A Circulatory State Runtime Architecture for Resilient and Safe Autonomous Systems

This work introduces a conceptual runtime architecture for autonomous systems based on constrained circulation through a finite set of abstract system states. Rather than issuing direct commands or optimizing toward objectives, system behavior is governed by regulated transitions, intrinsic reflection stages, and mandatory return paths to a baseline condition. The architecture embeds safety, recovery, and bounded action directly into the runtime structure, allowing pathological behaviors such as looping or escalation to be detected as structural patterns rather than semantic errors. Presented without implementation details, this paper establishes a generalizable design paradigm for stability-first autonomy applicable across artificial intelligence, governance software, and long-running interactive systems.