Voice-first morning planner
A morning planner with voice input and AI that turns a stream of speech into a structured plan. Tap the mic, say the day as it is — and an AI secretary shapes it into clean, prioritised task cards.
The to-do space is crowded, and almost every app dies on retention: capturing tasks is friction, and sitting down each morning to “lay out” the day is a mental block. The goal was to reduce that friction to zero — no typing, no manual structuring, just speak.
Daybreak isn’t a task list. In the morning you pick up your phone, tap the mic, and say everything as it is — priorities and asides included. The AI acts as a personal secretary: it parses your speech into clean tasks, extracts durations (“about fifteen minutes”), priority (“the most important thing”), even explicit positions (“put this third”), and assembles a finished plan.
Before any cards are created, a “Did I get this right?” step lets you confirm — and the day is ready. One tap: Start.
On-device speech recognition in 12 languages, with optional AI parsing via Claude Haiku 4.5 (enabled with your own Anthropic key). When on, the secretary orders the day by importance and honours explicit positions; when off, each spoken phrase simply becomes a card — no input is ever lost.
Morning (“What’s today?”) → Day (checklist + progress ring) → Evening (wrap-up). Bottom navigation: Day / Tasks / Night.
Priority, time estimate, drag-to-reorder, a text note and a Telegram-style voice note — with a subtle “Voice / Typed” source label.
The main motivator rewards the act of planning, not just hitting 100% — retention built around the ritual itself.
Before creating cards and before marking a task done, the system asks rather than acting on your behalf. Calm, never presumptuous.
A calm morning ritual instead of yet another to-do app: a warm cream-sunrise background with almost-invisible grain, editorial serif typography (Playfair Display / Cormorant Garamond), “glass” cards, and a unified, emoji-free SVG icon set.
At the centre is a custom mark — a “D” monogram with a rising sun, rays and a horizon line — extended into a full vector logo suite: mark, widely-tracked wordmark, tagline and app icons.
React 18 + TypeScript + Vite — fully client-side, no backend. Local storage
(localStorage) for state and IndexedDB for audio; the Web Speech API and
MediaRecorder for capture; drag-and-drop via @dnd-kit; an
optional Anthropic SDK (Claude Haiku 4.5) parsing layer, gated behind your
own API key. The habit logic (streaks, carry-over, progress) is pure and
unit-tested. Shipped as a PWA — installable, offline-capable, with a
self-updating service worker and CI/CD deployment.
MVP in personal testing. Next up: