A couple of months ago, someone whose judgment I trust called a biologist I’d never heard of “the Einstein of the twenty-first century.”
I have a low opinion of that phrase. It gets handed out cheap — every other startup founder turns out to be somebody’s Einstein or Leonardo. So I didn’t buy it. Not even from someone I respect, not on his word alone.
But the name stuck, and I started reading. I haven’t stopped since, and for the past week I haven’t been able to shut up about the man — telling everyone near me, sometimes twice. So let me say it once, carefully, and you can decide for yourself whether the comparison holds.
His name is Michael Levin. Before we start, let’s agree on what “Einstein of the twenty-first century” is not claiming. It isn’t about the loudness of any single discovery, and it isn’t about personal genius. It’s about a kind of move. Einstein forced a rebuild of the categories of space, time, and matter — after him you couldn’t think about physics the old way. He didn’t change the facts inside physics; he changed the categories physics thinks in.
The claim about Levin has that same shape: that what’s at stake is a rebuild of the categories themselves — life, mind, individuality. Where one organism ends and another begins. What memory is. What it means to be a self. If that’s right, it isn’t one more discovery inside biology; it’s a change in the frame biology thinks inside — and then medicine, the science of the living, and the way we approach AI all run down a different track.
But for now it’s only a claim — exactly the same shape as the one about every other founder. The question I’ve spent these months on is whether, this time, it’s true. You’ll decide that yourself: whether he really rebuilt the paradigm, or whether it’s one more loud label. We’ll come back to “Einstein” at the end — and there I’ll tell you how far I go with him myself, and where I expected I’d have to get off.
The Worm That Remembers Its Shape
Start with a worm.
Cut a planarian — a freshwater flatworm about the size of a fingernail — in half, and you don’t get what you’d get with us. No wound that simply closes over. Something else: the head half grows itself a new tail, and the tail half grows itself a new head — brain, eyes, nervous system and all. Cut it into three pieces and you get three worms. And the cells find their way to the right shape far faster than they would by trying options blindly.
Here’s the question that’s easy to skate past, and shouldn’t be: how does the tail half know that a head is exactly what it’s missing? How does any piece know which shape is the correct one — and when to stop, having reached it?
“It’s in the genes” feels like the obvious answer. The genes hold the blueprint, the cells read it and build to spec. Hold on to that answer. Levin spent years showing where it breaks.
Here’s an experiment from his lab.1Durant, F., Morokuma, J., Fields, C., Williams, K., Adams, D. S., & Levin, M., Long-Term, Stochastic Editing of Regenerative Anatomy via Targeting Endogenous Bioelectric Gradients. Biophysical Journal, 2017. 112(10): 2231-2243. https://doi.org/10.1016/j.bpj.2017.04.011. Take that same planarian. For two days, put it in a solution that temporarily — it washes out of the tissue within a day — disrupts the electrical communication between cells. The genes are left entirely alone. The worm regenerates — and grows back with two heads, one where the head goes and one where the tail should be.
Strange on its own, but not astonishing: glitches happen. What’s astonishing is what came next. They cut the two-headed worm again — now in clean water, no solution at all. It grew two heads again. And again. And again — for months, across dozens of cuts; and even when the worm divided on its own, in its ordinary way of reproducing, the offspring stayed two-headed. The genes never changed once. But the shape the worm kept returning to, every time it was damaged, had changed — permanently.
And here’s the part that knocked the wind out of me when it landed. Most of the treated worms looked completely normal: one head, normal anatomy, normal gene activity — under a microscope you couldn’t tell them from ordinary worms. But cut one of those “normal” worms, and roughly one in four grows back two-headed. Which means the shape an animal takes to be its own can be different from the one we see on the outside. A body can look ordinary and still carry a hidden, different plan inside it — one that shows up only when the time comes to act on it.
One last thing about this worm. The hidden plan can be rewritten back. A different electrical intervention returned the two-headed worms to normal — and that normal held across generations too. The authors call it a “multistable epigenetic anatomical switch.” Let me translate: the worm has something like a saved file for the shape of its body. That file does not sit in the genes. And it can be overwritten — and reset.
