Persistent Geometric Memory in Langevin Responders: A Phase Diagram for Vanished-Source Localization
Thomas S. Mitchell · Zenodo (CERN European Organization for Nuclear Research) · 2026
This archive contains the frozen simulation instrument, parameter manifest, raw results, analysis scripts, and figures associated with: Persistent Geometric Memory in Langevin Responders: A Phase Diagram for Vanished-Source Localization The project tests a single computational question: Does a statistically significant region of parameter space exist in which a vanished source remains localizable from the post-removal state of responding agents? A population of 2,000 agents evolves on a 512×512 domain under Langevin dynamics while responding to a transient Gaussian source. The source is removed after a fixed lifetime, and source recovery is attempted using only the post-removal responder state. The study employs: Frozen simulation code Predefined success and failure criteria Randomized source ensembles Empirical chance baselines Noise robustness sweeps Source lifetime sweeps Results identify a contiguous below-chance region ("recoverable island") within parameter space where vanished sources remain localizable from post-removal responder velocities. Included in this archive: Frozen simulation instrument (pgm_simulation.py) Parameter manifest (FROZEN_MANIFEST.md) Raw trial results (pgm_full_results.json) Figure-generation scripts Phase-diagram visualizations Written conclusion generated from recorded results This work reports a computational phase-diagram result within the tested Langevin framework. It does not claim biological validation, universality, or the existence of a general law beyond the examined system. Development involved iterative adversarial review and hypothesis refinement using ChatGPT ("Edwin"), DeepSeek, Meta AI, and Zo Computer as analytical tools. Final study design, interpretation, and publication decisions remain the responsibility of the human author.