# Position: It is Time to Virtualize Foundation Models with a Self-evolving Operating System Layer

Published: 2026-09-18 · Updated: 2026-09-18 · Category: Research

Source page: https://www.sipply.dev/story/position-it-is-time-to-virtualize-foundation-models-with-a-a01a0d8a · 中文: https://www.sipply.dev/zh/story/position-it-is-time-to-virtualize-foundation-models-with-a-a01a0d8a

## Summary

A recent paper proposes the creation of a Foundation Model Operating System (FMOS) to tackle the existing fragmentation in AI applications that leverage foundation models. The authors argue that by virtualizing interactions among these models, FMOS can provide a unified system layer that enhances portability, governance, and operational efficiency. This approach aims to streamline the deployment and management of AI applications, making them more accessible and manageable across various platforms.

## Key takeaways

- The proposed FMOS could significantly reduce fragmentation in AI applications.
- Virtualization of foundation models may lead to improved governance and efficiency.
- A cohesive system layer could enhance the portability of AI applications.

## Why it matters

The development of FMOS could revolutionize how AI applications are built and managed, leading to more integrated and efficient systems. This is crucial as the demand for scalable AI solutions continues to grow.

- **Product managers:** For product managers, the FMOS concept offers a potential framework to streamline product development and deployment, ensuring that AI solutions can be easily integrated and scaled across different environments.
- **Developers:** Developers and engineers may benefit from a standardized operating system for foundation models, which could simplify the complexities of integrating various AI components and enhance collaboration across teams.
- **Students and job seekers:** Students and job-seekers should pay attention to FMOS as it represents a cutting-edge area of research that could define future job opportunities in AI development and application management.

## Sources

- [Position: It is Time to Virtualize Foundation Models with a Self-evolving Operating System Layer](https://arxiv.org/abs/2609.19203) (arXiv cs.AI, 2026-09-18)

## 中文

### 职位：是时候用自我演化的操作系统层虚拟化基础模型了

最近的一篇论文提出创建一个基础模型操作系统（FMOS），以解决利用基础模型的AI应用中的现有碎片化问题。作者认为，通过虚拟化这些模型之间的交互，FMOS可以提供一个统一的系统层，增强可移植性、治理和操作效率。这种方法旨在简化AI应用的部署和管理，使其在各种平台上更易于访问和管理。

**要点：**

- 提议的FMOS可以显著减少AI应用中的碎片化。
- 基础模型的虚拟化可能导致治理和效率的改善。
- 一个统一的系统层可以增强AI应用的可移植性。

**为什么重要：** FMOS的开发可能会彻底改变AI应用的构建和管理方式，导致更集成和高效的系统。这在对可扩展AI解决方案的需求不断增长的背景下至关重要。

---
Sipply: Sip AI News Simply. https://www.sipply.dev