Aaron Tan's robotic laundry-folding lamp went viral — here's what's real and what's rendered

our last guest of the day hopping on the stream, Eric. Went viral uh yesterday for creating a robot that can uh that that looks like a lamp. So, let's bring him in from the reream waiting room. There he is. Good to meet you. Welcome to the stream. How you doing? Great. Great. Thanks for having me on.

Uh break down the video. What are you building? Uh how much of that is CGI? How much of that is real? Where are we in the techn technology roll out of this type of robotics project? Yeah. Yeah. So, uh the video you saw probably yesterday, it was a it was rendered.

Um it's not AI generated unfortunately, but it was rendered. Sure. Um so, we have a prototype working. Yep. This one right here that you see in the background. Okay. Um so, we we we are building robotic lamps. Uh they're meant to sort of disappear into your home when you Oh, look at it working over there.

Yeah, it's working. Wow. That's working away. We're we're getting the laundry folding going. Yeah, it's a super cool uh like I don't know.

I mean, part of why it went so viral is like people have been talking about laundry folding robots for years and then we've seen uh companies take it very seriously, but never with a form factor that could just melt into your home.

I don't know how protectable that is, but it just felt like from a user experience, from a from a perspective of like actually integrating this into your life, it felt much more real to me. Um, how much of where did the idea come from and and what are the benefits of like building it into the bed frame basically?

Well, I mean to be clear, so it's not it's not actually built into the bed frame at all. It's a single floor lamp that you can technically place anywhere. Oh, you can put it anywhere. Okay. So, it's like by your couch, by your table, by the bed.

Uh, it's just through most people that we spoke to, the bed happens to be where you do your laundry. People just dump it there before they go to bed. Yeah. Um, but yeah, like uh we're familiar with like, you know, some of the previous attempts at laundry folding.

And I think the main thing that has always held it back is the form factor. Um, you know, like there's usually they come like sort of with these kind of like pre-installed rollers that you kind of have to feed in manually. It's like semi-autonomous. It only works for specific type of clothing.

Um, what we going have here, you know, mimics a little bit kind of like the human arms. So we can adapt to different types of clothing sizes, uh, different types of articles and styles. Um, and you know, with the recent breakthroughs in AI, we're finally able to generalize across different clothing.

How much of those recent breakthroughs in AI can you actually capitalize on? Because I, you know, like GPT 4. 5 is amazing. GPT4, you know, 03 Pro is amazing, but like it doesn't seem like it would translate to laundry folding just yet.

what exactly are you a beneficiary of and then what work do you have to do to kind of take it across the last mile? Yeah.

So I mean there's there's been a lot of tremendous work done in the vision aspect uh for robots to now really be able to understand uh what is seen in front of itself right and then also in the last year within robotics community there's also a lot of foundation models for uh vision language action models that are also available.

It's the BLMs that are very popular. Right. Exactly. But yeah, yeah, but we call them VAS because there's an action component to to these models. So they're they're they're doing pretty well. They can generalize the clothing items that, you know, they haven't seen.

um they're able to plan in sort of like the joint angle space of these robotic arms uh to to be able to fold them um into neat piles uh you know whatever the So do you have to build a uh I I imagine that like like recognizing that I'm looking at a t-shirt is somewhat commoditized like you can kind of get that out of the box for free with some of the foundation models but then there probably no off-the-shelf model that understands exactly how your particular set of joints work.

So is that where you're doing fine-tuning or post- training? Like how exactly are you translating into the specific robotic machinery that you've cho chosen to assemble in a particular way? Yeah. Yeah. So I mean it comes down to like I think you touched up on there uh it comes down to the data collection uh process.

Okay. So we right now we manually telly operate these arms that you see behind me sense collect the data and yeah we fold the laundry ourselves. Yeah. Depending on what you want to fold it can be pretty fast. Um then we post train um fine-tune fine-tune the the existing models essentially.

Teleoperation is it someone using like gloves basically? How are they actually doing the teleoperated folding? I imagine it' be very hard to do it on an Xbox controller. There is two two methods there. Can we go full screen on this video? Yeah. Hold up. Oh, you got some gloves on. Oh god. Whoa.

This is Ready Player One stuff. I love it. Okay. Oh wow. That's awesome. Oh yeah. Yeah. VR. Yeah. with the yeah you you hold it yourself and you sort of do the task. Okay. And then the inverse kinematics happens within the robot to decide where how it actually gets to that particular place in the 3D space.

Yeah, that makes sense. That's cool. How do you how do you you know where where do you expect obviously your number one competition is just people doing their own laundry.

I think it's smart to build a robot that has some intrinsic value just like constantly, you know, just even a a reading lamp, however, however somebody would look at it. But uh are you expecting to have most of your competition coming from humanoids? How do you think about that category broadly?

