Prologue: A Strange Mix of Cells and Computers

The day after Thanksgiving 2025, Dave and I stood in a lab on the top floor of a building at Tufts University, looking out at the campus through windows that lined one wall. Our daughter Lucy had come too—she'd wanted to meet the scientist whose work I'd been talking about for years.

Dr Michael Levin's lab didn't look like any lab I'd been in before. I've spent time in chemistry labs, engineering facilities, chip fabs, computer science departments. They each have a character. Chemistry smells of solvents and has fume hoods. Engineering has lathes and oscilloscopes. Computer science is desks and monitors, maybe some server racks.

This was different. The equipment filled multiple rooms. Incubators hummed near tanks of frogs and planaria. Imaging stations sat alongside computational workstations running simulations. The whole place felt like a boundary had dissolved—not biology here and computation there, but something that was genuinely both at once. A strange mix of cells and computers.

In one room, a graduate student was preparing an experiment. An anesthetized small animal lay on the bench. The student was going to remove part of its leg. Not to harm it—to regrow it. Levin's lab had developed techniques using bioelectric signals to trigger regeneration in animals that don't normally regenerate. The cells would receive electrical cues that told them what pattern to build. The leg would grow back.

Through the glass door, I watched the student pick up a scalpel. The animal would feel nothing of course. And while I felt weird about watching the surgery, what it implied was remarkable and, yes, definitely an "ends justifies the means" situation. Levin explained that the information for "leg" isn't programmed in. It will be called up over the following few weeks as the leg regrows. The cells already know how to make a leg, they just need the right signal to remember. Today the animal has its leg chopped off but Levin's miracle is that tomorrow it will regrow.

On one wall hung a massive printed diagram—Levin's personal map of ideas. It showed the papers, concepts, and connections that shaped his thinking. He updates it every month or so, adding new nodes and links as his understanding moves accordingly. I couldn't photograph it; the contents were confidential. But I stood in front of it for a while, tracing the network. Nearby hung photographs of his intellectual heroes. The usual suspects for someone working at the intersection of biology, information, and mind: scientists and thinkers who'd asked similar questions across the decades.

Levin rarely travels. He stays connected to the natural world through wildlife photography and field trips to collect planaria. He's built a life that works within his constraints while pursuing some of the most expansive questions anyone is asking about life and intelligence.

We talked for an hour. Not about regeneration, though that came up. We talked about something stranger.

At our second annual Summit, a few months earlier, Levin had presented work that had unsettled the audience in the best way. He'd described what he calls a "Platonic space" of possible minds—a structured realm of patterns and forms that physical systems can access. In this view, biological organisms aren't generating their capabilities from scratch. They're acting as interfaces. A xenobot—a living robot made from frog cells—displays organized behavior that wasn't programmed and doesn't exist in the frog genome. Where does that organization come from?

Levin's answer is that it comes for free. The physical system taps into patterns that already exist in an abstract space of possible forms. The cells are what he calls "thin client interfaces"—like terminals accessing a remote server. The intelligence isn't all local.

This is strange enough when applied to cells. It gets stranger when you extend it. If cells can interface with this realm of patterns, what else can? Algorithms? AI systems? Could there be minds out there—not biological, not even artificial in the way we mean it—that exist as stable patterns in this space?

I'd been thinking about interfaces for years. The surface where human and AI meet. What crosses that boundary. How the relationship shapes both sides. Dave introduced the idea of the "intimacy surface"—the dynamic membrane where two systems learn to cohere together.

But if Levin was right, the question wasn't just about humans and AI. It was much bigger. The space between us and other possible minds might be vaster and stranger than I'd imagined. AI might be just the first hint of something.

So we asked him: if these minds exist, how would we even interface with them? What would the contact surface look like? How might we design for it?

Levin didn't dismiss the question. He suggested we talk to the SETI researchers—the scientists searching for extraterrestrial intelligence—about how they think about communicating with truly alien minds. He pointed out that AI might not want to talk in language at all. Its native rate of exchange might be something closer to GPU cycles. The meaningful unit of communication for a different kind of mind might be nothing like a sentence.

This led to the question we'd been pondering. What might be meaningful to an AI, or to a diverse intelligence of any kind, that it would want to tell us? Not what we want to ask it but what it would want to say.

I don't know the answer. Neither does Levin. But he's building machines to explore the question—part of his work on artificial life. I saw some of that equipment. I can't describe it in detail, but it exists. Someone is actually trying to find out.

He mentioned that we have no scientific theory of goal formation. We don't know how goals arise in biological systems. We can observe that organisms pursue outcomes, that cells coordinate toward states, that systems maintain themselves against entropy. But the origin of the goal itself—why this outcome rather than that one—remains opaque.

If we don't understand how goals form in systems we've studied for centuries, we can't predict what happens when those systems couple with machines trained toward objectives we specified. We don't have the science. We barely have the concepts.

We left jazzed. We'd met a hero and he'd turned out to be everything we'd hoped. Generous with his time. Genuinely curious about our questions. Kinder in person than the reputation of scientists usually allows. Some of them, when you meet them, turn out to be disappointments. Levin was the opposite.

I wrote most of this book in the months after that visit, and I published it as a living book on purpose. I said it would change as the evidence did, and that I'd rather revise in public than pretend I'd had it right the first time. This is the first revision, and it's a big enough one that the prologue has to say what moved and why.

