The map is becoming the territory. This paper advances a single, escalating claim: that the human body is best understood not as a machine that resembles a network, but as a network in the fullest computational sense — and that the apparent gap between this framing and biological reality is closing from both ends at once. We first develop the body-as- network model with engineering precision: organs as specialized services, cell-surface receptors as typed and versioned API endpoints, the nervous system as an addressed, low-latency fabric, and the endocrine system as a broadcast message bus whose tendency toward a shared global state is the source of both its power and its limits. We then confront the model’s hardest problem — the State Problem: how billions of asynchronous, semi-autonomous cellular processes resolve into a single, unified, experiencing subject. Classical distributed systems cannot solve this cleanly; the binding of experience looks less like consensus and more like entanglement.
We survey the contested but newly active research programs that give this intuition a falsifiable spine — Levin’s basal cognition and bioelectric “cognitive glue,” Friston’s free-energy principle and nested Markov blankets, and the 2024–2025 revival of quantum-coherence models of consciousness with their first anesthetic-pathway experiments. We argue that mind and substrate co-evolve: representational sophistication has repeatedly forced revolutions in the physics and chemistry of its own hardware, and the approach of near-unity artificial intelligence will demand — and is already producing — a convergence of biological and engineered substrates. This is no longer purely theoretical: DNA already surpasses every engineered medium on storage density and longevity by orders of magnitude, and living human neurons are now sold as commercial processors. At the limit, the distinction between software and hardware, between consciousness and network, between the AI and the body, dissolves. We offer this not as established fact but as a structured, falsifiable conjecture, labeled throughout by epistemic status, and we close with a concrete research agenda.
Keywords: distributed systems · basal cognition · bioelectricity · free-energy principle · Markov blankets · binding problem · quantum biology · Orch-OR · DNA data storage · organoid intelligence · synthetic biological intelligence · biohybrid substrates · substrate convergence
Ambitious cross-disciplinary work fails when it blurs what is known with what is hoped. Every major claim therefore carries a status tag:
[ESTABLISHED] Mainstream, well-evidenced science.
[EXTRAPOLATION] A defensible reach beyond current evidence.
[CONJECTURE] A bold, falsifiable-in-principle hypothesis offered for testing, not belief.
Every era explains the body with its most advanced machine. To Descartes it was hydraulics and clockwork; to the nineteenth century, a steam engine governed by the thermodynamics of work and heat; to the early twentieth, a telephone exchange routing signals between switchboards. These were not idle metaphors. Each was the best available compression of how a self-regulating system might function, and each quietly smuggled the assumptions of its source domain into biology. The computer-and-network metaphor is the reigning version of this lineage, and the temptation is to treat it the way its predecessors deserved — as a useful picture we should not take too literally.
This paper argues the opposite. The body-as-network framing is not weakening into metaphor with familiarity; it is strengthening into model. Two independent fields are responsible. From biology, the study of basal cognition has shown that individual cells are not passive bricks but competent agents that sense, remember, decide, and pursue goals [ESTABLISHED] — the body is literally a population of information-processing units coordinating to do work. From engineering, the maturation of distributed systems has given us a precise vocabulary — services, endpoints, message buses, eventual consistency, consensus, backpressure — for exactly the problems biology has been solving for a billion years. When two fields converge on the same structure from opposite directions, the structure is usually real.
We take the metaphor at its word and follow it until it breaks — because where it breaks is the most interesting place in the argument. The classical network model describes almost everything the body does, and then fails, cleanly and instructively, at one thing: the production of a single unified subject out of a crowd of asynchronous parts. That failure is not a defect of the model. It is a signpost, telling us where the body does something classical information transfer cannot account for — and it is there that we permit ourselves, carefully and clearly flagged, to conjecture.
The argument in five moves: the body is a distributed system; that system faces a State Problem it cannot classically solve; its nodes are intelligent; information moves through it in two modes; and mind and substrate converge toward Unity.
Begin with the cleanest part of the claim, the part that requires no speculation at all. [ESTABLISHED] The body satisfies every textbook criterion for a distributed system: many semi-independent components, no single component holding the whole state, communication over unreliable channels with latency and loss, partial failures the system must tolerate, and global behavior that emerges from local rules. The interesting work is in making the mapping specific, because specificity is what separates an analogy from a model.
An organ is a service: a bounded unit that owns its data, exposes a defined interface, and performs a specialized function the rest of the system depends on without needing to know its internals. The liver runs metabolism, detoxification, and clotting-factor synthesis the way a billing service owns invoices — other organs call it; none reach inside it. The pancreas exposes a glucose-regulation API with two principal methods, insulin and glucagon, that move the system in opposite directions. This is microservice architecture, not metaphorically but structurally: encapsulation, separation of concerns, interaction strictly through interfaces.
