Two-Regime Control: Latent Coordination vs Compensation in Intelligent Systems

Intelligent embodied systems often operate in two distinct regimes: a low-energy latent coordination regime where distributed dynamics carry control, and a high-energy compensation regime characterized by corrective overrides, stiffness escalation, estimator churn, and frequent replanning. This paper introduces a regime-level framing, proposes measurable proxies for “compensation load” (a compensation index and duty-cycle style persistence), and outlines a slow baseline regulator layer that biases behavior back toward latent coordination without micromanaging task control. We provide falsifiable predictions and practical experiments for locomotion and manipulation.