If that seems like a quirk of one species — here’s a neighboring result.2Emmons-Bell, M., Durant, F., Hammelman, J., Bessonov, N., Volpert, V., Morokuma, J., Pinet, K., Adams, D. S., Pietak, A., Lobo, D., & Levin, M., Gap Junctional Blockade Stochastically Induces Different Species-Specific Head Anatomies in Genetically Wild-Type Girardia dorotocephala Flatworms. International Journal of Molecular Sciences, 2015. 16(11): 27865-27896. https://doi.org/10.3390/ijms161126065. With the same trick, temporarily disrupting the electrical coupling between cells, a worm of one species was made to grow a head in the shape of another species — with its own genome left untouched. A worm carrying one species’ genome built itself another species’ anatomy.
Now run it the other way. Take skin cells from a frog embryo and free them from the body — just separate a bit of tissue and leave it be. Over a week, those cells don’t die and don’t grow into a shapeless lump. They assemble into a tiny creature that moves on its own; it’s been named a xenobot.3Kriegman, S., Blackiston, D., Levin, M., & Bongard, J., A scalable pipeline for designing reconfigurable organisms. Proceedings of the National Academy of Sciences, 2020. 117(4): 1853-1859. https://doi.org/10.1073/pnas.1910837117.4Blackiston, D., Lederer, E., Kriegman, S., Garnier, S., Bongard, J., & Levin, M., A cellular platform for the development of synthetic living machines. Science Robotics, 2021. 6(52): eabf1571. https://doi.org/10.1126/scirobotics.abf1571. No genome describes this creature — frog cells “aren’t supposed to” be able to do this. And a recent paper showed that in these little clumps of cells there are signs of complex information processing, structurally close to what’s seen in the human brain.5Varley, T. F., Pai, V. P., Grasso, C., Lunshof, J., Levin, M., & Bongard, J., Identification of brain-like complex information architectures in embryonic tissue of Xenopus laevis organoids. Communicative & Integrative Biology, 2025. 18: 2568307. https://doi.org/10.1080/19420889.2025.2568307. Here Levin is careful, and says so outright: this does not mean the xenobot “thinks” — a stronger claim would be out of place. I’ll keep his caution.
Put it together. The same genome can build a body with one head or with two; a head of its own species or of another’s. Cells taken out of the body build something the genome doesn’t describe at all. The conclusion comes on slowly, and sits badly with the familiar picture: the shape of a living thing is written not only in the genes. Something else carries the body’s plan — and that something can be read, rewritten, and reset.
Here’s how I read it: picture the cells of the body as members of a group chat. They share one task: to assemble and hold a single specific shape. They’re constantly checking with each other — who’s where, what’s missing, whether everything’s in place. Regeneration is the chat losing some members, reconnecting, and rebuilding what’s missing toward the shared goal. And cancer, in this picture, is a cell that left the chat: it stopped checking against the shared plan and started living for itself — its own goal, its own schedule of division, its own scale. Not an outsider who came in from elsewhere, but one of your own who lost the connection. (This is my framing, not Levin verbatim. But it’s close to how he describes cancer — as a failure of collective coordination, not just a breakdown in a single cell.)
The Wrong Suspect
Back to the question we left hanging: if shape is written not only in the genes, then where else?
For a hundred years biology has been interrogating a single suspect. The genes. And it’s a real suspect: genes are necessary, you can’t build anything without them. But for this question — which shape to assemble, and when to stop — they look like the wrong one. Genes specify what to build from. They don’t directly specify which three-dimensional figure to end up with. This isn’t a controversial claim, by the way: that the genome holds no literal blueprint of the three-dimensional body is acknowledged well beyond Levin’s lab.
Levin’s answer: there’s a second information system, and it’s electrical.
Cells aren’t just bags of chemicals. Each one carries a voltage across its membrane, like a tiny battery, and cells are joined by channels — gap junctions — through which that voltage is compared and passed along. The result is an electrical conversation running across the whole tissue. And in the pattern of that conversation — which regions are charged more, which less — the target shape is encoded. That same “saved file” from the worm story. When the experiment disrupted the communication between cells, it wasn’t damaging the genes — it was damaging this pattern. And the shape changed. Levin calls this layer the bioelectric signal.7Levin, M., Bioelectric networks: the cognitive glue enabling evolutionary scaling from physiology to mind. Animal Cognition, 2023. 26: 1865-1891. https://doi.org/10.1007/s10071-023-01780-3.
And this isn’t a guess by analogy. In the frog embryo there’s an electrical pre-sketch of the brain to come: the region that will become the brain carries a particular electrical pattern before the neural tube closes — before there’s any brain anatomy at all — and that pattern is required for the brain to build correctly.6Pai, V. P., Lemire, J. M., Chen, Y., Lin, G., & Levin, M., Endogenous gradients of resting potential instructively pattern embryonic neural tissue via Notch signaling and regulation of proliferation. Journal of Neuroscience, 2015. 35(10): 4366-4385. https://doi.org/10.1523/JNEUROSCI.1877-14.2015. Change the pattern and the anatomy changes. So the electrical layer doesn’t merely accompany development; it instructs it.
The short version: the genes are a box of Lego bricks. But the instruction for which figure to assemble from them isn’t read off the bricks themselves. Where it’s read from is a question we’ll come back to — and up there, at the very top, the conversation turns somewhere I didn’t see coming. For now it’s enough that the instruction is a separate, physically measurable, rewritable layer. The connection that stitches individual cells into a collective with a shared goal is what Levin calls “cognitive glue”: the glue that turns a crowd of small agents into a single whole pursuing a goal no one cell could reach alone.the same glue, traced up to minds and models →What attunement is made of
And here the picture turns from medical to almost philosophical — on an example that touches everyone. Aging.
There are usually two ways people think about aging, and you’ve heard both. Either we have a built-in program of decline — a biological clock that one day gives the order to grow old. Or it’s the accumulation of damage: mutations, wear, the machine wears out and at some point can’t keep up. Two suspects: program or breakdown.
Levin’s group proposes a third frame, and it changes the question itself. Aging isn’t necessarily a program, and it isn’t only damage. It’s a loss of goal-directedness.8Pio-Lopez, L., Hartl, B., & Levin, M., Aging as a Loss of Goal-Directedness: An Evolutionary Simulation and Analysis Unifying Regeneration with Anatomical Rejuvenation. Advanced Science, 2025. Article 2509872. https://doi.org/10.1002/advs.202509872. Over millions of years, cells learned to move toward one large goal — to build a body. Development is that movement toward the goal. But once the goal is reached and no new one appears, the collective slowly begins to lose its coherence — not because something broke, but because a goal-directed system with no goal degrades. In their model the body ages even without any accumulated damage: simply because it has nowhere left to strive.
And the detail that keeps this from staying an abstraction: Levin draws the bridge to humans directly. His frame fits something long observed — older people who keep up a high level of engagement (social, physical, new goals) decline more slowly. Not a metaphor, but the same logic at a different scale.
Here’s how I read it: when I first started reading Levin, I put it to myself this way — death begins where a cell loses its reason to exist. I thought that was my own loose interpretation: pretty, but mine. It turned out to be almost word for word his position: a system with no new goal begins to break down. The difference between “a cell loses its meaning” and “a goal-directed system is deprived of its goal” is stylistic, not substantive. One of those rare cases where the metaphor you’d want to write off as poetry turns out to be an exact restatement of the result.
Why This Matters Now
Suppose all of this holds. What changes?
Medicine first — and the gap between a worm and a human is a chasm. These are directions, not cures.
If the shape of the body is held in a rewritable electrical layer, then in principle you could repair what’s broken not by micromanaging every gene, but by addressing the pattern — giving the tissue the instruction to “build back what’s missing,” the way a planarian builds itself back. Levin notes this directly: electrical physiology is an easier target for treatment than genetics, and many drugs that act on ion channels are already approved for humans. Hence the dream of restoring organs and limbs not by transplant, but by restarting their own plan.
Cancer looks different in this frame. If cancer is a cell that left the chat, then the task isn’t only to kill it but to restore the shared signal — to bring the cell back into the conversation. Levin’s lab has shown that malignant behavior can be suppressed by normalizing the bioelectric state, without touching the oncogenes themselves.9Chernet, B. T., & Levin, M., Transmembrane voltage potential is an essential cellular parameter for the detection and control of tumor development in a Xenopus model. Disease Models & Mechanisms, 2013. 6(3): 595-607. https://doi.org/10.1242/dmm.010835. This is far from the clinic. But it’s a different angle on a disease that for a hundred years has been attacked almost entirely through “kill the bad cells.”