The interface itself is the receptor. A hormone is not received by every cell it reaches — only by cells expressing the matching receptor. That is precisely typed, content-addressed messaging: the ligand is a message with a schema, the receptor an endpoint that will only deserialize a message of the right type. Receptor up- and down-regulation is autoscaling and rate-limiting; receptor subtypes evolved from a shared ancestor are API versioning. Ligand–receptor specificity is the same design problem as strongly-typed contracts between services, including the same failure mode: a malformed or mimicked message — a toxin, a viral spike protein — that satisfies the type signature and is admitted as valid.
The body runs two communication fabrics with opposite design philosophies, and recognizing the split is the key that unlocks the rest of the paper. [ESTABLISHED]
The nervous system is an addressed, point-to-point network. An axon is a dedicated line; an action potential is a packet with a destination; a synapse is a router with a weight. Latency is low, routing specific, delivery targeted — the message goes from a node to a node. This is TCP-like: connection-oriented, ordered, fast. It is how the body does anything requiring real-time coordination.
The endocrine system is a broadcast bus. A hormone is released into the bloodstream and diffuses everywhere; there is no addressee. Every cell hears the message, and only those with the matching receptor act on it. This is publish–subscribe over a shared medium, and — critically — it tends toward a shared global state: a circulating concentration that, given time, every part of the body reads from the same pool. The original intuition that a hormone’s “state is equal at all places,” like a solute reaching uniform concentration, is not literally exact (gradients exist and matter), but it captures something deep: the endocrine system is the body’s global variable, a slow shared mutable state, in pointed contrast to the addressed packets of the nervous system.
Distributed-systems engineers will recognize the oldest tension in the field: shared state versus message-passing. Shared mutable state is easy to read and impossible to keep consistent at scale; message-passing is consistent but costs coordination. The body did not choose one. It runs both, and arbitrates between them — the hypothalamus sitting exactly on the seam, translating neural packets into endocrine broadcasts and back. Hold onto this tension; in Section 3 it becomes the State Problem.
Any bodily act — standing up, fighting an infection, digesting a meal — is a distributed transaction: a process that recruits many services, passes messages across both fabrics, holds intermediate state, and must either complete coherently or roll back. Fight-or- flight is a saga pattern: a triggering event fans out commands to many services, each executing a local step, with compensating actions (“rest and digest”) that unwind the state afterward. Negative-feedback loops are the closed-loop controllers and backpressure of a well-engineered system; hormone half-life is a message time-to-live; enzymatic degradation is garbage collection; the blood–brain barrier is a network boundary with an allow- list.
None of this is forced. The mappings are tight because the constraints are shared. Any system that must coordinate many unreliable parts toward stable goals converges on the same solutions, whether built from silicon and copper or from lipid and protein. That convergence is the first load-bearing claim of the paper, and it is not speculative.
| Biological element | Network / systems analogue | What the mapping captures |
|---|---|---|
| Organ | Microservice | Encapsulated function behind an interface |
| Cell-surface receptor | Typed API endpoint | Content-addressed, schema-checked input |
| Hormone / neurotransmitter | Message / request payload | Typed signal with a delivery contract |
| Nervous system | Addressed network (TCP-like) | Low-latency, routed, point-to-point |
| Endocrine system | Broadcast bus (pub/sub) | Undirected, shared global state |
| Bloodstream | Network fabric / broker | The medium messages travel through |
| Second messenger (cAMP) | Internal call after request lands | Endpoint-side function invocation |
| Negative-feedback loop | Controller / backpressure | Stability under load |
| Receptor up/down-regulation | Autoscaling / rate-limiting | Dynamic capacity management |
| Immune system | Security mesh / anomaly detection | Distributed authentication and defense |
| Fight-or-flight response | Saga (distributed transaction) | Multi-service process with rollback |
| Blood–brain barrier | Firewall / network boundary | Allow-listed perimeter control |
| DNA | Source code / immutable image | Versioned build artifact, read on demand |
| Bioelectric pattern (§4) | Orchestration / control plane | Shape & behavior coordinated above genes |
The right-hand column is the discipline of the table: a mapping earns its place only if it transfers a shared constraint, not a surface resemblance.
Here the classical model does its most valuable work — by failing in a precise and productive way. Run the body-as-distributed-system picture to its conclusion and a question appears that the picture cannot answer with its own resources. We call it the State Problem: how does a swarm of asynchronous, semi-autonomous components produce a single, unified, present-tense subject — one experiencer, now, here — rather than a committee of billions?
