Humans before feeds
Real people and time-sensitive work can pass through. Engagement bait, vanity metrics, and sales pressure get delayed.
Local-first AI protection layer
A calmer notification layer for phones: apps ask to interrupt, the shield decides what deserves your attention, and everything else waits for a digest.
Current focus modes mostly ask people to maintain static allow and block lists. Attention Shield starts from a different assumption: interruption is expensive, and the device should defend the human by default.
Real people and time-sensitive work can pass through. Engagement bait, vanity metrics, and sales pressure get delayed.
Deep work, driving, sleep, recovery, and family time need different thresholds without rebuilding every app setting.
The preference graph should live with the user. Cloud AI can be optional later, not the default gatekeeper.
Deep Work
Allow direct calls, important contacts, calendar changes, and critical business messages. Everything else becomes an hourly digest.
Human Mode
Favor actual people and suppress algorithmic noise, social vanity metrics, flash sales, and habit-loop notifications.
Nature / Yoga Mode
Emergency-only interruptions, grayscale-friendly UI, gentle sound defaults, and a low-stimulation notification surface.
Do not start with a full OS. Start with the narrow layer Android already permits, then let TommyOS inherit the lessons.
The product is not another assistant. It is an AI protection layer for sensory-friendly, open, user-owned computing.