How the current product works
This page explains the live browser-only v1: local grouping, autonomous tab management, feedback-driven learning, and the optional OpenAI summary layer.
The full experience still requires installing the Chrome extension from GitHub.
1. Popup grouping
The popup groups open tabs locally with Gemini Nano, caches the result, and gives the user manual control over sleep, wake, and close actions.
2. Stats page
The Stats page shows estimated memory saved, group state, grouping quality ratings, study submission, and the autonomous activity feed.
3. Autonomous activity
The background agent can auto-sleep low-need tabs, wake related contexts, and log every action with explanations plus Undo and Protect controls.
4. Optional OpenAI layer
When enabled on the backend, OpenAI adds advisory policy summaries and recommendations. The browser agent itself still makes local real-time decisions.
Architecture at a glance
The extension is the real product surface: popup, background service worker, local autonomous policy, Stats page, and feedback loop. The website/backend supports deployment, study collection, admin review, and optional OpenAI-powered summaries.
If you want to inspect live submissions, reward/regret trends, and training-style policy signals, go to /admin.