Pre-Stabilisation Dynamics: A Process-First Ontology for Drift, Formation, and Governance in Complex Systems

Several independent research traditions — distributed systems, machine learning, complexity science, biological development, control theory, and AI safety — have each documented structurally similar dynamics: systems exhibit directional, metastable behaviour before any stable entity exists. This paper suggests that these observations reflect a shared generative phase called pre-stabilisation — the interval in which behavioural patterns are forming but have not yet consolidated. We introduce Verbanatomy, a process-first ontological framework designed to identify and govern behavioural drift in complex autonomous systems. The framework offers six primitives — Signals, Tendencies, Constraints, Pre-Nodes, Nodes, and Clusters — as a computable telemetry layer for formation-layer interventions. By formalising the Pre-Node as a metastable saddle point (identified via indefinite Hessian geometry), we propose a shared vocabulary for governance that functions across system boundaries and multi-domain agentic pipelines. This paper serves as the theoretical foundation for two companion works: A Calculus of Governance: Formalising Pre-Stabilisation Dynamics, and Pre-Stabilisation Signals in Transformer Systems: An Empirical Investigation. Keywords: AI Safety, Complexity Science, AI Governance, Pre-Stabilisation, Verbanatomy, Systemic Drift, Process Ontology.