Some people dictate a sentence. Some people think out loud for three hours. Psst was built for the second kind.
No timer
Many dictation apps stop after thirty seconds, or a minute, because the cloud call behind them has a limit. Psst records until you tap the bot again. An hour is normal. The transcript grows in the panel as you go.
Why the fans stay quiet
On Apple silicon the speech model runs on the Neural Engine, a part of the chip built for exactly this kind of work at very low power. A large-v3-turbo model transcribing continuously on an M-series Mac uses less energy than a video call. The chassis stays cool; the battery lasts.
What happens while you talk
Audio is captured in short chunks and handed to the model as you speak, so the transcript is never far behind you. When you stop, the last chunk is finished and the file is written. If the Mac sleeps mid-recording, the recording pauses and resumes; nothing is lost.
Where it goes
Every recording becomes one text file, named by date and time, in the folder you chose. A day of talking is a handful of files you can open in anything. Search them with Spotlight or grep. Move the folder to iCloud Drive if you want them on your phone too. Psst does not keep a copy anywhere else.
The habit
People who use Psst this way tend to do the same thing: tap the bot at the start of a walk, a drive, a cooking session, and talk through whatever they are working on. Tap again at the end. Read it back later, copy the good parts. It is the cheapest thinking tool there is, and the words stay in the room.