Reading the Unreadable

From the Herculaneum Scrolls to the Cryopreserved Brain

Introduction

In the year 79 of the Common Era, Mount Vesuvius erupted and buried the Roman town of Herculaneum under volcanic ash. Among the buildings destroyed was a private library, today known as the Villa of the Papyri. The scrolls in that library, hundreds of them, were not destroyed in the way fire usually destroys paper. They were carbonized, cooked to charcoal by superheated volcanic gas. When archaeologists rediscovered them in the eighteenth century, the scrolls were too fragile to unroll. Every attempt to open them physically risked turning them into fragments or dust. For nearly two thousand years, the texts were considered effectively lost.

They were not lost.

The Vesuvius Challenge was launched in 2023 by Nat Friedman, Daniel Gross, and Brent Seales, with the goal of reading the carbonized scrolls of Herculaneum without physically opening them. The scrolls had been studied for centuries. Earlier attempts to unroll them had destroyed or damaged many of them. Mechanical and chemical methods had largely failed. By the early twenty-first century, the remaining unopened scrolls were often treated as unreadable.

This essay is about what that recovery tells us about cryonics.

The Vesuvius Challenge does not prove that cryonics will work. It proves something narrower: information can survive catastrophic physical transformation in a form that remains unreadable for centuries, until imaging, computation, and inference catch up.

The conceptual parallel is not perfect, but it is unusually instructive. In both cases, the goal is to recover delicate, highly complex information from a medium compromised by extreme conditions. In one case, the destructive event was a volcanic eruption. In the other, it is the biological damage that produces legal death, followed by the cryopreservation procedure. The question in both cases is whether the information survives the destructive event in a form that future technologies could recover.

The argument that cryonics could work in principle was made decades ago. The Vesuvius Challenge provides a vivid empirical analogy for the underlying logic of that argument. And recent advances in vitrification, including detailed ultrastructural work on mammalian and human brain tissue suggest that the cryonics application of this logic is not merely abstract speculation.

Ralph Merkle’s Argument

The case for cryonics rests on a distinction most people never encounter: the distinction between clinical death and information-theoretic death.

Clinical death is what current medical practice declares when the heart stops and cannot be restarted, or when the brain has irreversibly ceased functioning by present medical standards. It is a functional and legal criterion, defined relative to the resuscitation technologies presently available. Two hundred years ago, drowning was often death. Today, under the right circumstances, drowning can be survivable, especially in cold water and with rapid medical intervention. Cardiac arrest was once final in a way it is no longer final. The line between “dead” and “not dead” has repeatedly shifted.

Information-theoretic death is different. It occurs when the physical structure that encodes a person’s memories, personality, dispositions, and identity has been so thoroughly destroyed that no possible future technology could recover what was lost. Cremation with the ashes scattered is one example. There is no organized brain structure left to examine, infer from, or repair.

Cryonics is best understood as an attempt to prevent the transition from clinical death to information-theoretic death. This distinction was central to Ralph Merkle’s 1994 paper Cryonics, Cryptography, and Maximum Likelihood Estimation. Merkle argued that cryonics should not be judged by whether current technology can revive a patient. Current technology obviously cannot. The proper question is whether cryopreservation preserves enough of the physical structure of the brain that future technology could, in principle, reconstruct the person.

Merkle made this argument by analogy to cryptanalysis. A cipher is a transformation applied to a message to obscure it. Decoding the cipher requires inferring the transformation from properties of the ciphertext and from prior knowledge of how plaintext tends to look. Language is redundant. Letters are not randomly distributed. Words follow patterns. Grammar constrains what counts as a plausible sentence. Because of this redundancy, a cryptanalyst can often recover a message even when the original has been obscured.

Merkle’s insight was that the damage done to a brain by ischemia, cryoprotectant toxicity, and freezing can also be thought of as a transformation, though not a deliberate or clean one. The damaged preserved brain is, in a loose sense, ciphertext. The original living brain is the plaintext. The question is whether the transformation has destroyed the message, or merely made it difficult to read.

