This Home Robot Clears Tables and Loads the Dishwasher All by Itself
The future of household chores may very well reside in the gleaming white chassis of Memo, a revolutionary home robot from the innovative company Sunday Robotics. While it might not be the quickest barista on the block, Memo’s capabilities are undeniably impressive, marking a significant leap forward in the quest for truly autonomous domestic assistants. Envisioned to liberate individuals from the mundane routines of daily life, Memo promises a future where the burden of laundry, dishes, and other household tasks is gracefully shouldered by intelligent machines.
During a recent demonstration in an open-plan kitchen nestled in Mountain View, California, Memo showcased its prowess by flawlessly executing a series of complex tasks, from crafting a perfect espresso to meticulously clearing a table and loading items into a dishwasher. Its appearance is reminiscent of a character plucked from the animated film Wall-E, featuring a pristine white body, two highly articulated arms, and a friendly, almost cartoonish face, topped with a jaunty red baseball cap. Unlike fully humanoid robots that rely on legs for locomotion, Memo navigates its environment with remarkable fluidity, utilizing a robust wheeled platform. Its height is dynamically adjustable, allowing it to effortlessly slide up and down a central column, adapting to various counter heights and tasks.
The demonstration began with a request for an espresso, a seemingly simple act for a human but a monumental challenge for a robot. Memo responded instantly, smoothly rolling towards the countertop. Its two pincer-like hands, equipped with sophisticated sensors and actuators, then embarked on a slow, deliberate, yet precise sequence of actions required to operate a standard espresso machine. It carefully filled the portafilter with coffee grounds, applied just the right amount of pressure to tamp them down, slotted the portafilter securely into place, positioned a coffee cup beneath the spouts, initiated the brewing cycle with a precise press of the buttons, and finally, retrieved the steaming hot drink. This intricate dance of object identification, delicate grasping, and sequential operation in an unstructured environment highlights the immense complexity that Sunday Robotics has managed to overcome.
"We want to build robots that free people from laundry, from the dishes, from all chores," articulated Tony Zhao, cofounder and CEO of Sunday Robotics, as Memo elegantly delivered the freshly brewed coffee to the person who requested it. His statement encapsulates the ambitious vision driving the company: to transcend the limitations of current robotics and bring truly helpful, adaptable machines into our homes. The act of making an espresso, while appearing unspectacular to the casual observer, represents a colossal achievement in the field of robotics. It demands an advanced ability to identify a diverse array of objects—from coffee beans to cups and machine components—understand their properties, figure out how to grasp them reliably without causing damage, and then use them properly in a dynamic, unpredictable setting like a typical home kitchen. Sunday Robotics isn’t just assembling parts; they are building their own hardware from the ground up and, crucially, developing and training the sophisticated AI models that enable their system to learn and perform these complex tasks. Zhao emphasizes their "full-stack" and "vertically integrated" approach, acknowledging that this comprehensive strategy, encompassing everything from mechanical design to AI development, is "a very ambitious thing to do."
The vast majority of robots prevalent today are relegated to performing highly precise, repetitive tasks within tightly controlled industrial environments. Think of robotic arms in factories, tirelessly moving the same item from one fixed position to another, over and over again. These industrial workhorses, while efficient, lack the adaptability and improvisational skills inherent in humans. They struggle with changes, unexpected obstacles, or unfamiliar situations. The last decade has seen incremental progress, with some companies developing robots that leverage AI for simpler tasks, such as identifying objects on a conveyor belt and determining the optimal way to grasp them. However, operating in an environment as inherently varied, cluttered, and unpredictable as a real home presents a challenge of an entirely different magnitude. A home is a dynamic landscape, constantly changing with human activity, shifting objects, and unpredictable elements like pets or children. For a robot, navigating this chaos and performing meaningful tasks requires a level of intelligence, dexterity, and adaptability that has, until now, largely remained within the realm of science fiction.
Of course, the dazzling performance of a robot in a controlled demonstration environment does not always accurately predict its utility in the real world. The ultimate litmus test for Memo will be its ability to consistently and reliably perform its programmed tasks across a wide variety of homes, each with its unique layout, lighting conditions, and potential for unexpected scenarios, and crucially, without the constant supervision or intervention of Sunday Robotics’ engineers. This transition from a meticulously set-up demo to the chaotic reality of everyday life is where many promising robotic endeavors falter.
Nevertheless, Memo has already demonstrated a suite of truly impressive skills beyond coffee preparation. During the same viewing, the robot adeptly cleared glasses from a table and loaded them into a dishwasher. This particular feat was especially noteworthy due to its nuanced dexterity: Memo managed to grasp two glasses simultaneously in a single hand. It achieved this by holding one glass securely between its thumb and pointer finger, while ingeniously utilizing the remaining fingers and palm to grip the second. Such a sophisticated, multi-object grasp is a testament to the advanced training and manipulation capabilities embedded within Memo’s system.
This remarkable dexterity and human-like manipulation are not accidental; they are the direct result of Sunday Robotics’ core innovation: a novel and highly effective method for training robots. Rather than relying solely on traditional teleoperation, where a human directly controls a robot’s movements, Sunday employs remote workers who wear specialized gloves designed to mimic Memo’s hands. These gloves, costing approximately $400 a pair, provide a significantly more accurate and nuanced training signal. By directly capturing human hand movements and forces as workers perform household chores, Sunday generates a rich, high-fidelity dataset that translates into more natural and effective robot actions. This detailed training data, gathered from human workers performing real-world tasks, is then fed into a sophisticated AI model. This model, in turn, processes input from the robot’s array of sensors (cameras, force sensors, etc.) and translates it into precise control signals for Memo’s motors and grippers. Ken Goldberg, a renowned roboticist at UC Berkeley and cofounder of Ambi Robotics, lauded this approach, describing it as "a very exciting variant on home robots," praising Memo’s "beautiful design, and a much smarter kind of data capture."
