The model is interchangeable, but the accountability of its actions is not - this is where the concept of agency liability stack comes into play, defining the operational layers of AI systems with a focus on liability and control. I built the Active MirrorOS stack to implement this concept through various layers, including reality, evidence, memory, context, model, interface, narrative, consent, agency, receipt, liability, and learning. This stack is designed to ensure that every AI action can be grounded, bounded, consented, auditable, reversible, and owned. The emphasis is on building a control layer that guarantees accountability and transparency in AI operations. As I see it, "the bus is identity," and this identity must be rooted in a robust agency liability stack that prioritizes accountability. The architecture of the Active MirrorOS stack is deliberate and layered. It starts with reality and evidence, establishing a foundation for memory and context.…