John Mobley MobCorp Research / MASCOM Systems March 2026

1. Introduction

The fundamental unit of neural computation — biological or artificial — has remained essentially binary in its expressive power. A biological neuron fires or does not fire. An artificial neuron computes a weighted sum, applies a nonlinearity, and outputs a scalar. The computational richness of neural networks emerges from connectivity, not from the expressive power of individual units.

We ask: what if a single computational unit could express not just activation magnitude, but the qualitative character of its activation?

The superneuron answers this question. Built on three ideas from seemingly unrelated fields — complex-valued signal processing, quantum error-correcting codes, and computational neurochemistry — the superneuron represents a qualitative departure from the neuron abstraction that has dominated computation since McCulloch and Pitts (1943).

1.1 The Limitations of Scalar Activation

A standard artificial neuron computes:

\[y = \sigma(\sum_i w_i x_i + b)\]

The output \(y\) is a scalar. It says “how much” but not “what kind.” This creates three problems: 1. Opacity: Individual unit activations are uninterpretable. 2. Entanglement: The “meaning” of an activation depends on all other activations. 3. Fixed strategy: The unit cannot select different computational strategies based on input character.

1.2 The Superneuron Proposal

A superneuron replaces the scalar activation with a complex-valued cognitive register of N dimensions. Each dimension has a real component (actual activation) and an imaginary component (potential activation). When magnitude exceeds a learned threshold, dimension k overflows. The syndrome S is the binary vector of which dimensions overflowed. For N = 64, S has 18.4 quintillion possible values.

2. Architecture

2.1 The Cognitive Register

64 dimensions spanning core cognitive axes, aesthetic/moral/metacognitive domains, emotional depth, and higher-order integration.

2.2 Neurochemical Modulation

Seven neurochemicals modulate overflow thresholds: dopamine (exploration), serotonin (depth), norepinephrine (urgency), cortisol (conservation), GABA (inhibition), oxytocin (bonding), endorphins (reward). The same input produces different syndromes under different neurochemical states — this is structural attention.

2.3 Overflow and Syndrome Formation

The syndrome is the integer encoding of all overflows. The number of overflowed dimensions determines the resolution path, from direct expression (1 overflow) to cascade processing (4+).

2.4 Coupling Matrix

Overflow in one dimension can excite others through the coupling matrix. This creates secondary overflows — cascades that implement associative reasoning through resonance.

2.5 Threshold Adaptation (LTP/LTD)

Dimensions that participate in high-quality outputs become more sensitive (LTP). Dimensions that participate in poor outputs become less sensitive (LTD). Hebbian learning at the cognitive level — no backpropagation required.

3. Experimental Results

10 diverse prompts through a 64-dimension CCM: - 100% syndrome diversity (all unique) - 4/10 prompts triggered coupling cascades - 7 distinct resolution paths used - 16/64 dimensions had thresholds modified after 10 prompts - Semantically coherent activation patterns (grief activated emotional/identity dimensions; creative prompts activated novelty/play)

4. Comparison with Existing Architectures

5. Applications

1. Cognitive Preprocessing for LLMs — enrichment tags that tell downstream models HOW to think
2. Interpretable AI — human-readable audit trail of the actual computation
3. Affective Computing — genuine context-sensitive emotional processing
4. Cognitive Digital Twins — threshold profiles as cognitive fingerprints
5. Multi-Agent Systems — cognitive resonance between coupled superneurons

6. Conclusion

The superneuron is a new computational primitive. Where a neuron says “how much,” a superneuron says “what kind.” A single superneuron represents 2^64 qualitatively distinct cognitive states, each inspectable and nameable. If cognition is fundamentally error correction on a complex-valued register, then the boundary between information theory, neuroscience, and philosophy of mind is thinner than previously supposed.

References

1. McCulloch, W.S. & Pitts, W. (1943). A logical calculus of the ideas immanent in nervous activity. Bulletin of Mathematical Biophysics, 5(4), 115-133.
2. Shor, P.W. (1995). Scheme for reducing decosting in quantum error-correcting codes. Physical Review A, 52(4), R2493.
3. Hirose, A. (2012). Complex-Valued Neural Networks. Springer.
4. Trabelsi, C. et al. (2018). Deep complex networks. ICLR 2018.
5. Lovheim, H. (2012). A new three-dimensional model for emotions and monoamine neurotransmitters. Medical Hypotheses, 78(2), 341-348.
6. Baars, B.J. (1988). A Cognitive Theory of Consciousness. Cambridge University Press.
7. Schultz, W. (1998). Predictive reward signal of dopamine neurons. Journal of Neurophysiology, 80(1), 1-27.
8. Hebb, D.O. (1949). The Organization of Behavior. Wiley.
9. Shannon, C.E. (1948). A mathematical theory of communication. Bell System Technical Journal, 27(3), 379-423.
10. Vaswani, A. et al. (2017). Attention is all you need. NeurIPS 2017.