The assistant becomes the first screen on all three platforms — but it opens by telling you your day, not by showing an empty chat box. Same layout everywhere: brief on top, Start Next untouched, one input with a mic. Typing or talking expands the full conversation in place. Every existing screen stays exactly where it is — the tabs and rail don't move.
Composed from your real tasks, schedule, and profile — rendered instantly from local data (zero tokens, works offline). The LLM only wakes when you type or talk.
From patterns it noticed (gym every Wednesday…). Max one per day, one-tap answers, dismiss never nags twice.
The hero stays exactly as it is. Open app → press one button → focusing. The brief sits above it, never in its way.
One input, mic one tap away. Sending expands the full thread in place — brain-dumps, questions, the interview, everything the lab does today.
Tabs, FAB, rail, widgets — all unchanged. Every old screen is one tap away, and the whole app still works if the AI is off or offline.
New Settings screen listing every learned fact — edit and delete anything. Every fact it saves shows a receipt with Undo, like the lab.
Static mockup — nothing on this page is wired. The live conversation engine is on the lab page; what you feel there is what ships inside frame ④.