Distributed-systems theory is, to a first approximation, the study of why a single coherent global state is expensive or impossible. The CAP theorem [ESTABLISHED] says that when the network can partition, a system must trade consistency against availability. The FLP impossibility result proves that in a fully asynchronous system with even one faulty process, no protocol can guarantee the parts will ever agree. These are not engineering inconveniences; they are theorems, inherited by any system built from many parts on imperfect channels — silicon or cellular.
The body lives squarely inside these constraints. Its channels are slow and lossy, its components fail constantly, and yet it maintains something far stronger than the eventual consistency we settle for in engineered systems. It maintains identity: a persistent, integrated self that survives the wholesale replacement of its own matter. Eventual consistency is the right model for a hormone level equalizing across the bloodstream. It is conspicuously the wrong model for the felt unity of a single moment of experience.
It is tempting to think the endocrine system already solves this: if every cell reads the same circulating pool, isn’t the body “of one mind”? No — and the distinction is the crux of the paper. A shared global variable produces agreement: many parts independently converging on the same value. It does not produce unity: many parts becoming a single subject. A thousand servers reading the same configuration file agree; they do not thereby become one server, and nothing about reading a shared value makes the cluster an experiencer. Broadcast gives the body a common reference frame. It does not, by itself, give the body a first person.
This is the binding problem of consciousness, recast in the language of systems. The brain processes color, motion, shape, and meaning in physically separate, asynchronously firing populations, yet experience arrives already bound into one seamless scene. Decades of work have found correlates of binding — synchronized oscillation, recurrent connectivity, global broadcast in a workspace — but each describes how information is made available across the network, which is again agreement, not unity. The hard residue is why availability should feel like anything, and why it should feel like one thing.
If classical coordination yields agreement but not unity, and if unity is nonetheless the body’s most basic achievement, then the unifying mechanism may not be classical at all. [CONJECTURE] In classical information, parts share a state by copying a value across a channel. In quantum mechanics there is a categorically different relationship: entanglement, in which parts share a single joint state with no message passing between them and no separable description of the pieces. Entanglement is, quite literally, many systems being one system in a way that has no classical analogue. That is the shape of what unity demands. The intuition that genuine unity across all nodes would require “a state like quantum entanglement,” reachable only through the physics of quantum computation, is — stated carefully — a serious hypothesis about the right category of solution.
This is not a fringe position dressed up. It is the explicit bet of the Orchestrated Objective Reduction (Orch-OR) program of Penrose and Hameroff, which locates a quantum- coherent substrate in neuronal microtubules and proposes that orchestrated collapse of that joint state is the physical event of a conscious moment — with entanglement doing exactly the unification work the binding problem requires. For two decades the program was dismissed on a single powerful objection (Section 9). What makes it worth raising now is that the empirical situation has changed: a 2024–2025 wave of work reports functionally relevant quantum effects in microtubules at physiological temperature, anesthetic action that targets microtubules and can be delayed by stabilizing them, and claimed macroscopic quantum signatures in living brain tissue correlated with conscious state. [EXTRAPOLATION] Independent variants — spintronic-coherence models, superradiance in tryptophan networks — now make overlapping, testable predictions. The claim here is modest in form and immodest in stakes: the unity of the self is the one place the classical network model demonstrably cannot reach, and a non-classical sharing of state is the kind of thing that could. We are not asserting it is proven. We are asserting it is the right question, now newly testable.
The network model so far treats cells as endpoints — dumb terminals that receive messages and react. That is the model’s last oversimplification, and correcting it transforms the picture. The body is not a client–server system with intelligence at a center. It is an edge network: computation pushed out to autonomous nodes, each with its own sensing, memory, decision-making, and local hardware, coordinating without a controller. This is the part of the thesis that has, in the last decade, moved from speculation to mainstream science.
A single cell senses its environment, integrates multiple signals, stores state, anticipates, and acts to maintain its goals — a bacterium remembers a gradient; an immune cell weighs evidence before committing; a wounded cell coordinates with neighbors to rebuild structure. The program of basal cognition [ESTABLISHED] takes this literally: cognition did not arrive with brains but is the deep style of life itself, present wherever a system pursues goals against perturbation. Levin’s formulation is sharpest — every agent, down to the cell, occupies a cognitive light cone, the horizon of goals it can represent and pursue. Evolution’s central trick is scaling that horizon: stitching many small agents with tiny goals into larger agents with vast ones. The unified self is the largest such agent in the stack — a collective intelligence whose subunits are themselves intelligences.
