MARE Meta-Architecture: Anti-Obsolescence by Design for Sovereign and Sustainable AI Systems

Artificial intelligence systems are increasingly exposed to technological volatility. Models, libraries, APIs, runtimes, cloud services, and hardware platforms evolve, disappear, or become economically unsustainable at a pace that many organizations cannot follow. As a result, a component may remain functionally adequate while changes in the surrounding technological ecosystem make its continued use, maintenance, or replacement increasingly difficult. For well-resourced institutions, this is a recurring engineering cost; for non-profits, schools, libraries, municipalities, and other public-interest organizations, it can become a barrier to technological participation. This paper argues that technological obsolescence in AI systems is not only a component-level problem, but also an architectural one. We therefore introduce the MARE Meta-Architecture, as a technology-agnostic model for preserving functional continuity under technological change. MARE separates stable architectural responsibilities from the mechanisms that currently realize them. Systems are described through Roles, which define what must remain architecturally stable, and Executors, which enact those responsibilities through replaceable technological, organizational, or human mechanisms. Interactions among Roles are organized through Transfer, Activation, and Observation (TAO) semantics, allowing coordination to remain meaningful across heterogeneous implementation technologies. Recursive Role decomposition further enables complex systems to be represented at different levels of abstraction. The paper grounds MARE in MGVT (Multimodal Generative Virtual Teacher), an implemented AI tutoring system, from which it abstracts a broader meta-architectural model for AI systems exposed to ecosystem volatility. The contribution of MARE is the coordinated definition of Role stability, Executor replaceability, TAO interaction semantics, and recursive encapsulation as mechanisms for preserving architectural continuity across technological change. Anti-obsolescence is thereby pursued not through continuous corrective maintenance, but by design. Together, these mechanisms provide a coherent vocabulary for maintainability, heterogeneity, and temporal interoperability in AI-intensive systems.