The very fact that a company like Sunday Robotics believes it can build a truly useful and functional home robot is a powerful indicator of the skyrocketing optimism and rapid progress occurring within the field of robotics. This optimism is not unfounded. The last few years have witnessed groundbreaking advancements, particularly in the integration of large language models (LLMs)—the sophisticated AI brains behind today’s popular chatbots—into robotic systems. Researchers have demonstrated that robots can now tap into the capabilities of LLMs to interpret complex natural language commands, understand the context of a scene, and even plan multi-step actions, moving beyond mere reactive programming to more intelligent, adaptive behavior. This new paradigm promises to unlock unprecedented levels of understanding and autonomy for robotic agents.
Furthermore, a significant research thrust is focused on collecting massive quantities of diverse data that illustrate how humans perform various actions. This includes everything from picking up cups to folding shirts and loading laundry. The hope is that by exposing robots to an "internet for robotics"—a vast repository of human demonstrations and interactions with the physical world—a more general, adaptable form of robotic intelligence can emerge. This ambitious goal aims to move away from task-specific programming towards systems that can generalize learned skills to novel situations, much like humans do. Tony Zhao articulates this challenge eloquently: "If you think about the most powerful AIs, ChatGPT or image-generation models, they are trained on the whole internet. We just don’t have the internet for robotics." This highlights the immense data gap that Sunday Robotics and others are striving to fill with innovative data collection methods.
The foundational work for Sunday Robotics’ innovations is deeply rooted in the groundbreaking contributions of its cofounders. Tony Zhao’s prior work at Stanford University included the Mobile ALOHA project, which focused on training robots using a low-cost mobile teleoperation system. This project explored efficient ways for humans to teach robots complex manipulation tasks in dynamic environments. Similarly, Cheng Chi contributed to a collaborative project involving Stanford, Columbia University, and the Toyota Research Institute. This research demonstrated how a relatively inexpensive claw-like device could be effectively used to gather valuable data from humans performing tasks such as cleaning dishes, paving the way for more accessible and scalable data collection methodologies. These experiences provided crucial insights into robot training, human-robot interaction, and the challenges of real-world manipulation, directly informing Sunday Robotics’ full-stack approach and their unique glove-based training system.
The landscape of home robotics is rapidly becoming a hotbed of innovation, with a handful of other startups also hustling to develop and deploy more capable robots, specifically designed for the complexities of ordinary homes. Companies like Physical Intelligence, Skild, and Generalist are all actively working on developing advanced robot models that can adapt to new and unforeseen situations, leveraging similar data-driven, AI-centric approaches. Even 1x, a company that recently unveiled a humanoid home robot, acknowledges that its system still requires human teleoperation for certain tasks, underscoring the formidable challenges of achieving full autonomy in a domestic setting. This vibrant competitive environment signifies a collective belief in the imminent arrival of a new generation of practical, intelligent home robots.
Industry experts and investors are taking notice. Sarah Guo, founder and managing partner of Conviction, expresses strong confidence in Sunday Robotics, citing its team, which includes veterans from Tesla and Google DeepMind. "Tony and Cheng are incredibly good," she states, adding that "They’ve since recruited and enabled an all-star team that can uniquely do hardware-AI co-design, and deliver a full-stack product." This endorsement highlights the critical importance of a team capable of seamlessly integrating both hardware engineering and cutting-edge AI development. Eric Vishria, a general partner at venture capital firm Benchmark, which is backing Sunday, echoed this sentiment in a statement. He emphasized that Sunday’s "practical approach is the way to make robots more useful." Vishria asserted that "The promise of AI robotics isn’t doing a backflip or dancing demos, but robots that work in messy, real-world situations," concluding that Sunday’s "breakthroughs mark the start of an exponential curve toward a future where robots actually work in our day-to-day lives." These endorsements from seasoned investors and industry figures underscore the significant potential they see in Sunday Robotics’ pragmatic and comprehensive strategy.
Sunday Robotics plans to roll out Memo to beta testers next year, a crucial phase that will truly test the robot’s mettle in diverse, real-world home environments. This pilot program will provide invaluable feedback on how people interact with and respond to having a home robot capable of performing certain chores—even if slowly, and perhaps not with absolute perfection every single time. A key question that the beta testing will address is Memo’s reliability and efficiency when confronted with the guaranteed complications of a typical home: the unpredictable movements of kids and pets, shifting clutter, and the inherent messiness of daily life. These are the variables that separate a successful demo from a truly useful product.
Following the beta testing phase, Zhao indicates that Sunday will begin rolling Memo out to its first commercial users. He draws a parallel to the early days of home computing, where the initial machines were complex and primarily appealed to enthusiasts who were willing to tolerate some "rough edges" in exchange for being at the forefront of technological advancement. Zhao believes Memo might initially find its audience among those who are eager to embrace a robotic future and are prepared to engage with the technology, potentially even showing their robots how to perform new tasks. "I do think that people should be able to teach their own robots," Zhao affirms, envisioning a collaborative future between humans and their robotic assistants.
The journey towards truly capable and ubiquitous home robots is undoubtedly long and fraught with challenges. Yet, with innovations like Memo from Sunday Robotics, driven by a full-stack approach, novel training methodologies, and a clear vision, the era of intelligent domestic assistance seems to be drawing closer than ever before. For now, while the promise of a chore-free future beckons, many might simply settle for a consistently decent espresso delivered by a robot.