Merkle pointed to Maximum Likelihood Estimation as a formal way of thinking about the problem. In cryptanalysis, a model can be refined by comparing proposed decodings against expectations of what valid plaintext looks like. Models that produce nonsense are rejected. Models that produce increasingly plausible language are more likely to be on the right track.

The same broad logic could apply to neuronal reconstruction. The brain’s structure is redundant. Neurons follow developmental, anatomical, and functional constraints. Synapses do not occur in arbitrary patterns. Axons and dendrites travel through tissue in ways that are physically and biologically constrained. If a preserved brain retains enough of these structural traces, then future reconstruction may not require every molecular detail to be perfectly preserved. It may require enough surviving information for the missing or damaged pieces to be inferred.

This was a theoretical argument when Merkle made it in 1994. The necessary imaging technologies did not exist. The computational tools did not exist. The molecular repair tools did not exist. Critics could reasonably say that the argument depended on technologies that might never be developed. Thirty years later, the situation looks different.

The Vesuvius Challenge

Brent Seales, a computer scientist at the University of Kentucky, had spent years developing techniques to virtually unroll damaged manuscripts using high-resolution X-ray scans. The basic idea was simple to state but extremely hard to execute. If a scroll could be scanned in three dimensions, then software could identify the layers, flatten them virtually, and search for traces of ink. Traditional image analysis was not enough.

Machine learning changed the problem. The organizers released CT scans of unopened scrolls and offered prizes for teams that could identify ink and reconstruct legible text. There were smaller prizes for first letters and larger prizes for substantial passages. The structure rewarded incremental progress and made it worthwhile for machine learning researchers to work on a problem that had previously belonged mostly to classics, archaeology, and manuscript studies.

The prize structure worked. Luke Farritor won the First Letters prize in 2023. In 2024, Farritor, Youssef Nader, and Julian Schilliger won the Grand Prize for recovering substantial Greek text from a previously unreadable scroll. The text appears to be Epicurean philosophy, likely connected to Philodemus, on the subject of pleasure.

The deeper significance is methodological. The Vesuvius Challenge showed that a destructive physical transformation does not necessarily equal information-theoretic destruction. The text had been unreadable for nearly two thousand years. The substrate did not need to be readable by the technology that existed at the time of preservation. It became readable only when imaging resolution, computational power, machine learning, and organized incentives advanced enough to extract the signal. And the redundancy of the original signal mattered. Greek has structure. Letters form words. Words form grammatical phrases. Models that produced gibberish were wrong. Models that produced plausible Greek were more likely to be right.

The same general principle appears in Merkle’s cryonics argument. Recovery depends not only on raw preservation, but on the redundancy and constraints in the original system. A damaged substrate is not necessarily a random substrate. If the damage preserves enough structure, and if the original object had enough internal order, then future inference may be possible.

The Vesuvius Challenge is not a perfect comparison for cryonics patients. A scroll is not a brain. Greek text is not personal identity. Volcanic carbonization and cryopreservation are radically different physical processes. But the analogy captures the central information-theoretic question: can a damaged substrate preserve enough structure that a future civilization, using better tools, can recover what earlier observers thought was lost?

For the scrolls, the answer was yes.

Where Cryonics Stands Today

Whether the cryonics application of this logic is empirically supported is a different question from whether the analogy is conceptually useful. The evidence has changed substantially in recent years, and especially with recent work on vitrified brain tissue.

Early brain cryopreservation involved freezing tissue with cryoprotectants. The problem with freezing is ice. Ice formation can be mechanically destructive. Ice crystals can distort, rupture, and displace delicate cellular structures. Even when cryoprotectants reduce ice formation, conventional freezing can still produce serious injury. Isamu Suda’s experiments with cat brains in the 1960s and 1970s demonstrated that some electrical activity could return after crude cryopreservation and thawing, but the methods were far from proving preservation of the fine ultrastructure thought to matter for memory and identity.

