Claude Duet 发布：后台主动干活的助理

Tolik @tolik12308原推发布于 2026.10.27

在 X 上打开原推

YOUR AI ASSISTANT IS ABOUT TO START MESSAGING YOU FIRST. Anthropic just introduced Claude Duet, an AI agent designed to work in the background instead of waiting for you to give it instructions. And that changes the relationship between people and AI. Duet runs on Claude Opus 5.5 inside its own cloud computer. You can close the app, walk away, and it keeps working. Every morning, it sends you a briefing based on your connected apps. Throughout the day, it reaches out only when something needs your attention. But here is the part I find most interesting. The real challenge is not getting AI to do more tasks. It is teaching AI when NOT to bother you. A useful assistant needs to distinguish between something that matters right now and something that can wait. And Anthropic is putting that judgment at the center of Duet. There are also safeguards. Before sending a message, making a payment, signing a document or deleting something, Duet shows you what it plans to do and waits for your approval. You stay in control while the agent handles the background work. But there is one catch. Those background checks count toward your usage limits, meaning Duet can consume your plan allowance faster than using chat alone. And Anthropic is entering an increasingly crowded race. xAI has Grok Bot. Meta has Muse. OpenAI has dots. The direction is becoming clear. AI companies are no longer competing only to build smarter chatbots. They are competing to build agents that know what is happening, take initiative and act on your behalf. The next big question is not whether AI can answer you. It is whether AI can figure out what you need before you even ask.

这条是引用着另一条写的：x.com/tolik12308/status/2114799703413989663