thermodynamic-substrate 1.0.1

Non-von Neumann runtime with thermodynamic execution physics

Thermodynamic Computing Substrate (TCS)

A non-von Neumann runtime abstraction where execution behavior emerges from simulated thermodynamic laws rather than boolean logic.

Overview

TCS introduces computational physics to AI runtime systems. Every entity (file, process, operation) has temperature, entropy, and phase state that mathematically couple to execution parameters.

Key Features

• Mathematical Coupling: Phase coefficients directly modify execution (not if/else)
• Heat Transfer: Inter-entity thermal conductivity for autonomous workload distribution
• Self-Regulation: System maintains equilibrium without external control
• Thermal Reflection: AI adapts behavior based on environment it creates
• Phase Transitions: Execution characteristics change with thermal state

Quick Start

from thermal_substrate import ThermalSubstrate

# Initialize
tcs = ThermalSubstrate('thermal.db')

# Register entity
tcs.register_entity('my_process', 'operation', initial_temp=300)

# Execute with thermal coupling
result = tcs.execute_with_coupling('compute', 'my_process')

# Heat transfers automatically between interacting entities
tcs.heat_transfer('process_a', 'process_b')

# Get thermal state for AI reflection
state = tcs.get_system_state_prompt()

Installation

pip install thermodynamic-substrate

Or from source:

git clone https://github.com/prettybusysolutions-eng/thermodynamic-substrate
cd thermodynamic-substrate
pip install -e .

Use Cases

Autonomous Cost Control

Hot operations (>1000K) automatically throttle, preventing runaway compute costs.

Load Balancing

Heat transfer naturally distributes workload across entities.

Agent Coordination

Multi-agent systems self-regulate through shared thermal environment.

Resource Management

Frozen entities (<100K) automatically prune, freeing resources.

Documentation

• Integration Guide
• Academic Abstract

Architecture

┌─────────────────────────────────────┐
│         AI Agent / Runtime          │
└──────────────┬──────────────────────┘
               │ Operations
               ▼
┌─────────────────────────────────────┐
│   Thermodynamic Substrate (DB)      │
│  ┌─────────┐      ┌─────────┐      │
│  │Entity A │──────│Entity B │      │
│  │  800K   │ Heat │  300K   │      │
│  └─────────┘Transfer└────────┘      │
└──────────────┬──────────────────────┘
               │ Phase Changes
               ▼
┌─────────────────────────────────────┐
│    Emergent System Behavior         │
│  (Throttling / Pruning / Balance)   │
└─────────────────────────────────────┘

Performance

This is currently best framed as an experimental runtime and research prototype.

Stress test highlights:

• Initial: 397K average
• Peak: 968K (stayed below 1000K throttle)
• Final: 548K (self-regulated)
• Zero manual intervention required

Current benchmark tradeoffs for small operations:

• Overhead: +950.8%
• Total time: +532.5%

That overhead is expected at small scales where thermal bookkeeping dominates. TCS is not claiming production performance superiority yet; it is exploring whether physics-inspired coupling can unlock better load management under more realistic agent workloads.

License

MIT License - see LICENSE file

Citation

@software{thermodynamic_substrate_2026,
  title={Thermodynamic Computing Substrate: A Non-Von Neumann Runtime for AI Systems},
  author={[Authors]},
  year={2026},
  url={https://github.com/prettybusysolutions-eng/thermodynamic-substrate}
}

Contributing

Contributions welcome! Please:

• Open an issue for bugs or feature requests
• Submit PRs for improvements
• Add tests for new features
• Follow existing code style

Status

✅ Working implementation
✅ Tested and validated
✅ Experimental / research-stage
✅ Open source

---

A new kind of computer. One that runs on computational friction.