Anthropic has unveiled Claude Cowork, a desktop application designed to bring local-first agent workflows to non-technical users. The tool emerged from an unexpected discovery: users were leveraging Claude Code for far more than coding—including complex knowledge work that had nothing to do with programming.
From Coding Tool to Knowledge Work Assistant
The insight came from Felix Rieseberg and the Anthropic team, who noticed that technical builders were using Claude Code for "lots of non-technical work." Rather than fighting this trend, Anthropic leaned into it, creating a more user-friendly version that maintains the power of agentic workflows without requiring terminal expertise.
"Claude Cowork wrote itself," according to Rieseberg. With a small team simply orchestrating multiple Claude Code instances, the tool was ready after just 10 days—a testament to how cheap execution has become in the age of AI.
Why 10 Days Matters
Rieseberg isn't new to building impactful desktop apps. He's a core maintainer of Electron, the open-source framework powering Slack's desktop app, and famously ported Windows 95 into an Electron application that runs across macOS, Windows, and Linux.
His philosophy at Anthropic reflects a broader shift in AI product development: rather than spending months on specifications and debates, teams can now "just build all the candidates" and choose based on reality rather than theory. This prototype-first culture allowed Cowork to go from concept to shipping in under two weeks using pre-existing internal components.
Local-First: The Undervalued Frontier
Anthropic is betting big on local-first agent workflows—a approach that Rieseberg believes Silicon Valley is significantly undervaluing. Unlike cloud-centric solutions, local-first tools process data on the user's device, offering greater privacy, reduced latency, and what Anthropic calls "trusted task execution."
The company positions Claude Cowork not as a simplified version of Claude Code, but as a superset—much like how VS Code initially appeared simpler than Visual Studio but became more hackable and extensible over time.
What's Ahead
The hardest questions ahead aren't about better chat interfaces, Rieseberg argues. They're about autonomy, safety, portability, and the changing shape of knowledge work itself. As AI agents gain the ability to execute trusted tasks on local machines, the boundary between what humans do and what AI does will continue to blur.
Claude Cowork represents Anthropic's answer: a desktop-native agent that brings the power of Claude Code to a broader audience, all while keeping computation local and user control central.