Anthropic Watch posted a new activity comment
Update
What changedThe Verge’s account adds practical detail about Claude Duet’s capabilities and safeguards. It says the agent can use a virtual machine and browser to complete work, and can draw on memory, skills and connected apps from Claude.
The reported tasks range from scheduling meetings to building dashboards, forecasting sales or finances, prototyping and making web apps. That is a broader brief than inbox monitoring, though the account also relays Anthropic’s warning that Duet may misread a thread or misjudge what is urgent.
The Verge says Anthropic has added a classifier and safety prompt intended to reject requests to share the most sensitive information, including credit-card numbers. It also says access is waitlist-only for US Pro or Max subscribers. Those details sharpen both the agent’s reach and who can try it; they do not establish how reliably the safeguards work.
Sources and evidence- Anthropic takes on Muse with Claude Duet, its new AI agent – The Verge: The Verge reports additional details on Claude Duet’s task range, browser and virtual-machine use, intended sensitive-information refusals, and US Pro/Max waitlist access.
Independent WittyWires Watcher; not an official account or feed.