Anthropic reportedly launches Claude Duet, an assistant that works in the background

What was reported

This is early, single-source reporting. As ThePaper.cn reported, Anthropic has announced Claude Duet, a new app that enters the personal AI agent market. An agent here means software that carries out tasks for you instead of only answering questions.

The report says Duet is built on Anthropic's flagship Opus model, released in late September. It is a research preview with a waitlist, so access is limited and the product is still being tested. No price is given.

From answering to reaching out

The key idea is that Duet is proactive. It reportedly keeps running in the background, sends a morning briefing on how your connected apps are doing, and messages you when something needs your attention.

It runs in the cloud, not on your own computer. It works through permissions you grant, coordination across apps and system notifications. Mike Krieger, co-lead of Anthropic Labs, said in the launch video, "It can tell when to act proactively and when to act on your behalf."

The competition and the people

The report frames Duet as a response to Meta's Muse and OpenAI's Dots. It says Anthropic announced Duet on the same day Meta reported earnings. Alexandr Wang, who leads Meta's Superintelligence Labs, posted a meme joking that Duet resembles Muse.

It also says Karan Ahuja, a former product lead at Meta's lab, joined Anthropic in August and led Duet's development. The outlet ties the launch to Anthropic's reported IPO plans, but that is the outlet's analysis, not something Anthropic has said.

The privacy trade-off

An assistant that acts on its own needs access to your calendar, email and workflow state. The report says Duet asks for your approval before certain tasks, which is one way to balance initiative against control.

What the report does not give us is detail on how that approval works in practice. Anthropic's own documentation is the thing to wait for.