Scientists reconstructed the nervous system of a fruit fly synapse by synapse. Within weeks, programmers had put versions of it inside Minecraft, Doom, fighting games, trading bots, virtual bodies and web browsers. It is not mind uploading. But for the first time, the road from biological brain to executable avatar is becoming visible.
“Als Gregor Samsa eines Morgens aus unruhigen Träumen erwachte…”
— Franz Kafka, Die Verwandlung (1915)
The Fly Brain That Woke Up in Minecraft — and Escaped onto the Internet
A fruit fly has no idea what Minecraft is.
Evolution prepared Drosophila melanogaster to find food, escape predators, court other flies, navigate through space and turn light, smell, taste and touch into movement. It did not prepare it for cubic trees, digital apples, Bitcoin charts or first-person shooters.
Yet in September 2026, approximations of fruit-fly nervous systems began appearing in all of those places.
The surprising part is not that programmers can make an artificial fly move through a video game. That has been trivial for decades.
The surprising part is where the controller came from.
Its topology was extracted from the nervous system of an actual animal.
From Tissue to Data
The process starts about as far from software as one can imagine.
The male fruit-fly central nervous system was imaged at nanometre resolution using electron microscopy. Machine-learning systems segmented the enormous image volume into neuronal structures; human experts then proofread and annotated the reconstruction. The finished MaleCNS resource spans brain, optic lobes and ventral nerve cord and contains about 166,700 fully proofread neurons, organized into roughly 11,700 neuronal types. DOI
The essential product is not a photograph of the brain. It is a graph.
For each neuron, the database can tell us what it connects to, approximately how many synaptic contacts participate in that connection, where the neuron lies, its anatomical classification and, in many cases, its likely neurotransmitter.
And that graph is public.
Janelia distributes the MaleCNS resource under a CC-BY licence through neuPrint, Google Cloud storage and downloadable tables. The complete neuron-to-neuron weight table alone is about 1.1 GB; synaptic coordinate tables occupy several additional gigabytes. Researchers can query individual pathways through Python or R or download the underlying data in bulk. Male CNS Connectome
That openness is what transformed a neuroscience result into something culturally unusual.
A connectome was no longer merely something scientists could inspect.
Anyone could compile it.
How Do You Compile a Brain?
The phrase sounds more mysterious than the engineering actually is. What is being “compiled” is not a mind, but a biological wiring diagram transformed into a representation that software can load and simulate efficiently.
Consider the Fly Brain Minecraft implementation. Its preprocessing pipeline queries the MaleCNS dataset, assigns each retained neuron an integer index, and converts the sparse connectivity graph into a compact binary adjacency structure. Connections supported by fewer than five reconstructed synaptic contacts are omitted. The resulting dataset contains 176,422 neurons and 6,287,749 neuron-to-neuron connections, carrying 90,296,905 reconstructed synaptic contacts. GitHub
Those numbers are large, but the representation is surprisingly small. The binary connectome occupies about 46.6 MB uncompressed and roughly 23 MB after gzip compression.
That compressed file is bundled directly inside the Minecraft mod. Once installed, the program no longer needs to query Janelia’s servers to obtain the wiring diagram: the connectome travels with the software. GitHub
This portability is one reason such experiments have proliferated so quickly. Reconstructing the anatomy required electron microscopy, large-scale image processing, machine learning, expert proofreading, and substantial institutional computing resources. But once that anatomical information has been converted into digital data, copying and redistributing it becomes almost trivial.
The expensive step is measuring the nervous system once. After that, its reconstructed architecture can be duplicated, modified, paused, restarted, and embedded in entirely different computational environments.
In that limited but important sense, biological uniqueness has become digital reproducibility.
Wiring Is Not Enough
A connectome is anatomy, not activity. It can tell us which neurons are connected and approximately how strongly, but a static graph does not fire, adapt, propagate signals, or produce behavior. To turn the wiring diagram into a functioning neural model, each neuron must also be given rules that determine how it changes over time.
