The engine underneath
The 4D framework
Anthropic's four dimensions of AI fluency. Reflect pulls examples from your own conversations for each one. Tap a card for what it reveals - and the fix.
Most people will open Reflect, screenshot it like a Spotify Wrapped, and move on. That is the mistake. Underneath the friendly recap, Reflect quietly turns your chat history into a coaching dashboard: an honest read of how you actually work with AI, and where your habits are costing you results.
Reflect is a new feature that lives in Settings on Claude for web and desktop. With Memory turned on, it looks back over your conversations from the last 1, 3, 6, or 12 months and writes you a summary: your key topics, your usage patterns, the kinds of tasks you work through, your most active day and time, and the share each theme takes.
It is in beta for Free, Pro, and Max users. The part that matters is not the minutes - it reads the content of your work, and scores how you collaborate with AI across four dimensions.
Anthropic's four dimensions of AI fluency. Reflect pulls examples from your own conversations for each one. Tap a card for what it reveals - and the fix.
Reflect only works with Memory enabled. Open Settings → Memory and check that it is on. Incognito chats are always left out.
Go to Settings → Reflect on the web or desktop app and generate your report.
Choose 1, 3, 6, or 12 months. If you only recently turned Memory on, the first report will be thin - that is expected.
On privacy: Reflect needs Memory on, incognito conversations are excluded, sensitive integrations like health are left out of the read, and Reflect's data stays inside the feature.
Pick the habit that sounds like you. I'll name the dimension it points to, and the one thing to change.