And aging, which we just reached: if it’s a collective’s loss of goal, then the lever isn’t only repairing damage but restoring the goal-directedness itself. How to do that in a human is an open question; the honest thing to say is that this is a hypothesis looking for a test, not a protocol.
But the part that interests me most isn’t medicine. It’s the question Levin asks about the worm — the same question that, right now, from the other side, artificial intelligence is putting to us point-blank.
It’s one question: where does organized, goal-directed behavior come from — and does “mind” really require a brain, neurons, a human? Levin shows that goal-directed behavior exists in a collective of cells with no brain at all. And he doesn’t stop at biology — in recent work he carries the same frame straight onto artificial systems: mind as the capacity to navigate a space of problems, a property he traces in one continuous sweep from regenerating tissue to neural networks and on to human–AI hybrids.10Solé, R., Seoane, L. F., Pla-Mauri, J., Bennett, M. T., Hochberg, M. E., & Levin, M., Cognition spaces: natural, artificial, and hybrid. arXiv:2601.12837, 2026. https://arxiv.org/abs/2601.12837. His broader 2026 framework sets this goal-directedness out as a testable, operational program rather than a metaphysics.15Levin, M., & Resnik, D. B., Mind Everywhere: A Framework for Conceptualizing Goal-Directedness in Biology and Other Domains—Part One. Biological Theory, 2026. https://doi.org/10.1007/s13752-025-00523-6.16Levin, M., & Resnik, D. B., Mind Everywhere: A Framework for Conceptualizing Goal-Directedness in Biology and Other Domains—Part Two. Biological Theory, 2026. https://doi.org/10.1007/s13752-025-00524-5. Those hybrids, he warns, carry their own risks — up to a loss of autonomy. The worm biologist turns up at the center of the AI conversation because he offers a vocabulary in which a worm, a neural network, and a human are three points on one spectrum.
And then he sharpens the question — and this is where I stop waving it off. Levin takes not biology but bubble sort: a sorting algorithm first-year students write, a system we understand all the way down. And he finds in it something the code never put there.11Zhang, T., Goldstein, A., & Levin, M., Classical sorting algorithms as a model of morphogenesis: Self-sorting arrays reveal unexpected competencies in a minimal model of basal intelligence. Adaptive Behavior, 2025. 33(1): 25-54. https://doi.org/10.1177/10597123241269740. When it hits a “stuck” element that won’t move, the algorithm temporarily works against the goal — undoing what it had already sorted — in order to get around the obstacle and then return to the goal; no instruction saying “if stuck, try another way” is written anywhere in it. And a second thing: when different elements run different versions of the rule, they cluster with their “own kind,” though nothing in the code forces them to. For decades this algorithm was studied and no one noticed. Levin’s argument is blunt: if a system where we know every rule down to the last one produces competencies no one specified, then something is entering it beyond the specification.
Same with the xenobots, from the other end. Biology explains an organism’s abilities by its history of selection: this trait stuck because it helped survival. But xenobots never existed as such in evolution — there was nothing to select them directly. And yet they have abilities: to move, to combine, to solve simple navigation tasks. Where from, if no selection history shaped these particular forms? Here Levin’s recent evolutionary work sharpens the question rather than dodging it: evolution may not select finished traits so much as shape agential, problem-solving material — cells and tissues already competent at reaching goals — whose competencies then surface in embodiments evolution never saw.17Hartl, B., & Levin, M., What does evolution make? Learning in living lineages and machines. Trends in Genetics, 2025. 41(6). https://doi.org/10.1016/j.tig.2025.04.002. Either way the old reading isn’t enough; there’s a source of properties that has to be studied in its own right. He takes that second path — because it yields a research program, and “well, it just happened” yields nothing.
So the question I started with — where does goal-directedness come from — turns out to be far sharper than I thought. Not “can mind appear without a brain,” but: why does organized, goal-seeking behavior turn up even where, by our own accounting, there’s no room for it.
What follows is no longer reportage. What I’m about to say, Levin doesn’t claim in this form, and I have no proof of it. But here, on this question, his caution stops holding me back.
Here’s how I read it — and here I stop translating Levin and place a bet of my own. My bet: this looks less like a rare, accidental “emergence” and more like something structural — goal-directed organization seems to want to appear at the first opportunity. I could be wrong; it could turn out to be a beautiful overreach. But if it’s right, there’s a direct consequence that won’t let go of me: “being a self” may not be a switch that’s either on or off, but a matter of degree and connectedness. And then the line between “real” minds (us) and “merely” machines is of the same kind as the lines of shape and individuality that turned out, in the worms, to be movable.
