Persistent Agent Architectures Based on Regime Dynamics. Applying General Network Theory to the Design of Artificial Intelligence Systems

This work presents an application of the TGR-Ω framework (General Network Theory – Omega formulation) to the design of artificial intelligence architectures capable of maintaining persistent internal state, continuous internal dynamics, and meta-stable operational organization. Starting from the fundamental postulates of TGR-Ω—dynamic network organization, operational unity, structural continuity, endogeneity, and meta-stable regimes—architectural conditions for persistent artificial systems are derived. Based on these conditions, a reference architecture termed WM-SSM (World Model – State Space Model) is proposed. The architecture is characterized by a persistent global latent state, continuous state-space dynamics, endogenous multi-scale memory, and regulation through dynamic regime variables. The document also introduces an initial engineering specification together with experimental protocols designed to evaluate dynamic properties such as operational continuity, functional unity, and meta-stability in artificial agents. In addition, an exploratory variant based on homeostatic principles (AI-HOME) is presented as a possible direction for future research. The overall aim of this work is to establish a structural framework for the design of persistent artificial agents by connecting systems theory, network dynamics, and artificial intelligence architecture within a unified conceptual scheme.