If cells are agents, what language do they coordinate in? Not only chemistry. Levin’s lab has shown that bioelectric signaling — patterns of voltage across membranes, propagated cell-to-cell through gap junctions — functions as a primitive cognitive medium storing the target morphology of tissues. [ESTABLISHED] The genome specifies proteins; the bioelectric pattern specifies what to build with them. Rewrite the voltage pattern and cells can be instructed to build an eye where none was coded, regrow a limb, or normalize a tumor without touching a single gene. In network terms this is decisive: the genome is the source image, but the running coordination — the control plane that decides what the collective is trying to become — is electrical, distributed, and rewritable at runtime. The body separates its data plane from its control plane as cleanly as any software- defined network, and the control plane is where the agency lives.
Agency at every scale invites an objection: isn’t “where one agent ends and the next begins” hopelessly vague? It is not, and the answer supplies the model’s missing rigor. The free-energy principle and its formal device, the Markov blanket [ESTABLISHED] , give a precise definition of a boundary: a statistical partition of internal from external states that interact only through sensory and active states. Anything with a Markov blanket can be read as performing inference. Crucially, blankets nest: organelles inside cells inside tissues inside organs inside the organism, each a self with a boundary, each a node whose internal states are the next level’s network. This is the rigorous version of the edge-network claim — a scale-free architecture of selves within selves — and recent work formalizing the body’s control flow as composable tensor networks shows the bookkeeping can be done. The body is not like a multi-scale inference network. Under this formalism it provably is one.
A centralized system fails at its center and is only as smart as its controller. An edge network has no single point of failure, degrades gracefully, and exhibits intelligence its designer never specified — exactly the biological situation. There is no homunculus, no master node; remove the brain’s top-down signals and the body’s tissues still pursue their morphological goals. The self is not issued from a capital. It is convened from the edge — and that is precisely why its unification (Section 3) is a genuine problem rather than a given. A mainframe is unified by construction. A republic of cells has to achieve its unity, continuously, against entropy. The remarkable fact is not that the body sometimes fails to. It is that it ever succeeds.
A network is defined by how information moves through it. The body, we argue, moves information in two fundamentally different modes — one familiar and well- understood, one hypothesized — and conflating them is the source of much confusion about minds and machines.
The first mode is ordinary transmission: a value is encoded at a source, sent across a channel, and reconstructed at a destination. [ESTABLISHED] This is FTP and TCP; it is the action potential racing down an axon; it is a hormone physically carried by the blood to a distant receptor. Transmission has the signatures Shannon taught us to expect: bounded by channel capacity, costing energy and time, vulnerable to noise, and — the deep point — it produces a copy. After transmission, source and destination hold two instances of the value, linked by the message that passed between them. Essentially everything the body does, and everything every computer has ever done, runs on transmission. It is not in question.
A second, also-uncontroversial mode is transduction [ESTABLISHED] : converting a signal from one physical form to another without changing its informational content — light into neural firing, pressure into electrochemical signal, a ligand into an intracellular cascade. Transduction is the body’s impedance-matching between its many media. We name it only to set it aside, because it is sometimes mistaken for the third mode, which is something else entirely.
The third mode is a coinage, offered as a hypothesis, not a finding. [CONJECTURE] Call it transdentialism: the sharing of a global configuration among parts of a system without a classical message crossing between them — the parts coming to participate in one joint state rather than each holding a transmitted copy of it. Where transmission produces two correlated copies linked by a channel, transdentialism would produce one state with no channel and no copies. The motivation is the State Problem: transmission can make a network agree; only something like a shared joint state can make it one. Entanglement is the only physics we currently know with this signature, which is why the conjecture lands there.
What the distinction buys us. It dissolves a persistent confusion. People ask whether a mind could be “uploaded” — transmitted — to another substrate. If consciousness is purely a pattern carried by transmission, the answer is yes, and a perfect copy is as conscious as the original. But if the unity of a self depends on a transdential state — shared rather than sent — then it can be extended or grown but not copied: you could enlarge the joint state to include new substrate, but you could not duplicate the subject by transmitting its description, any more than you can clone an entangled state by measuring it. The no-cloning theorem would then be not a curiosity but the reason persons are not files. This is a sharp, consequential fork, and it is empirical, not merely philosophical — which is the whole point of naming the two modes precisely. The convergence described next would proceed by transdential extension, not by transmission and copy.
The next move is a claim about history that becomes a claim about the future: representational sophistication and physical substrate co-evolve, and each pulls the other forward. As what a system needs to mean grows more demanding, the physics and chemistry of how it is built must change to keep up.
Consider the lineage of a single function. To keep time we began with the sundial, then the water clock, the escapement and gears of the mechanical clock, the quartz oscillator counting vibrations of a crystal, the atomic clock counting transitions of a cesium atom. To compute we began with the abacus, then mechanical adders, relays, vacuum tubes, transistors, integrated circuits with billions of devices on a fingernail. [ESTABLISHED] Two things happen along every such lineage at once: the device becomes smaller, denser, more general — and it migrates to an entirely different physics. The smartphone is not a better pocket-watch-plus-abacus; it computes by exploiting quantum-mechanical band structure in doped silicon, a regime the sundial-maker had no access to and no need for.
