Some moments you can't talk — you're in a meeting, on a quiet train, sitting next to someone. Type the thought instead. The composer's text field takes the same input as the mic, and the AI reads it the same way.
How it differs from voice
It doesn't, really. The only difference is the upload step. A typed message skips it, so the result lands immediately — there's no "transcribing" beat. Everything else (people, dates, lists, reminders, searchable history) is parsed exactly the same way.
Examples that work
These all do what you'd expect:
- "remind me to call mom friday at 10" — schedules a reminder.
- "the wifi password at the airbnb is windows92" — captured as a fact you can ask back later.
- "groceries: bread, milk, eggs" — saved as a list.
- "nice quote from raj — 'product is what you ship, not what you plan'" — attached to the person and searchable.
No formatting required. Lowercase works. Typos work. The AI is forgiving.
When to type, when to talk
Type when the moment is quiet, or when the thing is short enough that thumbing it out is faster. Talk when you're walking, driving, mid-thought, or when there's a lot to get out at once. See Voice messages for what to expect from longer voice notes.