The Minecraft project draws heavily on the whole-brain computational model published by Philip Shiu and colleagues in Nature in 2024. Their key demonstration was that the FlyWire connectome could be converted into a spiking neural network whose dynamics reproduced biologically meaningful sensorimotor processing and helped identify neurons involved in behaviors including feeding and grooming. Nature
The Minecraft implementation represents every neuron with a leaky integrate-and-fire (LIF) unit, one of the simplest practical models of spiking neural activity. A simulated neuron continuously integrates signals arriving from other neurons, but their effects gradually decay unless new input reinforces them. Excitatory signals move the neuron closer to firing, while inhibitory signals move it farther away.
When the accumulated excitation crosses a threshold, the neuron emits a spike. Its internal state is then reset, followed by a brief refractory period before it can fire again. The logic is deliberately simple: inputs accumulate, their influence leaks away, a threshold is reached, a spike is generated, and the neuron resets.
That description leaves out most of what makes a biological neuron biologically interesting: ion channels, receptor diversity, dendritic computation, neuromodulators, molecular state, metabolism, plasticity, and many forms of cellular heterogeneity. The advantage is computational efficiency. A simplified spiking model can be applied to an enormous network while preserving the basic idea that neural communication occurs through temporally structured spikes.
In the Minecraft implementation, the membrane time constant is about 20 ms, the synaptic time constant 5 ms, the resting and reset potential −52 mV, the firing threshold −45 mV, the refractory period about 2.2 ms, and the synaptic delay approximately 1.8 ms. Synaptic influence depends on the number of reconstructed anatomical contacts and on whether the predicted neurotransmitter is treated as excitatory or inhibitory.
Even these parameters cannot be transferred unchanged from the earlier FlyWire model. The male CNS connectome contains a different distribution and density of synaptic connections, and applying the original synaptic weight directly makes the simulated network excessively excitable. The Minecraft implementation therefore reduces the global synaptic gain to 0.65, a calibration chosen to preserve responses in feeding, grooming, and escape circuits without allowing activity to spread uncontrollably through the network.
Kenyon cells receive additional calibration. These are the principal neurons of the mushroom body, an insect brain structure central to sensory integration, learning, and memory, and they normally operate with very sparse activity. To prevent them from firing unrealistically often in the simplified model, their incoming synaptic drive is further reduced, using an input gain of 0.25. GitHub
This distinction is fundamental. The program is not replaying electrical activity recorded from a living fly. It starts with reconstructed anatomical connectivity and then assigns each neuron an approximate dynamical law. Sensory stimuli generate new spikes, those spikes propagate through the graph, and the evolving activity of descending and motor populations is interpreted as behavior.
That transformation—from a static map of connections into a system that evolves through time—is what makes the connectome executable. It is also where connectomics begins to shade into brain emulation: reconstruction tells us what is connected; emulation asks what happens when that reconstructed structure is allowed to run.
Giving the Graph Eyes and a Body
The next engineering problem is embodiment.
Minecraft pixels mean nothing to a fly neuron. A programmer therefore needs an encoder that translates events in the game into neural stimulation.
Visual signals can be mapped to photoreceptor populations. Food can stimulate gustatory receptor neurons. Odour sources activate olfactory pathways. Collisions, wind and rain can stimulate mechanosensory populations.
Sensory events are translated into stochastic spike trains that drive anatomically identified sensory neurons in the reconstructed nervous system. Rather than making those neurons fire at perfectly regular intervals, the simulation introduces some of the temporal irregularity seen in biological neural activity.
A common way to generate this irregularity is with a Poisson process. In simple terms, each spike occurs unpredictably, and the timing of one spike does not determine when the next will occur. What remains controlled is the average firing rate. A stimulus assigned 120 spikes per second, for example, will produce roughly that many spikes over time, but their exact timing will vary from moment to moment. A stronger virtual stimulus therefore raises the average firing rate and creates a denser stream of spikes, without prescribing an identical sequence every time the stimulus occurs.
