When an AI suggests slowing down its use

Anthropic launches Reflect, a Claude beta dashboard developed with MIT to track user activity, set quiet hours, and encourage digital well-being.

Anthropic is adding a beta feature to Claude that allows users to review their own usage of the assistant. Named Reflect and located in the settings on the web and desktop app, it stems from a question frequently raised in the company's interviews with its users: how often to use AI, and for what tasks.

The dashboard summarizes activity over the last one, three, six, or twelve months: topics discussed, times when the assistant is most frequently used, types of tasks handled, with a view of time spent announced for later. It also brings up reflective questions, such as "what would you like to continue doing yourself, even if Claude could do it faster?", and allows for discussion. Two settings complete the suite: quiet hours and a break reminder after a certain period of use, both presented as simple reminders that can be dismissed.

The feature is based on an in-house framework, the AI Fluency Framework, structured around delegation, objective description, discernment regarding responses, and responsibility in usage, with examples for each axis drawn from how the person works with the assistant.

Regarding privacy, the company specifies that the summary ignores conversations in incognito mode and does not draw from files of connected tools: a summary of an email inbox may appear, but not the original emails, and any conversation related to a health tool is excluded, with sensitive topics appearing only at a general level. Anthropic states it has worked with digital well-being specialists from the MIT Media Lab, the Digital Wellness Lab at Boston Children's Hospital, and the Family Online Safety Institute. The feature is available for free, Pro, and Max plans, provided memory is activated, with expansion to Cowork conversations planned.