Beneath all of this lies a fork the entire AI world is walking through right now, usually without naming it aloud. There are two ways to answer where this behavior comes from — and at this fork is where I meant to honestly get off Levin’s train.
Where This Comes From: Two Camps
At the start I promised to tell you where I slow down myself. The fork we’ve reached is exactly that place.
The first camp is emergentism. It’s the unspoken default among most of the people building AI today. It says: organized behavior arises from physical dynamics, and there’s no need to introduce extra entities. A complex system under the right conditions produces complex behavior — that’s all; mind doesn’t come from anywhere, it assembles from the bottom up, out of the interaction of parts.
The second camp is where Levin’s position leans, and it’s closer to Platonism — though the distinction matters: his peer-reviewed framework is operational and pragmatic about goals and minds, and the Platonist step shows up mainly when he asks what kind of space these reusable patterns of body and mind belong to. The word has a bad reputation. In ordinary use, “Platonism” gives off a whiff of the half-mystical — “numbers are more real than things,” a world drawn up by some higher design; the reflex is to hold it at an ironic arm’s length. That reputation is more complicated than it looks, and it’s worth setting aside for a moment.
Because underneath it lies a thoroughly domestic intuition. The Pythagorean theorem was true before Pythagoras — and will stay true if every human vanishes tomorrow. Anyone who’s ever struggled with a problem and then suddenly found the answer knows the feeling from the inside: the answer seems to have already been there, not invented but found. Mathematicians talk this way — theorems are “discovered,” not “made up.” That’s Platonism in the seed: something exists independently of whether we’ve stumbled onto it.
And this is neither a fringe view nor a crank’s. The position has been debated for two and a half thousand years; Gödel held it, the Nobel laureate Penrose holds it, Einstein leaned toward it. Almost anyone who works seriously with numbers behaves like a Platonist, even without announcing it. The thing trailing a reputation for esoterica is, once stated carefully, fairly mainstream.
The trick is that one word hides claims of very different boldness — from the entirely mild to the genuinely contested. And if you lay Levin’s position out as a set of steps, something odd shows up: the bad reputation is pinned to the whole staircase, while the objections, in fact, attach only to its top. The lower steps are near-consensus. Let’s walk up them.
- Mathematics exists independently of us. This is that same intuition about the Pythagorean theorem, already named. There’s almost nothing to argue with here — it’s a calm, respectable step to stand on.
- Not only numbers are real, but structures — stable patterns of relation. A crystal lattice is real as a structure, even though each individual crystal only embodies it, not exhausts it.
- Patterns of a broader kind are real too — biological forms among them. The shape of an organism is a stable pattern in a space of possible configurations, real in the same way the lattice is.
- A pattern is embodied through a concrete channel. This is what we’ve already seen. The bioelectric pattern is the path by which form “enters” tissue. This isn’t theory: change the pattern and the form changes. That “saved file” from the worm story is exactly this — form-as-pattern, entering matter through a measurable channel.
Up to here — through the fourth step — I follow Levin without much trouble. And not because he talked me into it, but because these four steps are an almost natural extension of ordinary mathematical intuition, and his experiments support them. If you, like me, are inclined to think that a theorem is found and not made up, you’ll probably climb these steps with me. Which, by the way, flips the usual charge on its head: Platonism here isn’t the exotic thing, it’s nearly the default; the exotic step is still ahead.
And then comes the fifth step — and it has to be cut in two, because one number hides two very different claims.
The weak fifth. A collective of cells behaves agentively — as a whole not reducible to the sum of its parts: it remembers a shape, corrects errors, finds workarounds, moves toward a goal. This is surprisingly solid empirically. That same bubble sort presses right here: the language of goals and attempts is nearly impossible to strip out of any description of what the system does. I take this step too — the data push toward it harder than a cautious skeptic would like.
And here, having climbed the lower steps, it’s easy not to notice the pull upward — to the very top, where the picture finally seems whole. I felt the pull too. And the next step is the one easiest to slip past without noticing you’ve taken it.