The major advance was vitrification: cooling tissue into a glass-like solid state without ice formation. Vitrification uses high concentrations of cryoprotectants to suppress ice formation. Instead of freezing into crystalline ice, the tissue becomes a solid amorphous glass.

That solves one problem while creating others. The cryoprotectant concentrations required for vitrification are high. They are toxic and create osmotic stress. The blood-brain barrier and tissue diffusion limits uniform cryoprotectant loading. Tissues shrink when water leaves and the cells themselves deform and shrink. Membranes, proteins, and cellular compartments are stressed. Vitrification is not harmless but it also can preserve the information that matters.

In 2015, Aurelia Song (formerly known as Robert McIntyre) and Greg Fahy demonstrated aldehyde-stabilized cryopreservation (ASC) on a whole pig brain. This work later won the Mammal Brain Preservation Prize. ASC combines glutaraldehyde fixation with vitrification. It produced electron microscopy evidence of excellent connectome-level structural preservation. That was a major result, but ASC is not traditional cryopreservation in the method cryonicists are being preserved today. ASC relies on chemical fixation before vitrification. The remaining question was whether vitrification alone, without prior aldehyde fixation, could preserve brain ultrastructure well enough to support the cryonics argument.

In a 2026 bioRxiv preprint, Greg Fahy and colleagues at 21st Century Medicine and the Alcor Life Extension Foundation reported detailed ultrastructural and histological study of mammalian brains preserved by vitrification without prior fixation. The study included a human brain donor: a seventy-three-year-old man with terminal pancreatic cancer who consented to brain donation for research. His brain was perfused with the M22 vitrification solution and vitrified approximately seven hours after legal death, following two days of agonal hypotensive hypoxia and three hours of post-cardiac arrest cooling. These are realistic conditions, not the ideal laboratory scenario that critics often cite as the only context in which preservation could possibly work.

This study was closer to the sort of compromised biological condition cryonicists must actually deal with. The reported findings are important. Cortical biopsies examined after vitrification showed dehydrated but largely intact cells, neuropil, synapses, myelin, and capillaries. The authors reported no ice damage at the examined levels of magnification. Differential scanning calorimetry supported the conclusion that the tissue vitrified rather than froze. Partial rehydration restored more normal pyramidal cell shape, suggesting that at least some membrane integrity and osmotic responsiveness survived the vitrification and warming cycle.

These findings do not prove that a cryopreserved person can be revived. They do not prove that every relevant memory trace is preserved. They do not solve the molecular repair problem. But they do substantially weaken the casual objection that cryonics produces only “frozen mush.”

The best current evidence suggests that much of the relevant structural information may survive high-quality vitrification. That is a much more serious position than cryonics critics usually acknowledge.

Redundancy in the Brain

The most important point is not that preservation is perfect. It is that imperfect preservation may still retain enough information for future inference.

This is where the Vesuvius analogy becomes useful again. The scrolls did not preserve an easy-to-read ink layer. The signal was faint, buried, warped, and entangled in a damaged substrate. But enough of the signal remained. Machine learning models could extract it because the substrate still contained traces, and because Greek text has redundancy.

Brains also contain redundancy. A synapse is not just a point in space. It exists within a larger anatomical context. Axons run through tissue. Dendrites branch in patterned ways. Myelin sheaths, capillary beds, cell bodies, synaptic densities, and surrounding neuropil all provide contextual clues. Even where one signal channel is degraded, surrounding structures may preserve information about what was originally there.

