How drifty reads work

Classification is the core feature. Tracking captures what was open. Classification decides what the time was.

What gets sent

drifty does not send a screen recording or a raw timeline. Classification uses a reduced set of activity fields: app, site or window context, and enough surrounding signal to label the block. If Page address is enabled, its protocol, host, and path can be included after credentials, query strings, and fragments are removed. Raw page or message bodies, session dumps, and provider keys stay out of the request.

The local tracker remains the source of truth. Classification writes labels back onto captured blocks; it does not replace the history.

What you should see

After classification has run, a session can show:

• a category such as workspace, learning, communication, or entertainment
• a productivity label: focus, neutral, or drift
• enough detail to tell similar apps apart

Zero manual tagging is the default. You can still correct a block. Those corrections teach later classification.

Projects and corrections

Personal project classification is part of the production app. See Projects for named bodies of work, and Correct classifications when a label is wrong.