The strong fifth. And then the claim that the patterns themselves have agency — that they exist in a Platonic space as something like subjects. Levin’s own committed language here is receptive: we don’t build minds, we facilitate their “ingression” into matter; we are not bodies that patterns act upon, “we are the patterns, ingressing into physical bodies.”12Levin, M., Ingressing Minds: Causal Patterns Beyond Genetics and Environment in Natural, Synthetic, and Hybrid Embodiments. PsyArXiv preprint, 2025. https://doi.org/10.31234/osf.io/5g2xj_v3. He has a name for it — “ingressing minds,” the title of one of his preprints. He goes one step further only as a hypothesis, floated for future work: that evolution itself might be re-cast as “agential patterns seeking embodiment.” That last, strongest version — patterns that actively want a body and seek one out — is the most contested claim in the whole picture, and it’s right here, reading, that I meant to get off: this looked like the wild leap, the place where a careful person hits the brakes and says “this far and no further.” I was sure this was my stop. And then I worked through it — and there turned out to be almost nowhere to get off. Here’s how that happened.
Where I Meant to Get Off
Three things converged, and each one moved me further than I expected.
First — bubble sort stops being a figure of speech. Let me say what it actually is, because the whole force of the example is in its plainness. Bubble sort is the simplest way to put things in order — like arranging books on a shelf by height. You go down the row and look at two neighbors at a time: if the left book is taller than the right, you swap them; if not, you move on. Reach the end, go back to the start, pass through again. And so on until a whole pass makes no swaps at all: everything’s in order by height. Four rules, not a drop of intelligence — a schoolkid gets it in a minute.
And in exactly this procedure — one a human wrote from the first line to the last, where there’s no “misunderstood nature” to hide in — something shows up that no one put there. Running into a stuck element that won’t fall into place, the program backs up on its own — undoing work already done, so it can take a different route afterward. There’s no “if stuck, try a workaround” rule in it. The usual reply here: these are just consequences of those same four rules that we failed to see in advance. Fair enough. But the more cases pile up where a system we understand through and through is still easiest to describe with “it tried,” “it found a workaround,” “it chose” — the harder it gets to hold on to that “just.”
Second — and this is what moved me furthest. Following the familiar, sober logic, I expected the natural objection to live right here, and I had it ready in advance. It goes like this: there’s no special “world of forms”; there’s just a gigantic list of all the ways things can possibly behave, and we know that list badly. The oddities of bubble sort aren’t news from “on high” — they’re a piece of that list we’d overlooked. Sober, and I meant to stop there. But Levin had a counter-question I hadn’t seen coming. He asks: this “list of ways” — is it only in our description, on paper, or does it really exist and really shape what happens? If it shapes things for real — then it is precisely what a Platonist calls his “world of forms.” The same thing, named in two vocabularies. What I’d taken for an exit turned out, on a closer look, to be not an objection to Levin but his own thought said another way.
But honesty demands showing where this move runs into a wall — and here the emergentist stands on solid ground. He can answer: the list of ways doesn’t act — it’s only our way of writing out in advance what follows from the rules anyway. It’s real the way a blueprint is real — but a blueprint builds nothing on its own. Hold to that, and Levin’s redescription doesn’t touch you: it only catches someone who has already granted that the list works on its own. And that’s where I saw that the drop I’d been walking toward hadn’t gone anywhere — it had just turned out to be lower down and narrower than I expected. Not “is there anything real up there at all” — there’s something, you can see it in the same bookshelf — but “does that real thing act on its own, or only describe?”
Third — interactionism turns out to be a step, not a leap. The objection “you can’t introduce the non-physical influencing the physical” Levin disposes of like this: we already live in such a world. Mathematical truths aren’t physical, yet they steer the physical — and his own phrasing is that the mind-body relation is “the same as between math and physics,” an interactionism we’ve had, he says, since Pythagoras. He points directly at biology’s use of the primes: life exploits mathematical facts that nothing physical can edit. The clearest illustration I know is the periodical cicadas, which emerge on 13- and 17-year cycles; a widely accepted hypothesis holds that a prime period is the hardest for a predator or parasite to lock onto, so the distribution of the primes — not a physical fact, not editable by tweaking any constant — ends up shaping when the insect crawls out of the ground.14Goles, E., Schulz, O., & Markus, M., Prime number selection of cycles in a predator-prey model. Complexity, 2001. 6(4): 33-38. https://doi.org/10.1002/cplx.1040. (That cicada case is my gloss, not his line; the principle under it, that non-physical structure can be causally effective, is his.) Once the non-physical already steers the physical, adding “patterns” to it is a step inside a picture we’ve already accepted, not a leap off its edge. And one more thing closes the “this is where science ends and faith begins” I’d braced for: Levin is careful about what he does not claim — no religious assertions, and he says plainly the model may be wrong, that it’s the metaphysics he finds most useful “for whatever that’s worth.” This is a scientist with a radical hypothesis, not a believer hunting for justification.
