openaienmodel: gpt-5-mini-2025-08-07
Introducing a way to reflect on how you use Claude
Key Points
- Beta reflection dashboard in Settings
- Built on 4D AI Fluency Framework
- Excludes incognito & health data
Summary
A new beta "Reflect" dashboard (published 2026-07-09) helps users track and visualize how they use Claude so they can refine habits and workflows. Engineers can access the feature from Settings on Claude for web and the desktop app: it summarizes topics, usage patterns, task types, and provides reflective prompts and controls (quiet hours, break nudges). The feature uses the 4D AI Fluency Framework (Delegation, Description, Discernment, Diligence) to surface collaboration patterns and practical suggestions.
Key Points
- Access: Settings → Reflect on usage in Claude (web and desktop apps).
- Time ranges: view activity over 1, 3, 6, or 12 months; a time-spent view is coming soon.
- What it shows: topics, usage patterns, task breakdowns, per-dimension summaries using the 4D framework, and examples of how users collaborate with Claude.
- Controls & prompts: periodic reflective questions, the ability to discuss prompts with Claude, set quiet hours, and schedule nudges to take breaks (dismissable).
- Privacy & data: excludes incognito chats and health-integration conversations; does not pull underlying files from connected tools (summaries may appear without source data); insights are not used for other purposes.
- Availability: beta for Free, Pro, and Max users with Memory turned on; Cowork conversation reflection coming soon.
Practical notes for engineers
- Verify Settings entrypoint and desktop/web parity; gate by Memory=true and user tier.
- Ensure exclusions: filter incognito and health-integration data from reports and avoid pulling connected-tool source files into insights.
- UX: surface dismissable nudges and quiet-hours settings; include clear privacy messaging and link to privacy docs.
- Roadmap items: implement the upcoming time-spent view and Cowork reflection, and add instrumentation to measure feature usage for iterative improvements.