Quoting Felix Rieseberg

In the previous version of Claude Cowork, the model’s inference ran in the cloud, but tool calls were handled inside a virtual machine (VM) provided by Anthropic and downloaded to the user’s device. That setup allowed only the data the user explicitly added to the session to be mapped, improving security and isolation. However, users complained that the VM consumed local resources, drained battery life, and lowered computer performance. Closing the laptop also interrupted work, limiting continuity.

The new Cowork iteration moves both model inference and the VM to the cloud, assigning each session its own sandbox that does not share state with other sessions. When the VM needs to access a file or another resource on the user’s device, the desktop app handles the tool call. This architecture removes local processing load, enables tool use from mobile devices, and keeps background activity without draining the battery.

This update matters because it enhances Claude Cowork’s usability and efficiency, making it easier to adopt in mobile work environments and reducing hardware costs for users. By delivering a smoother, more sustainable experience, the solution strengthens AI assistants’ competitiveness in the productivity market.

Read the original article on Simon Willison

This summary is an informational synthesis produced by dataqbs.com. All rights to the original content belong to its author and the cited media outlet. We act solely as curators of technology news and claim no authorship.