Voice to-do list

A to-do list you talk to.

Press the Action Button, say what you have to do, tap Save. A model on your iPhone reads the note and files each task under Morning, Afternoon, Tonight or Anytime. If you named an hour, the task gets it.

It needs an iPhone 15 Pro or newer, because the model that reads the note runs on the phone and nowhere else. On an iPhone 14 or 15 it records, transcribes and files one task per note. Model by model.

In
One spoken note
Out
Separate tasks, timed as you said
Parts of a day
Morning, Afternoon, Tonight, Anytime
Languages
English, Simplified Mandarin, Spanish or Brazilian Portuguese, one at a time
Account
None

What one note becomes

§1

How often it gets it right

§2

We measure this rather than promise it. On our everyday English set, 261 written notes of the kind people actually say, 224 of 261 come out entirely right: the number of tasks, the day and the time. A note that asks for one task comes out as one task 149 times in 151. The named day lands 114 times in 123; a stated clock time 30 times in 33. And no task was ever given a clock time its note never said: 0 of 980 across the whole launch study.

Numbers, dates, times, currency and spelled-out codes it gets right: 60 of 60 in the set built to test them. Every figure here was decoded on a Mac, not on an iPhone, at temperature zero, in late August 2026; the corpus and the conditions for each are on the blog.

Where the tasks live

§3

A day has four parts: Morning, Afternoon, Tonight, Anytime. Enough structure to plan with, not a calendar to maintain. Pick a card up and hold it at the edge of the screen; the day changes underneath it, and the card can be dropped on a different one.

Not everything you say is a task. A thought with nothing to do stays a note, kept by day. Search reads the whole note, not just the title, so a task can be found by words you spoke but never typed. Lists shows everything in one place: open tasks, completed tasks, notes, and a month of days.

It listens in English, Simplified Mandarin, Spanish or Brazilian Portuguese, one at a time; you pick which in Settings. It does not work out which language you are speaking, and it does not hear two at once.