What happened
Anthropic is expanding Claude Cowork, its agent for knowledge work, to web and mobile. Previously a desktop-only application, the tool is now available to Max subscribers across devices, as reported by TechCrunch AI. The update allows users to start tasks on one device and monitor them on another, with the agent continuing to work in the background.
Anthropic also released usage data from May, showing that business process tasks (33.4%) and content creation (16.4%) far outstripped software development (8.7%) as the primary use cases.
How the room's reading it
The move is widely seen as Anthropic's push to position Cowork as a general-purpose agent, not just a niche coding tool. Analysts are framing this as part of a larger war among labs to own the surfaces where work actually happens — moving beyond simple chatbots. The consensus, echoed in reporting from TechCrunch AI, is that the battle is shifting from model performance to workflow integration. Anthropic's own data release reinforces this, suggesting the real market is in automating administrative and operational tasks — the 'work around the work'. This positions Cowork directly against similar efforts from OpenAI.
Sailfish's take
The usage data is the real story here. We've suspected for a while that the biggest market for agents isn't replacing developers, but automating the tedious 'work around the work'. Anthropic's numbers — just 8.7% for software development — confirm it. The valuable problems aren't complex coding challenges; they're reconciling spreadsheets and drafting reports. For builders, this is a clear signal. Chasing a perfect AI developer is less important than shipping a reliable AI admin assistant. We believe the agent wars will be won not by the smartest model, but by the most dependable background worker. This makes us much more interested in agentic workflow tools that solve for drudgery.