Daily Planner App

The product opinion is that only today matters, so the app needed a task surface scoped strictly to the current date, with each day starting clean, while still giving users something valuable from the days behind them.

That ambition came with constraints. An account system would be pure friction for a single-user utility, so identity had to work with zero onboarding. And because an AI reflection is a daily-cadence feature, naive per-request model calls would make the economics untenable; the AI usage had to be engineered down to roughly one call per day.

We kept the navigation footprint deliberately tiny: rather than a navigation library, the app runs a four-state screen machine, splash, onboarding, home, AI summary, driven by plain React state. For an app whose whole premise is one focused surface, this keeps both the UX and the dependency count small.

We split the data by sensitivity. Tasks, the personal, minute-to-minute data, never leave the device; they live in AsyncStorage under per-day keys. Only the AI-generated summaries, and the titles of completed tasks that feed them, are written to Supabase, so users get a cloud-backed history of reflections while day-to-day task data stays local.

Identity is frictionless: the app synthesizes a stable per-device user ID from device metadata and a timestamp, caches it locally, and uses it to scope all cloud reads and writes. And AI usage is cache-first, every summary request checks Supabase, then a local cache, and only calls Gemini when no summary exists for the day or the user explicitly regenerates.

The home screen shows the current date and today's tasks, nothing else. Tasks are added from a single input, checked off in place, and deleted behind a confirmation dialog. Long-pressing a task lifts it into a drag state with elevation and shadow feedback so the day can be re-prioritized by hand; the resulting order is stored per task and survives restarts. Because storage is keyed by date, each day starts clean.

The AI Summary screen offers three layers of perspective: yesterday's reflection, generated by sending the previous day's completed tasks to Gemini with a prompt tuned for concise, encouraging retrospectives; an on-demand progress summary of today's completed versus pending work; and a scrollable history of past reflections pulled from Supabase, each card showing the date, summary text, and the completed tasks behind it, with pull-to-refresh. Generated summaries are written to both Supabase and local storage, so past reflections remain readable offline.

A first-launch guide explains the three product ideas, single screen, quick tasks, AI summaries, and is shown exactly once, tracked by a persisted flag. The splash screen uses spring-based fade and scale animation, and the interface is built on a consistent gradient-and-glass visual language with keyboard-avoiding layouts for one-handed entry.