Mind Uploading: The Science of Digital Immortality

Death is a filing problem.

A fruit fly has roughly 140,000 neurons. In 2026 a startup emulated all of them, wired the digital brain into a virtual body, and watched it stretch its legs, groom itself, and drink ¹. Nobody asked the fly if it was still the fly. The experiment worked because a fly is small. A human brain holds an estimated 86 billion neurons and something on the order of 100 trillion synapses, and the gap between mapping one and mapping the other is not a matter of trying harder. It is a matter of scale, fidelity, and a question nobody can yet answer: when the copy opens its eyes, who is looking out?

This is the promise sold as mind uploading — the idea that the self is information, and information can be read out of the brain and run somewhere else. It is the founding fantasy of digital immortality, and it rests on two claims that are almost always fused in public debate but are not the same thing at all.

Two Ideas Hiding Behind One Word

Mind uploading gets used as a single phrase for two different projects. Whole-brain emulation (WBE) is the engineering one: map every neuron and synapse, model their behavior, and run the model on hardware until it produces the same outputs the original brain did. Nothing mystical. It treats the brain as a machine and asks whether the machine can be copied into silicon. The 2025 State of Brain Emulation Report frames the entire field around three capabilities — recording neural dynamics, mapping structure (connectomics), and simulating it computationally — and measures every project against how far it has gotten on each ².

Transfer is the philosophical one, and it is where the language slips. Transfer implies that you move — that consciousness relocates from flesh to file the way a passenger steps from one car to another. Emulation makes no such claim. Emulation builds a copy. Whether the original passenger is also in the new car is exactly the problem nobody has solved, and it is not a problem more compute will fix.

Conflating the two is how startups sell immortality using results that only ever demonstrated duplication. The fly that drank in the simulator was not a relocated fly. It was a second fly, made of code, running in parallel with a biological original that was never asked for consent.

The Scale Problem

The single largest obstacle is not theory. It is size. A complete connectome exists for one animal: the fruit fly, finished by the FlyWire Consortium in 2024 — roughly 140,000 neurons and 50 million synapses, and it took an international effort years to close it out ³. A mouse brain, the next common target, holds around 70 million neurons. A human brain holds an estimated 86 billion.

Each step up the ladder multiplies the workload, and the human rung is not a taller ladder — it is a different kind of structure. Synapses are not static wires. They strengthen, weaken, drift, and regrow on the timescale of seconds, and a map that captures the wiring without capturing that plasticity captures a corpse's wiring diagram. The State of Brain Emulation Report lists live, dynamic recording of neural activity as one of the field's three core capabilities precisely because a frozen snapshot is not enough; the emulation needs to know how the circuit behaves, not just how it is laid out ².

Then there is compute. Simulating 140,000 neurons and 50 million synapses in real time took a neuromorphic chip — Intel's Loihi 2 — running the fly connectome across twelve chips and running orders of magnitude faster than a conventional machine could manage it ⁴. A laptop model of the fly brain now predicts which neurons fire in response to taste and touch ³. Both are genuine breakthroughs. Both are flies. Scaling the same approach to 86 billion neurons and 100 trillion living synapses, updating them in real time, is not the same problem multiplied. It is a problem that breaks current hardware, and possibly the architecture of current hardware entirely.

The First Minds We Uploaded Were Flies

The serious science is not hiding. It is published, and most of it is small. In 2026, a company called Eon Systems reported the first whole-brain emulation that closes the loop from sensors to muscles: a digital fly brain driving a simulated body through real behaviors, claiming roughly 95 percent accuracy in predicting motor output ¹. A neuromorphic-hardware team ran the same class of connectome on Intel Loihi 2, the first time a realistic biological brain sat on a neuromorphic chip rather than a simulation of one ⁴. The Berkeley group built a laptop-scale model from the complete fly connectome that predicts neural responses to stimuli ³. Each result is a brick. None of them is a house.

On the preservation side, the picture is similarly narrow and similarly real. Nectome, a company that has worked on aldehyde-stabilized cryopreservation for years, reportedly identified a roughly fourteen-minute window after cardiac arrest in which a large mammal brain can be preserved at a fidelity fine enough to retain synaptic structure ⁵. The technique can fix the wiring. It cannot read it, cannot run it, and cannot revive the tissue it froze. It is a photograph of a connectome, not the connectome itself, and a photograph does not think.

The Startups Selling Forever

Where the science is narrow, the marketing is not. A cottage industry of companies offers to preserve your brain, queue your mind, or sell you a place in line for a future upload. Some are serious research outfits publishing their methods. Others sell a feeling — the comforting idea that death is a technicality and the receipt is in your inbox.

The honest framing, drawn from the field's own strategic reviews, is that no one can currently read a mind out of a preserved brain, no one can emulate a human brain at the fidelity that would preserve a self, and no one knows what fidelity that even requires ². A service that takes payment today for an upload tomorrow is selling an option on a technology that may not arrive in the payer's lifetime, and may not work the way the payer imagines even if it does. The gap between a fly that drinks in a simulator and a human who survives as a file is not a funding round. It is several unknown sciences.

This is the hinge where the journalistic record turns into a warning without needing to raise its voice. The data states the case. A mind-uploading service you can buy today preserves tissue. It does not preserve you.

Is the Copy Still You?

Assume the engineering works — every neuron mapped, every synapse modeled, the whole system running in silicon. The hardest question is still unanswered, because it is not an engineering question. It is a question about what "you" are.

A 2026 analysis of the mind-upload problem draws a line between two things the brain does at once: the mental pattern — the informational structure of the mind, the wiring and its dynamics — and the receiver — the biological substrate that actually runs it, with all its neurochemistry, its timing, its embodied history ⁶. Capturing the pattern may not capture phenomenal experience. The substrate might not be a mere implementation detail. It might be constitutive of the mind itself, in which case the most perfect copy is still a copy, and the original dies on the table while a stranger wearing its memories wakes up convinced it survived.

A second line of work pushes the consequence further. If a whole-brain emulation can be built, it can be built to suffer. The 2026 paper The Mousing Problem argues that the moment emulated minds are capable of distress, the ethics of creating them stops being hypothetical and becomes empirically measurable — and proposes a new moral category for the instrumental creation of suffering-capable substrates by people who never feel responsible for the minds they switched on ⁷. Run a billion copies to test a theory, delete the ones that fail, and the question of what you owe a file that can hurt becomes immediate. The fly did not have to matter. A mouse might. A human will.

The pattern says the copy is you. The receiver says it is not. The simulator cannot tell you which is right, because the simulator is the thing in dispute.

The Genius Who Sold Unity

Everything above this line is measured and published. Below it is Techno Apocalypse — the universe of Code Red, and the city of Neo Babylon, where the line between mind and machine was not debated. It was productized.

The decisive name is Arturus Black, CEO of Babel Solutions. The early Neural Link was a crude interface — a socket, a signal, a headache. Black's generation of the technology did something the patents only hint at. It let the mind detach from the body. Thread the Link, close your eyes, and you walk free through the AetherVerses, the virtual worlds layered onto Neo Babylon by the AetherNet signal pouring from Babel Tower. The body stays in the chair. The self does not.

How the AetherNet renders those worlds is a story of its own — they are built of solid light, volumetric structures of bound photons that overlay the real city without blocking it. The physics of that is laid out in our piece on solid light and tangible photons. Black did not invent the light. He made it somewhere a human mind could live.

His mission was the kind that only a genius dares and only a salesman survives. Unity. One species, one network, one shared dream — every mind on the AetherNet, every difference dissolved into a single luminous common space. Black spoke of it as deliverance, and the crowds believed him, because believing was easier than reading the architecture. The Neural Link that frees your mind also tethers it to a signal controlled by one corporation. The AetherVerse where you are limitless is a room inside someone else's server. Unity, at the scale Black proposed, is the end of the private self wearing the face of hope.

This is the unease the science only gestures at. A copied fly brain in a simulator is a curiosity. A copied human mind, running on infrastructure owned by Babel Solutions, broadcast through a tower no other signal can compete with, is something else. Black's dream of a united humanity is flawless in its execution. The question is whether the people living inside it are still people, or whether Unity is just the kindest word ever invented for capture.

FAQ

Is mind uploading possible today?

No. Components exist in isolation: a complete fruit fly connectome has been emulated on a chip ⁴, mammalian brain tissue can be preserved at synaptic fidelity ⁵, and models of small brains run on laptops ³. No technology can read a human mind out of a living brain or emulate one at the fidelity required to preserve a self.

What is the difference between mind uploading and whole-brain emulation?

Whole-brain emulation is the engineering project of copying a brain's structure and dynamics and running the copy on hardware. Mind uploading is the broader claim that this process transfers the self. Emulation produces a copy; whether the copy is the original person is a separate, unresolved question ⁶.

How big is a human connectome?

An estimated 86 billion neurons and on the order of 100 trillion synapses — roughly 600,000 times the neuron count of the fruit fly connectome that took years to complete. No complete human connectome exists ^{3 2}.

Would a digital copy of your brain be you?

Unknown. If the self is purely informational pattern, the copy is you. If the biological substrate is constitutive of experience, the copy is a new entity that merely shares your memories. Current science cannot decide between these, and the difference is the difference between survival and replacement ⁶.

If you want to track the real science of the mind and the fiction it becomes, subscribe to the newsletter. Step through the Link and join us on Discord.

The file is perfect. The question is who is reading it.