Conceptual scientific white paper
Date: 28 August 2026 Status:
Exploratory proposal. This article describes a research direction, not an existing technology.
Executive Summary
What if a computer were not only manufactured, but partly grown?
This paper examines a speculative class of biological computing system that integrates six elements: human- or animal-derived neural cells, living fungal mycelium, DNA-based molecular memory and logic, microelectrode arrays and sensors, artificial neural networks, and conventional digital hardware.
The originating vision is a living hybrid platform in which neurons supply learning and adaptive memory, mycelium provides distributed sensing and physical connectivity, DNA performs molecular logic and archival storage, and digital AI coordinates the system as a whole.
Individually, several of these strands are already under active investigation. Brain organoids have exhibited nonlinear dynamics, fading-memory properties and limited task-related adaptation when interfaced with electrode arrays. Fungal mycelium generates measurable electrical activity and impedance changes. DNA circuits can execute molecular logic and store information at extraordinary density. What does not yet exist is any established system uniting all three living substrates into a single functional computer.
The paper’s central conclusion follows from that gap:
A neuron–mycelium–DNA computer is scientifically imaginable. Integrating all three substrates at the outset would almost certainly generate more problems than capability.
The viable path does not begin with a fused human–fungal organism. It begins with a modular biohybrid platform in which each living component remains physically separated, electronically coupled, independently monitored and safely replaceable.
Several simpler architectures may deliver much of the same value at a fraction of the biological cost: neural organoids interfaced directly with electronics; mycelium-based environmental sensor networks; DNA archival memory paired with silicon; neuromorphic processors; memristor arrays; engineered living materials built from non-neural cells; and hybrid designs that simulate the mycelial layer rather than cultivating it.
The long-term opportunity is genuine. The strongest expression of it is likely an ecosystem of specialised processors rather than a single merged organism.
1. The Core Concept
The proposed system is best described as a NeuroMycelial Biocomputing Platform.
Its purpose is not to construct a miniature brain or a thinking mushroom. Those framings are vivid but unhelpful and the scientific literature is already amply supplied with dramatic fungi. A more defensible objective is this:
To determine whether neural tissue, fungal networks, DNA circuits and digital electronics can contribute genuinely distinct computational capabilities within one controlled system.
Each layer carries a defined role:
| Component | Proposed function |
| Neural cells or organoids | Adaptation, pattern recognition, biological memory |
| Mycelium | Distributed sensing, environmental responsiveness, structural growth |
| DNA circuits | Molecular logic, biochemical state recording, archival storage |
| Electronics | High-speed communication, stimulation, measurement, control |
| Artificial intelligence | Signal interpretation, training, coordination, safety monitoring |
The project therefore investigates multi-substrate computing: computation distributed across materials that process information by fundamentally different physical mechanisms.
2. Why Combine Neurons and Mycelium?
Neural tissue and mycelium are both living networks, but they are not variations on a common theme.
Neurons specialise in rapid electrochemical signalling, adaptive connectivity and information processing. Mycelium comprises branching filaments called hyphae that grow through physical environments, respond to stimuli and alter their electrical properties under changing conditions.
Mycelial networks demonstrably produce electrical spikes, impedance shifts and sensor-like responses. Yet fungal electrical activity remains difficult to measure and interpret with consistency. Researchers cannot yet reliably distinguish which signals represent communication, growth, nutrient transport, stress response or unrelated metabolic processes.
Mycelium should not, therefore, be treated as a second brain. Its realistic near-term roles are narrower and more credible:
- A living environmental sensor
- A slowly modulating conductive material
- A self-growing physical network
- A damage-sensitive structural component
- A source of complex analogue signals
- An experimental biological reservoir
The neural layer could, in principle, learn to classify the patterns the mycelium produces. Fungal responses to moisture, chemical exposure, temperature shift or physical damage would be translated into stimulation patterns delivered to a neural culture.
The neural tissue would not “understand” the fungus in any meaningful sense. It would learn statistical structure in the signal.
3. The Contribution of DNA
DNA can serve two distinct and largely independent functions in this architecture.
3.1 DNA as data storage
Synthetic DNA encodes digital information at exceptional density. Suitable payloads include experimental histories, calibration records, biological lineage data, long-term training summaries, system configurations, environmental event logs and authentication credentials.
DNA storage is not, however, a substitute for RAM or solid-state media. Writing, sequencing, retrieval, editing and error correction all remain slower and more cumbersome than conventional electronic storage. Its natural niche is long-term archival memory, not working memory.
3.2 DNA as molecular logic
Engineered DNA circuits respond to molecular inputs and produce molecular outputs. They can be designed to perform AND and OR operations, threshold detection, signal amplification, event counting, biochemical classification, state transitions and safety interlocks.