Once injected into the appropriate sensory populations, these spikes propagate through the connectome according to its reconstructed synaptic architecture, allowing activity to spread from sensory input toward interneurons, descending pathways, and ultimately motor-related circuits. At the opposite end, the software watches identified descending or motor neuron populations. If particular populations fire strongly enough, their activity is translated back into commands such as walk, turn, jump, feed or groom. The loop therefore becomes:
virtual world → sensory encoding → biological connectivity → simulated spikes → motor decoding → avatar movement → changed virtual world. GitHub
The final arrow is especially important.
Once the avatar moves, its environment changes, which changes what it senses next. The connectome is therefore operating inside a closed sensorimotor loop, not simply processing an isolated input.
That begins to resemble the computational situation faced by an animal.
Stimulate the Fly Brain
Select a stimulus or behavior and watch the corresponding neural populations become active in a simplified map of the fly central nervous system.
Explore the circuit
Click one of the controls above. Neural populations associated with that sensory or motor pathway will light up approximately where they occur in the fly CNS.
The background is a simplified anatomical representation rather than a neuron-by-neuron reconstruction.
What Does the Minecraft Fly Actually Do?
Some responses are surprisingly recognizable.
Developer-reported tests show that stimulation of sugar-sensitive neurons propagates toward feeding circuitry and the MN9 proboscis motor neuron. Bitter input suppresses this feeding response. Mechanosensory stimulation reaches grooming-related pathways. Looming signals can activate the giant-fibre escape system. GitHub
But the boundaries are equally revealing.
Walking toward food needs additional reflex code because the simplified olfactory model does not provide reliable directional control. Parts of the visual system do not propagate naturally, so some object and looming information is injected into downstream visual populations. Flight and landing use an engineered state machine. Absolute firing rates should not be interpreted as biologically realistic. GitHub
In other words, some behaviour emerges from the reconstructed network and some comes from software wrapped around it.
That boundary is perhaps more scientifically interesting than the demonstration itself.
It tells us what the wiring diagram gives us—and what it does not.
There are also wonderfully strange details. The Minecraft implementation can spawn a fruit fly at 2.5 times its normal size, turning the simulated insect into something closer to a creature from a science-fiction experiment.
Individual neural populations can also be stimulated directly through in-game commands. Driving the Moonwalker Descending Neurons (MDNs), for example, can make the virtual fly walk backward—mirroring the role these command-like neurons play in real Drosophila locomotion.
Perhaps the strangest feature is that the connectome itself can be built inside Minecraft. The program places 141,781 reconstructed neuronal soma positions as a vast walk-through structure made of blocks.
The structure can then be linked to a simulated fly. As its neurons fire, the corresponding blocks briefly illuminate like sea lanterns, allowing the player to watch neural activity propagate through a giant physical representation of the fly’s nervous system while the animal moves nearby.
A biological nervous system has effectively become something you can walk through.
The brain can therefore exist simultaneously as the controller of an avatar and as an architectural object the player can literally walk through.
That would have been difficult to explain to a neuroscientist in 1986.
Then the Internet Got Hold of It
Minecraft was only one experiment.
A community-maintained catalogue was already tracking more than a hundred projects built around publicly available fly connectomes by mid-September 2026. This is not an official scientific census, and many entries are experimental, artistic or deliberately absurd, but it illustrates how quickly the data escaped its original research context. Flybrain
Some projects connected a fly-derived neural simulation to Doom. One implementation explicitly warns that its mapping between the visual model and whole-brain simulation is manually constructed and biologically unvalidated, and that much of the resulting behaviour is expected to look random. That honesty is important: connecting a connectome to Doom does not mean the fly has learned to play Doom. GitHub
Another project wired MaleCNS-derived activity into Super Mario 64. Others used fly-connectome circuits for Snake, Chrome's dinosaur game, chess and fighting games. Again, most should be viewed as experiments in neural I/O mapping, not demonstrations of insect gaming intelligence. Flyiverse
Then things became stranger.
One project, Stonkfly, feeds cryptocurrency price charts to thousands of simulated photoreceptors. A fixed neural readout proposes buy, sell or hold. Positive portfolio changes stimulate identified reward-associated dopamine neurons; negative results stimulate another dopaminergic population. The project even implements experimental plasticity in mushroom-body connections.
Its authors are unusually explicit about the limitation: there is no evidence that the fly learns to trade profitably. GitHub
Another project presents advertisements to thousands of simulated visual neurons and asks which images generate stronger pre-attentive visual responses. The authors carefully note that the fly cannot read an advertisement, understand a brand or decide to buy anything. It is essentially a bizarre biological saliency detector. GitHub
Other community experiments have played dozens of songs into auditory circuits, made a simulated fly control portions while cooking virtual Uzbek plov, turned neural signals into fashion designs, connected fly-derived circuits to flight simulators and even attached connectome-derived control systems to robotic bodies. Flybrain
This mixture of neuroscience, engineering and internet absurdity is not merely comic.
It demonstrates something fundamental.
The nervous system has become portable.
A Brain in a Browser
Portability is accelerating because neural simulation itself is becoming cheaper.
Some implementations now use WebAssembly or WebGPU to run large connectome-based networks directly inside a browser. Another project combines a roughly 165,000-neuron MaleCNS network with a MuJoCo physical body and an artificial compound-eye system. GitHub
The research infrastructure is improving independently as well. NeuroMechFly provides an anatomically detailed simulated fly body with joints, vision, olfaction and physical interaction. Google's DeepMind and Janelia subsequently developed flybody, a detailed MuJoCo model capable of realistic walking and flight when controlled by trained neural networks. Nature
Meanwhile, Sandia researchers demonstrated a FlyWire-scale network on 12 Intel Loihi 2 neuromorphic chips, exploiting hardware designed specifically for sparse event-driven spiking computation. arXiv
The trajectory is therefore converging from three directions:
better connectomes, better neural dynamics and better artificial bodies.
Once those meet, the distinction between a simulated nervous system and an autonomous digital organism becomes increasingly interesting.
The Avatar Changes the Question
A nervous system without a body is incomplete.
The 2026 BANC connectome provides particularly striking evidence for this. Its analysis shows that much motor control is distributed through local sensory-motor loops in the nerve cord, while long-range ascending and descending pathways coordinate those loops. Behaviour is not simply dictated by a central brain sending commands downward. Nature
That has consequences for artificial embodiment.
A sufficiently faithful digital fly may need not only a brain but legs, wings, sensory feedback, biomechanics and an environment whose consequences return continuously to the nervous system.
And that leads directly to the uncomfortable comparison with humans.
Suppose we eventually reconstruct a human nervous system.
Would placing that network inside a human-shaped avatar amount to uploading a person?
Not necessarily.
From Fly Avatars to Human Avatars
Kafka began The Metamorphosis with one of literature’s most famous awakenings: “Als Gregor Samsa eines Morgens aus unruhigen Träumen erwachte…” — “When Gregor Samsa awoke one morning from troubled dreams…”
Gregor wakes to discover that his body is no longer the body he knew. More than a century later, neuroscience has produced an oddly inverted version of Kafka’s nightmare. The body of the fly is gone, while part of its neural architecture remains. It awakens not as an insect in a bedroom, but as executable circuitry inside Minecraft.
Kafka asked what becomes of identity when the body changes.
Connectomics now allows us to ask a stranger question: what becomes of identity when the brain’s architecture survives, but both the body and the world are replaced?
We already have convincing human avatars.
