On July 9, Anthropic introduced a feature that sounds odd at first: Claude wants to help you talk to Claude a little less mindlessly. It’s called ‘Reflect’, it’s still in beta, and you’ll find it in Settings on Claude for web and the desktop app.
What Reflect shows
At its core, this is a usage dashboard. It summarizes how you’ve used Claude over the past 1, 3, 6, or 12 months: which topics, which patterns, which kinds of tasks. You can see when you use Claude most and what you worked on during that time. A view of total time spent is coming soon.
The press quickly reached for the obvious comparison: ‘Claude Wrapped’, the Spotify-Wrapped idea applied to your AI usage. That nails the look — but not quite the intent.
The interesting part: the uncomfortable question
Reflect doesn’t just show you stats. Every now and then it surfaces a question — for example: ‘What’s one thing you want to keep doing yourself, even if Claude could do it faster?’ And you can talk it through with Claude right there.
On top of that come two wellbeing features. You can set quiet hours when nothing nudges you, and you can schedule a reminder to take a break after a certain amount of use. Both are reminders of your own intentions — and both can be dismissed.
Anthropic hangs all of this on its ‘4D AI Fluency Framework’: delegation, description, discernment, diligence. The report maps how you collaborate with Claude onto those four dimensions and offers practical tips — like starting a Project instead of re-explaining context every time.
Privacy
One point that matters to me: Reflect doesn’t pull from incognito chats and doesn’t reach into files from your connected tools. If you ask Claude to summarize your inbox, the summary might show up in the report — but the source emails won’t. Conversations tied to health integrations are left out entirely. The feature needs Memory switched on, or it won’t work.
My take
I find the direction remarkable. An AI company handing you a tool to look critically at your own usage — and even asking what you deliberately don’t want to delegate. That’s the opposite of ‘use us as much as possible.’
Of course you can be cynical: a dashboard showing how deep you’re already in might just lock you in harder. TechCrunch called it ‘a feature that quietly sells you on AI.’ Both things are true at once. I’m trying it anyway — if only for that one question about what I want to keep doing myself. That part is worth it.
Sources: