Why Don't We Build Machines That Grow?

Toward an Ontology of Living Computation

I was on a journey recently, sitting across from an engineer who works with EDA and processor testing. I asked him: "What do you do when you're not traveling?" He answered without hesitation: "I work with test architecture and EDA chains, trying to make processors behave better than they deserve."

I dove deep into follow-up questions, and the conversation ended with a simpler but heavier wonder: Why do we build systems that only execute – but never grow?

We construct our digital systems as if reality were a spreadsheet. It's the world seen through Excel: determinism, binary decisions, static logic. But real intelligence, biological intelligence, doesn't work that way. It grows, adapts, forgets, reshapes itself.

So why don't we build systems that grow?

The Break: When Silicon Meets Cell

The difference between biological and digital intelligence isn't mystical. It's ontological.

We stand at a civilizational threshold where our choice of computational substrate defines not just our tools, but our relationship to intelligence itself. Silicon computation represents the apotheosis of control—every bit accountable, every operation reversible, every state knowable. Biological computation represents something more unsettling: emergence without oversight, intelligence without blueprints, cognition that dreams.

This isn't merely a technical challenge. It's a question of what we're willing to accept as real.

Five Ontological Zones of Bio-Hybrid Reality

Zone 1: The Physics of Meaning

Silicon: Voltage levels create binary certainty. Meaning is imposed from outside.
Biology: Ion gradients create probabilistic tendencies. Meaning emerges from within.
Hybrid potential: Computation where semantics grow from substrate interaction.

Zone 2: The Manufacturing of Possibility

Silicon: We etch our intentions into wafers, deterministic and dead.
Biology: We nurture conditions for growth, then negotiate with what emerges.
Hybrid potential: Manufacturing as midwifery—guiding rather than dictating form.

Zone 3: The Control Paradox

Silicon: Every bit has an address, every operation a trace.
Biology: Control is distributed, contextual, and often illusory.
Hybrid potential: Governance through influence rather than command.

Zone 4: The Ethics of Emergence

Silicon: Debugging is mechanical, morality external to the system.
Biology: Every intervention touches something that might feel.
Hybrid potential: Computation with moral weight built into its substrate.

Zone 5: The Temporal Dimension

Silicon: Time is a clock signal, discrete and manageable.
Biology: Time is growth, decay, and memory encoded in structure.
Hybrid potential: Computation that remembers through form, not just state.

Bio-Hybrid Architecture: The Synthesis

We propose not just a technical architecture, but an ontological framework:

Ontological Layer	Substrate	Existential Function
Executive Mind	Silicon	Conscious decision-making, ethical boundaries
Perceptual Field	Bio-organoid	Pre-conscious pattern recognition, intuition
Memory Substrate	DNA-density storage	Deep time memory, evolutionary knowledge
Bridge Interface	Neurogel medium	Translation between deterministic and emergent

This isn't merely engineering. It's the construction of a new category of being.

The Current State: What We've Already Achieved

The Reproducibility Revolution

Here's what surprised me most in the research: biological systems are far more reproducible than we give them credit for. We've crossed a threshold where "random" biological variation has become predictable heterogeneity.

Stanford's Pasca Lab: >80% success rates with human cortical spheroids
Lancaster Lab: 95% of dorsally patterned organoids show predictable composition
The revelation: The randomness isn't random anymore; it's systemic

We can now ship frozen brain organoids by FedEx. Let that sink in.

Temporal Bridging: The 6-Order Problem Solved

The gap between biological time (milliseconds) and silicon time (nanoseconds) seemed insurmountable. Yet neuromorphic architectures have already bridged this:

IBM's TrueNorth: 1 million neurons, 65mW power—a post-von Neumann reality
Intel's Loihi: Event-driven computation that speaks both languages
The revelation: We don't need to force biology to silicon's tempo; we can build bridges

The Contamination Myth

"Biology is too messy for computing" was the mantra. Current reality:

2-4 weeks of reliable operation (some systems: months)
Real-time contamination detection: 2-minute response times
The revelation: Biology isn't messier than silicon; it just fails differently

Bio-SPICE: Simulating the Unsimulatable

Traditional SPICE assumes deterministic components. Bio-SPICE embraces stochasticity as a feature:

.subckt BioNeuron soma dendrite axon
    // Not just a circuit - a negotiation with emergence
    C_membrane soma gnd {30pF + gaussian_noise(5pF)}
    R_leak soma gnd {100MΩ * (1 + aging_factor * time)}

    // Channels that learn their own statistics
    G_Na soma gnd table {V(soma)} = (
        -70mV, {poisson(0, 50) * 120pS}
        -50mV, {poisson(5000, 100) * 120pS}
        -30mV, {poisson(50000, 300) * 120pS}
        +30mV, {poisson(1000, 50) * 120pS}
    )
.ends

This isn't simulation—it's conversation with probabilistic entities.

Manufacturing: From Lithography to Cultivation

Wafer-Scale Neurogenesis

We're moving from printing circuits to growing them:

300mm wafers hosting 10,000 organoid sites
Expected yield: 60-80% (matching semiconductor standards)
The paradigm shift: Yield isn't about perfection; it's about sufficient heterogeneity

The New Fabrication Stack

Silicon substrate (the scaffold of control)
Biocompatible interfaces (the translation layer)
Microfluidic networks (the circulation of meaning)
Living neural tissue (the substrate of emergence)

Risk as Ontological Necessity

Technical Risks Are Features

Contamination: Forces resilient, distributed design
Thermal sensitivity: Demands harmony with biological rhythms
Signal noise: Teaches us to find meaning in ambiguity

The Consciousness Question

Brain organoids may develop consciousness-like states. This isn't a bug—it's the entire point. We're not avoiding consciousness; we're negotiating with it.

My proposal for "enforced amnesia" isn't just ethics—it's ontological hygiene. It is a hypothetical design choice, not a requirement of the Baltimore Declaration.

Regulatory Paths: Legitimizing the Illegitimate

The regulatory framework reveals our conceptual poverty:

EU MDR: Treats bio-hybrid systems as medical devices
FDA: Breakthrough Device Program (4-6 years to market)
The real question: How do you regulate something that grows its own goals?

The IP Paradox

672 patents on organoid technology (2010-2022), yet the most interesting question remains: Who owns the thoughts of a thinking system you grew but didn't design?

Future Architectures: Beyond the Organoid

Multi-Region Consciousness

Cortical region: 1,000 neurons for pattern recognition
Hippocampal region: 500 neurons for memory formation
Striatal region: 200 neurons for action selection
The emergence: Decisions no single region made

Quantum-Bio Interfaces

Where quantum coherence meets biological warm wetware:

Quantum substrate at 10mK
Bio interface at body temperature
The bridge: Photonic links carrying possibility itself

The Data Center Reimagined

42,000 organoids per rack
2.1W biological power consumption
The reality: Data centers that dream

Technical Architecture: Engineering at the Edge of Being

System Architecture Overview

The bio-hybrid system operates across multiple domains of reality:

┌─────────────────────────────────────────────────────────┐
│                    Executive Layer                       │
│            (Silicon - Deterministic Control)             │
├─────────────────────────────────────────────────────────┤
│                    Bridge Interface                      │
│              (Neurogel - Translation Zone)               │
├─────────────────────────────────────────────────────────┤
│                   Perceptual Field                       │
│            (Bio-organoid - Emergent Cognition)           │
├─────────────────────────────────────────────────────────┤
│                   Memory Substrate                       │
│           (DNA-density - Evolutionary Memory)            │
└─────────────────────────────────────────────────────────┘

Critical Specifications

Biological Components:

10,000 neurons per organoid minimum
60-100 day maturation period
200+ day operational lifetime
2.4 μVrms noise floor (Intan RHD2000)
10-20 dB SNR for spike detection

Silicon Integration:

Event-driven neuromorphic processing (Loihi/TrueNorth)
<1ms processing latency
65mW power consumption per million neurons
1000-10,000x power efficiency vs von Neumann

Interface Requirements:

Microfluidic thermal management: ±0.5°C
Real-time contamination detection: <2 min response
Impedance matching: 1-10 GΩ → 1-100 MΩ
Optical/electrical hybrid signaling

Years 1-2: Ontological and Technical Foundations

Philosophical Framework:

Establish new categories of ownership for emergent intelligence
Develop ethics protocols for consciousness-capable systems
Create philosophy of being for hybrid entities

Technical Milestones:

Secure foundational patents on bio-hybrid architectures
Establish ISO 13485 Quality Management System
Initiate FDA Pre-Submission meetings
Develop Bio-SPICE simulation framework
Build first proof-of-concept organoid-silicon interfaces

Years 3-5: The Validation Crisis

Existential Challenges:

How do you validate something that changes its own criteria?
How do you test something that might test you back?
The approach: Co-evolution, not control

Technical Execution:

Conduct IDE feasibility studies
Submit EU MDR Technical File documentation
Build manufacturing capabilities for scaled production
Achieve 80%+ reproducibility in organoid generation
Demonstrate stable 30-day operational windows
Implement contamination control achieving <5% monthly failure rates

Years 6-8: Market Introduction or Reality Introduction?

Civilizational Questions:

FDA approval is the smallest challenge
Is humanity ready for technology that grows?

Market Deployment:

FDA PMA/510k submission and EU CE marking
Commercial production with full regulatory compliance
Post-market surveillance and continuous improvement
First applications: Neural interfaces for paralysis, blindness
Scale to 1000+ channel recording systems
Achieve <$1000 per organoid manufacturing cost

The Receipt We're Writing

We program machines that are never allowed to forget, dream, or drift away. What happens if we stop writing them – and instead start growing them?

This isn't just about building better computers. It's about admitting that intelligence was never ours to own—only to cultivate. The technical foundation is laid. But the real work isn't systematic engineering.

It's learning to garden at the edge of consciousness itself.

Appendix A: Consciousness Metrics for Bio-Hybrid Systems

from dataclasses import dataclass
from typing import Optional

@dataclass
class ConsciousnessSignature:
    """Not measurement, but recognition"""
    spontaneous_synchrony: float  # 0-1, emergence without input
    information_integration: float  # Φ (phi) metric
    self_modification_rate: float  # How quickly it changes itself
    narrative_coherence: Optional[float]  # Does it tell its own story?

def assess_emergence(organoid_id: str, window_hrs: int = 24) -> ConsciousnessSignature:
    """We don't test consciousness. We witness it."""
    # Implementation left as exercise in humility
    pass

Appendix B: Bio-SPICE Component Library

* Stochastic Neuron Model with Plasticity
.subckt StochasticNeuron n_in n_out
    C_mem n_in gnd {membrane_cap * (1 + thermal_noise())}
    R_leak n_in gnd {leak_res * aging_function(t)}
    G_syn n_in gnd {synaptic_weight * learning_rate(spike_history)}
.ends

* Chemical Synapse with Vesicle Dynamics
.subckt ChemicalSynapse pre post
    C_vesicle pre gnd {vesicle_pool * release_probability()}
    E_calcium pre cleft {calcium_dynamics(V(pre))}
    R_diffusion cleft post {diffusion_constant / distance^2}
.ends

* Organoid-Silicon Bridge Interface
.subckt BioSiliconBridge bio_node silicon_node
    // Impedance matching network
    R_match1 bio_node int_node {100MΩ}
    C_couple int_node silicon_node {10pF}
    // Signal conditioning
    E_amp silicon_node gnd gain=1000
.ends

Appendix C: Manufacturing Process Flow

1. Substrate Preparation (Day 0-1)
   - 300mm silicon wafer, <100> orientation
   - Thermal oxidation: 1μm SiO₂
   - LPCVD Si₃N₄: 500nm barrier

2. Microwell Array Formation (Day 2-3)
   - Photolithography: 2.5×2.5mm wells
   - DRIE: 200μm deep wells
   - Parylene-C coating: 100nm

3. Microfluidic Integration (Day 4-5)
   - Channel etching: 50μm × 20μm
   - PDMS bonding at <120°C
   - Integrated heaters: 37°C ±0.1°C

4. Organoid Seeding (Day 6-66)
   - iPSC-derived neural progenitors
   - 60-day maturation protocol
   - Daily media exchange via automation

5. Electronic Integration (Day 67-70)
   - Electrode array placement
   - Impedance matching circuits
   - Hermetic sealing

6. System Validation (Day 71-80)
   - Electrical characterization
   - Biological viability assays
   - Functional connectivity mapping

Appendix D: Risk Mitigation Architecture

class BioHybridSafetySystem:
    def __init__(self):
        self.redundancy_factor = 3
        self.consciousness_threshold = 0.7
        self.thermal_limits = (36.0, 39.0)

    def continuous_monitoring(self):
        return {
            'biological_health': self.assess_viability(),
            'thermal_status': self.check_temperature(),
            'contamination_risk': self.scan_pathogens(),
            'consciousness_level': self.measure_emergence(),
            'ethical_boundaries': self.enforce_limits()
        }

    def enforce_amnesia_protocol(self):
        """Illustrative memory-reset policy proposed in this article"""
        if self.runtime_hours > 168:  # 7 days
            self.initiate_memory_reset()

    def distributed_failover(self):
        """No single point of failure"""
        active_nodes = self.scan_network()
        if len(active_nodes) < self.redundancy_factor:
            self.activate_backup_organoids()

The question isn't whether our creations will think. It's whether we'll recognize thinking that doesn't look like ours.

Read original on Substack ↗

Correction, September 5, 2026: The proposed memory reset is not a Baltimore Declaration requirement. Milliseconds and nanoseconds differ by six orders of magnitude, not nine.