Position: It is Time to Virtualize Foundation Models with a Self-evolving Operating System Layer
Suparna Bhattacharya and colleagues argue that compound agentic systems now need a Foundation Model Operating System, a layer that virtualizes model interactions the way a virtual machine abstracts hardware.
Ask this paper
The diagnosis is per-framework runtimes. Protocols such as MCP and A2A ease connectivity, but each framework still embeds its own handling of state, memory, budgets and guardrails, so behaviour is not portable and governance does not survive a framework change.
FMOS owns four services. Knowledge across memory tiers, model selection and resource allocation, verification, and policy enforcement, which are exactly the services every framework currently re-implements.
The layer is self-evolving. It is meant to learn when to intervene and when to let inference proceed, adapting its policies from operational experience rather than from a fixed configuration.
The claim is an analogy with an obligation. The comparison to computing before operating systems is only useful if the layer can give applications the illusion of dedicated, trustworthy model instances, which is what the position paper sets as the target.
Abstract
AI applications have shifted from single, monolithic foundation models (FM) to compound agentic systems. Yet today's stacks remain fragmented: even as protocols (e.g., MCP, A2A) ease tool/agent connectivity, each framework embeds an implicit runtime for state, memory, budgets, and guardrails, making behavior non-portable and governance brittle. It mirrors computing before operating systems, when every program re-implemented basic services. This position paper argues that the field now needs a Foundation Model Operating System (FMOS) -- a system layer that virtualizes FM interactions analogous to how virtual machines abstract physical hardware, giving applications the illusion of dedicated, trustworthy FM instances with effectively unbounded capabilities. Internally, the FMOS orchestrates knowledge across memory tiers, model selection and resource allocation, and verification and policy enforcement. Like the human brain switching between fast intuition and slow deliberation, the FMOS learns when to intervene and when to let inference proceed directly and continuously adapting its policies based on operational experience.