ANTHROPIC LAUNCHES CLAUDE REFLECT

Anthropic just gave Claude something no other major AI assistant has yet: a built-in dashboard that shows you what you actually do with it. The feature is called Claude Reflect, and it is available in beta for free, Pro, and Max users who have turned on memory. It lives in Settings on the web and desktop apps, and it generates a summary of your Claude activity for one, three, six, or twelve months.

The idea is not to feed your ego with a highlight reel. Reflect breaks down the topics you engage with most, identifies patterns in your usage, and categorises your interactions using Anthropic's 4D AI Fluency Framework: delegation, description, discernment, and diligence. It surfaces questions like "What’s one thing you want to keep doing yourself, even if Claude could do it faster?" — a prompt that may land differently depending on how attached you already are to outsourcing your thinking.

you may like

ANTHROPIC WANTS YOUR DRIVER'S LICENSE FOR SOME CLAUDE FEATURES

FABLE 5 WONT TELL YOU WHAT A MITOCHONDRION IS

GEMINI FOR HOME GETS WAY MORE CONTROVERSIAL

WHAT "REFLECT" ACTUALLY SHOWS

Open the settings, select the option to reflect on your usage, and you get a visual dashboard with overall usage metrics, peak activity times, conversation line charts, and task breakdown charts. The timeframes are the same four options: one month, three months, six months, twelve months. A dedicated time-spent view is coming soon, which suggests the current build shows prompt counts and topic clusters but not yet a simple clock tally.

The dashboard also provides suggestions on how to improve your workflows with Claude. If you keep re-explaining the same context, it might nudge you to start a Project instead. If you are delegating without verifying the output, it could flag a need for more diligence. The framework's four pillars give a vocabulary for thinking about your relationship with the tool: are you using it as a description engine, a discernment partner, a delegation clerk, a diligence checker, or some mix?

For now, Cowork conversations are excluded from Reflect, though that is coming soon. Incognito chats and conversations connected to a health integration tool are kept out entirely. If you asked Claude to summarise your email inbox, the summary itself might appear in Reflect,ect but the specific emails would not. Sensitive conversations can surface at a high level, but the company says the data remains in the dashboard and is not used for any other purpose.

BUILT WITH WELLNESS RESEARCHERS

Anthropic did not build this alone. The feature was developed with MIT Media Lab (specifically its Advancing Humans with AI program), the Digital Wellness Lab at Boston Children's Hospital, and the Family Online Safety Institute. That is an unusual set of collaborators for a product feature from an AI company. It signals that Reflect was designed with care for how people form habits around these tools, rather than as a vacuum to pull more engagement data.

The tool also lets you set quiet hours and schedule nudges to take a break after a certain amount of time using Claude. That is a genuinely odd thing for a commercial product to offer: a built-in off switch for its own use. Most platforms want you staring at the screen longer. Anthropic is betting that making people feel in control of their usage is a long-term trust play that beats short-term session metrics.

THE PRIVACY PICTURE

Reflect only sees what memory already sees. If you have not enabled memory, the feature is not available. Conversations in incognito mode never appear, and health-integration chats are excluded entirely. The data stays local to the dashboard, not fed back into training or advertising or anything else. That is a meaningful boundary. The reflection dashboard also draws from a clear rule set: what you see is generalisation, not a transcript.

The company describes the tool as a way to build skills around the 4D AI Fluency Framework. That framing is a little paternalistic, but it is also honest about the reality that most people using AI do not have a mental model for whether they are using it well or poorly. Reflect is an attempt to give them one.

WHY THIS SHOULD MATTER TO YOU

A feature that asks you to consider what you want to keep doing yourself, even if the machine could do it faster, is a genuine philosophical statement.

The risk is that Reflect ends up being a novelty. Users might glance at it once and never return. The time-spent view is the kind of metric that could become obsessive or guilt-inducing, or it could become genuinely useful for people who want to calibrate their tool use. The team's collaboration with academic wellness researchers suggests they are aware of those dynamics and built for the second case.

what to read next

UNO REVERSE CARD: ALIBABA BANS CLAUDE CODE

ANTHROPIC IPO COULD ECLIPSE EVERY SOFTWARE COMPANY IN HISTORY

HOW THE TURN TABLES... XBOX MIGHT BE THE ANSWER TO THE DISCPOCALYPSE

What comes next is whether Anthropic will extend the same reflective tooling to the enterprise tier, where usage surveillance by employers creates a different set of trust problems. For now, consumer users get a dashboard that breaks down what they ask Claude to do and categorises the work, and quietly suggests they take a break. Whether users want to think that hard about their prompting habits is another question.