The table-to-dishwasher task is the classic nightmare scenario for roboticists: Long-horizon, highly dexterous, precise, whole-body manipulation combined with delicate, transparent, reflective, and deformable objects. Yet Memo handles it so naturally and elegantly.

It took us over a year to engineer the core infrastructure. We then spent the past 3 months to produce all the autonomous results above. Below, I highlight some of my favorite parts of this release.

Skill Capture Glove aligns the hands, but what about the rest of the body? Human collectors vary in height and arm length, and are also visually different. We developed Skill Transform, a method that converts glove data into equivalent robot data with a 90%+ success rate.

The Skill Capture Glove gives us two orders of magnitude higher capital efficiency compared to teleoperation ($200 vs $20,000) It also allows us to scale diversity faster. You can collect data anywhere without needing to move robots around.

Instead of teleoperation, we train solely on data from our Skill Capture Glove. The glove is co-designed with Memo's hand, meaning they share the exact same geometry and sensor suite. If you can do it wearing the glove, Memo can learn it.

Today, we present a step-change in robotic AI @sundayrobotics. Introducing ACT-1: A frontier robot foundation model trained on zero robot data. - Ultra long-horizon tasks - Zero-shot generalization - Advanced dexterity 🧵->

Specifically, wine glass loading is the most delicate subtask: Push down with too much force? Shatter. Insert the wrong prong? Shatter. We broke many during development, but zero over 20+ live demo sessions.

One less-known fact about glove-based data collection: it produces higher quality data than teleop on contact-rich tasks. Remote teleop can’t provide good force feedback, but gloves do naturally, making tasks like sock folding, which rely on feel, far easier to capture.

It is even more fun to see how Memo reacts to unseen environments. We deploy it to 6 unseen Airbnbs and task the robot with fine-grained tasks such as picking up utensils from the plate. Because we train on data from over 500 homes, the new home is instantly familiar to Memo.

If you want to learn more about ACT-1 and Memo, read our first research blog post: Follow @sundayrobotics for all future updates.

The ultimate nemesis: the step

This is incredible @tonyzzhao Mind sharing if the table to dishwasher task was collected end to end including navigation parts and if robot is executing the entire skill sequence from the learned policy?

end-to-end including navigation, everything learned.

looks ai generated tbh

omg the little cap

Nice!! One thing i was wondering, after this task the hands will be greasy. What's the plan for washing them before going to the next task like folding my white shirts?

hands will be waterproof and memo will wash its hands under the faucet ; )

also soap-proof? Would be cool, but a bit dangerous if there's minor damage? I was also thinking either some kind of throw-away gloves, or (but maybe much too complex) detachable hands he can put into a hand-washer station and pick up again, a bit like a charging station.

can you confirm whether these are real videos with real physical robots? it's hard to tell, but they look a lot like 3d renders from sim.

this is the most impressive robotics demo I've seen in a long time - assuming it's fully autonomous. even the long horizon part alone is extremely impressive. congrats on this breakthrough!! 🚀

will sell 10x better than one with legs. Golden retriever vibes, well done.

How does it handle picking up dirty dishes? Particularly dishes and silverware with grease/oil/sauces. Does it wash its hands with soap before touching more surfaces?

Great job, but this environment seems too perfect. None of these dishes are wet, how does it get handled? How can it handle glasses with different material?

Why did you speed the video up? What else you hiding about this?

until it throws a lit roach in the trash

Amazing, congratulations. Can it wash its hands if dirty

Why is this video a 3D animation and not a real reel?

Skeptical of the wheels but this actually turned out quite nice. The dexterity is impressive

How does it climb little stairs if it divides the kitchen from dining room?

Fast and efficient. Robots can do better without legs. Can it vacuum while moving?

the way that my wife would make him pre-wash the plates and load them very (very) specifically... he doesn't have enough battery for that

First non-Optimus bot that I’ve felt has an edge that ensures its survival. Neat.

To be honest, dishwasher design kinda sucks. You could start designing all of society around robots and not humans.

It put the utensils in facing up. Total deal-breaker. I've ended long term relationships over this.

>dishwashing robot >still relies on the dishwasher

My partner @sulaimanghori and I would LOVE to be a part of the memo beta! We are always working and hardly have time to do laundry or dishes. Happy to help Memo learn and collaborate with you however needed!

very cool. Does Memo wash its hands after this?

How does this climb the Dutch stairs? Or cleanup my daughter's room with lots of toys on the ground?

@Mattlama92

its a classic nightmare scenario for me as well

Crazy how effortless this looks given how hard the task actually is.

Nice render lol

how does it operate on uneven flooring, bad floorboards, gaps, rivets, bumps e.t.c?

A normal home will not be having that clean floor I want some toys and clothes lying around and the robot to go around doing job.

Every robot company needs to be trying to tackle the world's 2 biggest problems first: dishes and laundry

I would trust it even more if the body was centered on the base. I understand it would decrease the reach but still..

What if I’m the dishwasher