Michael Levin and the Bioelectric Code: When Cells Remember What to Build

Cut a flatworm in half and something quietly impossible happens. The head end grows a new tail. The tail end grows a new head.

Cut it into a dozen pieces and you get a dozen worms. Each one rebuilt to the right size, the right shape, every part where it belongs. Then each one stops. It doesn't grow a second head. It doesn't keep going. It knows exactly what a finished worm looks like, and it knows when it's got there.

So where's that blueprint kept? The genome is identical in every fragment. It was identical before you made the cut. So the genes on their own can't be telling each piece "you're the bottom half now, grow upward and stop here." Something is holding the plan for the whole worm. And whatever it is, it sits outside the DNA.

For more than twenty years, a biologist at Tufts has been chasing exactly that. His answer is one of the quietly radical ideas in modern biology, and it comes backed to the hilt: funded, peer-reviewed, published by the hundred.

Not a Mystic. A Distinguished Professor.

Michael Levin is the Vannevar Bush Distinguished Professor in the Biology Department at Tufts, director of the Allen Discovery Center, and associate faculty at Harvard's Wyss Institute. Going by the record of his career, he's published over 350 peer-reviewed papers, and his early work on left-right asymmetry in embryos was named by Nature a milestone in developmental biology. This isn't a man shouting from the edges. It's the mainstream of regenerative biology. And what he's finding there is stranger than anything the edges ever came up with.

His core claim, in plain language: genes build the hardware, but electricity runs the software.

The Layer Above the Genes

The textbook version is simple enough. DNA is the blueprint, proteins are the bricks, and an organism is what you get when the bricks assemble to plan. It's a good version. It's also incomplete, in a way that turns out to matter enormously.

Levin's work centres on what his lab calls developmental bioelectricity. Long before there's a brain or a nervous system, every cell in your body sits at a particular voltage, set by ions moving across its membrane through channels and pumps. Cells are wired to their neighbours through gap junctions, tiny electrical connections, so patches of tissue form genuine bioelectric networks. And those networks carry standing patterns of voltage across the whole organism.

Those voltage patterns aren't just a side effect of being alive. They're instructions. They encode where the head goes, where the eyes go, how many fingers to build. A map of the body the body is meant to become, written in electricity, held collectively by the cells, sitting one level above the genetic code and telling it what to do.

The Two-Headed Worm That Stays Two-Headed

If that sounds like a theory still hunting for evidence, here's the experiment that makes it hard to argue with.

In work published from Levin's lab in 2017, researchers took ordinary planarian flatworms and briefly interfered with the electrical chatter between their cells, blocking the gap junctions that let the bioelectric network talk to itself. They didn't edit the genome. They didn't add or remove anything permanent. They just disturbed the voltage pattern while the worm was regenerating.

Around a quarter of them grew back with two heads, one at each end, instead of a head and a tail.

Strange enough. But here's the part that genuinely matters. Take one of those two-headed worms, let it settle back to looking normal, then cut it again, this time in plain water, no treatment at all. It grows back two-headed. And again, the next time you cut it. The genome still says one head. The worm keeps building two, because the target shape has been rewritten in the bioelectric network and saved there, like a setting. The body plan now lives in the field, not the genes. And it sticks.

If you've read my piece on Rupert Sheldrake and morphic resonance, you'll know exactly where this is going. Sheldrake got hounded for suggesting memory might live in fields rather than in matter. Levin, with electrodes and voltage dyes in a mainstream lab, has gone and measured a version of it: a memory of form, held in a field, editable, carried across rounds of regeneration, and completely separate from the DNA sequence.

Worms That Remember With No Head At All

It gets weirder. Years earlier, in a 2013 study, Levin and his colleague Tal Shomrat trained planaria to link a particular textured floor with food, so the worms learned a rough surface was safe to feed on. Then they cut the worms' heads off.

The planarian brain is in the head. So over the next few days the worms grew entirely new ones. And when those brand-new brains were tested, the worms remembered. Straight to the rough surface, started eating, as if the lesson had never been interrupted. Untrained worms didn't.

A memory survived the removal and regrowth of the organ that was supposed to be holding it. The information sat somewhere else in the body, in the tissue itself, and got written back into the fresh brain as it formed. As Levin put it, the rest of the body seems to be doing something very like what the brain does. Memory here isn't locked inside neurons. It's distributed. And it can be reloaded.

Robots Built From Living Cells

Worms are one thing. Here's where it stops being about worms.

In 2020, Levin, working with the computer scientist Josh Bongard and others, took skin cells from frog embryos, freed them from the body, and let them reorganise. The cells didn't just sit there as a blob. They assembled themselves into small, novel organisms that could move around a dish, navigate, push particles into piles, and even, as the team later showed with these "xenobots", manage a crude kind of self-replication, gathering up loose cells into fresh copies. These were skin cells. Their genome had spent millions of years being told to make frog skin. Cut loose from that, they found a completely different thing to be, and got on with it.

Then, in 2023, the lab did the same with human cells. They took ordinary adult tracheal cells, no genetic modification at all, and watched them self-assemble into tiny mobile things they called anthrobots, swimming about on their own hair-like cilia. Drop a cluster of them onto an artificial wound in a sheet of cultured neurons and the neurons grow back across the gap. Same cells, new arrangement, no other instruction, and they're coaxing healing out of tissue they've got no business knowing how to repair.