A DNA circuit could monitor the biological condition of the wider platform — detecting a specific combination of stress markers, for instance, and triggering an optical signal, chemical response or electronic alarm.
DNA is thus best positioned as a slow biochemical control and memory layer, not as the system’s real-time processor.
4. What Is Already Scientifically Plausible
4.1 Neural organoid computing
This is the most directly relevant body of existing work.
The emerging field of organoid intelligence examines whether three-dimensional neural cultures can be interfaced with computers, trained through feedback, and used to study learning and biological information processing. Proposed architectures combine organoids with microelectrode arrays, microfluidic life support and machine-learning interfaces.
A 2023 study introduced a platform called Brainoware, coupling a brain organoid to a multielectrode array as a biological reservoir computer. The system displayed nonlinear dynamics and fading-memory properties and was evaluated on speech-recognition and mathematical prediction tasks. It was a research demonstration rather than a general-purpose computer — but it established that neural organoids can participate meaningfully in computational experiments.
Viability assessment: moderately plausible, highly experimental.
The science is real; the obstacles are substantial. Organoids vary considerably between samples. They demand precise nutrient and environmental control. Reading and writing neural activity remain technically difficult. Useful computational capacity is unquantified, training methods are primitive, long-term stability is unresolved, and scaling compounds both biological and ethical complexity.
For now, neural organoids are more valuable as experimental platforms than as processor replacements.
4.2 Mycelial electronics
Mycelium can be incorporated into sensing materials and experimental electronic devices, changing impedance and generating electrical potential spikes in response to environmental conditions.
The operative word remains experimental. A fungal network may produce electrically rich activity, but richness is not computation. A thunderstorm generates elaborate electrical signals too, and no one has yet persuaded one to run a spreadsheet.
The research question is whether mycelial responses are repeatable, controllable, distinguishable from noise, stable across samples, suitable for encoding information, and useful enough to justify maintaining a living fungal system at all.
Viability assessment: plausible for sensing; unproven for computing.
4.3 DNA computing and storage
DNA circuits rest on a substantial scientific foundation, performing molecular operations in massively parallel chemical environments and storing large volumes of encoded information.
The difficulty is interfacing. A functional platform requires reliable methods for writing information, triggering reactions, retrieving results, correcting errors and converting molecular outputs into electronic signals.
Viability assessment: plausible for specialised logic and archival storage.
DNA is most compelling when the data already exist in molecular form — biomarker detection within a biochemical environment, for example. It is least compelling when ordinary digital data must be repeatedly transcoded into DNA and back again.
5. What Is Not Yet Viable
The following interpretation of the concept is not currently realistic:
A genetically fused human–animal–fungal organism operating as an autonomous conscious supercomputer.
The reasons are structural rather than incidental.
5.1 Biological incompatibility. Neural cells and fungi require materially different growth conditions. Direct physical contact invites fungal invasion of neural tissue, inflammatory response, toxic metabolic interactions, nutrient competition, contamination, tissue death and unpredictable biochemistry. A fused system would likely prove less stable and less useful than a separated one.
5.2 Communication incompatibility. The components operate across incompatible mechanisms and timescales — neurons in milliseconds, mycelium across hours or days, DNA circuits at reaction kinetics, silicon at nanoseconds. Connecting them is not a matter of biological wiring. It demands deliberate translation layers spanning electrical, chemical, optical and digital domains.
5.3 Maintenance overhead. Living computers are often described as energy-efficient, and the claim deserves qualification. An isolated biological network may draw little power directly, but the complete laboratory system still requires heating, sterile fluid circulation, oxygenation, nutrient delivery, waste removal, imaging, electrode amplification, contamination monitoring and conventional computers for interpretation. The fair comparison is not organoid versus microchip. It is the full organoid life-support system versus the full electronic system performing the same task.
5.4 Biological variability. Silicon is manufactured to narrow tolerances. Living tissue develops unevenly, ages, adapts, becomes stressed and sometimes dies. Occasionally that variability supplies useful complexity. More often it constitutes a serious engineering liability.
5.5 Absence of demonstrated advantage. Combining fascinating materials does not produce a superior computer. A hybrid must outperform simpler alternatives on at least one measurable axis — energy efficiency, adaptability, environmental sensing, fault tolerance, few-shot learning, self-repair, information density, or operation within biochemical environments. Absent that, it is an impressive experiment and a poor technology.
6. The Most Viable Architecture
The strongest formulation of the concept is modular and compartmentalised.
Environmental conditions -> Living mycelial sensor -> Electronic signal interface -> Neural culture or organoid -> Machine-learning decoder -> Controlled digital or physical output
A separate DNA chamber monitors chemical conditions and records selected long-term information:
Biochemical state -> DNA molecular circuit -> Optical or electrical readout -> Digital monitoring system
Critically, the biological components never touch. Each occupies its own controlled chamber.
The benefits are considerable: contaminated modules can be excised; the system can be shut down safely; components can be replaced individually; causal attribution across layers becomes tractable; neural and fungal conditions can be tuned independently; unnecessary genetic mixing is avoided; and ethical review becomes far more manageable.
The future system, in short, should resemble a biological computing laboratory connected by interfaces — not a creature suspended in a tank.
7. Viability Scorecard
| Element | Near-term viability | Potential value | Principal limitation |
| Neural cells on electrode arrays | Medium–high | Adaptive biological processing | Stability and reproducibility |
| Brain organoid reservoir computing | Medium | Pattern recognition, neuroscience | Limited control, unclear scaling |
| Mycelium as sensor | Medium–high | Environmental monitoring | Signal interpretation |
| Mycelium as general processor | Low | Unconventional analogue computing | No demonstrated advantage |
| DNA molecular logic | Medium | Biochemical decision-making | Slow kinetics, difficult interfacing |
| DNA archival memory | Medium | Very dense long-term storage | Costly, slow read/write |
| Direct neural–fungal fusion | Very low | Unclear | Incompatibility, biosafety |
| Electronically mediated neural–fungal system | Medium | Research, specialised sensing | Complex signal translation |
| Conscious human–fungal entity | Not scientifically established | No justified application | Scientific, ethical, conceptual |
8. Stronger Alternatives to the Full Hybrid
A credible research programme benchmarks the founding idea against simpler options. The objective is not to protect the most dramatic version of the concept, but to identify the version that works.
Alternative 1 — Neural organoid plus conventional electronics. Removes the fungal and DNA layers entirely: organoid, microelectrode array, microfluidic support, digital decoder, feedback loop. Advantage: it investigates biological learning without importing a second kingdom of life or an additional molecular language, and existing organoid-computing work makes it the clearest entry point. Cost: forfeits the self-growing environmental interface. Best use: biological reservoir computing, learning experiments, drug research, adaptive signal classification.
Alternative 2 — Mycelium plus conventional machine learning. Electrodes record fungal activity; software classifies it. Advantage: avoids human neural tissue altogether — safer, more scalable, deployable in the field. Cost: no biological neural learning; adaptation happens in software. Best use: soil and agricultural monitoring, moisture and contamination detection, responsive building materials, structural damage detection. This is likely the most commercially realistic route derived from the original vision.
Alternative 3 — Mycelium plus neuromorphic hardware. Fungal signals feed brain-inspired silicon rather than living neurons. Advantage: faster, reproducible, manufacturable, free of tissue-welfare concerns, deployable outside the lab. Cost: no genuine synapses or cellular plasticity. Best use: a practical “fungal nervous system” for buildings, farms and ecological monitoring — without growing a nervous system.
Alternative 4 — Memristor networks. Components whose resistance depends on prior electrical history, imitating synaptic memory. They retain adaptive behaviour, analogue processing, physical memory, low-energy event handling and reservoir-computing capability. Advantage: much of what makes neurons attractive, without nutrients, contamination or ethical ambiguity. Best use: edge AI, adaptive sensors, pattern recognition, low-power robotics.
Alternative 5 — DNA memory plus silicon computing. Silicon does the processing; DNA handles archival storage and biological identification. Advantage: assigns DNA to its most promising role. Cost: not a living computer, and transcoding remains difficult. Best use: long-duration archives, sample identification, tamper-evident records.
Alternative 6 — Engineered bacterial or yeast systems. Microorganisms engineered to detect chemicals, record molecular events and produce measurable outputs. Advantage: microbial synthetic biology offers far clearer genetic engineering methods than a mixed neural–fungal system. Cost: no synaptic plasticity or neural dynamics. Best use: biosensing, diagnostics, environmental monitoring, engineered living materials.
Alternative 7 — Digital twins of biological systems. Use living tissue for discovery, then transfer the learned dynamics to a simulated or neuromorphic implementation. Advantage: reproducible, scalable, faster to test, no ongoing biological maintenance, safer to commercialise. Cost: simulation may miss unknown emergent properties. Best use: the strongest bridge from laboratory discovery to deployable product.
9. Recommended Development Strategy
Phase 1 — Define the problem, not the organism. Begin with a task. Can a mycelial network detect environmental change more effectively than a conventional sensor? Can a neural culture classify complex fungal signals? Does a biological reservoir adapt from fewer examples than a software model? Can DNA circuits reliably monitor platform health? Does the hybrid deliver a measurable energy or sensing advantage? A programme without a benchmark becomes an elaborate demonstration in search of a purpose.
Phase 2 — Characterise each module independently. Neural: response consistency, short-term memory, learning behaviour, long-term stability, electrode signal quality, tissue health. Mycelial: stimulus response, cross-species variation, sample reproducibility, signal propagation, impedance, recovery after damage. DNA: logic accuracy, reaction speed, leakage and false activation, stability, reset capability, compatibility with biological fluids.
