Mindful Scroll Interruption

Screen Time breaks the scroll gesture mid-flick and asks one quiet question. The starting point: blockers reliably provoke reactance, and dashboards rarely change behavior. I designed a system that interrupts gently, lets people answer honestly, and uses on-device AI to reframe what they came in for.

Three intensities — Gentle, Standard, and Firm — each chosen per app. Gentle nudges; Standard pauses; Firm requires an explicit override. Letting users self-select the friction is the difference between a tool they keep and a tool they uninstall after one bad afternoon.

After a set threshold the app surfaces four quick options for the easy cases, plus a longer path for the messy ones — bored, stressed, avoiding — with a five-tag emotional picker. The point is to make honesty cheaper than dismissal.

Apple's Foundation Models run on the Neural Engine to reframe a user's stated goal from avoidance to approach language. Text never leaves the device, never persists, never transmits. Local reframing is the only version of this feature I'd ship.

For users who want a harder stop, a system-level ManagedSettings shield breaks the gesture itself. It's offered, not recommended, and locked behind a 30-minute cooldown so the choice can't be taken back in a single weak moment.

SwiftUI for UI. DeviceActivity and ManagedSettings for system-level interruption and shielding. FamilyControls for the authorization model. Foundation Models for the on-device reframing layer. Core Data for local persistence. Every piece chosen so nothing sensitive leaves the phone.