Dr. Michael Levin’s research at Tufts University reveals that bioelectric signaling—electrical communication between cells via ion channels and gap junctions—serves as a cognitive medium that stores anatomical memories, guides development, and coordinates tissue behavior. This reframes medicine from a bottom-up molecular approach to a top-down reprogramming paradigm, where diseases like cancer, birth defects, and limb loss can be addressed by resetting the electrical pattern memories of cell collectives rather than micromanaging genes or proteins.
The Anatomical Compiler: A Vision for Future Medicine
Levin’s long-term goal is an “anatomical compiler”—a system where a user draws a desired anatomical structure on a computer, and the system compiles that specification into a set of bioelectric stimuli that instruct cells to build it. This would eliminate birth defects, traumatic injury, cancer, aging, and degenerative disease by convincing cells to construct whatever anatomy is desired.
Current molecular medicine excels at understanding which proteins bind, which genes turn on and off, and what individual cells are doing, but it cannot restore limbs or repair complex structural defects. Levin argues this is analogous to 1940s–50s computer science, where reprogramming required physically rewiring hardware. Today’s medicine similarly manipulates biological “hardware” because it has not yet fully exploited the reprogrammable, intelligent nature of living material.
Intelligence at Every Biological Level
Living systems exhibit problem-solving competence at every scale, not just the organismal level. Single cells (like the protist Lacrymaria) handle all physiological and metabolic tasks without a brain or nervous system. Even below the cellular level, molecular signaling pathways within single cells can form six types of memories, including Pavlovian conditioning—no neurons or even intact cells are required.
Bodies are composed of a “multiscale competency architecture”: molecular networks, cells, tissues, and organs each navigate their own problem spaces. The challenge is understanding how these layers integrate into coherent large-scale outcomes.
Evidence of Developmental Intelligence: The Picasso Tadpole Experiment
To test whether embryonic development is hardwired or adaptive, Levin’s lab created “Picasso tadpoles” (Xenopus laevis) with facial organs scrambled into abnormal positions—eyes on top of the head, mouth off to the side—resembling a Mr. Potato Head arrangement.
Despite these scrambled starting positions, the tadpoles developed into normal frog faces. Each organ moved through novel, unpredicted paths to reach its correct final position, demonstrating that cells and tissues pursue a goal (the correct frog face) rather than executing a fixed genetic script. They stop when the goal is achieved, implying they store a memory of the target anatomy.
Bioelectricity as the Cognitive Glue
Levin draws an analogy to neuroscience: just as neural networks use electrical signaling (via ion channels setting membrane voltages and propagating signals through synaptic connections) to store memories and guide behavior, tissues throughout the body use bioelectric networks to store anatomical memories and coordinate construction.
This system is far older than brains—it evolved around the time of bacterial biofilms. Every cell in the body has ion channels, and most cells have electrical synapses (gap junctions) to neighbors, forming tissue-scale electrophysiological networks.
Using voltage-sensitive fluorescent dyes, Levin’s lab has imaged these bioelectrical conversations in living frog embryos. One key pattern they identified is the “electric face”—a bioelectric map that appears before craniofacial organs form, specifying where the eyes, mouth, and placodes will go. This map literally reads out the electrical memory of what a correct face should look like.
These signals also integrate across multiple embryos: injury to one embryo propagates a bioelectrical wave that neighboring embryos detect, showing that bioelectrical communication merges subunits into coherent wholes.
Reading and Rewriting Electrical Memories
Bioelectric patterns can be read (diagnosing birth defects) and, crucially, rewritten for therapeutics. Levin’s lab does not use external electrodes, fields, or magnets. Instead, they manipulate the natural interface cells use to communicate: pharmacology to open or close ion channels, and optogenetics to control voltage and intercellular communication.
Three experimental stories demonstrate the power of this approach:
Cancer as a Bioelectric Disconnection Disorder
When human oncogenes (e.g., KRAS, p53 mutations) are injected into tadpoles, cells develop aberrant voltages before tumors form. This aberrant voltage causes them to disconnect from the bioelectrical network. Once disconnected, a cell loses access to the collective’s large-scale anatomical goals and reverts to a primitive, amoeba-like state—its “cognitive light cone” shrinks to single-cell concerns. Cancer is framed as a dissociative identity disorder of the cellular collective intelligence.
Instead of killing cancerous cells, Levin’s lab forced them to reconnect by co-injecting an ion channel that restores the correct electrical state. Even with strong oncoprotein expression, no tumor formed because the cells remained connected to the collective and continued working on normal tissue goals. The outcome is driven by physiology and cellular decision-making, not by the mutation alone.
Fixing Birth Defects via Computational Bioelectric Design
Introducing a mutation in the neurogenesis gene notch in tadpoles causes severe brain abnormalities: the forebrain is missing, and the midbrain and hindbrain form a large bubble. These animals have virtually no behavior.
Levin’s lab built a computational model of the normal bioelectric pattern that tells the brain its correct size and shape. The model predicted which ion channels to open or close to restore the correct pattern despite the mutation. They identified existing FDA-approved drugs (used for other conditions in patients) that implement this intervention.