Phase 3 — Couple mycelium to software first. Classify fungal signals with conventional machine learning to establish a baseline. Any organoid introduced later must outperform — or at minimum meaningfully differ from — that baseline.
Phase 4 — Add an electronically mediated neural module. Record mycelial activity, convert it into controlled stimulation patterns, deliver it to the neural culture, decode the response electronically. No direct tissue contact is required. This phase poses the most scientifically interesting question in the programme: can one living network learn useful patterns generated by another, biologically unrelated living network?
Phase 5 — Introduce DNA logic only where it earns its place. Not because it sounds futuristic, but because there is a defined molecular task: detecting metabolic stress, recording chemical exposure, confirming biological identity, triggering a contamination alarm, or archiving an experimental record.
Phase 6 — Benchmark relentlessly. Compare against ordinary sensors, standard ML models, neuromorphic processors, memristor networks, mycelium-only systems, neural-only systems and DNA-free systems. The full hybrid is justified only where the combination delivers a capability its parts cannot provide more simply.
10. Ethical Boundary: Entity or Instrument?
A system containing living human-derived neurons may be biologically alive without being a person, an organism or a conscious entity. These categories must not be casually conflated.
No universally accepted test for consciousness in an organoid currently exists. Electrical activity, learning, memory and complex responsiveness are not, individually or collectively, proof of subjective experience.
That uncertainty nonetheless creates an obligation of caution. Organoid-intelligence researchers have recommended embedding ethics into the research process from the outset, rather than deferring it until advanced systems already exist.
The platform should therefore operate under explicit limits: use the least complex neural model sufficient for the experiment; avoid mixing human and animal neural cells absent a specific scientific rationale; set maximum bounds on organoid size and maturation; monitor for unexpected increases in coordinated neural activity; adopt humane termination procedures; secure informed consent for human-derived cells; prohibit reproductive use; prevent environmental release; make no claims of consciousness without evidence; and halt development should morally relevant capacities become plausible.
The goal is a useful computational instrument — not biological complexity pursued to see how far it will go.
11. Overall Assessment
Is the concept scientifically possible? In parts. Neural organoid computing, fungal electrical sensing and DNA computing each rest on genuine foundations, and electronic mediation between them is conceptually achievable.
Is a complete living hybrid viable today? Not as a unified organism. Fusing human neurons, animal neurons, mycelium and DNA computing into a single living entity would generate biological, engineering, safety and ethical difficulties out of all proportion to the expected benefit.
Is a modular version viable? Potentially, as an experimental research platform. The defensible architecture keeps components separate and uses electronics to translate between them.
Would it outperform conventional computing? Not demonstrated, and unlikely on general-purpose terms. It will not compete with silicon at arithmetic, high-speed logic or general computation. Its plausible strengths lie elsewhere: biological sensing, adaptation, interaction with complex environments, self-growing interfaces, molecular information processing, low-frequency event-driven computation and self-repairing materials.
Strongest near-term alternative: a mycelium-based sensor network coupled to conventional or neuromorphic AI.
Strongest biological computing experiment: a neural organoid interfaced with electronics, with fungal data supplied digitally rather than through tissue contact.
12. Closing Vision
The future of biological computing may not be a single human–fungal mind. It may look far more like a carefully governed ecosystem — mycelium sensing the physical environment, neural tissue discovering adaptive patterns, DNA detecting and recording molecular events, electronics supplying speed and precision, artificial intelligence translating among them, and human oversight defining purpose and limits.
The value of this concept does not lie in combining as many living materials as possible. It lies in the question it forces:
Can different forms of biological intelligence and information processing be organised into a useful computational system without compelling them to become one organism?
The answer may eventually be yes. But the practical breakthrough will almost certainly arrive through selective integration, not total biological fusion.
The best future machine may be neither fully alive nor fully artificial — a modular partnership between living systems and engineered technology, each contributing the part it does best.
References
Frontiers in Science — Organoid intelligence (OI): the new frontier in biocomputing and intelligence-in-a-dish. https://www.frontiersin.org/journals/science/articles/10.3389/fsci.2023.1017235/full
Nature Electronics — Brain organoid reservoir computing for artificial intelligence. https://www.nature.com/articles/s41928-023-01069-w
Nature Reviews Chemistry — DNA data storage and molecular computing. https://www.nature.com/articles/s41570-024-00576-4
PubMed — Fungal electrical activity and mycelium-based sensing. https://pubmed.ncbi.nlm.nih.gov/40118505/
PubMed — Electrical response of fungi to changing moisture and stimuli. https://pubmed.ncbi.nlm.nih.gov/34979157/