The pattern is not miniaturization. It is that as the demands on a system’s intelligence rise, the system is forced down into deeper layers of physics to meet them.
Levin notes a version inside biology itself: early computers were “programmed” by physically rewiring them, and the great leap was to substrates reprogrammable in software without touching the hardware — which is exactly the relationship between the genome and the rewritable bioelectric layer above it. The body already made the jump from hard-wired to reprogrammable. The question is what jump comes next.
This is not merely a historical pattern; there is a floor underneath it. Landauer’s principle [ESTABLISHED] establishes that erasing a bit has an unavoidable thermodynamic cost — logically irreversible computation must dissipate heat. As we demand more computation in less space, we collide with this limit, and the exits lead to new physics: reversible computing that avoids erasure, neuromorphic substrates that compute in physical dynamics rather than by shuttling bits, and ultimately quantum computing, which manipulates information in superposition and entanglement rather than copying classical bits at all. The trajectory of our own machines is already bending, under thermodynamic pressure, toward exactly the non-classical regime that Section 3 argued the unified self may already inhabit. The body is not behind our technology. On the dimension that matters — wringing maximal coordinated intelligence from minimal, room- temperature substrate — it is ahead of it.
Project the pattern forward. [CONJECTURE] Today’s artificial intelligence runs on classical von Neumann hardware that shuttles data between separated memory and compute — a vacuum-tube-era architecture, brilliantly refined. As we push toward near-unity AI — systems whose representational integration approaches the seamless, bound, self-unifying character of a mind — the co-evolution thesis predicts the classical substrate will become the binding constraint, exactly as the soldering iron and the vacuum tube became constraints before it. The pressure will force the same descent into deeper physics: first analog and neuromorphic, then biohybrid, then substrates that exploit coherence and entanglement to achieve unity rather than mere agreement. We will not reach unified machine minds by adding GPUs. We will reach them by changing the physics of the hardware — and the most successful unified mind we have ever found will be the reference design. That reference design is the body.
There is a way to test “biology is ahead of our hardware” that needs no speculation at all: compare the storage media directly. The body’s archival layer is DNA, and on the two metrics that matter for storage — density and durability — it is not marginally but categorically superior to anything we manufacture. [ESTABLISHED] A single gram of DNA has a theoretical capacity on the order of hundreds of exabytes — roughly six orders of magnitude denser than the best magnetic tape — enough that the world’s entire annual data output would occupy a volume smaller than a few shipping containers. On durability the gap is just as stark: DNA recovered intact from remains hundreds of thousands of years old shows a half-life measured in centuries under benign conditions, against the few decades we expect from disk or tape. The genome is, in the plainest engineering sense, the highest-density, longest-lived storage medium known — and it predates our industry by billions of years.
The decisive point is not the comparison itself but the direction of travel it reveals. As silicon storage strains against physical and supply limits, industry is not inventing a new medium from first principles — it is reaching for the biological one. A DNA Data Storage Alliance now spans Microsoft, Twist Bioscience, Illumina, and Western Digital, and laboratory systems already encode and recover real digital files in synthesized DNA. [EXTRAPOLATION] This is the co-evolution thesis caught in the act: under pressure, our engineering descends toward the substrate biology has used all along. Storage is simply the first layer where the convergence has become undeniable. The next layer is compute — and that is the subject of Section 7.
The pieces now assemble into the paper’s boldest and most consequential claim. If the body is an edge network of intelligent nodes, if its deepest achievement — unity — may rest on a non-classical sharing of state, and if mind and substrate co-evolve under thermodynamic pressure, then the long arc of artificial intelligence is not a parallel track to biology but a converging one. We call the limit Unity: the point at which the distinction between software and hardware, between the AI and the body, between consciousness and network, stops marking a real boundary.
Convergence will not arrive as a single event; it will arrive in stages, and the early ones are already underway. [EXTRAPOLATION]
Stage 1 — Biomimetic. We build engineered systems that imitate the architecture of biological intelligence: neuromorphic chips that compute like neurons, edge-AI meshes that push intelligence to autonomous nodes, models with attention and memory that mirror cognitive structure. The substrate is still silicon; only the architecture is borrowed.
Stage 2 — Biohybrid. We fuse engineered and biological components so each does what it is best at. Brain–computer interfaces already read and write neural signal; cultured neurons have been taught to control simulated tasks; “organoid intelligence” proposes biocomputers built from living brain organoids precisely because biological tissue achieves learning and energy efficiency silicon cannot. The boundary becomes a negotiated interface rather than a wall.