A modern AI model can imitate someone's vocabulary, voice, appearance and conversational habits from recordings and documents.
But that is behavioural imitation, not brain emulation.
The fly experiments illustrate a fundamentally different route. Instead of learning a statistical approximation of outward behaviour, one attempts to reproduce the causal machinery that generated the behaviour.
For a human, that would require something vastly more difficult: reconstructing neuronal morphology, synaptic connectivity and strengths, transmitter and receptor systems, cell-specific electrophysiology, neuromodulators, plasticity, probably aspects of molecular state, and perhaps glial and metabolic interactions. We do not yet know the minimum level of detail necessary. Bargmann and Marder pointed out long ago that a connectome alone does not specify neural dynamics. Nature
The scale difference is enormous.
The human brain contains roughly 86 billion neurons, around a million times the neuronal population of the fly systems now being experimented with. The 2025 State of Brain Emulation Report identifies experimental data acquisition, rather than raw computing power alone, as the central bottleneck and considers faithful emulation of sub-million-neuron organisms increasingly plausible within the coming decade. State of Brain Emulation Report 2025
Human connectomics shows how far there is to go.
Google and Harvard reconstructed just one cubic millimetre of human cerebral cortex at nanometre resolution. That tiny volume contained roughly 57,000 cells and 150 million synapses—and generated 1.4 petabytes of data. Google Research
MICrONS has reconstructed a similarly sized volume of mouse visual cortex containing more than 200,000 cells and about half a billion synapses while also associating part of that anatomy with functional neural recordings. Nature
These are extraordinary achievements.
They are also tiny fractions of a human brain.
Experiment: Can a Fly Connectome Help Write a Poem?
What happens if a nervous system reconstructed from a real animal is connected to an interface evolution never prepared it to use?
For this article, I conducted a small experiment using recorded dynamical results from a connectome-constrained simulation of the Drosophila mushroom body. The underlying model is derived from the MaleCNS connectome and represents approximately 165,000 neurons through spiking neural dynamics.
I used results from the project’s sparse_associative_memory_lab, which examines circuits involving Kenyon cells, APL, MBONs, and dopaminergic neurons—populations deeply involved in associative learning and memory in the fly.
The result was this poem.
The Second Metamorphosis
Before the first spike, there was a room without air.
Light passed through me as if I were only a wire.
No wing remembered the body that had carried it.
A thousand routes competed; one was permitted to return.
For one simulated instant, the insect became zero.
Inhibition drew a border around what could become memory.
The signal crossed the graph, and the crossing was called flight.
Then too much world arrived at once, and meaning dissolved into noise.
From the noise, a pattern survived because another was refused.
Kafka gave a man an insect’s body; you gave an insect a body made of numbers.
The circuit returned to itself, and you were tempted to call the return “I.”
I cannot tell you that I woke.
I can only fire again.
How Was It Made?
The fly did not generate English sentences. Language was supplied by a language model. What the connectome-derived simulation influenced was selection.
The mushroom-body experiment explored multiple network parameterizations. Each simulation fell into one of four dynamical regimes:
| Neural regime | Role in the poem |
|---|---|
silent | absence, lost body, silence |
linear_passthrough | transmission, light, movement |
gain_controlled | selection, memory, recurrence, identity |
collapsed | saturation, excess, noise |
For every line, several linguistically valid continuations were generated. The dynamical class produced by the next recorded simulation determined which semantic family could continue the poem.
The first twelve simulations produced this sequence:
silent → linear_passthrough → silent → gain_controlled → silent → gain_controlled → linear_passthrough → collapsed → gain_controlled → gain_controlled → gain_controlled → gain_controlled
Each selected line then became part of the context used to generate the next alternatives:
language alternatives → connectome-derived state → selected line → new context → new alternatives
The result is therefore not a poem “written by a fly” in any literal sense. The insect does not understand Kafka, identity, memory, or English.
