> Core intuition: We assume our body's shape is determined by genes—like a blueprint specifying a house. But Michael Levin argues that the signals coordinating the construction crew are not the DNA blueprint, but the network of electrical signals between cells. Rewriting these bioelectric signals is like changing the construction crew's walkie-talkie channels—it can make them build entirely different structures, even regrow a broken wall on its own.
---
1. How a Computer Programmer Sees Biology
Michael Levin's background is unusually non-biological. He earned a double degree in computer science and biology, then a Harvard PhD in genetics—but his career has been driven by one question:
> "If intelligence can run on different physical substrates, why assume it exists only in the brain?"
This led him from code to cells. In 2000, he founded his lab at Tufts University with research directions that sound radical:
- Organ regeneration—regrowing lost limbs
- Cancer reprogramming—not killing cancer cells, but "persuading" them to return to normal
- Non-genetic body-plan modification—making flatworms grow two heads by changing only electrical signals
- Synthetic living machines—programmable micro-robots built from frog cells
- No gene editing—only small-molecule drugs altering electrical signals
- Adult frogs—individuals that had already lost regenerative ability
- Functional recovery—the new limbs worked for swimming and movement
- 24-hour treatment—not continuous dosing, but a short bioelectric environment reset
- Autonomous movement—driven by rhythmic heart-cell contractions
- Environmental interaction—obstacle avoidance, particle aggregation
- Self-repair—cut in half, each half reorganizes into a smaller Xenobot and keeps moving
- Designability—AI designs different forms for different functions
- Altering tumor-cell voltage states can suppress tumor growth
- Certain bioelectric patterns can induce cancer cell differentiation—reversing runaway division back toward normal cell states
- Bioelectric changes can activate tumor-suppressor pathways without drugs
- No chemo side effects—no DNA changes, no cell killing, only persuasion
- Low resistance—no selective pressure, so cells needn't "evolve resistance"
- Systemic repair—bioelectric signals coordinate globally, potentially fixing multiple tissues at once
- Automated morphological inference: using AI to infer shape-control rules from bioelectric data
- Extended connectionism: cells as processing units in computational networks, not isolated nodes
- Embodied intelligence: brainless Xenobots solving problems—challenging whether intelligence requires neural networks
- Synthetic bio-design: AI-designed organisms for drug delivery, environmental remediation
- DNA is not program code—code executes sequentially; DNA is data and template. The real "computation" happens in cell networks and bioelectric signals
- Form is not the output of a blueprint—blueprints are static; biological form is dynamic, adaptive, self-repairing—an emergent property of distributed computing
- Regeneration is not "re-running the blueprint"—it is the cell collective recomputing the target morphology, and that computation can be redirected by external signals
- Not "how do genes control development" but "how do cells collectively decide what to become"
- Not "how to kill cancer cells" but "how to remind cancer cells who they are"
- Not "how to repair a severed limb" but "how to reactivate a dormant regeneration program"
- Not "what is life" but "how does information processing emerge across different physical substrates"
- Durant, F., et al. (2017). "Long-Term, Stochastic Editing of Regenerative Anatomy via Targeting Endogenous Bioelectric Gradients." *Biophysical Journal*.
- Pai, V.P., et al. (2012). "Transmembrane voltage potential controls embryonic eye patterning in Xenopus laevis." *Development*.
- Kriegman, S., et al. (2020). "A scalable pipeline for designing reconfigurable organisms." *PNAS*.
- Murugan, N.J., et al. (2022). "Acute multidrug delivery via a wearable bioreactor facilitates long-term limb regeneration and functional recovery in adult Xenopus laevis." *Science Advances*.
- The Levin Lab, Allen Discovery Center at Tufts University: https://drmichaellevin.org
- Michael Levin biography: Vannevar Bush Distinguished Professor, Tufts University; Associate Faculty, Harvard Wyss Institute; 400+ peer-reviewed publications.
His lab has published 400+ peer-reviewed papers, including work Nature called a developmental biology milestone. His most publicly famous creation is a sci-fi-sounding word: Xenobots—the world's first "living robots."
---
2. The Core Idea: Genes Are the Blueprint, Bioelectricity Is the Walkie-Talkie
Traditional biology's narrative: DNA encodes proteins, proteins determine cell structure and function, and ultimately the body's shape. Genes are the blueprint.
Levin's insight: DNA is just the blueprint; what coordinates the construction crew is the network of electrical signals between cells.