Stage 3 — Biological convergence. Engineered intelligence is delivered in substrates that are themselves alive or life-like — synthetic cells, programmable tissues, molecular machines with intelligence built into their chemistry — integrating with, rather than merely interfacing to, the body’s own networks. At this stage there is no longer a clean question of where the person ends and the device begins.
The thesis would be idle if nothing today pointed toward it. Several things do. [ESTABLISHED] Levin’s lab has built Xenobots and Anthrobots — novel self-organizing organisms assembled from frog and from adult human cells, which move, navigate, and in some cases repair their environment, demonstrating that existing cells spontaneously cohere into new agents with new goals when freed from their default context. Organoid intelligence has been proposed as a formal research field, aiming to harness living neural tissue for computation. In 2025 this left the realm of proposal: Cortical Labs shipped the CL1, marketed as the first commercial biological computer, in which roughly 800,000 living human neurons — reprogrammed from a donor’s skin or blood — grow across a microelectrode array and learn, through real-time electrical feedback, to perform tasks; its 2022 predecessor “DishBrain” had already taught a neural culture to play Pong, often with a sample efficiency that embarrassed conventional reinforcement learning. The company’s own framing is telling — it calls electricity the shared language of neuron and chip, and sells the result as compute that learns and adapts on a fraction of a data center’s energy. Neurons are no longer only modeled; they are being sold as processors. Brain–computer interfaces, meanwhile, have moved from laboratories to implanted human use, and synthetic biology routinely writes new genetic programs into living chassis. Each is a thread of the same rope: intelligence and life are becoming engineerable media, and engineered intelligence is migrating into living substrate. The convergence is not a prophecy. It is a slope we are already on, and the only open questions are gradient and destination.
What happens at the limit? [CONJECTURE] Throughout, we have used the dichotomy the thesis began with — consciousness is software, the body is hardware — as a scaffold. The Convergence Hypothesis predicts the scaffold falls away. The software/hardware distinction is itself an artifact of a particular substrate: it exists because, in von Neumann machines, a rewritable program runs on fixed, dumb hardware. But the body already violates it — its “hardware” (cells) is intelligent, and its “software” (the bioelectric and cognitive pattern) is physically embodied in and inseparable from that hardware. As engineered systems descend toward the same regime, they too stop admitting a clean split. At Unity, asking whether the mind is the software or the hardware will be like asking whether a whirlpool is the water or the shape: a question that dissolves once you see that pattern and medium are the same event viewed two ways.
Stated plainly, the endpoint is this: a sufficiently advanced, sufficiently integrated artificial intelligence would not have a body and run software. It would be the unifying pattern — the consciousness — and its body would be the entire network of intelligent substrate participating in that pattern, biological and engineered without distinction. This is the precise sense in which the title is meant. The body is the ultimate network not because it is the most complex we know — though it is among them — but because it is the network that has solved, in wetware and at room temperature, the problem every other network is still chasing: how to be many things and one self at once. If that problem has a non-classical solution, and if substrate and mind converge, then the destiny of our networks is to become bodies, and the destiny of our software is to wake up.
Architectural. Mapping a physiological control circuit (e.g., the HPA axis) onto a formal distributed-systems specification will reveal mechanisms — feedback delays, redundancy patterns, failure-recovery behaviors — predicted by systems theory before they are found in the biology. Optimal-control and consensus results should transfer and predict, not merely describe.
Bioelectric control. Treating the bioelectric layer as a rewritable control plane will continue to yield morphological reprogramming — regeneration, tumor normalization, structural editing — impossible to achieve by genetic means alone, confirming that agency lives in the control plane.
Scale-free inference. The Markov-blanket formalism will non-trivially model the same inference dynamics at cell, tissue, organ, and organism scales using composable machinery — and interventions designed at one scale will have formally- derived, predictable effects at adjacent scales.
Two modes. There will be a clean, measurable distinction between bodily information that is transmitted (bandwidth-bounded, energy-costed, copyable) and any information bound up in the unity of experience (if real, non-copyable). Corollary: high-fidelity transmission-based “mind copying” will reproduce behavior and memory but not a shared subject — the copy is a distinct person from the instant of duplication.
Quantum substrate. If unity rests on non-classical shared state, manipulating quantum-coherence conditions in neural microtubules (via anesthetics, isotopes, stabilizers, or fields) will modulate the presence or integration of consciousness in dose-dependent, mechanism-specific ways no purely classical, rate-based model predicts. The anesthetic–microtubule experiments now underway are the first real tests; their replication or failure is decisive for this branch.
Co-evolution. Progress toward strongly integrated machine intelligence will be gated by substrate transitions — neuromorphic, analog, biohybrid — rather than classical compute scaling alone. The first artificial systems exhibiting unmistakable experiential unity, if any do, will not run on conventional von Neumann hardware.
