Sunday predeployment data strategy
Sunday
Data collection is the real product bottleneck in home robotics, and Sunday is trying to solve it before the robot ships. Instead of putting remotely operated robots into private homes and waiting for failure cases to generate training data, Sunday pays people to wear glove and camera gear while doing chores, then converts those hand motions into robot actions. That front loads dataset creation, lowers early hardware burn, and avoids the privacy tradeoff that comes with teleoperation inside the home.
-
Sunday already has a concrete pre deployment pipeline. Its Skill Capture Gloves and camera hat collect demonstrations from workers in real homes, and the system translates those motions into trajectories for Memo. More than 1,000 gloves across 500 plus households means Sunday can gather varied home data before a robot enters a customer kitchen or bedroom.
-
The Bot Company is making the opposite bet. It rejects teleoperation in homes on privacy grounds and relies on in situ autonomy data, which means the robot has to start operating in the home to begin collecting the highest value training examples. That can produce richer robot native data, but only after hardware is already deployed.
-
This mirrors a broader split in robotics. Industrial robot companies often use intervention based teleoperation because factories tolerate remote human backstops and need uptime, but that playbook is harder to переносе into homes where constant remote observation is more sensitive. Sunday is effectively building a home safe version of the same data flywheel.
The next phase is likely to favor companies that can turn messy household behavior into training data without paying a human to sit behind every robot. If Sunday can keep expanding glove coverage and convert those demonstrations into reliable action policies, it can enter the home market with a larger head start on the hardest part, which is not mechanics, but learning how ordinary homes actually work.