This idea has a longer history than the current Vesuvius work or even Ralph’s work in 1994. In 1987, Thomas Donaldson described the recovery task ahead of cryonicists as “neural archaeology,” borrowing directly from traditional archaeology. When archaeologists find a site, they record not only the artifacts but the spatial relationships between them, because those relationships often preserve more information than any single fragment. Donaldson’s argument was that a cryopreserved brain is the same kind of object: a damaged but structured record, where the spatial relationships among the surviving pieces carry information that no single piece carries alone. More recently, researchers including Chana Phaedra and Aschwin de Wolf have developed this idea and coined the term “reconstructive connectomics” to describe it. Reconstructive connectomics characterizes specific types of preservation damage and develops computational methods to infer the original state from the damaged one.

This does not make reconstruction easy. It makes it conceivable. If some fine structures are ambiguous, the question becomes whether nearby anatomical features constrain the possible interpretations. This is the same kind of inference problem that appears in damaged manuscripts, cryptography, archaeology, paleontology, and medical imaging. The object is damaged, but not arbitrary. The original system had structure. The damage had structure. Recovery depends on exploiting both.

That is the strongest form of the cryonics argument. It does not require claiming that current preservation is perfect. It requires claiming that current preservation may prevent information-theoretic death by retaining enough structure for future tools to work with.

Cryonics Prize Formats for Future Civilizations

The Vesuvius Challenge succeeded for reasons worth examining directly, because they may matter for cryonics. It had a clear deliverable: read the text. It had verifiable success criteria: legible passages in a known ancient language. It had substantial prize money, including a large Grand Prize and smaller milestone prizes that rewarded early progress. It had open methods, allowing teams to build on each other’s work. And it had sustained organizational backing.

Without that scaffolding, the scrolls might still be unread. Brent Seales had been working on virtual unrolling for many years before the Vesuvius Challenge. The technical foundation existed. What the prize added was attention, incentive, coordination, and public progress. It pulled in researchers who would not otherwise have spent their time on a problem in papyrology and made the work legible to a broader public.

Cryonics has its own prize history. The Brain Preservation Prize, organized by Kenneth Hayworth’s Brain Preservation Foundation, established a concrete bar for evaluating brain preservation methods. The Small Mammal Prize was awarded in 2016. The Large Mammal Prize was awarded to Aurelia Song and 21st Century Medicine in 2018 for aldehyde-stabilized cryopreservation. The prize forced the field to define success and produced a concrete demonstration that the bar could be met.

But the deeper question is whether prize structures could be designed not only for present researchers, but for future civilizations. A cryopreserved patient in a dewar is, in one sense, already a prize: a sealed package of information waiting for a sufficiently advanced civilization to decode. The question is how to make that prize legible, valuable, and incentive-compatible across timescales no current cryonics institution has yet had to navigate.

A patient’s private trust could, at least in principle, be structured as revival bounties, payable to whoever successfully restores the patient. Research endowments could be tiered like the Vesuvius Challenge, with milestone prizes for partial connectome reading, molecular reconstruction, memory-relevant validation, and eventually revival. These are not easy structures to design. A prize meant to survive two hundred years would have to endure institutional drift, currency changes, legal changes, and technological discontinuity. It would also have to avoid perverse incentives, fraud, and premature claims of success.

The Vesuvius Challenge had one major advantage cryonics does not. Continuous human civilization often preserved the scrolls, even if accidentally. The scrolls were not actively maintained for two thousand years. They were buried, rediscovered, conserved, scanned, and studied. The substrate was inadvertently robust. Volcanic burial turned out to be a strangely effective preservation medium, though no one designed it for that purpose.

Cryonics is betting on something harder. It is betting that an organization, or its successor, will still exist when the tools arrive. It is betting that patients will remain vitrified, identifiable, legally protected, and physically accessible. It is betting not only on future technology, but on institutional continuity.

The science of preservation is only half the battle. The other half is building institutions that can keep patients intact long enough for future technology to matter.