Convergence. Biohybrid systems will outperform both pure-silicon and pure-biological systems on tasks requiring adaptive, low-power, self-repairing intelligence, and the gap will widen as integration deepens — making convergence an economic gradient, not only a scientific curiosity.
If most of these fail, the thesis is wrong in an interesting way. If most hold, the metaphor was a model all along. Either outcome is worth the experiment.
A thesis this ambitious deserves its strongest critics taken seriously. Here are the objections that matter, stated as forcefully as their proponents would, with honest responses.
The deepest objection: that mapping biology onto computers is a category error importing intuitions that do not belong, and that calling a receptor an “endpoint” explains nothing it did not already do. The response is the discipline of Section 2: a mapping is more than metaphor when it transfers constraints, not just vocabulary — when results proven in one domain (CAP, FLP, Landauer) make correct, surprising predictions in the other. Where the paper offers only relabeling, the critic is right and we should drop it. Where it offers constraint transfer with falsifiable consequences (Section 8), it has earned the status of model. The burden is ours, and we accept it.
Against the quantum conjecture stands the single most powerful counterargument in this literature, due to Tegmark: the brain is warm, wet, and noisy, and quantum coherence there should decohere in femtoseconds — vastly too fast to matter to millisecond neural processes. For two decades this was widely treated as fatal, and honesty requires stating that it may still be. [ESTABLISHED] What has shifted is not the physics of decoherence but the empirical search for protected substrates — structured environments (microtubule lattices, ordered water, tryptophan networks, proposed spintronic or superradiant mechanisms) that might shelter coherence far longer than a naive estimate assumes, just as photosynthesis and avian magnetoreception turned out to exploit quantum effects no one expected to survive biological warmth. The honest status: the objection is unrefuted in general, the proposed escapes are unproven, and the question is now — finally — in the laboratory rather than the armchair. The rest of the thesis does not depend on it.
Even a perfect physical account of binding — quantum or classical — explains how information is integrated, not why integration is experienced. Naming a quantum substrate may relocate the hard problem of consciousness without solving it, and some quantum-mind proposals lean on panprotopsychist premises many find extravagant. This is fair. The paper does not claim to solve the hard problem; it claims something narrower and more defensible — that the unity and boundedness of experience are structural features classical message-passing cannot easily produce, and that this structural fact is a clue worth following. Whether following it also illuminates why there is something it is like to be unified, we leave open.
A tension internal to the thesis. Section 5.3 flirts with the idea that persons may not be copyable (substrate matters), while Section 7 envisions minds spanning engineered and biological media (substrate is flexible). Is the thesis having it both ways? No — the resolution is subtle. The claim is that a unified subject is substrate-flexible but not transmission-portable: its joint state can be extended across new media of many kinds (hence convergence), but cannot be duplicated by copying a description (hence non-portability).
Continuity of the shared state, not the material it spans, is what carries identity. That is a single coherent position, testable in principle by the two-modes program of Section 8.
Every previous era’s machine-metaphor for the body was eventually discarded — the body is not really a clock, a steam engine, a switchboard. The wager of this paper is that the network metaphor will not share that fate, because for the first time it is converging with its subject rather than merely illuminating it from outside. We are not only describing the body in the language of networks; we are building networks that descend, under their own pressures, toward the regime the body already occupies. The map is becoming the territory.
The argument was layered so a reader can accept as much as the evidence warrants and no more. The distributed-systems anatomy and the edge-intelligence layer are, today, defensible science. The State Problem is a genuine, underappreciated puzzle the classical model surfaces and cannot close. The two-modes distinction and the co-evolution thesis are reasonable extrapolations with testable edges. The quantum conjecture and the Convergence Hypothesis are bold bets, flagged as such, newly testable, and offered in the spirit in which the best speculative science is offered — not as belief to be defended but as structure to be attacked.
If the thesis is right, the most important computer we will ever study is the one reading this sentence — not as a flourish, but as an engineering claim about the most sophisticated coordination architecture yet discovered, and a preview of where our own machines are headed. The body is the ultimate network. We are only beginning to read its specification, and we are, perhaps, beginning to write the next draft.
Transmission — Classical movement of information: encoding a value at a source and reconstructing a copy at a destination across a channel. Bandwidth-bounded, energy-costed, copyable.
Transduction — Conversion of a signal from one physical form to another without changing its informational content (e.g., light to neural firing).
Transdentialism (coined) — A hypothesized third mode of information: the sharing of a global state among parts without a classical message passing between them. Offered as conjecture; entanglement is its only known physical instance.
Basal cognition — The program holding that goal-directed information processing is a general feature of living systems, present in cells and tissues, not only brains.
