Anthropic just introduced Claude Duet, its answer to Meta’s Muse and OpenAI’s Dots. It’s a proactive agent that “reaches out before you ask.”
What it does
- It’s built on Opus 5.5, connects to your apps, sends a morning brief, and checks in every few hours when something needs attention.
- It has its own cloud VM and browser, so it handles tasks from simple admin to building web apps without your computer.
- It asks for approval before sending a message, making a payment, signing a document or deleting a file.
- It uses your Claude memory, skills and connected apps, all in one ongoing thread.
- It doesn’t see passwords, and a dedicated classifier guards against leaking sensitive info.
- It’s a waitlist research preview for Pro and Max subscribers.
My take
Proactive agents are the big trend this fall. Approval before anything irreversible is exactly the right default.
Have a blessed day! And Happy Hacking!
