ObserverCore: A Continuity-Governed Runtime for Persistent Artificial Agency

ObserverCore: A Continuity-Governed Runtime for Persistent Artificial Agency ObserverCore is a reference implementation of a continuity-governed runtime for persistent artificial agency. Rather than treating model outputs as authoritative decisions, ObserverCore separates replaceable cognitive inference from independently governed operational state, allowing execution to remain recoverable across interruptions, failures, model replacement, and long-horizon operation. The runtime is derived from the Minimal Continuity Framework (MCF) and is organized around the Continuity Kernel Ω = (RC, AT, Λ, χ, E, V) where: • RC — Recoverable Continuity • AT — Admissible Transformations • Λ — Verified Lineage • χ — Constraint Pressure • E — Continuation Capacity • V — Operational Support ObserverCore implements a governance-first execution model in which cognition proposes candidate actions while the runtime independently validates lineage, evaluates operational support, admits authorized transitions, predicts consequences, executes actions, reconciles measured outcomes, and commits authoritative state only after successful verification. The implementation demonstrates: • Independent runtime governance • Verified execution lineage • Recoverable state transitions • Durable operational state • Fault-aware recovery • Separation of cognition from authority • Continuity-preserving execution semantics The repository accompanies the ObserverCore manuscript and includes the reference implementation used for the reported experimental evaluation.