MeridianSeal: Making AI-Agent Boundaries Architectural

Why it surfaced

MeridianSeal turns its argument into a worked exhibit: a 60-class action catalogue, a boundary function called L(a,G), and twelve layers that ask what would actually stop an agent from exceeding its mandate. The distinction between rules an agent is merely told to follow and controls it cannot reach around gives this research project a concrete, inspectable shape.

A bilingual public showroom for a PhD dissertation on governing autonomous AI agents through enforceable system architecture rather than instructions alone.