The race accelerated in recent weeks as Meta launched Muse and OpenAI introduced Dots, framing always-on agents as assistants for consumers as well as businesses. Meta said more than 10 million people in the US and Canada had set up Muse agents in roughly a month and a half, a sign that the category is moving beyond demos.¹

On Wednesday, Anthropic entered that contest with a research preview of Claude Duet, a standalone and web-based assistant built on Claude Opus 5.5. Rather than chase every personal errand, the company is targeting professionals and small-business owners: Duet can monitor connected apps, deliver a morning brief, schedule meetings, build dashboards and produce forecasts.²

The distinction is strategic. Meta has marketed Muse principally to consumers, while OpenAI’s Dots spans consumer and enterprise use. Anthropic says its own product is designed for the workplace and is therefore “not quite as strong” at personal chores such as booking tickets.³ That is less a retreat than a bid to make usefulness—not novelty—the battleground.

Anthropic’s pitch is proactivity with a brake pedal. The company says Duet “comes to you with what needs your attention,” keeps working after a user steps away and checks before acting on their behalf.Addy Osmani (addyosmani)⁴ For consequential actions, including sending payments or deleting material, it seeks approval; Anthropic has also added a classifier and safety prompt intended to block requests for highly sensitive data.⁵

Still, the preview is deliberately hedged. Anthropic warns that Duet may misread a thread or misjudge what is urgent—exactly the kind of error that becomes more consequential when an agent is watching, deciding and acting in the background. Access is rolling out from a waitlist to eligible US Pro and Max subscribers, leaving the company time to test whether work-focused autonomy can justify the trust it demands.