#23: The newfangled vacuum salesman

All right, this is going to be fun when we're talking robotic vacuums and floor cleaners. And I have a thousand questions for you. First of all, who are you?

I'm Mehul. I'm a co founder of a company called Matic Robots. We're building fully autonomous indoor robots. Think of it as a Tesla FSD for the indoor world. And we're starting that with a floor cleaning robot. Because if you think about it, this is the only robot accepted in our home.

Yeah.

Last year, 18.5 million robot vacuums were sold. Now, you know, I haven't met anyone who says, I love my robot vacuum. No. Yet they are selling.

I've had a bunch of them.

Yes.

And I mean, let's start right? I have real expensive rugs with tassels that it gets caught on. Let's start with that. I have a lot of computers around, you know, with a lot of cords. They get caught on that.

Yeah.

Now my friend test these things out and he has cats, right? And they'll shit on the floor and then it'll roll through this shit and track it all through the house and make a big mess.

That's exactly right. Those are all the problems we experience ourselves. I had a golden retriever and our joke was that he sheds twice a year, six months each, and we're like, okay, there has to be a product that can get all his hair and not actually have to. I don't want to flip it and cut it and all those things out there. So can we not do it? Can we also make sure that it doesn't chew my kids logos or wires, you know, and wires. It's funny, a lot of people on the Internet these days are worried about robots rising and killer robots. And our joke internally is just throw wires in front of them and they'll be fine. None of the robots can see wires at the moment. That's what we wanted to fix. And we started with this idea that look, it's first Roombus came out in 2002. Professor Brooks. Rodney Brooks and the team did amazing job of bringing Inventing Robots vacuums as a category. Thinking through the use.

Now, those didn't have cameras, right? Those had. They didn't even have lidar.

They didn't even have anything. They were.

They were just like a bump sensor, didn't they?

Yeah, exactly. It's like a blind robot where you just kind of walk around and keep bumping and if you bump, you say, oh, I'm there. And then the idea was that if it bumps enough, it will Cover. But then it doesn't know where the cords are. It doesn't know where things are. So very quickly we realized that every row of one of these robots got stuck all the time. And it required a lot of attention. And sometimes it went over dog poop and cat vomit and all kinds of things. Then they evolved and eventually a company called Neato, which did an amazing job, they also started out of Silicon Valley. They'll put a single pixel.

I got one of those somewhere in the garage.

Yeah, exactly. And those guys did a really good job. But single pixel lidar is basically still having a blindfold around your eyes, but you're just pointing a hand out and say if I touch anything, I shouldn't bump into it. But if it's lower than my hand or lower than my height, it will still do it. And that still doesn't allow it to see where the wires are, where the rock tassels are. And if you're constantly vacuuming, it doesn't work. The other part about circular robots is if you flip them, you realize that actual cleaning is 2 inches away from the wheels.

Yeah.

So sides don't get clean, corners don't get clean. So they don't even do as good of a job of a cleaning. And then the bin sizes are so small, and every time you take it out, the dirt plume kind of blows at you. And with our golden retriever, it got full super fast. Within like 200 square feet. Like half a room. It will get full, but it would just keep going. And then we would have this rolled up, a hair left on the floor, which it couldn't pick it up. So there were all these issues that were there. And then one of the first robot vacuums you, you talked about expensive ones. So one of the first robot vacuums I tried was the 360i Dyson robotic vacuum back in 2016. I pre ordered it for thousand dollars. It came, it got onto one of our nicer, nice rugs. And Dysons have amazing suction. That's their ammo. But it got onto a rug and got stuck. When you have too much suction, it pulls off a rug. And then it kept hiding for 45 minutes without moving. And rock shed, so the entire patch of rug was gone.

These rugs, we just had them cleaned and they're $15,000 a piece. These are from Iran. Right. They're, they're handmade silk rugs. And if it destroys this rug, my wife might divorce me.

