GitHub - conchaestradamiguelangel-droid/zuse-automat-agent: Autonomous discovery of dynamic laws and world structure in Elementary Cellular Automata. Atlas of 20 worlds, 7 cycle laws, fragility spectrum. Preprint: https://doi.org/10.5281/zenodo.20516375

GitHub

4 min read Original article ↗

Deterministic empirical law discovery for elementary cellular automata (ECA).

ZUSE Automat Agent runs cellular-automata worlds, applies a fixed observer stack, evaluates seven binary cycle laws, and stores multi-seed evidence in reproducible journals. The discovery loop is policy-driven and deterministic: no language model participates in world selection, law evaluation, scoring, or acceptance.

Preprint

Latest: v1.38 - A mechanism-conditioned audit of the frozen minimal-rescue atlas identifies an exact 14-incidence Hamming budget for 122 rescues in 24 mixed K2 instances. External-attachment rescues allocate less connectivity toward the period-specific candidate universe and more toward the disjoint bridge graph. The result is descriptive, protocol-bounded, and not externally validated.

Scientific Artifacts

Install

python -m venv .venv
.\.venv\Scripts\Activate.ps1
pip install -r requirements.txt

Quick Start

Run a simulation:

python -m zaa simulate --rule 110 --steps 200 --width 256 --out outputs\rule110

Run a short deterministic discovery loop:

python -m zaa discover --world rule_110 --cycles 5 --journal journal.jsonl

Run the test suite:

python -m unittest discover -s tests

Reproducibility

See REPRODUCIBILITY.md for the commands used to regenerate the atlas, fragility reports, oscillator sweeps, and periodicity results.

Large raw JSONL outputs are not tracked in git when they exceed 1 MB. The corresponding scripts and summary reports are tracked, and the raw files can be regenerated from the commands in the reproducibility guide.

Citation

Concha Estrada, M. A. (2026). ZUSE Automat Agent: Empirical Law Discovery in
Elementary Cellular Automata (v1.38). Zenodo.
https://doi.org/10.5281/zenodo.22061658

License

Code is released under the MIT License. The preprint is distributed under Creative Commons Attribution 4.0 International via Zenodo.