跳转至

文章背景与核心概要

随着人工智能技术的飞速发展，AI 智能体正逐渐演变为能够调用工具、维护记忆并在各类应用间完成复杂任务的长期运行计算实体。然而，在实际应用场景中，任务的完成往往需要人类与 AI 交替主导，这种协作模式高度依赖于工作状态在人机之间的高效无缝交接。针对当前主流智能体系统由于人机工作环境隔离而导致的交接成本高、扩展性差的问题，本文作者提出了“人机共栖（Human-AI Co-inhabitation）”这一全新概念，并设计实现了操作系统级别的解决方案——CrabOS。

CrabOS 创新性地将工作状态表示为人类与 AI 均可直接读写的自然语言文本对象，并通过统一的界面使双方能够直接访问和操作工作状态，从而彻底摒弃了繁琐的外部桥接机制。通过将复杂任务的支持从依赖特定应用的解决方案提升为原生操作系统级能力，CrabOS 为未来 AI 智能体的开发和运行奠定了坚实的基础，开启了人机协同计算的新范式。

CrabOS: An Operating System for Human-AI Co-inhabitation

Summary

  • Authors: Qi Yang, Yun Ma
  • Submitted: August 28, 2026
  • Primary Subject: Artificial Intelligence (cs.AI)
  • Additional Subjects: Human-Computer Interaction (cs.HC), Operating Systems (cs.OS)
  • Links: arXiv:2608.28165 | View PDF | TeX Source

Abstract

AI agents are evolving into long-running computational entities that can invoke tools, maintain memory, and complete complex tasks across applications. In real-world settings, completing a task often requires humans and AI to take turns leading its execution. Such alternation depends on the seamless handoff of the work state of the task between humans and AI.

AI agents are evolving into long-running computational entities that can invoke tools, maintain memory, and complete complex tasks across applications. In real-world settings, completing a task often requires humans and AI to take turns leading its execution. Such alternation depends on the seamless handoff of the work state of the task between humans and AI.

现有智能体系统为人类和 AI 提供了各自独立的工作环境。因此，AI 智能体在接力工作时必须依赖额外的桥接手段：要么由开发者构建特定任务的接口来访问工作状态，要么由用户通过截图或文本描述手动传输相关部分。这两种方法都导致交接成本高昂且难以扩展。

Existing agent systems, however, provide humans and AI with separate work environments. AI agents must therefore rely on additional bridges to continue work: either developers build task-specific interfaces to access the work state, or users manually transfer relevant parts of it through screenshots or textual descriptions. Both approaches make handoffs costly and scale poorly.

为了解决这一问题，作者提出了人机共栖（Human-AI Co-inhabitation）——一种能够让人类与 AI 无缝交替继续处理同一任务的工作环境。他们设计并实现了 CrabOS 来实现这一概念： * 共享状态（Shared State）： CrabOS 将工作状态表示为人类和 AI 均可阅读的自然语言文本对象，供双方共享。 * 可审计界面（Auditable Interface）： 双方无需外部桥接，即可通过同一界面直接访问和操作工作状态。 * 操作系统级支持（Operating System Level Support）： 案例研究表明，CrabOS 将对复杂任务的支持从依赖桥接的应用级解决方案提升为了原生操作系统级能力，为开发和运行 AI 智能体奠定了新的基础。

To address this, the authors propose Human-AI Co-inhabitation—a type of work environment that enables humans and AI to seamlessly take turns continuing work on the same task. They design and implement CrabOS to realize this concept: * Shared State: CrabOS represents the work state as natural-language-readable text objects shared by humans and AI. * Auditable Interface: Both parties can access and manipulate the work state directly through the same interface without needing external bridges. * Operating System Level Support: Case studies demonstrate that CrabOS elevates support for complex tasks from bridge-dependent application-level solutions to native operating-system capabilities, establishing a new foundation for developing and running AI agents.

Submission History

  • [v1] Fri, 28 Aug 2026 10:28:54 UTC (98 KB)

Submission History

  • [v1] Fri, 28 Aug 2026 10:28:54 UTC (98 KB)