Pioneering AI, pragmatic engineering.
Powered by people.
We’re building a future where technology finally works the way it should: removing distraction instead of creating it. We own the entire stack: software, hardware, data, training, and model evaluation, so we can iterate faster than anyone else.
Watch Memo autonomously clear the dining table, fold socks, and pull a real espresso — a glimpse of what general-purpose robots can do.
A breakthrough in Robotic AI
Memo learns without teleoperation
While other robots learn from teleoperated data—collected from the robot itself—Memo learns directly from humans wearing the Skill Capture Glove™.
The bottleneck in robotics is training data. Learning a single complex task can take thousands of hours, and traditional data gathering—like teleoperation—is slow, costly, and intrusive. Sunday’s goal is different: we are building true autonomy in a scalable way.
Enter the Skill Capture Glove. This technology allows our team of Memory Developers to generate data at scale and train Memo rapidly, long before it reaches your living room. We co-designed Memo’s hand to be a perfect mirror of the glove’s shape and sensors. As a result, any skill you demonstrate, Memo can master.
The world’s richest movement library
Data capture
We capture hundreds of hours of human movement from real homes every single day, building the world’s most diverse library of motion data.
9700 Eps
To date, we’ve shipped over 2,000 Memory Gloves to contributors across the country. These Memory Developers transform their everyday movements into training data, allowing Memo to learn faster than any other robot.
This massive, diverse dataset is the foundation of Memo’s intelligence. Higher quality data directly translates to faster skill acquisition and more natural movements. Unlike traditional programmed robots, Memo learns from actual human experience—understanding not just what to do, but how real people do it.
The infrastructure behind the intelligence
Transforming data into skills
Memo gets smarter every week. We’ve built the shortest iteration cycle in robotics: from data capture to deployed capability in less than a week.
Recruit
We start by carefully recruiting an ever-growing community of data contributors we call Memory Developers (MDs). This community represents a wide diversity of US households who are curious and eager to shape the future of helpful robotics.
Recruit
48 Hour Cycle
Our infrastructure seamlessly connects every step: recruiting Memory Developers, capturing movements, training models, and evaluating performance. Memory Developers perform carefully choreographed actions while our Memory Glove captures precise motion data through force sensors, cameras, joint potentiometers.
Each cycle builds on the last. Our data operations team evaluates performance, identifies gaps, and refines collection strategies. With every iteration, we expand recovery from failure and accelerate Memo’s path to real usefulness.
Real-world ready, technically proven
Specifications
Designed for real lives and trained in real homes across the US, Memo works with you, not around you. From studio apartments to family houses, it adapts to your space and routines from day one. No special setup, no learning curve: just a helper that fits naturally into your household.
Hardware Specs
Overview
Height: 1.7 meters
Weight: 170 pounds
Horizontal reach: 0.8 meters
Vertical reach: 2.1 meters
Degrees of Freedom
Arms: 2×7
Hands: 2×4
Torso: 1
Lower Body: 4
Speed
Navigation Max: 1 meter/s
Manipulation: Memory Developer do demonstrations at about half of natural human speed during a task, and our team runs models at about 60-80% of data speed.
Battery
Runtime: 4 hours
Charge time: 1 hour
Self-charging will be available for the Beta Program.
Materials
A blend of rigid and elastic polymers for robustness and safety. The soft pieces of the body's cladding are available in a wide array of colors.
Ingress Protection
Hand: IP67
Lower Arm: IP66
Design and dimensions
Safety and security
Hardware safety
We designed Memo to be compliant and passively safe. If you bump into it, it naturally yields, and it remains stable even if power is cut at any configuration.
Software safety
Memo has built-in collision avoidance for both static and dynamic obstacles. It only performs tasks we’ve explicitly taught it—tasks that are useful and safe. Every behavior is governed by rigorous data review.
Data security
With Skill Capture Glove and our Memory Developers, we do not need to depend on data from your home to train Memo. If you ever choose to share feedback about Memo’s behavior, it will only happen with your explicit consent.
Glossary
Actuator
A component that converts energy (electrical, hydraulic, pneumatic) into motion, enabling movement of a robot’s joints or mechanisms.
Teleoperation
Teleoperation is the control of a robot by a human operator at a distance. The human sees what the robot sees and directly sends commands to move the robot’s arms, wheels, hands, etc.
Skill Capture Glove
Our patented Skill Capture Glove is a one-to-one representation of Memo's hand and enables anyone to teach Memo nearly any skill. What you can do wearing the Skill Capture Glove, Memo can learn.
Compliant Control
In robotics, compliant control refers to any control strategy that allows a robot to intentionally yield, flex, or adapt its motion in response to external forces — instead of behaving like a rigid, unstoppable machine.
Skill Library
Our ever-growing dataset of in-the-wild data, generated by Memory Developers using our Skill Capture Gloves. As of November 2025, it's approximately 10 million examples.
Passive Stability
A system is passively stable if it naturally settles into a stable equilibrium (e.g. not tipping over) due to its mechanical design, and does not require ongoing actuation to maintain that configuration.
Memory Developer
Memory Developers (MDs) are data collectors who use our Skill Capture Gloves to train our models on a wide diversity of tasks. Our MDs represent households across the US and collect from their lived-in homes, rather than warehouses.
Autonomous
When a robot operates autonomously, it acts on its own. It senses the world and accomplishes tasks without moment-to-moment guidance. This contrasts teleoperation, in which humans remote-control robots.
ACT-1
ACT-1 (pronounced act-one) is the robot foundation model we’ve built in-house at Sunday. It learns from data captured by Memory Developers using our Skill Capture Gloves.
Sunday
The helpful robotics company
Learn more