The result: normal brain structure and learning rates indistinguishable from controls, despite the persistent mutation. This demonstrates fixing a “hardware” error (genetic mutation) through “software” intervention (a brief biochemical reset of electrical patterns).
Limb Regeneration in Adult Frogs
Adult frogs do not regenerate legs (unlike salamanders). Levin’s lab designed a drug cocktail applied for only 24 hours that triggered approximately 18 months of leg growth. The cocktail activated pro-regenerative genes (e.g., MSX1) and produced a touch-sensitive, motile leg with toes and toenails.
Critically, no ongoing management was required—no stem cell manipulation, no scaffolds, no micromanagement. A single early signal instructed cells to enter the “leg building” pathway in anatomical space rather than the default scarring pathway. The cells’ own competence did the rest.
Inducing Novel Organs: Eyes on the Gut
Levin’s lab reproduced the bioelectric eye spot pattern from the “electric face” in an abnormal location (the gut) by injecting RNA encoding a specific ion channel into early embryos. The cells built a complete eye—with retina, lens, and optic nerve—on the gut.
This demonstrates that bioelectric signals are instructive at the organ level: no gene-level instructions or stem cell directives were needed. A high-level “subroutine call” (“build an eye here”) was sufficient. The cells are highly competent at executing complex responses to simple signals.
Furthermore, the injected cells recruited neighboring non-injected cells to help build the eye, showing the material can scale itself to match the task—similar to how ants recruit colony members for large tasks.
Anthrobots: Synthetic Living Therapeutics
Levin’s lab has created “anthrobots”—synthetic living constructs made entirely from human adult tracheal epithelial cells. Through a process that reboots their multicellularity, these self-motile creatures form without any genetic modification. Their existence would not be predicted from the human genome alone.
In wound-healing assays, anthrobots placed in a field of human neurons with a scratch wound formed “super bot clusters” and knitted the two sides of the wound together within days.
Potential applications include personalized autonomous therapeutics: made from a patient’s own cells (no immune rejection, no immunosuppressant drugs), biodegrading within weeks, and capable of cleaning joints, seeking cancer cells, delivering pro-regenerative molecules, or repairing neural connections.
The Relationship Between Bioelectricity and Epigenetics
Genetics specifies the hardware—which ion channels and voltage-transducing machinery a cell can have. Everything after that is physiological “software,” which is bioelectrical, biochemical, and biomechanical.
In many cases, tracking genetics, transcriptomics, or proteomics gives the wrong answer about anatomical outcomes. The tadpole brain defect and cancer examples both show that genetic information is insufficient to predict what will happen; the physiological state (bioelectric pattern) is the decisive factor.
Toward a Unified Theory of Biology
Levin argues that biology lacks a unifying theory comparable to physics, and that the correct framework will come not from physics or chemistry (equations, emergence, complexity theory) but from behavioral science—concepts like goals, memories, preferences, and problem-solving.
The fundamental story of life is the scaling of intelligence: how tiny goals of single cells (metabolic, proliferative) are integrated into grandiose goals (building a limb, a face) through cognitive glue like bioelectric networks. A mature theory of biology will be a multi-scale theory of intelligence, and it is already enabling transformative therapeutics.
The Future: Somatic Psychiatry, Not Chemistry
Future medicine will resemble “somatic psychiatry” more than chemistry—communicating with and reprogramming the intelligent subsystems of the body rather than micromanaging molecular states. AI will be essential for interfacing with the body’s different layers.
Levin envisions a software system with an AI front end that communicates in natural human language, fed by wearable sensors and scans. It would allow doctors (or patients) to communicate with organs and tissues, training cells, resetting set points, and delivering information through electroceuticals (ion channel drugs) or optogenetics.
Because of the plasticity of life, virtually any combination of evolved material, engineered material, and software constitutes a possible embodied mind. Cyborgs, hybrids, and augmented humans are coming, requiring new frameworks that move beyond the living-versus-machine distinction.
Levin also raises the possibility of interfacing with tissues linguistically—not in human terms, but in the terms of the physiological state spaces those tissues inhabit (e.g., having a conversation with a liver about potassium flux and metabolic balance).
Evolutionary Implications
Levin’s work effectively accelerates processes that took millions of years in nature (e.g., the wasp that hacks plant morphogenesis to build galls). He envisions a future where humans are no longer constrained by the accidental body they were born with—freeing people from the limitations, susceptibilities, and diseases that result from trial-and-error evolution.
Cancer cannot be eliminated entirely because it is a fundamental failure mode of multicellular cooperation, but effective treatments and preventative strategies will exist. Some conditions with organic bases (certain brain structure and physiology disorders) will be fixable, while others rooted in existential or experiential thought patterns will remain.
Practical Information
Levin’s lab website: drmichaellevin.org
Personal blog: thoughtforms.life
Levin co-founded a company called Morphoceuticals to push bioelectric therapies toward biomedical use, including wearable bioreactors for limb and organ regeneration in mammals.
