At the heart of Memo’s intelligence are 10 million real household chore episodes, captured from 500 families. This is not a lab-born dataset with neatly staged actions, but a sprawling archive of how people actually clear tables, fold laundry, sort shoes, and brew espresso under unpredictable settings at home. It was collected by Mountain View startup Sunday, which was founded by Stanford-trained roboticists Tony Zhao and Cheng Chi using their patented Skill Capture Glove-a wearable that captures fine-grained human motions in cleaning, organizing, and manipulating objects.
But the approach directly solves one of the biggest bottlenecks that has dogged domestic robotics: a lack of large-scale, varied, real-world training data. As Zhao said, “Most home robots start as adaptations of industrial machines, and those trained in labs rarely succeed in unpredictable, real-world environments.” Benchmark’s Eric Vishria echoed the statement, saying that we have about one-millionth of the data we need for robots to thrive in messy homes.
Memo’s capabilities extend to what robotics researchers term “long-horizon” tasks: multistep, context-rich sequences needing adaptive decision-making. In live demonstrations, Memo performed a Table-to-Dishwasher routine involving 33 distinct interactions with 21 objects over 130 feet of navigation while breaking zero wine glasses in over 20 sessions. It can fold socks, which requires precision in handling deformable fabric, and pull a perfect espresso shot, locking a portafilter into the machine with even force.
Underlying these skills is ACT‑1, a foundation model trained fully on human movement data, not robot demonstrations. This “zero robot data” paradigm mirrors developments in systems such as the Real‑to‑Sim‑to‑Real pipelines developed by MIT CSAIL, which capture capabilities first in the real world to create digital twins for simulation and then transfer policies back to physical robots. Thus, by rooting learning in realistic household contexts, Memo avoids the fragile performance common when robots, which have been trained in synthetic environments, encounter the variability of real homes.
Memo’s physical form also breaks with the humanoid trend, having a wheeled base with a telescoping central column to allow the robot to reach from floor level up to 2.1 meters while maintaining stability, even during power loss. Such a design draws on decades of wheeled mobile robot stability research, whereby kinematic and dynamic analyses demonstrate that a low center-of-gravity design offers lower risk of falls and increased precision of control. The use of wheels for locomotion also enables navigation algorithms that can efficiently plan routes and avoid obstacles without the challenge of bipedal gait control.
The exterior is clad in a soft, silicone-like material; principles of soft robotics design provide a safe interaction between human and robot. Rounded contours, a friendly cartoon-like face, and interchangeable caps continue the retro-futuristic aesthetic of Baymax or early Nintendo hardware, making Memo visually approachable in domestic spaces.
Safety and adaptability are not only aesthetic considerations. The soft exteriors reduce injury risk in case of accidental contact, while the stable wheeled base ensures controlled motion across a variety of flooring. Applying principles found in soft-wheeled locomotion systems, Memo’s chassis and actuation are tuned for consistent torque delivery, enabling transitions from carpet to tile to hardwood without destabilizing payloads. The massive real-world datasets, competence over long horizons for tasks, stability-focused design, and soft-material safety make Memo a significant leap in home robotics.
Sunday’s early access program, opening on November 19, 2025, will place 50 numbered units into selected homes for beta testing ahead of a broader release in 2026. Those “Founding Families” will have a direct impact on Memo’s refinement, feeding back data that could further expand its already-unprecedented chore repertoire. To the tech-savvy consumer and robotics enthusiast, Memo represents a shift from spectacle to service-robots built not to impress in staged demos but to integrate into the daily ebb and flow of real life, adapting to the unpredictable choreography of human homes.