The dirty secret, like if you, if you, you know, back in 60s used to have these manual sweepers. And we still have Lego toys where you just kind of drive and it picks up all the Legos and they don't have motors. So actually we realize that the analogy is that imagine dusty car drive it. Even if you drive it at 100 miles per hour, it's not gonna get cleaned. Yeah, you have to nudge it. So in the same exact way. It's not just about the suction when it comes to floor cleaning. It's really about the brush roll. And we learned all these things. So actually if you have a really high suction power in your vacuum, there's a good chance it's ruining your rug. It's balding your lug without you realizing it because it just sucks up all the hairs with too much of a power. So what you really need is an amazing brush roll. So we design the whole thing from the scratch. Have a first of its kind tangle free brush roll which we'll talk about it. We've also added a self, self cleaning mopping system. Again, first of its kind. No one has this. So we don't just spray the water and drag dirt with us. As you wipe it. It actually cleans and picks up everything there. Then as we're looking into it, realize that a lot of people have a thick pile rocks and hard surfaces. So mopping is a critical feature. Well, what do you do with dirty water? Yeah, right. And a lot of the solutions like mop facts or even latest robots have these big docks where there is a dirty water tank and there is a clean water tank. Yeah. But this dirty water tank gets actually really smelly, moldy. If you leave it alone for a week, you'll probably have a bacteria inside it. Where do you throw it? It's not necessarily good to throw it in a sink or a toilet. So we actually built first of its kind HEPA bags which is both collectible dry as well as wet waste. And we also have a version of diaper salt in it so it turns into a gel so it doesn't slosh around and leak. But we also added antimicrobial powder so it kills all the bacteria. Never smells. Wow. And then when the bag is full, it just tells you.

That's some of the questions. My friend. My friend has a bunch of these robots and was asking me questions like how do you clean it?

Yes.

Because it's a big pain in the ass.

It's a big pain. So it just. That's the whole point that it's a very simple system. Tells you when the bag is full, you throw it away. And then we also realized that, look, we are humans before forget things. We absolutely forget how to, how to change filters in our vacuums. You know, raise your hand if you, if you've changed filters in your vacuum, everyone forgets it.

My coffee maker has a filter. I'm supposed to change. Have I changed? I think I've changed it a few times.

There you go. So we realized that, you know, people forget to change. So we actually built in a HEPA filter right in the bag. So every time you change a bag, which is it can last for a month.

You're really a great product thinker, man. How much. How long have you been working on this thing?

We've been working on this thing for about four years now.

And how many robots have you gone through?

We've gone through a lot of them.

But do you have like a test house?

Yeah, we have a test house, but our internal barometer was that if I can get my wife to fall in love with it, we're there. Or if my founder, Navneet has a test. Exactly.

I used to work in a consumer electronics store. Right, yeah, that was a must. If you wanted, if you wanted to buy a refrigerator, the spouse had to pass it, right?

Yeah, exactly. So had to.

You have my telescope. Your spouse had to prove it.

Exactly, exactly. So. So to go back to this idea, you know, we started with this point of view that all these robots haven't really improved. They're adding sensors to it, they're adding cameras to it, but they have a ant like vantage point, which is not necessarily good to see what's in front of it. It doesn't.

Occlusions are a problem.

Occlusions. What is behind it, whether it's a shoes, whether it's something else, or whether it's a threshold, if you just have a tiny vantage point, you can't really tell what's going on. Also, they're very low to the ground, so they can't go over shag rugs, they can't go over the threshold. So that. And a lot of people have shag rugs and thresholds now. So redesigned whole mobility system. And then a lot of these robots have sensors at the bottom and sensors. Cameras are also low. So over time, if you're cleaning, they get dirty, they get caked with dirt. And as they get caked with dirt, we have to remember how to change it. If you don't remember it, then it falls over and it starts doing all kinds of messy things. So we consistently heard from Users that no matter how advanced robots they bought, within three months, they started doing all kinds of weird things. So we just reinvented the whole thing from scratch, just using computer vision. That's our background. My co founder.

So you only have cameras?

We only have cameras. It's just fine.

You like Tesla. That's why you're like the Tesla.

That's right. That's exactly right. Tesla for indoor requirement. All the processing happens on the device. Even when you send information to the app, it only comes through local WI fi. There is no cloud trip. Your data actually never leaves your home. So it's private, safe first. And because it's.

My wife's walking around naked and she gets picked up by the camera. Nothing.

No, no privacy issues. It's never going anywhere there. It just stays in.

That's important that we're introducing a lot more cameras into the home.

Exactly, exactly.

