girl after twelve hours, that might have been the scarier result.
6. Why are “fruit-fly brain controls X” projects suddenly everywhere?
There is a real reason they feel more common now.
In 2024, researchers published a whole adult female Drosophila brain connectome with 139,255 neurons and 54.5 million synapses.[5]
In 2026, the MaleCNS project extended the frontier to the complete male central nervous system, including both brain and ventral nerve cord. Google Research described it as the largest brain map to date.[1]
Crucially, the data are not locked inside a paper.
They are available for computational use.
That means engineers can touch them too.
Someone connects the model to a game.
Someone gives it a 3D fly body.
Someone puts it into a virtual environment.
And eventually somebody gives it an anime girl.
Open data distributes research material not only to scientists, but also to people who have extremely strange ideas.
That is how uses appear that were never in the original grant proposal.
7. The bigger lesson is not “we recreated a fly,” but “even a fly is unbelievably difficult”
At first the story sounds like:
“Wow, we can put a fly brain in a computer now.”
Then you look more closely and the feeling reverses:
“Wait, even a fly is this hard?”
A fruit fly has a tiny nervous system compared with a human, yet it still contains well over one hundred thousand neurons and tens of millions of synapses. After mapping the wiring, researchers still have to determine how structure becomes realistic dynamics and behavior.[5][4]
A brain is not merely a bundle of cables.
It is a dynamic system of
structure × electrical activity × chemistry × plasticity × body × environment × time.
The fact that “copy the wiring and press Run” does not solve the problem is itself evidence of how much is happening in a nervous system.
8. Could we eventually do the same thing with a human brain?
In a limited sense, that work has already begun.
Google Research and Harvard collaborators reconstructed roughly one cubic millimeter of human temporal cortex at nanoscale resolution.[6]
That tiny sample contained about 57,000 cells, including about 16,000 neurons, and roughly 150 million synapses. The dataset occupied about 1.4 petabytes.[6]
One cubic millimeter sounds trivial.
It generated 1.4 million gigabytes.
Google Research notes that H01 represents only about one-millionth of the volume of the entire human brain. Scaling current high-resolution methods to a whole human brain could require as much as a zettabyte—one billion terabytes—of data acquisition and analysis, beyond current technological capabilities.[7]
So it is difficult to claim that a human connectome is forbidden by physics.
But it is equally wrong to say:
“We did a fly, so now we just scan a human.”
The scale gap is enormous.
9. The harder questions begin after the wiring map is complete
Suppose a future machine somehow produces a complete human connectome.
The problem is still not finished.
What neuron model should run on that wiring?
How precisely must synaptic strengths be preserved?
Do we need receptor distributions and ion channels?
What about neuromodulation, glia, gene expression, metabolism, and hormones?
Must we preserve the electrical state at the instant of scanning?
Does the model need a body?
If the body changes, is the resulting system still the same person?
And then comes the nastiest question:
when the simulation starts, is it “you”?
Being able to copy a connectome is not the same as proving that personality, memory, or subjective consciousness has been copied.
Current science has no experiment or accepted theory that guarantees that jump.
“Brain as data” and “digital immortality” remain very far apart.
10. The final lesson may simply be that brains are ridiculous
The story eventually stops being about how impressive computers are and becomes a story about how absurdly complicated brains are.
Humans thought:
“Maybe if we read every wire, we will basically understand it.”
So we sliced tissue, imaged it, used AI to trace neurons, and built gigantic graphs.
We actually reached the point where detailed wiring maps are possible.
And what we found was roughly:
“The wiring diagram alone is gigantic, and there are still dynamic layers on top.”
This is true even for a fly.
A human brain is many orders of magnitude larger.
As the science improves, the conclusion is not necessarily “the brain turned out to be simple.”
It is increasingly:
“How many layers do you have?”
11. Back to the beautiful avatar
That is why the original story feels so satisfying.
Scientists spent years using electron microscopy, AI, and neuroanatomy to map a fruit fly nervous system in extraordinary detail.
The data became public.
Someone made it run in a computer.
Someone else connected it to a virtual body.
Then the original avatar artist looked online and discovered:
their beautiful character had become a fly’s body.
“I never expected my work to be used like this.”
Of course not.
But if the unexpected use is harmless, technically fascinating, and funny enough to make the creator laugh, it may be one of the best possible second lives for a creative work.
The work leaves the creator’s hands and enters a world the creator could never have planned.
Here, user innovation and open science collided at exactly one point, producing:
a beautiful anime avatar powered by a fruit-fly connectome.
The future arrived earlier than expected.
It just came from the wrong direction.
Sources
- Google Research, A connectomics milestone: Mapping the complete male fruit fly brain, 2026-09-03 research.google
- Berg et al., Sexual dimorphism in the complete connectome of the Drosophila male central nervous system, Cell, 2026 research.google
- Chan & Lim, The Emergence of Novel Product Uses: An Investigation of Exaptations in IKEA Hacks, Management Science pubsonline.informs.org
- Shiu et al., A Drosophila computational brain model reveals sensorimotor processing, Nature, 2024 nature.com
- Dorkenwald et al., Neuronal wiring diagram of an adult brain, Nature, 2024 nature.com
- Shapson-Coe et al., A petavoxel fragment of human cerebral cortex reconstructed at nanoscale resolution, Science, 2024 research.google
- Google Research, Google Research embarks on effort to map a mouse brain research.google