Because obviously a big pro a lot of people like the first demo they'll or the the promise you know a humanoid robot manufacturer might make is that oh it's going to do your laundry. First promise is dancing. It's always dancing. But then yes the second promise is laundry. That's right. Yeah. Yeah.

I mean, I think human noise are, you know, they're they're going to be there and I'm sure that there's going to be a, you know, market segment of the market that that would welcome those things into their homes.

I think uh where we are sort of the differentiator is that we are really sort of going after uh people that simply just don't want to share their living space with like robotic humans. Um and they they want to be in full control of their their own place.

And what that means is that you know in your home you have furniture, you have appliances, right? and they sort of serve a single purpose. They have a place that they belong. They don't do anything when when you don't want them to and they only activate when you want them to.

So, you're you're very much sort of in control of your space. And that's that's sort of where we come in um to try to also I mean just the cost standpoint if you were going to buy Yeah.

If you're going to buy a humanoid and let's say it costs $20,000 and you really wanted to do laundry and then you can come out with an option that's like it feels like it's a fifth of the cost because it's a fifth as many machine like parts, right? It's a lot. Yeah. Yeah. It's got to be a lot less.

So, I I would imagine that you're, you know, closer to Roomba territory than Tesla Optimus territory. And that just pulls you forward a couple. Final final boss for robotics folding uh socks feels feels pretty difficult.

Maybe you could like line them up and sit them down on top, but you need to kind of human's last job will just be like, you know, folding them. Maybe. Maybe. Have you have you cracked that problem yet? Uh we're working on it. Uh I think we can fold most articles of clothing right now.

The the main challenge we're going on right now is anything that's inverted. Uh that might require you to sort of like push the stock out the other way. Yeah, that's hard. That's a bit more challenging without fingers. So, yeah. What's the What's the price target?

You you got $50 pre-orders right now, fully refundable, but I imagine that the final price will be some multiple of that. Yeah. Yeah. So, right now we're looking at trying to get the price to under uh $2,000. Wow. Uh per per unit. Yeah, that'd be definitely doable. Would you need two?

I know in the demo video there were two arms working together. That feels like folding laundry with one hand would be uh kind of an IMO level challenge for me at least. Uh do you need two of these things? Yeah. Yeah. I mean it's doable with one, but you're just going to have to wait a little bit longer.

So if you're watching a movie or something like that, hopefully it's done by done the movie. Uh but I think the ideal setup is is probably to have two. Um but again, it depends, you know, whether your space can afford it, you know.

Um if if it's the couch, if it's a single chair, if a smaller table, you might only have room uh for one, right? So um is there a world where you sell the robot with a teleoperation package?

So there's someone somewhere else who's folding my laundry remotely and I still don't have to fold my laundry and I pay $2,000 for the install and then $10 an hour for someone to tell operate it or something like that. Yeah.

Well, I think I mean our goal I think is to really try to get to ship with autonomous voting out the box because one thing that we're really cautious about is like the privacy aspect. Sure.

Um we we are trying to avoid having to have people that you don't know basically now be given a set of arms and a pair of eyes into your into your private do some like blurring of the I think I I I don't see consumers being down to to know you could say that about whales though.

Those are teleoper operated and they have cameras inside like but But it's also like a Google product and Aaron Aaron's smart but not not the same level of brand but good luck. Yeah.

I guess like how how are you thinking of mitigating people's concerns of like putting robot arms right above their beds and like the you know sort of like black mirror scenario where the robots start. We saw a lot of that in the comments a lot of that in the quote tweets.

You know it happens when you set the timeline on fire. Yeah. Yeah. And I mean I think I think it's key that for these arms that you know when they're not activated they have to be and function like a regular lamp. So that that means like a hardware switch that turns them off.

The lamp hoods naturally sort of covers up the grippers and the cameras. So they act as sort of a mechanical blockage already as is. Um and you know I think you know do when when we do things like that and then we communicate clearly uh to our customers on how to actually operate these things.

Um that's when we can really set the expectations pretty clear. Um it shouldn't work when there's people on the bed. Um they should only f laundry when there's laundry on the bed. Um and yeah, it's very Yeah. The other thing I I Yeah, be curious.

It's very possible that people don't actually need to put these by their beds in the long run. You could put it somewhere else in the room as well if you don't want to have the the laundry room. Yeah, basically. There you go. Lots of options. Yeah. Well, I'm excited to see how things Good luck with the next phase.

Congrats on the launch. We'll talk to you soon. Have a good one. Good stuff. And if you're looking for top tier cleaning and 24/7 concier service, go to wander, find your happy place, book a wander with inspiring views, hotel great amenities, dreamy beds.