Wait until you figure out Mark Zuckerberg has a bunch of them on a headset. Yeah, it's like that's a new problem. Right. Or even Apple has like this Vision Pro 19 camera. I don't know how many cameras. It's like a lot of cameras that I shoot videos with it.

It's great. So that was the thing that. Look, in indoor space, privacy is an issue. Latency is an issue. If you're, you know, you don't want a robot that says, oops, I fell down the stairs because Internet was weak or cloud connection was weak. Right. Also, indoor space changes a lot. So even though we built this system, what we have to do is continuously map because self driving cars and all cars have this advantage that they have a Google Maps and gps.

In fact, the kids just brought these backpacks home, right? Yeah. There's a kid. Hey, kid. And they have. And they have straps. Right. And if something rolls over that it gets caught.

Starts pulling out. It also has those things. So pull the threshold. So we kind of take care of all of that stuff. And because we use vision based system, it's. It sees everything. And the idea is that if it can see it, it can clean it. If it can see it, it can avoid it. If it can see it, it can maybe even manipulate it.

All my son is doing is stimming. Hey, Mulan.

Yes.

Can you leave us alone for a few minutes?

I can pee up in there.

Okay, but go in your bedroom.

Thanks.

I'm doing a video.

All right. So that's the idea behind it. That look a vision based system. So what we do is a robot comes out just like us. They Will first explore your home. Build a full blown Google street view like a map of your home. And because everything happens on the device that's continuous. So anytime the furniture moves, anytime a table moves, even if there is a live thing, it knows. Even if there is a wire in the middle, it knows. And then it remembers that I didn't clean this area last time. So it actually comes back and do it. And that was one of the biggest pieces thing that we thought about as well, that why don't these robots have memories?

Yeah.

If you use Roomba or any of these devices, they get off the dock, they clean. If you if 15 minutes into its cycle, if you pick it up, put it back on the dock and start again, it will clean the same stupid area for another 15 minutes. It doesn't remember that I just cleaned it.

Yeah.

So can we give it an ability to remember that what it cleaned recognize and if it sees like a Google street view, like a map, if it has that map, then it can know that here's the kitchen, here's the dining area, here's the living room. Then it can also, you can also tell it exactly where to clean that. I just want this kitchen area to.

Be right here underneath the cereal. The kids come in and get, you know, it gets really dirty right around here and then you track it through the house, you know.

Exactly. And when you have this, you don't want to say hey robot, go home, clean my entire home. You just want to say just clean this area and go back on the dock.

Yeah.

So giving being able to give that instructions to a robot and telling it exactly when to clean is the priority. That hey, get off the dock and don't clean. Besides my dog, actually that's like us taking the vacuum out of the closet and cleaning besides the closet. That doesn't make sense. We wanted to go and clean the kitchen area or dining area and eventually learned that after we eat meals, kitchen and dining area is going to be dirty. So most of the families get up and start cleaning. So why doesn't it just come and join? You know, why doesn't it clean near the backyard and dining area, Backyard doors and doggy doors 10 times a day because that's where kids and dogs go in and out. All these things just felt like a very low hanging fruit. To use one of Elon's favorite quotes. Like no one had thought to the product for a long, long time from first principles. Yeah. And we just kind of took a step back and we said, okay, if we had to build robot vacuum or floor cleaning robot today, what would it look like? How would you even think about it? So we threw all the pre, preconceived notions of that entire system out, and we said that, look, we realized that a lot of families, they don't even care about cleaning underneath the couch or bathroom or any of that stuff. They just care about cleaning kitchen area, dining area, and living room.

When you're walking around here and somebod spilled some cereal on the floor, it's crunchy underneath your feet, and then you're tracking stuff all over the place.

Right?

So clean this.

Right, Exactly.

I never walk in the bedroom. You know, that's exactly right.

Yes.

You know, you have to clean and clean other places, but, you know, I don't have a cat. My friend has cat.

Which is. The point is to give users a control. Point is to say, hey, robot, learn users preferences. Each home is different. Each home has a different style. And what we want is not necessarily robot. Like, if you had a cleaner and they came and cleaned on their own in a weird way that you didn't like, you'll probably fire them eventually. They adopt your home in the same way. Can we tell robot what to clean, when to clean, how to clean? And it just does it for you. That's really about users, because for each home, cleanliness is in the eye of the beholder.

One reason I don't use them is the noise, because I do podcasts. I'm on Twitter Spaces.

