In a recent LinkedIn post, Linas Beliūnas highlights the significant implications of Anthropic’s new Claude Cowork tool, suggesting it represents a pivotal moment for AI-native work. Beliūnas frames Claude Cowork not just as another AI application, but as a potential precursor to an “LLM as your operating system.” He emphasizes that this development could fundamentally change how individuals interact with their digital environments, moving beyond traditional app-based workflows.
The Emergence of Agentic AI for Ordinary Tasks
Linas Beliūnas explains that Claude Cowork extends the capabilities of Claude Code to non-developers, allowing users to grant the AI access to a folder and a specific goal. The AI then autonomously handles the planning, execution, error correction, and continuation of tasks without constant human intervention. This is a stark departure from current AI interactions that often require step-by-step prompting and manual data transfer.
According to Beliūnas:
Claude Cowork is basically Claude Code, but for non-developers.
You give Claude access to a folder.
You give it a goal.
It plans, executes, fixes mistakes, and keeps going.No prompts every step.
No copy-pasting between apps.
No babysitting.Just: “Handle this.” And it actually does 🤖
Beliūnas points out that this level of autonomy and capability is a “real glimpse of agentic AI for normal work,” suggesting it surpasses expectations set by tools like Microsoft Copilot and ChatGPT Desktop.
AI Building AI: A New Development Paradigm
A particularly striking aspect highlighted by Linas Beliūnas is the origin of Claude Cowork itself. He reveals that the tool was reportedly built entirely by Anthropic’s Claude AI using Claude Code in approximately 1.5 weeks. This self-building capability signifies a new era where AI is not only performing tasks but also creating the tools that enhance AI’s utility.
As Linas Beliūnas notes, this development is groundbreaking:
AI is now building AI tools to make AI even more useful.
And this marks a turning point.
Beliūnas contrasts this with previous challenges associated with autonomous agents, which were often deemed too fragile, risky, or technically demanding for widespread adoption. He underscores that Claude Cowork addresses these concerns through features like sandboxed virtual machine operation, a default human-in-the-loop design, and a focus on practical application rather than mere demonstrations.
The Future: An LLM as an Operating System
The core thesis presented by Linas Beliūnas is that Claude Cowork offers a preview of a future where Large Language Models (LLMs) function as an operating system. In this paradigm, users would define desired outcomes rather than manually operating individual applications.
Beliūnas elaborates on this vision:
An LLM as your operating system.
You don’t open apps.
You describe outcomes.Claude orchestrates:
Drive → Docs → Sheets → Slides → Web → Back again.
He argues that this shift would make the desktop environment AI-native, relegating traditional applications to mere “implementation details.” This evolution, in Beliūnas’s view, represents the phase where AI becomes an unavoidable and integrated part of daily professional life.
📝 About This Content
This article is based on insights shared by Linas Beliūnas on LinkedIn.
📅 Originally posted on January 13, 2026 | View original post on LinkedIn →