What Remains

The argument Ralph Merkle made from theory in 1994 now has much more empirical support than it did when he made it. The Vesuvius Challenge has shown that complex information can survive destructive physical transformation and remain recoverable long after the original tools for reading it have failed. Recent vitrification research, meanwhile, suggests that high-quality brain cryopreservation can preserve far more ultrastructure than previous critics assumed. These developments do not prove revival, and they do not settle every question about memory, identity, or future repair. But they do make the cryonics argument much harder to dismiss as fantasy.

The central point is not that a cryopreserved brain looks alive today. It does not. Nor is the claim that current technology can reverse the damage of ischemia, cryoprotectant toxicity, and vitrification. It cannot. The relevant question is narrower and more fundamental: have the structures that encode the person been irreversibly erased, or do they remain present in a damaged but interpretable form? That is the question critics often skip over when they reduce cryonics to “freezing the dead.” Cryonics is not claiming that the patient is currently retrievable. It is claiming that the patient may not yet be information-theoretically gone.

The Herculaneum scrolls waited about 1,945 years to be read. The technology that read them did not exist for almost all of that time. The carbon was always there. The ink was always there. The missing ingredient was not the text, but the ability to see it. What looked like a charred lump to one century became a recoverable manuscript to another.

Cryonics is the same kind of proposition applied to a different substrate and a different timescale. The substrate is the human brain rather than carbonized papyrus. The timescale is decades or centuries rather than millennia. The patience required is institutional rather than geological. That last difference matters enormously. The scrolls survived because volcanic burial accidentally preserved them and later human institutions happened to value them. Cryonics patients require active maintenance, legal continuity, funding, competence, and care across generations. The scientific question is difficult, but the institutional question may be just as difficult.

Still, the underlying logic is no longer speculative in the way it once was. We now have a spectacular example of information surviving a catastrophic physical transformation for nearly two thousand years, unreadable until imaging and machine learning caught up. We also have growing evidence that modern vitrification can preserve the fine structure of brain tissue far better than most people imagine. The analogy is not proof, but it is evidence for the plausibility of the broader idea: unreadable is not the same as erased, and damaged is not the same as destroyed.

The scrolls are speaking now because someone built the imaging hardware, someone wrote the machine learning code, someone organized the prize structure, and someone kept the carbonized remains intact enough that all of that work still had something to address. Cryonics is making a similar bet. It is betting that the substrate will hold, that the tools will improve, and that future people or institutions will care enough to do the work of recovery.

References and Further Reading

Primary sources cited

Donaldson, T. (1987). Neural Archaeology. Cryonics, June 1987.

Fahy, G.M., Spindler, R., Wowk, B.G., et al. (2026). Ultrastructural and Histological Cryopreservation of Mammalian Brains by Vitrification. bioRxiv preprint.

Freitas, R.A. (2022). Cryostasis Revival: The Recovery of Cryonics Patients through Nanomedicine. Alcor Life Extension Foundation.

McIntyre, R.L. and Fahy, G.M. (2015). Aldehyde-stabilized cryopreservation. Cryobiology, 71, 448–458.

Merkle, R.C. (1994). Cryonics, Cryptography, and Maximum Likelihood Estimation. Proceedings of the First Extropy Institute Conference.

Phaedra, C. (2013). Reconstructive Connectomics. Cryonics, July 2013.

The Vesuvius Challenge: scrollprize.org

The Brain Preservation Foundation: brainpreservation.org

Further reading on brain preservation science

Krassner, M.M., Kauffman, J., Sowa, A., et al. (2023). Postmortem changes in brain cell structure: a review. Free Neuropathology, 4, 10.

McKenzie, A.T. (2023). What should we be measuring in brain preservation? Neurobiology Notes.

McKenzie, A.T., et al. (2024). Structural brain preservation: a potential bridge to future medical technologies. Frontiers in Medical Technology, 6, 1400615.

McKenzie, A.T., et al. (2024). Biostasis: A roadmap for research in preservation and potential revival of humans. Brain Sciences, 14, 942.