What is bioelectricity?
All living cells have a voltage difference—ion concentration gradients across the membrane create a tiny electric field. This voltage is not exclusive to neurons; every cell has one.
When many cells organize into tissue, their voltages form a spatial voltage gradient. This gradient is not random—it precisely maps the location and form of future organs.
> Levin's analogy: bioelectric signals are the cell's "software"; genes are the "hardware." > > The same hardware running different software can do entirely different things.
Key experiment: two-headed planarians
Planarians can regrow heads: cut one in two, and the tail segment grows a new head because its anterior voltage gradient gets recoded.
Levin's team went further: without touching genes, just altering voltage—using drugs to change cell voltage at the cut surface:
> The tail segment grew two heads—not two heads sharing one body, but two complete head structures, including brains and eyes.
Paper: Durant et al. (2017). "Long-Term, Stochastic Editing of Regenerative Anatomy via Targeting Endogenous Bioelectric Gradients." *Biophysical Journal*.
This is not "gene mutation causing deformity." The voltage network was rewritten, so the construction crew's instructions changed.
More astonishing: when these two-headed worms were cut into pieces, every piece regrew as two-headed. The new phenotype was stably inherited—with zero DNA change.
Levin calls this "morphological memory": the bioelectric network forms a stable computational state that self-maintains and self-replicates, entirely independent of the genome.
---
3. Ectopic Eyes: Growing Eyes Where They Shouldn't Grow
Another signature experiment: ectopic eyes—inducing complete eyes on frog embryo gut tissue.
In normal development, eyes form only at specific head locations, determined by the spatiotemporal expression of genes like Pax6. Levin found:
> By electrically altering the voltage state of gut cells, those cells begin expressing Pax6, then form complete retinas, lenses, and optic nerves—everything an eye needs.
Paper: Pai et al. (2012). "Transmembrane voltage potential controls embryonic eye patterning in Xenopus laevis." *Development*.
The key: gut cells and eye cells have identical DNA. If genes were the only determinant, gut cells could never become eyes. Bioelectric signals changed the cells' gene expression pattern—not the genes themselves, but "which gene turns on when."
This supports Levin's core theory: bioelectricity is an upstream controller of gene expression patterns, not a downstream byproduct.
---
4. Limb Regeneration: The Wearable Bioreactor
This is Levin's most human-relevant work.
Adult African clawed frogs (Xenopus laevis) cannot regenerate limbs—tadpoles can, adults cannot. That is the fate of most vertebrates.
Levin's team designed a wearable bioreactor—a sleeve-like device fitted over the amputation stump, loaded with five small-molecule drugs that diffuse into tissue and rewrite the bioelectric environment of the wound.
The result:
> Within 24 hours, the wound's voltage gradient was reprogrammed. By day 7, clear skeletal structures appeared. After 9 months, the frog had regrown a nearly complete hindlimb—bone, muscle, nerves, and skin.
Paper: Murugan et al. (2022). "Acute multidrug delivery via a wearable bioreactor facilitates long-term limb regeneration and functional recovery in adult Xenopus laevis." *Science Advances*.
Key points:
Levin's interpretation: the frog's DNA always contained a "regeneration program," switched off in adulthood. Altering bioelectric signals reactivates this dormant program.
---
5. Xenobots: Programmable Living Machines from Frog Cells
In 2020, Levin and Josh Bongard (University of Vermont) unveiled Xenobots—the first fully designable micro-machines made of biological cells.
The process: 1. Extract skin and heart cells from frog embryos 2. Use AI to simulate millions of cell combinations, predicting which would perform a task (movement, carrying, aggregation) 3. Assemble cell clusters according to the AI-designed blueprint 4. The clusters self-organize into living structures that move autonomously and self-repair
Paper: Kriegman et al. (2020). "A scalable pipeline for designing reconfigurable organisms." *PNAS*.
Xenobots' capabilities:
In 2021, Xenobots 2.0—capable of self-replication. Not by growing, but by collecting loose cells from the environment and assembling new Xenobots, like a Karel-style puzzle shape.
Levin calls these not "robots" but "designed organisms"—made of life itself, not steel and silicon. No nervous system, no brain, yet the cell collective exhibits goal-directed behavior.
This raises a deep philosophical question: if a group of cells without neurons can solve problems, adapt, and self-repair, is our definition of intelligence too narrow?
---
6. Cancer Reprogramming: Persuade, Don't Kill
Levin's view of cancer is also contrarian.
