Design @sundayrobotics
Sunnyvale, CA
Joined November 2009
- Tweets2.2K
- Following623
- Followers1.9K
- Likes3.3K
SHY retweeted
I’m hiring a Senior Product Manager for Infra @sundayrobotics.
I promise a tight-knit early team that works very hard, fast, and autonomously. If you are passionate about scaling deployment of consumer robots, apply here: jobs.ashbyhq.com/sunday/c864…
I predict a vast majority of compute in space will be dedicated to simulating synthetic data for training
SpaceX has just released a new look at what their 1 million Starmind satellites constellation might look like (dots do not represent true size of sats).
SpaceX VP @michaelnicollsx said today: "We need to operate clusters of satellites in tight formation to get enough coherent compute to run AI models efficiently. A cluster will be 10-ish satellites connected with 10 terabits or so of bandwidth between them, and interconnected to the broader constellation."
“I think if you do something and it turns out pretty good, then you should go do something else wonderful, not dwell on it for too long. Just figure out what’s next.”
SHY retweeted
Stunning demo from @tonyzzhao showing off Memo and the new ACT-2 model.
A few great things I saw in this live demo:
- Memo was able to operate even with a shirt obstructing its view from the hat camera
- Memo was able to quickly recover when Tony threw more shirts on top of what Memo was folding
- Memo can very precisely fix twists in pant legs and shirts
- Memo can be given new skills just with software updates
Incredible demo!
Sunday Robotics' Tony Zhao runs a fully autonomous robot demo live and unscripted, and it corrects itself every time you mess with it:
@brexton: "I want to confirm for the audience right now that this is fully autonomous. There's no teleoperation, no one behind the curtain, and this is real time. And we did not prepare this demo before. Why do you think seeing real-time demos is so rare for robotics? What the audience is seeing right now is not that common."
@tonyzzhao: "In order for a live demo to work, it's really stressing all parts of the system. From the bottom is the reliability of the hardware, then the software system, then the model. Because we're a full-stack company, we're able to do excellent work across all of them. We have these robots running pretty much every day, and it just keeps getting better as they practice."
@brexton: "So kids can bump into Memo, people can drop clothes, and it will correct itself just on the fly."
@tonyzzhao: "Act Two is the first robotic policy that can handle the diversity of homes in a way that is this reliable. You can see as I'm twisting it, if I twist it in again, it will keep doing it. So Memo has infinite patience."
@sundayrobotics
SHY retweeted
Sunday Robotics CEO @tonyzzhao believes home robots only need to master a handful of chores before they become indispensable.
"What is the minimum number of tasks that a robot needs to do to justify its own existence so that people love having it in their homes? I actually think the list is pretty short. The tasks are pretty repetitive. You just need to learn it and just keep doing that over and over again, like folding shirts or loading dishwashers."
SHY retweeted
home robotics is unique in having research-market fit: every step toward a useful product is also a step toward embodied general intelligence.
Product perspective on “Research-market fit” as described in our recent research update:
The hardest problems in physical AI get solved closest to the user, not furthest from them. I’ve spent the past couple weeks visiting user homes and watching practice deployments on airbnbs.
One thing is clear: general-purpose robots won't be won in a lab or in simulation- it’ll be won with real-world data. And there's no harder environment to generalize to than the home. Every home is different, messy, and constantly changing as people move through it- it’s the truest measure of general intelligence we have.
Sunday is preparing to deploy zero-shot in new homes with 99% reliability. Excited to test us against the long tail of edge cases in real deployment next.
So grateful for the thousand Memory Developers who are collectively developing the muscle memory for Memo
SHY retweeted
This video is a great snapshot of our progress, and the blog post offers an exciting glimpse into the scaling laws we’re discovering. For me though, some small, unintentional, unscripted moments have done the most to make our progress feel real in recent months. I'll share anecdotes from a couple of those moments in this 🧵
Onboard inference is the only way to build robots that protect the privacy of our most intimate spaces
We just hit a weird milestone: our model became more reliable than your average home WiFi.
Just like everybody else, we thought cloud inference was the obvious choice. Yet 2 days into the ACT-2 eval, our mind completely changed.
If our hero @ArpitKalla didn’t cook, this video wouldn’t exist 🧵
SHY retweeted
one of the remarkable things about Sunday is how tight the in-house iteration loop is. we curate datasets as intentionally as we design models, then run systematic evals to see what actually improved.
as the models get better, the eval problem gets harder in ways that aren’t obvious at first. a few thoughts:
SHY retweeted
3 hours of Memo folds to relax/study to
youtube.com/watch?v=a2HZyURU…
