TCS Engine - Temporal Constraint Systems for Multi-Agent AI

The first open-core framework for quantifying irreversible temporal identity dynamics in multi-agent systems.

---

Quick Install

pip install tcs-engine

---

What is TCS Engine?

Temporal Computational Selfhood (TCS) is a formal framework for modeling irreversible temporal constraints in multi-agent AI systems.

Unlike traditional agent architectures that treat identity as stateless, TCS introduces time as a computational constraint-not merely a sequence index.

Traditional RL Agent:    state -> action -> reward -> reset
TCS Agent:               state -> action -> PERMANENT CONSTRAINT -> no reset

---

Core Features (Open Source - MIT)

Mechanism	Description
Temporal Drift	Deviation from historical identity mean
Commitment Ledger	Non-erasable decision records with asymptotic decay
Regret Accumulator	Bounded monotonic counterfactual regret
Self-Prediction Loss	Self-model consistency measurement
Existential Load	Entropy-based identity complexity
Lock-In Engine	Population-level irreversible state transitions
SII	Sigma-Integration Index (statistical coupling metric)

---

Quick Start

Python API

from tcs_engine.core.agent import TemporalConstraintPopulation

# Create population of 20 agents
population = TemporalConstraintPopulation(num_agents=20)

# Run simulation - agents accumulate irreversible history
for step in range(500):
    population.step()

# Get metrics
summary = population.get_summary()
print(f"SII: {summary['sii']:.2f}")
print(f"Mean Drift: {summary['mean_drift']:.4f}")
print(f"Population Locked: {summary['locked']}")

Command Line

# Run simulation
tcs run --agents 50 --steps 500

# Run with SII analysis
tcs run --agents 50 --steps 500 --scan sii

# Show info with demo
tcs info --demo

---

The Three Principles

1. Identity Inertia - Current identity gravitates toward historical mean
2. Policy Path-Dependence - Optimal policy set shrinks with accumulated commitments
3. Historical Weight - System complexity grows monotonically with history

---

Applications

• AI Safety Research - Agents that cannot arbitrarily reset value commitments
• Institutional Modeling - Organizations with path-dependent constraints
• Long-Horizon Planning - Systems that must honor past decisions
• Multi-Agent Coordination - Populations with collective lock-in dynamics

---

Upgrade to Pro

TCS Engine Pro extends the open-source core with advanced capabilities:

Feature	Description
Phase Transition Scanner	Detect bifurcation points and critical transitions
Alignment Auditor	Risk diagnostics for temporal drift violations
Real-Time Visualizer	Live matplotlib dashboard for metrics
Experiment Manager	Batch runs with parameter sweeps
Dashboard API	REST API for integration with external systems

Coming soon - Contact for early access

---

Examples

See the examples/ directory:

• basic_simulation.py - Complete simulation demo
• temporal_drift_analysis.py - Identity drift tracking
• regret_accumulation.py - Bounded monotonic regret
• lockin_detection.py - Population lock-in dynamics
• sii_measurement.py - Sigma-Integration Index analysis

---

Citation

@software{tcs_engine,
  title = {TCS Engine: Temporal Computational Selfhood},
  author = {Mahangi, Nicolai},
  year = {2025},
  url = {https://github.com/nickmahangi/tcs-engine}
}

@article{mahangi2025tcs,
  title = {Temporal Computational Selfhood: Irreversible Identity
           Constraints in Multi-Agent Architectures},
  author = {Mahangi, Nicolai},
  year = {2025},
  journal = {arXiv preprint}
}

---

Ethics Statement

TCS Engine is a computational architecture for modeling temporal constraints.

What it is:

• Numerical mechanisms for path-dependent decision modeling
• Irreversible commitment tracking
• Population dynamics with lock-in

What it is NOT:

• A consciousness detector or measure
• A claim about machine sentience
• A proxy for phenomenal experience

The Sigma-Integration Index (SII) is a statistical coupling metric only. It is not derived from Integrated Information Theory and makes no claims about subjective experience.

---

Contributing

See CONTRIBUTING.md for guidelines. Core modules are MIT licensed and open to contributions.

---

License

• Core: MIT License - LICENSE-CORE.txt
• Pro: Commercial (coming soon)

---

Links

• Documentation
• Issues
• arXiv Paper (coming soon)

---

TCS Engine
Time as a computational constraint, not a sequence index.