Traditional treatment kills cancer cells—chemotherapy, radiation, targeted drugs: identify and destroy.
Levin's approach: cancer cells are not "evil," they are "lost"—they've forgotten instructions to stop dividing, maintain normal form, and participate in the collective. If electrical signals can "remind" them, they might normalize.
Experimental evidence from his lab:
Levin calls this "electroceuticals"—precisely controlled electrical signals that "reprogram" cell behavior instead of force-modifying it with drugs or gene editing.
If successful, this would be revolutionary:
---
7. The Bigger Picture: From Molecular Biology to Biological Computation
Levin's work challenges biology's foundational paradigm.
Traditional paradigm (molecular reductionism)
DNA → RNA → protein → cell structure → tissue → organ → organism
A bottom-up reductionist chain: understand the bottom (genes), understand the top (organism).
Levin's paradigm (bio-computation)
An organism = nested multi-layer information-processing systems
| Level | Information carrier | Timescale | Tools | |-------|--------------------|-----------|-------| | Molecular | DNA, RNA, proteins | ms–s | Sequencing, mass spec | | Cellular | Bioelectric signals, chemical gradients | s–min | Voltage dyes, ion-channel imaging | | Tissue | Voltage gradients, mechanical forces | min–h | Multi-electrode arrays, optogenetics | | Organ | Morphogenetic fields | h–days | Morphological computation models | | Organism | Behavior, cognition | days–years | Behavioral and cognitive experiments |
Levin's core claim: every level has its own "computation" and "decision-making." The bioelectric network is a distributed computing system that processes information, makes decisions, and adapts—essentially cognition.
He calls this "basal cognition"—cognition is not the brain's monopoly but a fundamental property of life, from single cells to human societies.
---
8. From Biology Back to AI
Levin's CS background is more than a résumé line—his team works at the AI-biology intersection:
Levin's grand vision:
> The "Anatomical Compiler"—a software platform where scientists input a target form (e.g., "I need a right hand"), and the system computes the required bioelectric patterns and outputs a treatment protocol (e.g., "apply these voltage gradients at the wound for 24 hours").
If realized, regenerative medicine shifts from "repair" to "design"—compiling organs like code.
---
9. Limitations and Controversies
Levin's work is exciting, but deserves a sober look:
1. The gap from frogs to humans
Planarians and frogs are relatively simple organisms. Whether their mechanisms apply to mammals, including humans, remains a long road. Mouse and pig experiments are underway; human clinical trials have not begun.2. Precision control
Bioelectric signals are global and diffuse. How to precisely control specific tissues and cell types without affecting others? The wearable bioreactor works on a frog stump—but what about internal human organs?3. Long-term safety
Bioelectric networks are complex and coupled. Changing one node's voltage may affect distant tissues. Long-term effects need extensive study.4. From reproducible experiments to standardized therapies
Lab results replicate, but every wound and tumor microenvironment differs. Standardizing personalized electroceutical protocols is a huge engineering challenge.5. Philosophical boundaries
Xenobots' "self-replication" triggered debates about "life" and "machine." They are neither traditional organisms (no evolutionary history) nor traditional machines (no human-made hardware). Society and science have yet to reach consensus on the ethical status of such "designed organisms."---
10. A Computer Scientist's Insight on Biology
> "I was a computer programmer before college. I was interested in the fact that intelligence can run on different physical substrates."
This perspective let him see what biologists might miss:
A new framework: life = matter (hardware) + information (software) + computation (process). DNA stores data, proteins process, bioelectricity communicates, morphology is the computational output. In this framework, the body isn't just grown—it's computed.
---
Conclusion: Rewriting Bioelectricity, Rewriting What's Possible
The shock Levin's work delivers to biology rivals the discovery of the DNA double helix—not because his experiments are more complex, but because his problem framing is entirely different:
Two-headed worms, ectopic eyes, Xenobots, limb regeneration—experiments that sound mythical are reproducible, peer-reviewed papers in Levin's lab.
They point to one conclusion:
> Life is not drawn by the genetic blueprint. Life is "computed" by the bioelectric network between cells. Rewrite the signals, and you rewrite the output.
This is a revolution in medicine (electroceuticals replacing chemo), engineering (living machines replacing silicon), and perhaps philosophy (intelligence as a property of life, not just brains).
The Anatomical Compiler remains distant. But the direction is clear: we may stop "repairing" bodies and start "recompiling" them.
---
References: