An AI baby monitor that understands the day. Ask anything about your baby.
Why this exists
You see the morning and the night. Whoever else is there sees the middle. Your baby sees all of it and cannot tell you any of it. The camera in the room already records every hour. It just does not tell you anything.
The app
LIVE
Hold to talk. Drag to move the camera.
Recent activity
Detection runs on vision-language models rather than a fixed computer-vision pipeline, which is why it transfers to a room it has never seen without retraining for that room. The camera itself is off-the-shelf and already certified.
What you get
Sleep, feeds, diaper changes, rollovers and caregiver time, read off the camera and laid on one timeline. Nobody typed any of it in, and where the camera did not see, the row stays empty.
Caregiver time is a duration, not a verdict. It is measured and reported the same objective way as everything else on this timeline.
How you use it
No scrubbing back through hours of footage to find one moment.
How it behaves
Seen and it didn't happen is an answer. Not seen is a different one. A zero would be a lie.
Where this goes
Children from birth to about five come first, because that is when a child cannot yet tell you how the day went. Pets and elders run on the same loop.
Where we are
The loop runs end to end on real hardware: camera, detection, timeline, phone, answers. Next up is a Korea beta with five households, August 2 to 9, to test accuracy against what parents actually logged, and whether the answers are worth paying for.
Twenty in-depth parent interviews shaped the product into automatic logging rather than another alert. The regulatory path is cleared in advance, which is unusual this early: the data handling is built for Korea's privacy act, and the beta fleet runs under a confirmed conformity route.
Not done yet: real-world accuracy at scale, retention, acquisition cost. We would rather show you exactly what we do not know than fill it in.
Who