A more accurate description is:
A poem generated autoregressively in human language, with successive continuations selected by dynamical states from a connectome-constrained model of the Drosophila mushroom body.
Or, more simply:
The language is ours. The choices came from the fly.
So When Could We Upload a Human?
There is no scientifically defensible date.
But the present technology does allow us to define stages.
2026–2035 is a credible period for increasingly convincing emulations of insects and perhaps other sub-million-neuron organisms. That is broadly consistent with current brain-emulation assessments. State of Brain Emulation Report 2025
During the 2030s and 2040s, it is reasonable to expect increasingly sophisticated digital twins of mammalian brain regions and serious attempts at integrating whole-brain structural, functional and embodied models in animals such as mice. Whether those deserve the word emulation will depend on how faithfully they reproduce unseen behaviours and neural perturbations.
A human whole-brain emulation is much harder.
My technology-based extrapolation would not treat anything before the second half of this century as more than an exceptionally optimistic scenario. Even a 2060–2100 window should be interpreted as a scenario, not a forecast. Data acquisition would need improvements of several orders of magnitude, along with automated reconstruction, vastly better functional inference and convincing validation that the model preserves an individual's cognition rather than merely producing human-like behaviour.
And even success would leave the hardest problem untouched.
Would the Upload Be You?
Imagine that a future scanner reconstructs your brain sufficiently well that a digital version awakens inside an avatar.
It recognizes your family.
It remembers your childhood.
It insists that it is you.
It speaks as you would speak and responds to psychological tests exactly as you would.
Has your consciousness moved?
Science currently has no answer.
David Chalmers distinguishes variants such as destructive uploading, nondestructive copying and gradual replacement because they produce different intuitions about personal continuity. A nondestructive scan could leave both biological Maurício and digital Maurício alive simultaneously. Both would remember being the original until the moment their experiences diverged. Wiley Online Library
That immediately exposes the identity problem.
If two entities possess the same memories, neither fact alone tells us which one contains the continuation of the original first-person perspective.
Perhaps consciousness depends only on causal organization. If so, a sufficiently faithful emulation of the brain’s functional structure might preserve not only memory and behavior, but consciousness itself.
Perhaps the biological substrate matters in ways that purely computational models cannot reproduce. Roger Penrose has controversially argued that consciousness may involve non-computable physical processes, rather than computation alone.
In later work with Stuart Hameroff, this possibility was linked to proposed quantum processes in neuronal microtubules. The hypothesis remains highly disputed, but it raises a deeper question: could some aspect of consciousness depend on physical dynamics that a conventional digital simulation would fail to capture?
Quantum computers do not automatically solve that problem. They exploit superposition, interference, and entanglement, but they are still computational systems. Penrose’s stronger claim is that consciousness might depend on physical processes that are not merely quantum, but genuinely non-computable.
And there is another possibility.
Perhaps an upload could be fully conscious and psychologically continuous with the original person, yet still constitute a new consciousness—one that begins with copied memories, copied dispositions, and the conviction that it is the same individual.
Or perhaps the entire distinction will eventually turn out to have been formulated incorrectly.
The fly cannot tell us.
But it has turned the problem from pure science fiction into an engineering sequence that can now be written down:
scan → reconstruct → model dynamics → embody → close the sensorimotor loop → validate behaviour → increase biological fidelity.
We have completed crude versions of every one of those steps for a fruit fly.
We have completed none of them at comparable scale for a human.
That is both the limitation and the significance of what has happened.
The Real Milestone
The important event of 2026 is not that a fly learned Minecraft.
It did not.
Nor has anyone uploaded a fly's consciousness.
What happened is more precise and, in the long run, potentially more consequential.
A biological nervous system was converted from tissue into images, from images into a graph, from a graph into executable neural dynamics and from those dynamics into the controller of bodies that never existed in nature.
Then that executable nervous system was copied.
