Model-Native Computing Architecture: Envisioning Future System Architecture Through the Lens of Computer Architecture
A new paper proposes the Intelligent Computing Architecture Model (ICAM), a six-layer framework for organizing the emerging field of model-native computing. ICAM aims to unify disparate research areas like LLM-as-OS, agent frameworks, and safety governance under a cohesive system architecture. The model introduces a dual-plane view (probabilistic execution and deterministic control) and three design laws to address challenges such as cache reuse, context management, and agent scheduling. AI
IMPACT Provides a conceptual framework for organizing research and development in model-native computing systems.