I'm.

My wife is doing a zoom call. She does a lot of zoom calls, right? And she's sitting here in the. In the living room or the dining room. And.

Yeah, yeah.

And if the robot starts up when she's giving a presentation, that's not good.

Right? You know, that was. And that was one of the worries we actually, when we started, we're like, okay, all vacuums has to be noisy. That's why they're noisy. So. So maybe it's the physics that's stopping us. So we started digging into it, and turns out we realized that a lot of these vacuum makers use noise as a metaphor for efficacy. So the louder the vacuum, the higher suction it must have. So more effective it must clean. It was the same metaphor that we used to have for car that the louder the car, the faster it must be. Until Tesla came and changed it Turns out most of the cleaning. You know, the analogy is that if you have a dusty car, you, no matter how fast you drive, unless you nudge the dirt on it, it's not gonna get clean. In the same exact way, the most effective part of the vacuum is actually the brush over. It has to nudge all the dirt and then pick it up. So we spend time figuring out how to build a really, really quiet vacuum. And we have the quietest vacuuming robot in the world. It cleans at 555dba. That's a decibel pressure meter. Yeah, that's about seven. And this is a log scale, so it's at seven decibel meters. Quieter than a human conversation.

Yeah.

And typical Robot vacuums around 65. If you use Dyson or any of the handheld devices, they're around 75 to 85.

Yeah.

And so that means you can not even have a conversation. So that was one of the biggest things that if you're going to have a robot that is constantly cleaning, can you do it in a way so it's so quiet that it just works? And one of the favorite features we are enabling now is that it actually is able to map and see and everything at night. So we call it a Zenrise feature that can you wake up to a clean home where it just cleans at night? So if it's quiet, you don't even have to worry about it. It can take two hours, but eventually at night, it can just clean your wall.

That's interesting because the dishwasher is running at night and you don't notice it. Well, you notice it if you're here.

If you're here but not in the.

Bedroom, you can hear it. I can't hear it in my bedroom.

That's exactly right. So dishwashers are usually around 52 DBAs. So we're just about 3 DB higher than them, actually. Okay. So. So we're very, very close to it. And if it's mopping, it's actually quieter than a dishwasher. So we have a first of its kind, self cleaning mopping system that collects all that. That has a microfiber brush roll that does a really. So the way it works is instead of spraying the water and having a pad that kind of drags all the dirt with it, we have a brush roll that we used and water. Or you can add vinegar in it, or you can add bona soap in it, which is not forming. It drips onto the brush roll, makes it wet, then we clean it, and as brush roll spins, we actually have a wringer that takes all the dirt and dirty water out and puts it into a bag. And this bag is where we spend quite a bit of time thinking about it.

Yeah, because it gets messy, it's smelly. You told me it can get bacteria in there.

Yes, exactly. All this dirty water thing. Like, you know, if you want to try, just mop, take a dirty water, just mop your floor, collect that dirty water and put it in the sink and see what happens. In three days it will have bacteria, will get moldy, it will get smelly. It's just nasty. Where do you do this? So instead what we have is we take this dirt and dirty water and dry waste and collect all of this into one bag, which is what we refer to as a HEPA bags. And that bag has a version of our diaper salt on it. We literally ask, hey, what is the dirtiest part thing we collect in our home? Well, it turns out that's usually baby poo and all that stuff. Yeah, so. And diapers are amazing at storing it and not letting it leak out. So that's why we used a version of diaper salts. So that turns into a gel. And we also have antimicrobial and antibacterial powder in it that kills all the microbes so it doesn't get smelly, moldy when the bag is full. It just tells you bag is full. And with each bag we also realize that people forget to change their filters. No one changes filters in their vacuum. Raise your hand if you change it. Light things. Right.

So I have a coffee maker that has a filter. I never change the filter in that I'm supposed to.

Yeah. And because we are humans, we forget things that we have busy lives, we don't have time to think about this thing. So we just kind of built HEPA filter right into the bag. So every time you change the bag every month, it just has a new filter you don't have to worry about. So air stays clean, not only keeps your clean floor, but keeps your air clean as well. And that was one of the thought process that, hey, can we just make it simple mess free cleaning system where you don't want to try to empty the bin and have entire dirt plume blow at you or a bunch of things fall down the thing. So, so that's where the bag is. A really bag is something we really spend time building and designing so that for users it's just a very simple thing. It tells you when the bag is full. And then, you know, we also designed it so that if you're just vacuuming, even with pet hair, it will last about 10 days to 11 month. If you're vacuuming and mopping lasts about a week to One month as well. So it's actually considerable size. And even how does it tell you.

