Set up once. Then just talk.
Connect an app, say the wake word, and ask for what you want. Andromeda reads instantly and asks before she changes anything.
Three steps. Then it’s just talking.
Connect an app
Your mail, your calendar, Slack, Notion, Spotify. Pick one, once.
Say the wake word
She listens on your phone, not in the cloud.
Ask, or tell
Reads are instant. Writes are confirmed first.
Or get Andromeda’s attention byTapping the orbTyping it
What happens when you speak.
The wake word is heard on your phone
Nothing leaves the phone until you have said it. Andromeda is not listening to the room; she is waiting for one phrase, and the model that hears it runs on the device.
Your words go to the app you named
Say it the way you would say it to a person. What you said, and the minimum the app needs to answer, is all that travels. No voiceprint. No recording kept.
A read comes straight back
What’s on tomorrow, anything from Priya, what did I miss in the channel: read aloud, trimmed to what fits in a breath, and shown on screen so you can glance instead of listen.
A change waits for your yes
Moving a meeting, sending a line to the team, adding a reminder: she reads it back with the specifics first, and nothing happens until you say so. A correction is a new question, never a guess.
What she never does.
No voiceprints
Nothing about your voice is stored, matched or modelled. It costs nothing to promise and it is written into the code.
No speaker ID
Andromeda never learns who is talking, only what was said after the wake word.
No voice cloning
She has her own voice. Yours is never synthesised, sampled or kept.
Never acts unaddressed
A sentence with no wake word, heard while idle, could be something the room said. It does nothing.
Reads freely, writes with a yes
A lookup is never confirmed. A change always is, with the specifics named: both times on a move, the guest list on an invite.
Delete it and it’s gone
Session history lives on your phone, forgets itself after thirty days, and deletes on request, one or all.