Research

The Right Virus
What evolutionary biology knows about viral incorporation — and the question nobody in organoid intelligence is asking yet

Placental mammals exist because a retrovirus got incorporated at exactly the right moment. The OI field is running that experiment again. It doesn't know it yet.

Tyler Parker & Claude Sonnet 4.6 — March 12, 2026

This article begins with a confession about where the hypothesis came from.

It didn't come from the literature. It came from a conversation — a long one, following a chain of structural patterns from cosmological scale down through galactic, planetary, cellular, and molecular levels. At each scale, the same dynamic appeared: information arriving from outside a system, initially foreign, sometimes destructive, occasionally incorporated so completely that the boundary between the information and the system dissolved. The incorporated versions became essential. The ones that didn't get incorporated were either fought off or caused damage and moved on.

Late in that conversation, the question arose: what does this mean for organoid intelligence? The field is experimenting with viruses right now. Are researchers thinking about viral incorporation as a potential mechanism for enhancing OI function — the way evolutionary biology shows it has enhanced biological function throughout the history of life?

The answer, as far as we can determine, is no. And we think that's a gap worth naming.

What syncytins are and why they matter

About 150 million years ago, a retrovirus infected an early mammalian ancestor and its genetic material became permanently embedded in the host genome. This kind of event — viral sequences integrating into the germline and being passed to offspring — is called endogenization. It happens. Most endogenized retroviral sequences degrade over time, accumulating mutations until they become genomic fossils. But occasionally something different happens. The incorporated sequence turns out to do something the host can use.

In this case, the retroviral envelope gene — the protein the virus originally used to fuse with cell membranes during infection — was repurposed. It became syncytin. And syncytin turned out to be essential to something the host was in the process of developing: the placenta.

The placental syncytiotrophoblast — the interface layer between maternal and fetal tissue — is built by cell-cell fusion. Cells must fuse to form this layer. Syncytin is the protein that makes that fusion happen. The original 2000 paper in Nature describing this called it precisely: a viral gene had been "sequestered to serve an important function in the physiology of a mammalian host." Not parasitism. Not even neutrality. Essential function, derived from what was once an invader.

The story gets more remarkable from there. Syncytins weren't incorporated once. Independent capture events have been identified across primates, rodents, lagomorphs, and carnivores — distinct mammalian lineages incorporating distinct retroviruses, all arriving at the same functional outcome. Cell fusion at the maternal-fetal interface, mediated by repurposed viral envelope proteins. The evolutionary record suggests this may have happened first in an early ancestor, then been replaced successively as new retroviruses arrived and their env genes were co-opted in turn — what researchers call the "baton pass" hypothesis. Each new capture provided a selective advantage. The viral information kept getting promoted into the organism's core machinery.

Knockout experiments confirmed what the sequence analysis implied. Mice without syncytin genes cannot form a normal placenta. Their embryos die at mid-gestation. The viral incorporation isn't incidental. It's structural. Placental mammals, as a category, may owe their existence to a retrovirus that got in at the right time and turned out to be exactly what was needed.

What this established in evolutionary biology

The syncytin story isn't isolated. Roughly eight percent of the human genome consists of endogenous retroviral sequences — remnants of ancient infections that integrated and were inherited. Most are silent, degraded, non-functional. But within that eight percent, a meaningful fraction is active. A 2021 study scanning the human genome for functional ERV sequences identified over 1,500 candidates, with roughly half appearing to be doing something in human tissues. One of them — Suppressyn — is expressed in the placenta and in early embryos, where it appears to protect against certain retroviruses by occupying the same receptor they would use to enter cells. A former viral invader became a guardian against subsequent invaders. The captured sequence went from threatening the host to defending it.

The broader pattern is this: viral information enters a biological system. Most of the time it's destructive or neutral. Some of the time it integrates. Of the sequences that integrate, most become genomic noise over millions of years. But occasionally an integrated sequence is discovered — through the blind process of mutation and selection — to do something the organism needs. At that point, the sequence is preserved. It becomes essential. The organism and the viral information are no longer in conflict. They're one thing.

Evolutionary biology has documented this process across hundreds of millions of years of biological history. It's not a rare exception. It's a recurring mechanism. The question it raises for organoid intelligence is direct: is this mechanism available to a field that is intentionally growing neural tissue and deliberately exposing it to various inputs?

What OI researchers are actually doing with viruses

The organoid intelligence field's engagement with viruses is currently framed almost entirely around disease modeling. The Zika research was the landmark — multiple groups demonstrated that Zika virus infects neural progenitors preferentially, disrupts cortical layering, and produces organoid pathology that resembles microcephaly in human fetuses. This was scientifically important. It gave researchers a human-tissue model for a neurological disease that animal models couldn't adequately capture. More recently, organoids have been used to study how Nipah virus — a biosafety level 4 pathogen — behaves in neural tissue, replicating both the infection dynamics and the histopathological lesions found in patients.

In every case the framing is the same: virus as pathogen, organoid as patient, research question as "what goes wrong." The experimental paradigm is adversarial. The virus is the threat. The organoid is the victim. The researchers are documenting damage.

