This paper presents an architectural framework for artificial cognitive systems with persistent state, internal regulation, and structural explainability, called XAGI (Extended Explainable General Artificial Intelligence). Unlike conventional language models that operate as reactive generation systems, the proposed framework maintains cognitive continuity through explicit separation between linguistic processing and internal state mechanisms. The architecture integrates components of affective regulation, cumulative memory, and cognitive governance operating on a persistent core. The system was evaluated through prolonged operation tests under controlled conditions, with experimental records showing stability of internal patterns and an absence of abrupt variations in functional configuration within the analyzed time horizons. The findings suggest the viability of persistent and auditable cognitive architectures as a complement to the purely statistical paradigm in language models, providing initial evidence for the study of cognitive systems with structural temporal continuity. Specific implementation details, configuration parameters, and concrete algorithms constitute intellectual property in the process of being protected and are not disclosed in this document. Access to complete technical material is available under a confidentiality agreement for collaborative research or evaluation purposes.