In the future, there'll be a robotic lamp that folds your laundry in every wander. Who knows? It's possible. It's a vacation home, but better, folks. And in the new in the world of uh robots that could potentially fold your laundry, Luke Metro has been traded from Android to skilled AI.

Modern AI is confined to the digital world. At skilled AI, we are building AGI for the real world unconstrained by robot type or task. A single omnibody brain. Today we are sharing our journey. And um Luke Metro says some personnel news. I work on this now. I'm super excited about the future of robot AGI.

Well, no, he messed up. He said some personal news. This feels more like professional news or personel. Personnel news. Yeah, this is personnel news. This is not personal news. Yes, this is cool. Uh, big vote of confidence for skilled AI.

I'm sure there's a hundred companies in the Gundo alone that would have wanted to pick up the legend Luke Metro. Metro. Absolutely number one on the timeline and uh a great engineer as well. We we will close it out with this post from Alexis O'hane because we got to get to New York soon.

Let's go to Tyler Cosgrove though. He has a little bit of an update for us. Let's hear it. What? This is about something else. Okay, tell me. Um I have some unfortunate news. What happened in Bloomberg? Luke Farador got hit with a hit piece. Wait, what? Yeah. No way. Why? Might have to cover this tomorrow.

Okay, we'll cover it tomorrow. We're going to be in New York. We'll protest. We'll get to the bottom of it. We might have to go to Washington DC and defend our boy Luke Farador. One of the best ever. Do it. A scrolls enjoyer being attacked. unfortable. They the I'm I pulled up the image. He looks incredibly sick.

So, it might it might be one of those things that just builds his aura. You never know. I think I think they're they're Yeah, he's a farming business week. Let's see. Well, we'll dig into that tomorrow. But we will close out with this post about Figma because we're going to New York to hang out with the Figma crew.

Uh, Alexis Ohania says, "Sure, I've seated dozens of billions of dollars, billiondoll companies, but I absolutely have my embarrassing miss list. Here are my Figma notes from 2016. We passed. " Uh, congrats Dylan Field and the entire Figma team. Y'all absolutely made something people love.

And his notes say, "Figma, why now? " WebGL co-founder was uh Elf on the first was on Elf, the first people to make WebGL demos. Avary had a great team, good software, but tech wasn't there. Lack of focus. I actually knew someone who worked at Avary.

It was kind of a a Photoshop competitor for a while and then they served a uh an SDK that you could basically bake Instagram filters into your app. If you wanted to have that as a feature in your app, you could pay Aviary to to offer that service to you. Um they launched two months ago.

Uh four out of seven users of seven days in the week. How many people log on four days a week? 600 now a thousand. three uh three three days per seven users or 3,000 weekly active users at 9,000 80,000 signups. This was at a time when Sketch was completely dominant. Yes, that's that's right.

So mostly word of mouth, no paid content definitely drives a lot. Interested in learning better why someone signs up and invites collaborators primarily designers using this but sees engineers, PMs and marketing people who are in the orbit being absolutely correct. Absolutely correct.

A decade early but Dylan knew that in the long term engineers would be using it, PMs would be using Figma and marketing people would be using Figma. This was 2016. 2016. Uh it's crazy. It's crazy this if if you had this this a a team with this competency Yeah. and this kind of early traction. Yeah. Today. Yeah.

You'd probably get a series A done. 100%. And this was a this was a seed pass. Crazy. Crazy. So, not clear about pricing strategy yet, but excited to get started. The pricing strategy is fantastic. Obviously, they want to have people paying. Yeah. Users say they'll pay right now. Uh 40% of users are US.

60% international estimates are $25 per user per team per month. That's ended up also being I think pretty much on point. Yeah. Amazing. Uh we will close out with one last post from Joe Weisenthal. He says, "I'm still flabbergasted that people talk about pursuing their white whale in a proud manner.

Setting aside the risks the book may also makes it clear this that's immoral. " Uh well, I have a white whale. My white whale it would be doing a show in person, a podcast in person in New York with Joe Weisenthal. Yeah. I think it's possible. It might destroy me.

It might destroy me, but I'm willing to risk it all to catch my white whale, which is a an inerson show with Joe. He's been on our show multiple times. I've been on his show, but we've never done it in person. Is it possible? Anything's possible. Anything's possible. We will catch the white. Folks, we New York City.

We're going to hop on a plane. And our production team is like, "Hang up. Hang up the phone. They really want us to go off. Going off, you know. Let's keep going a little bit longer. He's actually trying to unplug all the cameras. They're going to shut us off. But we will have a regular show tomorrow.

We're figuring out the exact.