Governed vs Autonomous AI Coding: Structural Drift, Architectural Entropy, and Empirical Evidence from Three Longitudinal Case Studies

This companion study presents longitudinal empirical analysis of three open-source repositories using the EntropyX structural entropy framework. The results indicate that fully autonomous AI coding phases correlate with rapid entropy accumulation, whereas human-governed hybrid approaches demonstrate bounded structural evolution and measurable entropy cooling. The study introduces anomaly detection and entropy slope modeling for architectural governance. The work is reproducible using open tooling and public repositories.