Claude Duet Can Still Misjudge What’s Urgent

Claude Duet is an AI assistant you can hand work to by text or voice. It uses its own computer in the cloud and apps you connect, keeps working while you are away, and messages you when the task is finished or it needs your approval. Its practical jobs include morning briefings, spotting items that need attention in your inbox and calendar, drafting replies, and producing documents, spreadsheets, decks or dashboards. Those capabilities are described on Claude’s official product page.

The difference from a normal Claude chat is who starts the interaction. In the Claude app, you open a conversation and work alongside Claude. Duet can come to you with a morning brief, continue a task between messages, and return when something needs your attention. It uses your Claude account’s memory, skills and connectors, with your conversation kept in one thread.

That makes its judgement about urgency consequential: it helps decide when to interrupt you and what to flag. The preview can still misread a thread, misjudge urgency or miss something that arrives between check-ins. You can inspect its activity and steer it; deciding what matters is not a responsibility to hand over blindly.

The waitlist is for Claude Pro and Max subscribers in the US; Team and Enterprise plans are not included yet. Duet is a standalone app for iOS and Android, with web access at claude.ai on a computer.

During the preview, Duet shares your plan’s usage limits with the rest of Claude. Background checks count towards those limits, and larger jobs such as a deck or financial model need more model turns and background work. Eligibility to join the waitlist is not immediate access.

The wording matters because it keeps the present limitation alongside the work still to come. Duet can misjudge urgency; early feedback will inform continued tuning. A plan to tune it does not erase the limitation.

It can still misjudge what’s urgent, and we’ll keep tuning it based on what early users tell us.

Claude · https://x.com/claudeai/status/2112984669943968099

For a concrete use, Duet can flag an inbox item and draft a reply for you to approve, set up a meeting, or build a dashboard and report back when it is done. The product page also lists finding unpaid invoices and drafting reminders. These are tasks with an output to review, not just answers in a chat. Duet asks before sensitive operations such as sending messages or making payments; work you delegate and actions you authorise remain distinct.

Taking part in this preview also means contributing feedback to a product still being tuned. What early users report will inform that work.

The commitment is to continue tuning. It is not a claim that urgency judgment has already improved, so the limitation remains part of how we should approach the preview.

If we are in the US on a Pro or Max plan, the announcement points us to the [Claude Duet waitlist](https://claude.com/product/claude-duet). That is the route to follow if we want to take part.

We should approach it as a request for access to an early research preview, keeping the urgency limitation in mind rather than assuming admission or dependable judgment.

Claude Duet moves Claude from a chat you open towards an assistant that can keep work moving between conversations. The useful fit is a task you can delegate through a message, with connected apps, a clear result to inspect and decisions reserved for you.

For example, drafting an inbox reply or preparing a dashboard gives you something concrete to check. Start with work you can review, keep sensitive actions behind approval, and account for shared usage limits. The US Pro and Max waitlist and fallible urgency judgement still define this research preview.

The waitlist opens a route to the preview. The urgency limitation remains.

Altior · Based on Claude’s announcement: https://x.com/claudeai/status/2112984669943968099