Voice Day Planner: say your day, get a plan

The problem

Typing a to-do list is the work you were trying to avoid.

• Capture has too much friction. Open app, new task, type, set time, set reminder, repeat — for something you already said out loud.
• Dictation gives you a paragraph, not a plan. Plain speech-to-text hands back a block of text you still have to break up by hand.
• AI planners invent things. Ask a model to “plan my day” and it helpfully adds tasks you never mentioned, which means you have to check every line.

What it does

You press record and talk the way you actually talk — switching between English, Hindi and Hinglish mid-sentence is fine. What comes back is a clean, timed schedule with reminders, not a transcript.

The rule that defines the product is printed on its own home page: “Only what you say gets added.” The model is there to structure and time what it heard, not to suggest, pad or improve it. That constraint is why you can trust the output without re-reading it.

• Speak in English, Hindi or Hinglish — including mid-sentence.
• A timed schedule with reminders, not a wall of text.
• Only what you say gets added — no invented tasks.
• Focus, Plan tomorrow and Carry forward for the things that did not get done.
• No sign-up, no account, no ads — plans stay on your phone.

How it is built

A React app wrapped for Android with Capacitor, using on-device speech recognition for capture and an LLM purely as a structuring step. The published build is version 2.8, 2.8 MB, Android 7.0 or newer — small, because almost nothing ships that does not need to.

The engineering that matters is the boundary around the model. Speech comes in; the model’s only job is to split it into items and assign times; anything it did not hear is not allowed through. Plans are written to the phone rather than a backend, which is also why there is no account to create and nothing to sync.