This startup bucked mainstream robotics training methods to teach a robot how to load the dishes (www.businessinsider.com)

Sunday Robotics this week emerged from stealth with Memo, a wheeled home robot that autonomously clears a dinner table and loads a dishwasher — even picking up two wine glasses in one hand — plus demonstrations of folding socks and loading an espresso machine. Founded in April 2024 by Tony Zhao and Cheng Chi, the startup says Memo completed these real-world tasks in over 20 live demos without breaking a glass, built after fewer than two years of development. The company’s key technical pivot is a proprietary glove that mimics Memo’s Lego-like gripper: humans wear the glove to perform household tasks and record contact forces, kinematics and intent directly in the robot’s action space. Sunday claims this produces task data far more cheaply and scalably than teleoperation or heavy simulation (Zhao cites roughly $200 vs $20,000 capital efficiency), and it already leverages 500+ human data collectors across the U.S. The approach tackles the core dexterity bottleneck—capturing fine contact and force information for fragile objects—while promising faster real-world generalization. Remaining questions include how well glove-derived data transfers across varied household geometries and objects, but the method could materially lower cost and time barriers for training compliant, manipulation-capable home robots.