Adversarial Co-evolving Execution Research Platform.
This repository is a local executable scaffold for TEK: a small research runtime that simulates an execution kernel under adaptive adversarial pressure, records stability metrics, and emits structured traces for later benchmark, stress, and visualization work.
- Local repository: ready
- Executable runtime: ready
- GitHub remote:
https://github.com/joy7758/tek-system - Public release:
v0.1.0
bash run.shRun with explicit options:
python3 -m cli.main --steps 64 --seed 11 --output artifacts/trace.jsonRun tests:
python3 -m unittest discover -s tests- Execution Kernel (TEK) with deterministic seeded evolution
- Co-evolution adversary system
- Stability monitoring
- Emergent behavior analysis
- Benchmark and stress scenario hooks
- Optional visualization helper
- Agent-readable manifest and module map
| Surface | Purpose |
|---|---|
agent_manifest.json |
Machine-readable project map, entry points, trace schema, and truth surfaces |
configs/default.yaml |
Default TEK runtime configuration |
cli/main.py |
CLI entry point |
tek_core/runtime.py |
Main run_system() orchestration |
tek_core/state.py |
Kernel state dataclass and transition helper |
coevolution/adversary.py |
Adversary action generator |
stability/monitor.py |
Stability score and phase classifier |
emergence/detector.py |
Emergence event detector |
benchmark/runner.py |
Multi-seed benchmark helper |
stress/scenarios.py |
Stress scenario presets |
visualization/plot.py |
Matplotlib trace plotting helper |
run_system() returns a list of JSON-compatible records. Each record includes:
step: integer step indexphase: stability phase labelstate: kernel state after the stepadversary_action: selected adversary actionstability: stability metricsemergence_events: detected event labels
Target release:
gh release create v0.1.0 \
--title "TEK System v0.1.0" \
--notes "Stable TEK runtime system: adversarial execution + coevolution + stability + emergence + benchmark + stress testing"v0.1.0.