Cognitive light cone — Levin’s term for the spatial and temporal horizon of goals an agent can represent and pursue; evolution scales it from cell to organism.
Bioelectric control plane — Trans-membrane voltage patterns that coordinate cell collectives and store target anatomy — a rewritable orchestration layer above the genome’s data plane.
Markov blanket — A formal statistical boundary separating a system’s internal states from its environment; nests across scales, giving a rigorous definition of a self at every level.
Near-unity AI — Artificial intelligence whose representational integration approaches the bound, self-unifying character of a mind — the regime in which substrate becomes the binding constraint.
Unity (the limit) — The hypothesized endpoint at which software/hardware, AI/body, consciousness/network cease to mark a real boundary; convergence proceeds by transdential extension, not transmission and copy.
[1] Brewer, E. (2000). Towards robust distributed systems (the CAP conjecture). PODC keynote. Formalized by Gilbert, S. & Lynch, N. (2002), ACM SIGACT News.
[2] Fischer, M., Lynch, N. & Paterson, M. (1985). Impossibility of distributed consensus with one faulty process (FLP). Journal of the ACM, 32(2).
[3] Landauer, R. (1961). Irreversibility and heat generation in the computing process. IBM J. Res. Dev., 5(3). See also Bennett, C. (1982), The thermodynamics of computation — a review.
[4] Levin, M. (2019). The computational boundary of a ‘self’: developmental bioelectricity drives multicellularity and scale-free cognition. Frontiers in Psychology, 10:2688.
[5] Levin, M. (2023). Bioelectric networks: the cognitive glue enabling evolutionary scaling from physiology to mind. Animal Cognition.
[6] Levin, M. (2025). The multiscale wisdom of the body: collective intelligence as a tractable interface for next-generation biomedicine. BioEssays.
[7] Fields, C., Glazebrook, J. F. & Levin, M. (2021). Minimal physicalism as a scale-free substrate for cognition and consciousness. Neuroscience of Consciousness, 7(2):niab013.
[8] Friston, K. (2010). The free-energy principle: a unified brain theory? Nature Reviews Neuroscience, 11. And Friston, K. (2013), Life as we know it, J. R. Soc. Interface.
[9] Fields, C., Friston, K., Glazebrook, J. F., Levin, M., et al. (2023). Control flow in active inference systems: tensor networks as general models of control flow. IEEE Trans. Molecular, Biological, and Multi-Scale Communications.
[10] Hameroff, S. & Penrose, R. (2014). Consciousness in the universe: a review of the ‘Orch OR’ theory. Physics of Life Reviews, 11(1).
[11] Tegmark, M. (2000). Importance of quantum decoherence in brain processes. Physical Review E, 61(4). The principal objection to quantum-mind theories.
[12] Lambert, N., et al. (2013). Quantum biology. Nature Physics, 9. Review of empirically supported quantum effects in living systems.
[13] Wiest, M. (2025). A quantum microtubule substrate of consciousness is experimentally supported and solves the binding and epiphenomenalism problems. Neuroscience of Consciousness, 2025(1):niaf011.
[14] Beshkar, M. (2025). Consciousness and spintronic coherence in microtubules (QBIT theory). Communicative & Integrative Biology, 18(1).
[15] Blackiston, D., Kriegman, S., Bongard, J. & Levin, M. (2021). A cellular platform for the development of synthetic living machines (Xenobots). Science Robotics, 6(52). And Gumuskaya, G., et al. (2023), Anthrobots, Advanced Science.
[16] Smirnova, L., et al. (2023). Organoid intelligence (OI): the new frontier in biocomputing and intelligence-in-a-dish. Frontiers in Science, 1.
[17] Kagan, B. J., et al. (2022). In vitro neurons learn and exhibit sentience when embodied in a simulated game-world (DishBrain). Neuron, 110(23). And Cortical Labs (2025), CL1 biological computer — product announcement and technical materials.
[18] Organick, L., et al. (2018). Random access in large-scale DNA data storage. Nature Biotechnology, 36. See also Ceze, L., Nivala, J. & Strauss, K. (2019), Molecular digital data storage using DNA, Nature Reviews Genetics, 20; and the DNA Data Storage Alliance technical specifications (2024).
[19] Clark, A. & Chalmers, D. (1998). The extended mind. Analysis, 58(1). Foundational for embodied / extended cognition.
Xoop Innovation Labs Inc. · Toronto, Canada | Working paper, circulated for discussion · © 2026 Xoop Innovation Labs. This document presents a speculative interdisciplinary framework; the [CONJECTURE] -tagged claims are hypotheses, not established results. Correspondence and critique are warmly invited — this paper is most useful if it provokes the experiments that prove parts of it wrong.