The substrate as the computational world
Modern computation spans incompatible physical models:
- Electronic logic
- Optical propagation
- Spike-timing dynamics
- Quantum evolution
Classical IRs force these into digital abstractions. Frameworks sit above hardware and inherit its limitations.
The substrate sits beneath all of them.
It defines the representational laws, the semantic contracts, the execution model, and the lowering rules. This is the foundation of heterogeneous intelligence.
Canonical graph structure
The substrate defines a canonical graph model that all architectures map into.
Node types
Nodes represent:
- Logical operations
- Optical transforms
- Spike dynamics
- Quantum operators
- Continuous fields
- Probabilistic processes
Edge semantics
Edges carry:
- Signals
- Amplitudes
- Probabilities
- Temporal events
- Continuous flows
Graph contracts
Every graph obeys substrate-level invariants:
- Semantic preservation
- Type coherence
- Temporal consistency
- Multi-physics compatibility
These rules ensure that heterogeneous systems can coexist without conflict.
Unified semantics
The substrate provides a single semantic space capable of expressing:
- Discrete logic — exact operations and instruction-level behavior.
- Continuous analog behavior — optical fields and analog transforms as native quantities.
- Probabilistic and stochastic systems — outcomes and distributions described directly.
- Temporal and event-driven dynamics — spikes, events, and causal timing as core semantics.
- Quantum amplitudes and measurement — superposition and collapse as fundamental state.
These semantics are not layered or emulated — they are intrinsic to the substrate.
This is the first system where logic gates, optical interference, spike trains, and quantum amplitudes share the same representational world.
Execution semantics
Execution in the substrate is governed by four principles.
- Multi-physics execution — electronic, optical, spiking, and quantum processes execute together.
- Heterogeneous scheduling — the substrate coordinates execution across multiple architectures without losing meaning.
- Semantic preservation — execution preserves the semantics declared in the graph — regardless of hardware.
- Dynamic graph evolution — graphs can restructure, adapt, evolve, and self-modify during execution.
This enables systems that are not static — they are alive within the substrate.
Lowering and mapping
Lowering is the process of translating substrate graphs into hardware execution. The substrate lowers into:
- CPUs, GPUs, TPUs
- Photonic processors
- Neuromorphic arrays
- Quantum devices
- Hybrid systems
Lowering is semantic-preserving, not a lossy translation.
The substrate ensures that meaning survives the transition from representation to execution.
Substrate invariants
All architectures must obey substrate-level invariants:
- Unified semantics
- Canonical graph structure
- Consistent temporal behavior
- Multi-physics compatibility
- Semantic preservation across lowering
These invariants guarantee coherence across heterogeneous systems.
They are the laws of the computational world.
Role in the ecosystem
The substrate is the foundation of the Pinnacle Ecosystem. It connects the whole doctrine into a single line of reasoning:
- Missing Layer — identifies the fracture.
- What HeteroIR Is — defines the substrate.
- Capabilities — describes what the substrate can do.
- The Substrate — formalizes how it works.
- Vision — shows where it leads.
This page is the technical backbone of the entire doctrine.