Coherence Stability Principle: Active Governance in Chaotic Systems v0.3

Added Py after publication. Failed to identify the file was not attached. Version 0.3 — corrected edition. Supersedes v0.2 (December 2025), which remains on this record for transparency. This exploratory simulation illustrates stabilization of a ramp-driven logistic map through a threshold-triggered reduction and subsequent hold of its control parameter. A sign-flip detector on the state increment fires inside the period-doubling regime (r ≈ 3.34), before the constant-parameter chaos threshold (r ≈ 3.57). In the tested configuration, cutting and holding the drive returns the system to a parameter range with an attracting fixed point; under continued drive, the same cut only delays re-entry into oscillation. The experiment does not independently validate the proposed control parameter Re_EOS as a coherence measure, establish general predictive performance for the oscillation detector, or demonstrate prevention of operational collapse. These remain hypotheses requiring independent definitions, calibration, and validation. What changed in v0.3. A re-audit of the v0.2 appendix code found that the intervention cut the drive and froze the ramp, making stabilization a consequence of the policy rather than evidence for the principle; that Re_EOS reduces to the logistic parameter r under the fixed constants, with C_bind and T_lat cancelling in the normalization; and that "reliably predicts phase transitions" and "prevented collapse" were not supported. Those claims are withdrawn. v0.3 adds a Correction Record, two continued-drive intervention variants, constant-r Lyapunov exponents, a trigger-sensitivity table, and a reproducibility script that recovers every v0.2 number exactly. The v0.2 numerical results were correct; the interpretation attached to them was not. Files: paper (PDF, TXT), csp_v03_sim.py (Python 3.8+, numpy, matplotlib), two figures. Running the script reproduces all reported values. Cross-model review (Claude, GPT) was used in the audit; all analysis decisions and errors are the author's.