Claude Duet Enters the Agent Race With Meta’s Muse Already Ahead
Claude Duet Enters the Agent Race With Meta’s Muse Already Ahead
Meta set the early pace in the personal-agent push. Muse launched on Sept. 8, followed by OpenAI’s Dots on Sept. 29, giving both companies a head start in the fight to become the assistant users turn to first. Meta said more than 10 million people in the U.S. and Canada had set up a Muse agent in roughly six weeks, with about a quarter using it on a typical day.
Anthropic had been moving toward that territory in stages. In June, its Claude Tag tool began tracking Slack discussions and unresolved tasks; in September, Claude was updated to keep working on assignments after a user closed a laptop. Those steps pointed to a broader goal: shifting Claude from a chatbot users consult into software that notices work and follows through.
On Wednesday, Anthropic made its answer explicit with a research preview of Claude Duet. The standalone app and web feature is aimed at professionals and small-business owners, not the consumer errands emphasized by some rivals. It can monitor connected apps, issue a morning brief, flag items needing attention and handle work ranging from meeting scheduling to dashboards and financial forecasts.
Anthropic’s pitch is autonomy with a brake pedal. Duet “keeps working even when you step away” but “checks with you before it acts on your behalf,” the company said. Before consequential moves such as payments or deletions, it asks for approval; the company says it has also added safeguards intended to block requests for highly sensitive data.
That caution underlines the gap between the marketing and the present product. Duet is waitlist-only for U.S. Pro and Max subscribers, shares usage limits with ordinary Claude chats, and will check accounts every few hours rather than continuously. Anthropic also concedes the early version can “misread a thread or miscalibrate what’s urgent,” and says it is not yet as capable at personal errands such as booking tickets.
The contest is therefore less about a single launch than competing visions of the agent future: Meta chasing broad consumer scale, OpenAI spanning personal and enterprise use, and Anthropic wagering that useful work is the stickiest daily task of all.
Write a comment