The era when ChatGPT functioned solely as a static AI interface on desktops feels like ancient history, dating back only to late last year. Just a few months on, confining artificial intelligence to a simple conversation window in a desktop program appears outdated. From early tools like Claude Code and Codex, AI has accelerated into broader applications, such as OpenClaw, where it extends beyond chats to manage personal computers directly, including access to users' documents and private information.
Among leading AI developers, Anthropic leads in transforming its Claude application into a hub for autonomous AI operations, featuring coordination systems for programming tasks through Claude Code and for general desktop functions via Claude Cowork.
In the last week alone, Anthropic released Claude Dispatch, enabling remote oversight of Claude desktop agents from the mobile version of the app. Additionally, Claude Channels now allows interaction with active Claude Code sessions using platforms like Discord and Telegram, mirroring features in OpenClaw for integration with social messaging services.
These rapid advancements by Anthropic have prompted responses from competitors Google and OpenAI, with reports indicating both are accelerating their efforts.
Google, lacking a dedicated desktop program for its Gemini AI beyond the Chrome browser's sidebar integration, is currently trialing such an application, beginning with Mac users. Although it may not include autonomous programming or desktop control features—those are handled separately by the Antigravity tool—this upcoming Gemini app is expected to introduce a 'desktop intelligence' capability, permitting the AI to view screen contents for contextual understanding, akin to Microsoft's Copilot Vision.
OpenAI is developing a comprehensive desktop application that combines ChatGPT, its autonomous programming tool Codex, and the Atlas AI browser. This integrated platform will deploy AI agents to operate independently on users' machines, performing activities like software development and data examination.
These developments occur amid the surging popularity of OpenClaw, an open-source platform that deploys groups of AI agents to function quietly on personal devices, guided sporadically through phone-based apps such as WhatsApp or Telegram.
OpenClaw, which emerged prominently just two months ago, continues to generate significant interest. It is increasingly viewed not as a single application but as a foundational system for AI agents, and Nvidia CEO Jensen Huang recently elevated this perception by dubbing it 'the new computer' during the announcement of NemoClaw, the company's version emphasizing security.
Huang's perspective echoes across the industry, with others like Perplexity introducing Personal Computer—a setup resembling OpenClaw that runs on a Mac mini—and Meta's Manus AI launching My Computer, another similar agent-based solution.
Such agent-driven AI platforms seek to surpass traditional software limitations by embedding themselves extensively into users' data stores and hardware. While assurances about data protection and isolation persist, granting extensive permissions to tools like Personal Computer, OpenAI's forthcoming integrated app, or Claude's desktop version often involves exposing critical personal files and system controls.
The core objective seems to involve providing AI agents with unrestricted computer access, fueling discussions about delegating full oversight to them. With recent desktop initiatives from Google, OpenAI, and particularly Anthropic emphasizing AI agent deployment, the industry appears to be advancing toward this paradigm.