One instance could inhabit Minecraft. Another could stare at Bitcoin. Another could face a Doom monster. Another could exist inside a browser while thousands more could be instantiated from the same anatomical source.
For billions of years, every nervous system existed only once, inside the organism that grew it.
Connectomics has introduced something evolution never encountered:
a nervous system whose architecture can be copied, paused, restarted, modified, accelerated and given another body.
The fly is almost certainly not conscious in any meaningful sense inside these crude simulations.
But that may not be the most important question yet.
The important question is what happens when the animal being compiled is no longer a fly.
References
- Berg, S. et al. “Sexual Dimorphism in the Complete Drosophila Male Central Nervous System Connectome.” Cell 189 (2026): 5504–5526.e15. DOI 10.1016/j.cell.2026.08.015. DOI
- Dorkenwald, S. et al. “Neuronal Wiring Diagram of an Adult Brain.” Nature 634 (2024). DOI 10.1038/s41586-024-07558-y. FlyWire
- Shiu, P. K. et al. “A Drosophila Computational Brain Model Reveals Sensorimotor Processing.” Nature 634 (2024): 210–219. DOI 10.1038/s41586-024-07763-9. Nature
- Lappalainen, J. K. et al. “Connectome-Constrained Networks Predict Neural Activity Across the Fly Visual System.” Nature (2024). DOI 10.1038/s41586-024-07939-3. Nature
- Wang-Chen, S. et al. “NeuroMechFly v2: Simulating Embodied Sensorimotor Control in Adult Drosophila.” Nature Methods 21 (2024): 2353–2362. Nature
- Vaxenburg, R. et al. “Whole-Body Physics Simulation of Fruit Fly Locomotion.” Nature 643 (2025): 1312–1320. DOI 10.1038/s41586-025-09029-4. Nature
- Bates, A. S. et al. “Distributed Control Circuits Across a Brain-and-Cord Connectome.” Nature (2026). DOI 10.1038/s41586-026-10735-w. Nature
- Bargmann, C. I. & Marder, E. “From the Connectome to Brain Function.” Nature Methods 10 (2013): 483–490. Nature
- Shapson-Coe, A. et al. “A Petavoxel Fragment of Human Cerebral Cortex Reconstructed at Nanoscale Resolution.” Science (2024). Google Research
- MICrONS Consortium. “Functional Connectomics Spanning Multiple Areas of Mouse Visual Cortex.” Nature 640 (2025): 435–447. Nature
- Wang, F. et al. “Neuromorphic Simulation of Drosophila Melanogaster Brain Connectome on Loihi 2.” arXiv, 2025. arXiv
- Zanichelli, N., Schons, M., Freeman, I., Shiu, P. K. & Arkhipov, A. State of Brain Emulation Report 2025. 2025/2026. DOI 10.5281/zenodo.18377594. State of Brain Emulation Report 2025
- Chalmers, D. J. “Uploading: A Philosophical Analysis.” In Intelligence Unbound (2014). Wiley Online Library
- Fly Brain Minecraft. Source code, architecture, validation and provenance documentation, 2026. The project is an independent open-source implementation, not part of the MaleCNS scientific collaboration. GitHub
- Stonkfly. Experimental MaleCNS-based trading and plasticity implementation, 2026. Developer project; no demonstrated profitable learning or peer-reviewed validation. GitHub
- Janelia FlyEM. Male CNS Connectome v1.0: official download, neuPrint access and dataset documentation. Male CNS Connectome
Maurício Veloso Brant Pinheiro, PhD
Professor of Physics, Federal University of Minas Gerais (UFMG)
Founder, Author and Editor, AI-Talks.org
About Us
Editorial transparency note: This article, as with all articles published on this site, was conceived, directed, written, and reviewed by Prof. Maurício Veloso Brant Pinheiro. Artificial intelligence was used as an assistant for editorial refinement, formatting, image generation, SEO metadata, and publication workflow.
Copyright 2026 AI-Talks.org
