The fastest way to get something out of your head and into Working Memory is to talk to it. Tap the mic, say the thing, release. There's no format to learn — no command words, no headers, no "Hey cobrain." Just say what you'd say to a friend who happens to remember everything.
What happens after you tap release
For short notes (under about a minute), the transcript and any structure show up almost instantly. The audio uploads, the AI reads it, and the result lands on your timeline.
For longer notes — a five-minute brain dump after a meeting, a long story you want to remember — there's a few extra seconds while the audio uploads and the transcript renders. The result is the same either way: the spoken words become text, and anything worth structuring is structured.
What the AI pulls out
It reads the transcript and decides on its own what's in there:
- Dates and times become reminders. "Call the dentist tomorrow at 2" schedules a notification.
- Names become people. Anything you said about a person attaches to them, so you can ask later.
- Comma-separated or numbered things become lists. "Groceries: bread, milk, eggs" is captured as a list.
- Everything else — opinions, observations, half-finished thoughts — becomes searchable history. You can ask for it back in your own words later.
You don't need to think about which bucket your note falls into. Talk first, the AI sorts it.
What happens to the audio
This is the main difference between the two plans:
- Defrag ($9/month) keeps the audio for 30 days. After that the transcript stays forever, but the original recording is deleted.
- Archive ($15/month) keeps the audio forever. You can re-listen to anything you've ever said.
Most people are fine with Defrag — once a thought is text, the audio is just the delivery method. Archive is for people who care about the voice itself: a story your mom told, the way a friend explained something, your own tone of voice when you first had an idea. See Plans explained and Choosing a plan for more.
When transcription fails
It almost never does, but: if the audio is totally silent, or the background noise is loud enough to drown out the speech, the AI may come back empty. When that happens, the original audio is still saved if you're on Archive (so you can re-listen and re-record from your end), and you can try again. Don't worry about this in everyday use — speech recognition is solid on modern phones and on the wired-in mic you're already used to.
If voice isn't an option
Some moments aren't for talking — you're in a meeting, on the train, around people. Type instead. The text field on the composer takes the same kind of input and the AI parses it the same way. See Text messages.