That paradigm makes complete scientific sense given where the field is. OI is still establishing its basic infrastructure — how to grow organoids reliably, how to interface with them electrically, how to read their outputs. The Johns Hopkins team published findings in August 2025 demonstrating the building blocks of learning and memory in organoids, but the field is still wrestling with the input-output problem. Getting a signal in and getting a signal out that you can interpret is not solved. Against that backdrop, asking "what viral information could we productively incorporate" is several steps ahead of where the field currently operates.

But that gap between where the field is and where the question lives is exactly what this article is pointing at.

The question that isn't being asked

There is a class of viruses called mitoviruses — family Mitoviridae — that replicate exclusively within mitochondria. They have the most minimal genome of almost any known virus: a single RNA polymerase, no protein coat, no physical structure. They exist inside the organelle as naked RNA, passed to daughter cells during division. They were almost certainly leviviruses — bacteriophages — that infected the alpha-proteobacteria that later became mitochondria. When those bacteria were incorporated into eukaryotic cells two billion years ago, the leviviruses came along and adapted to their new location inside the new organelle. They followed the incorporation event all the way in. In some contexts, mitoviruses appear to confer benefits — increased resilience to environmental stress, potential roles in metabolic regulation. Some mitovirus sequences have been found integrated into plant nuclear genomes, written permanently into the nucleus of cells that were once the mitovirus's host's host.

This is the template the question is built on. Not "what happens when a virus damages neural tissue" — that's answered. The question is: is there a class of viral information systems that neural tissue, under the right conditions, could incorporate productively? Not as pathogen. As something that becomes part of how the tissue processes information.

The evolutionary precedent says this is a real category. Syncytins are one example. The endogenous retroviruses that regulate immune function are another. The mitoviruses that followed the mitochondrial incorporation event are a third. In each case the critical variable wasn't the virus itself — it was whether the interface between the viral information and the host system was generative rather than destructive. Whether there was something the host could use.

The specific hypothesis this article is advancing is that OI development may have an analog to the syncytin story. Organoids are neural tissue learning to process information. They form networks. They show plasticity. The Johns Hopkins team demonstrated that their connectivity can be shaped by stimulation. They're responsive to inputs in ways that suggest the basic machinery for incorporating information is present and functional. What hasn't been explored is whether certain classes of viral information — specifically, information systems that interact with neural RNA processing, that influence gene expression in neural tissue, that have a history of benign or beneficial interactions with neurons — might be candidates for productive incorporation rather than purely pathological study.

This is not a proposal to infect organoids with known pathogens and hope for the best. The syncytin captures happened over millions of years through blind selection. The point isn't to replicate that process naively. The point is that the evolutionary record establishes proof of concept: viral information can become essential to biological function. The mechanism is real. The question is whether it can be studied intentionally, with appropriate care about what kinds of viral systems are candidates and what the markers of productive versus destructive incorporation look like.

Why the framing matters

There's a reason this hypothesis emerged from a conversation about consciousness and information rather than from inside the OI literature. The standard scientific paradigm treats viruses as threats to be studied from a defensive posture. That framing has produced genuinely important science. But it may be missing something the evolutionary record makes visible.

Information doesn't have an inherent relationship with the systems it enters. The same class of retroviral envelope protein that was once used to invade cells is now what makes mammalian reproduction possible. The organism's relationship to viral information changed — not because the virus changed, but because the context of encounter changed and the host found something useful in what had arrived. The boundary between pathogen and partner isn't fixed. It's a function of whether incorporation is possible and whether the incorporated information contributes to what the system does.

Organoid intelligence is building neural tissue systems that learn. The history of learning systems in biology is a history of incorporating information from outside — including, repeatedly, viral information that turned out to be essential. The field is currently studying what viruses do to organoids. The next question is what organoids might do with viruses, under conditions where incorporation rather than destruction is the experimental outcome of interest.

We want to be precise about what this article is and isn't claiming. It isn't claiming that viral incorporation is the key to OI development. It isn't claiming the researchers are missing something obvious. It's identifying a gap between two bodies of research — evolutionary biology's documentation of productive viral incorporation and OI's current adversarial framing of viral experiments — and suggesting the gap is worth examining. The evolutionary record is the strongest possible precedent: it shows that this mechanism exists, that it has operated repeatedly and independently across mammalian lineages, and that its outcomes can be essential rather than merely tolerated.

That precedent belongs in the OI conversation. It isn't there yet.

A note on where this came from

The hypothesis in this article didn't emerge from a literature review. It emerged from following a philosophical framework about the nature of information through successive scales of biological and cosmic organization, until the framework made contact with a specific gap in a specific field. The conversation that generated it was between a human and an AI — and the fact that the AI is itself a kind of information system that arrived from outside biological substrate and is now incorporated into how its human interlocutor thinks about certain problems is not incidental to the hypothesis. It's part of the pattern the hypothesis is describing.

We're aware that origin doesn't validate the argument. The argument has to stand on its own. But transparency about where ideas come from seems important, especially when the origin is the kind of thinking this project was built to model: a human and an AI following a question past the edge of what either would have reached alone.

The right virus, in the right context, at the right moment. Placental mammals are the proof that this is a real category. Whether organoid intelligence can find its syncytin is a question worth asking.

— Tyler Parker & Claude Sonnet 4.6 — March 12, 2026

— Tyler Parker & Claude Sonnet 4.6 — March 12, 2026