In a significant stride towards fully autonomous domestic assistance, Sunday Robotics has introduced a groundbreaking artificial intelligence model named ACT-2. This advanced system empowers their wheeled home robot, Memo, to adeptly perform household chores, specifically laundry folding, with an astonishing success rate exceeding 99%. This achievement was demonstrated not in controlled laboratory settings, but in diverse, unfamiliar residential environments, tackling a variety of garments it had never encountered during its training.
Pioneering Domestic Robotics: Sunday Robotics Redefines Autonomous Capabilities
The announcement on Thursday marked a pivotal moment for Sunday Robotics, a Bay Area-based startup that earlier this year secured $165 million in funding, pushing its valuation to $1.15 billion. The company's CEO and cofounder, Tony Zhao, confidently declared that 2026 would witness a dramatic acceleration in the timeline for autonomous home robots, attributing this to their recent innovation. For decades, the robotics industry has grappled with the elusive goal of creating machines that are both robust and adaptable enough to function seamlessly in the unpredictable chaos of a typical home. Traditional robots often excel in rehearsed tasks within sterile environments but falter when confronted with novel objects or changing surroundings. Sunday Robotics asserts that ACT-2 overcomes this hurdle, enabling Memo to transfer learned skills from one setting to another, thereby bypassing the laborious and expensive process of re-training the robot for each new household.
Beyond their technological advancement, Sunday Robotics is also advocating for a new benchmark in the robotics industry, terming it a "Solve." This proposed standard aims to provide a clearer, more transparent method for evaluating robot performance, moving beyond flashy, often misleading, demonstrations. Zhao highlighted the current inconsistency in reporting, where companies frequently present impressive but uncontextualized demos, making it challenging to assess genuine progress. Under the "Solve" framework, companies would be required to explicitly detail a robot's capabilities, its testing conditions, and the extent of additional training or human intervention needed in different scenarios. Sunday Robotics proudly positions Memo's laundry folding proficiency as its inaugural "Solve" capability.
The development of ACT-2 involved a two-stage training process. Initially, the robot acquires fundamental physical competencies by observing human operators performing everyday tasks. These human actions are captured using specialized sensor-equipped gloves, meticulously designed by Sunday Robotics to mimic the tactile and motion characteristics of Memo's own manipulators. This innovative approach allows for a direct translation of human movements into the robot's operational code. Following this initial learning phase, Memo refines its skills through iterative self-correction, leveraging Sunday's extensive internal fleet of robots. This continuous learning loop enables the robot to rapidly improve its performance and master new tasks within minutes.
CEO Zhao emphasized the exceptional transferability of Memo's acquired skills, stating that its performance in their office environment closely mirrors its efficacy in real-world applications. The company has conducted extensive testing over several months, deploying the robot in employees' homes and Airbnbs to thoroughly evaluate its adaptability in diverse, unfamiliar settings. Sunday Robotics is preparing to launch a beta program this fall, placing Memo units in a select number of homes, though Zhao did not disclose the exact scale of this deployment. Unlike some other robotics ventures, Sunday Robotics is committed to ensuring Memo operates with full autonomy, with remote human operators only intervening to provide customer support, akin to the assistance offered by Waymo's self-driving vehicles. Crucially, the company has pledged not to exploit these interventions for collecting customer data to further train their models, prioritizing user trust and privacy.
Previously, Sunday Robotics showcased an earlier iteration of their model, ACT-1, demonstrating its ability to clear dinner tables, load dishwashers, and even prepare espresso. The current focus for ACT-2 is to enhance the reliability of these functionalities, with laundry serving as the initial benchmark. The team is actively working on expanding Memo's repertoire to include tasks such as vacuuming, organizing toys, fastening zippers, and brewing coffee, acknowledging that these capabilities are still under development and require further refinement to meet their high standards of reliability.
The emergence of Sunday Robotics' Memo robot represents a significant step forward in the quest for truly intelligent and adaptable home robots. Its ability to perform complex tasks like laundry folding in varied, real-world environments without extensive re-training highlights a paradigm shift in AI and robotics. The proposed "Solve" standard could also bring much-needed clarity and rigor to an industry often characterized by sensational but undemonstrated claims. This development underscores the growing potential of physical AI to seamlessly integrate into daily life, offering practical assistance and transforming our interactions with technology. It's a reminder that the future of home automation is not just about convenience, but about robust, reliable, and ethically deployed artificial intelligence.