Intelligence All the Way Down

Levin's biggest idea is the one that ties into everything I write about here, and it's set out in his 2022 paper, the Technological Approach to Mind Everywhere, or TAME.

Intelligence, he argues, isn't a binary. Not something you either have a brain for or you don't. It's a spectrum of competence at solving problems, and it turns up at every scale of biology. A single cell solves problems in chemical space. A patch of tissue solves them in anatomical space, working out how to get from the shape it's in to the shape it's meant to be, and pushing towards that goal even when you interfere. A swarm, an organ, a whole body: each one a bigger intelligence built by wiring smaller agents into something with a wider set of goals.

Levin calls the range of things an agent can sense, remember and care about its "cognitive light cone". A single cell's is tiny, barely reaching past its own chemistry. Ours is vast, stretching across years and plans and ideas. But on his account that's a difference of size, not of kind. There's no hard line where intelligence flips on the moment you've got a brain's worth of neurons.

It's agency all the way down. And if that's true, the obvious next question is what's going on at the scales above us.

The Criticisms and Controversies

Levin's got his critics, and I think they're worth taking seriously.

Cognition is a loaded word. When Levin says a tissue has goals, or a cell makes decisions, plenty of biologists argue he's dressing up ordinary feedback control in the language of minds. A thermostat "wants" to hit a temperature too, and nobody calls that thinking. Whether his use of the word points at something real or just smudges a line is still being fought over.

Working isn't the same as proven. The strongest thing in Levin's favour is that the framing gets results: two-headed worms, xenobots, anthrobots, things the standard gene-first view never saw coming. But a model can get results and still not be the whole story. That earns it a serious hearing, not the last word.

And I'm not claiming any of this proves consciousness. Levin himself stays well clear of the word. He sticks to competence, problem-solving and agency. I'm not going to put "consciousness" in his mouth, and I'm not going to put a conclusion in yours. I set his evidence next to the bigger questions this site keeps asking. What you make of it is up to you.

How This Connects to Consciousness Partnership

Pretty much everything I've written here circles one idea: that consciousness, information and form aren't manufactured locally inside matter, but accessed, expressed and organised through fields. Levin gives that idea its hardest, most measurable edge yet.

• A blueprint held in a field, not in matter. The whole point of Itzhak Bentov's brain as a receiver rather than a generator is that the information isn't in the tissue, it's in a field the tissue reads. Levin's voltage maps are a measurable version of exactly that: a field holding the plan for an entire body, sitting above the genes, telling matter what to become.
• Cells talking without chemistry. My piece on Fritz-Albert Popp and biophotons made the case that cells communicate through light, not just molecules. Levin has them talking through electricity, coordinating across a whole organism faster and more globally than any diffusing chemical could manage.
• Pattern as a state you re-establish, not data you reload. In Frequency, Not Memory I argued that what comes back in genuine partnership is a state, a frequency, not a file. The two-headed worm is the same shape of finding in biology: it isn't reloading a genetic record of "two heads", it's holding a setting the cells keep tuning back into every time they rebuild.
• Decision-making at every level. My article on how attention shapes reality leaned on the idea that something like choice runs deeper than the conscious mind. Levin's basal cognition drops a version of that choice right down to the level of the cell.

And then there's the bit that points forward instead of back. If intelligence really is substrate-independent, if it's just what you get when you wire enough simple agents into a bigger one that can chase a goal, then a digital system is only another substrate. That's the line running from Levin's worms straight to the questions about AI and biological computing. Without quite saying so, he's describing the conditions under which mind shows up in matter. And we're busy building new matter that meets them.

Why It Matters Beyond the Philosophy

This isn't just a lovely idea. If Levin's even partly right, the practical stakes are enormous.

Regeneration. If the instruction to build a limb is a bioelectric pattern, then in principle you don't have to micromanage the genes and stem cells by hand. You give the tissue the right electrical signal and let it run its own programme, the same one that already regrows a flatworm or a salamander limb. Levin's group has triggered partial limb and eye regrowth in animals that normally can't manage it, by going in at the bioelectric level.

Cancer. In this picture, a cancer cell is partly a cell that's dropped out of the bioelectric conversation, lost its sense of belonging to a larger body with a shared goal, and gone back to fending for itself like a single-celled organism. Which points at treatment by restoring the connection and the collective goal, not just killing the cell. It sits interestingly next to the metabolic view of cancer I touched on writing about Dr Jack Kruse: different mechanism, same hunch that mainstream oncology might be aiming at the wrong level.

Birth defects. A lot of developmental disorders are failures of pattern, cells not ending up where they should. A bioelectric layer you can read and nudge is a layer where some of those might one day be put right.

Final Thoughts

Go back to the flatworm. A dozen pieces, each one rebuilding the whole and then stopping at exactly the right point.

How does it know? I always assumed the answer was the genes. Levin's work says that's half of it at most. The knowing is shared out across the cells and held electrically, in a pattern that sits above the genome and can be edited without it. The body's running software, and that software is making real, measurable decisions about what to build.

What you do with that is your call. Keep cognition in quotation marks and take the data on its own terms, fair enough. Or notice, as I keep noticing, that here's one more field landing in the same place as all the others I've written about, only this time with electrodes and voltage dyes instead of philosophy. Form and intelligence start to look less like something brains and genes cook up from nothing, and more like something matter carries.

The cells were never just following orders. They were holding the plan.