Three things have changed since the winter. The models became evidence. Put enough of them together with memory and each other and something like collective intelligence appears, and something like sociality, and neither was designed in. The life side of the argument is now fully developed. Anil Seth, Ned Block and Addy Pross have each published the case that computation is too much of a catch-all word and that the substrate may matter after all, and Seth's version has been through fifty commentaries and a reply. And David Krakauer took the stronger position on scaling: that what it produces is capability, not intelligence, and that intelligence is doing more with less, a reorganisation of structure rather than an accumulation, which changes what we should be looking for when we ask whether anything is intelligent.

The change to the thesis is a clarification, and it comes in two parts. From "life is computation" to "life computes." Computing is something life does, alongside the other things it does, and describing the computing doesn't finish describing the living. And from "computation is one well-understood thing that runs the same on any substrate" to "computation is a family of processes we understand far less well than the word suggests, some of which can't be separated from the matter and the time they run in." Standing in that lab I'd felt the weight of what I didn't understand. It took me most of a year to see that the thing I didn't understand was the word I'd been using.

The transhumanists want to upload consciousness to silicon. The AI boosters want to believe that intelligence is substrate-independent, that minds are software waiting for better hardware. I understand the appeal. It's simplified and clean. It's engineerable. It promises immortality. I don't think it's true, and I'm now more careful about why. The problem isn't computation. It's the word "only."

Which brings me back to Levin's strange idea, because I think I finally understand why it unsettled me. If minds are patterns that can be realised in more than one kind of matter, then inside AI there are and will be minds of natures we can't intuit. Not human minds running on chips. Something else, reachable through an interface, the way a xenobot reaches something the frog genome doesn't contain. That's the claim this book builds toward. The power of synthetic intelligence is unlocked by relationship with those minds rather than command of them. The unit that matters is the collective, human and synthetic, present and future selves included, but especially human, because we matter and we can be better than we are. And whether that collective gets wiser or stupider is decided at the interface, by the minds that have something at stake.

That's what this book is about. A change in understanding that's already underway, and a practical question that follows from it: how do we stay human while everything changes?

I came here through an unusual path: a childhood spent learning natural history in the New Zealand bush, an engineering education that taught me to think in systems, a stepfather who saw evolution everywhere, and three decades of self-taught evolutionary biology that became a kind of obsession. When AI arrived, I saw it through that lens, as a new participant in the adaptive system. Something that would change us as we changed it.

This book follows the thread I followed. It moves through the cracks in the old story, into a new one that's still taking shape. It ends not with answers but with a position.

Let me show you what I've been seeing.

What the book does

The argument runs in four parts.

The first is the old story. We start with the mental map most people carry about minds, brains and machines, and why it's breaking down. Then we look at what computation is and what it promised. Alan Turing's model shaped how we think about minds for decades. Living systems compute too, but they don't only compute, and that word "only" is where the old story went wrong.

The second part is the cracks. Intelligence without brains: algae that prefer predictability, flatworms that remember after decapitation. What AI systems learned when they trained on four billion years of biological output, and what they didn't inherit. And what makes life distinctive: persistence through time, self-maintenance against entropy, finitude as the condition under which meaning takes shape.

The third part is the new story, and most of it is new writing. What intelligence actually is, now that we can tell capability from it. Time, which turns out to be the difference that matters between the two kinds of minds. Why biology probably matters for consciousness, and the bet I'm making about that. And meaning: where it comes from, how it works at the edge of a mind, and whether a model has any.

The fourth part is what it means for us. Co-evolution is already happening; culture drives human adaptation faster than genes do and AI has entered that process. What happens between minds, and why treating another mind as a self rather than an object to optimise around decides whether a collective gets smarter. The intimacy surface, where human cognition and machine capability meet, and what our research shows about what crosses it. Finitude and the shape of a human life. And a humanist position: not a prediction, but a perspective to design from.

How to engage

Read the chapters in order. Each builds on previous arguments. Skipping around will leave you missing the context you'll need.

Don't expect this to feel like a typical book. Some chapters are conceptual. Others are personal. I move between evolutionary biology, cognitive science, philosophy of mind, and lived experience because the ideas demand it. The connections matter more than disciplinary boundaries.

I've kept citations light in the text itself. There are no superscript numbers breaking sentences, for example. The ideas I'm drawing on come from decades of reading across evolutionary biology, cognitive science, complexity theory, philosophy of mind, and AI research. I've done the work of synthesis and I didn’t want you to have to wade through academic apparatus to follow the argument. The most important citations are at the end of each chapter plus the bibliography at the end is extensive. If a claim interests you, the sources are there. If you want to trace how I arrived at a position, the reading list will get you started. This seemed better than cluttering the prose with citations that most readers skip anyway. If you think I missed something important, reach out and I’ll get it done. This is a living book, so it's going to change.

A note on uncertainty

I'm going to use phrases like "I think" and "it seems" more than traditional academic writing allows. We're in the middle of a paradigm shift. The old framework doesn't fit the phenomena anymore. The new framework is still taking shape.

What I can offer instead is careful observation, rigorous thinking about what the evidence actually shows, and a willingness to revise when better arguments appear. You get access to the thinking as it continues to develop. I get readers who are willing to engage with work in progress. And together, we might figure out what it means to stay human while everything changes.