The Universal Constraint Engine defines simple rules. Complex behaviors emerge automatically. No programming required. This is neuromorphic computing reimagined from first principles.
Every major neuromorphic chip — Intel Loihi, IBM TrueNorth, BrainChip Akida — requires engineers to explicitly program neuron behavior. The computational properties are designed in, not discovered.
Existing chips use fixed neuron models. Engineers must decide what each neuron does before fabrication. The behavior is baked in, not emergent.
1,000-2,500 transistors per neuron. A million neurons requires billions of transistors. This limits density and drives up cost.
There is no systematic method for finding which constraint configurations produce which computational behaviors. Until now.
UCE inverts the entire paradigm. Instead of designing neurons, you declare constraints. The engine automatically discovers every emergent behavior — memory, logic gates, oscillators, attractors — without writing a single line of behavioral code.
18 simple constraints
State enumeration + propagation
18 emergent properties found
FPGA, memristive, spintronic
The Constraint-Driven Self-Emergent Neuron is defined by seven binary conserved quantities, each mapping to a biological neural function. Their interactions produce all computational behavior.
The entire computational richness of the CSEN emerges from just three types of constraint rules applied to the seven quantities.
6 rules — AND conjunction
Defines when a quantity may change. Multiple conditions must ALL be true. Creates the gating network that chains quantities together.
8 rules — OR disjunction
Defines when a quantity cannot change. ANY satisfied condition blocks the toggle. Creates inhibitory constraints and safety locks.
4 rules — mutual exclusion
Eliminates states where two quantities are both active. Creates phase opposition: excitation vs. inhibition, firing vs. refractory.
When the UCE processes 18 constraint rules over 7 quantities, it automatically discovers 18 distinct emergent computational behaviors across four categories.
All seven quantities retain their value unless specific gating conditions are met. Each acts as an independent binary memory cell — 7 bits of per-neuron memory, no flip-flops required.
A dual-pathway cascade emerges: the excitatory chain L→E→T→F→P and the inhibitory branch L→R→I. Leak (L) is the master control for the entire neuron.
Four attractor cycles emerge as oscillatory patterns, including a potentiation oscillator and a sensitization cycle — self-sustaining excitation-learning feedback loops.
Nine stable equilibrium states form an attractor basin. The neuron settles into predictable resting configurations, providing the foundation for reliable computation.
| Feature | Intel Loihi 2 | IBM TrueNorth | CSEN (Ours) |
|---|---|---|---|
| Neuron Model | Pre-programmed | Fixed LIF | Emergent from constraints |
| Transistors/Neuron | ~2,300 | ~5,400 | Target: <100 |
| Per-Neuron Memory | Configurable | Limited | 7 bits intrinsic |
| Behavior Discovery | Manual | Manual | Automatic pipeline |
| Learning Rule | Programmable STDP | External | Emergent (potentiation) |
| Memory Scaling | Linear | Fixed | 3N-1 bits (N neurons) |
Patent applications filed with the United States Patent and Trademark Office, establishing priority dates for the core UCE technology stack.
Application No. 64/036,854
System and method for discovering emergent behaviors from declarative constraint rules over conserved quantities.
Application No. 64/039,741
Constraint-driven self-emergent neuron architecture for neuromorphic computing with 7-quantity, 18-rule design.
Join the early access list. Be the first to experiment with constraint-driven neuromorphic computing.