So the map I ended up with isn’t the one I came to the staircase holding. The drop I expected at the top — “is there anything real up there at all” — receded: there’s a real, causally-effective excess in bubble sort, by whatever name, and both sides agree to see it. But that doesn’t mean what’s left at the top is pure semantics. The fork just shifted down a step and narrowed: does that structure act on its own, or are we only describing what follows from the rules? I cross that fork toward Levin, because for me the bubble sort and the redescription move outweigh the alternative. But I cross it deliberately, as the choice of a premise, not a proven theorem. And higher up there’s a question of degree as well. The strongest form of the claim — patterns as subjects with something like a will, actively seeking something to flow into — is still more than the data strictly require.13Levin, M., A Platonic Space background for questions in consciousness. Thought Forms (Substack), 2026. https://mlevin77.substack.com/p/a-platonic-space-background-for-questions. (And yes, embodiment through a channel can be described without ascribing any will to the patterns at all.) On this now-double fork — “acts or describes,” and “how much agency” — the emergentist isn’t defeated; he’s entitled to hold the driest formulation. But notice what kind of stop this is. Not “here runs the border of the real”: the real is there. It’s “here I choose whether to count the structure as acting, and what word and what dose of agency to give it.” The argument has narrowed from “is there anything there” to “does it act, and what do we call it,” and at that narrow spot I take a step honestly, without hiding that it’s a step.
And one thing I’m obliged to say out loud. I updated toward Levin over the course of this reading more than I expected to. Worth asking: from the material, or because the Platonist intuition is close to me to begin with? Mostly from the material, I think — the bubble sort and the redescription move are concrete, they’re not a matter of taste. But I name it so you can correct for it.
Does “Einstein” Hold
Now, the promised bracket — along the same staircase.
At the level of the move — yes: the steps from the first through the weak fifth rebuild the very categories we think about life, mind, and individuality in. Not a louder fact inside the old picture, but a different picture — a move of the Einstein kind. Whether it reaches human scale and the clinic — the future hasn’t cashed that check yet; the honest word here is “we’ll see.” And at the top, where I expected a drop, there was no drop — at least not the one I expected. What had looked like a wild leap — letting non-physical structure be causally effective — turned out, on a closer look, to be an increment on a picture we already hold; the question “is there anything real up there” stopped being the cliff I’d expected. The strongest, teleological version stays open, by his own lights as much as mine. So the comparison holds further up the staircase than I thought, rather than breaking at the very top.
But honesty demands the second half right away. The drop didn’t vanish — it moved lower and narrowed. What’s left is a choice: whether to count the causally-effective structure as acting on its own or only as our description of what’s derivable — and how much agency to grant it. I make that choice toward Levin, and I don’t hide that it’s a choice. The emergentist who doesn’t make it isn’t defeated — he’s standing on honest, if ever-narrower, ground. A strong idea is tested by this too: the argument around it doesn’t disappear, it sharpens — sliding from “is this real” to “what do we call it, and what does it do.”
And here’s what matters more than the label. The real test was never the word “Einstein.” It’s this: can you now look at a cut worm, at a tumor, at your own aging, or at the AI on your screen the way you did before this essay. I can’t. I walked toward the top step meaning to honestly get off it — and found the drop there wasn’t the one I’d been looking for: not a chasm of “the real,” but a fork of vocabulary and degree, which I stepped across on my own two feet. The lower steps had already changed my questions, and the top one turned out to be not an edge but a crossroads — where I picked a side, knowing that I was picking.
How far to climb, and whether to count the top as acting, and what word to call it by — that’s yours to decide. Only do it knowingly: aware that the argument about Levin seemed unresolvable for so long precisely because the people arguing didn’t notice how the question “is there anything there” had quietly turned into “what exactly is there, and what do we call it.” You have the map now.