That it needs to be changed?

It just tells you there is a display right on the robots unlike and that part of the puzzle that you know, not only we added a camera based system which is slightly higher vantage point so you can see top down, but we added a display on it. So any given time it tells you, hey, I am about to clean a carpet or I'm about to clean your hard surface or I'm going to the kitchen. It gives you all these information so you don't have to pick up your phone and say what's going on? And then it just sends you alert on the phone or on the display saying that hey, my bag is full, can you change it? And when the bag is full, it also remembers where your trash can is and just comes by. So all you have to do is just pick it up and throw it to the trash. And that's automatic. That's automatic.

So it figures out where my trash can.

Yes, it can detect because it has a camera. It has a camera. So just like any other system, because it has a camera, it also, you can also, you know, and if your camera, your trash can is hidden behind cabinet, you can just point a location and it will remember so you can geotag it. And that's really what we built that with this street view map. We have given it a GPS of the indoor world. So you can geotag it and it will just come. It will also remember where your sink is. So when it needs the water for it to continue mopping, it just comes by there as well. So things very, very much, very simple. And then we're enabling very much a natural language instructions and gesture detection on top of it. That's really our background.

You're going to put an LLM in this, aren't you?

That's right.

Hey robot, there's a mess in the kitchen.

That's exactly right. We'll show you all of that today. Why do we have just a chat interface in GPT GPT3, you know, or chat GPD? Why can't we ever be able to just tell our robot that hey, I want to wake up to a clean home and it just figures everything out. That is the vision. That's where we're going. And we're just reinventing everything from scratch to actually build a fully autonomous level 5 indoor robot. And we can do that because we are really not a robots guy, we're actually computer vision guy. That's our background as I mentioned my co founder and CEO. His PhD Navneet Dalal. His PhD is a classical piece of computer, classical foundation layer of the classical computer vision. It's called histogram of oriented gradients. It has more than 45,000 citations. It won Test of Time award Longwood, his gin's prize at CVPR in 2015 for 10 years of impact in computer vision.

That's a big computer vision show.

That's the biggest one. Right. And then so we've been computer and working on computer vision oriented products since 2005. We both of us were early employees at this startup called RIA which became like.com and that was acquired by Google and we did all the, all the work you see in terms of face, photo tagging or any computer vision application. We've tried it at RIA and we were just sort of 10 years too early. We used to build Google Photos.

Could this start becoming a security camera?

Oh absolutely.

There's somebody in the house that's not supposed to be in the house.

Yes, yes. Security camera, monitoring camera, monitor your pets and kids.

In fact, because it's a full camera.

Like in fact, a lot of users we've demoed, we've done about 50 beta tests, trials. Now we've downloaded to about 150 people additional. And a lot of people have come to me that hey, I have a parents who are living in Florida. I don't want them to use this robot. I will control it remotely, I will clean it for them. But you have a camera, so can you use that camera to keep, can I use that camera to keep an eye on them? I just want to know and if it's a robot that is actually doing a job of cleaning, then it doesn't even feel like a big brother. Then it feels like, okay, there is.

A thing from down here. What does it see? So does it see you?

So, so we have.

Could it see your face?

It can. So what we have is, we have a, you'll see the robot has a, what we refer to as a black border.

Then it could watch grandma and see if she's taking her pills. Your vacuum is gonna out you.

Yeah. And fall detection. Just make sure they're okay. Just see that they're sleeping and all kinds of information and just kind of give you a summary at the end of the day as a kid that hey, here's the 5 second or 10 second sort of a time lapse of what?

Let's talk about the future because I want a little humanoid robot.

Yes, yes.

Is this where this is going where you're going to have a variety of different robots to do things in the house.

Precisely. That's why we're building the way. What we realized in 2017 is that robots are not blocked by hardware. Hardware is hard, but first humanoid was built by Honda in 2000, which was Honda smo. Yeah, it was amazing. It was expensive. It was amazing. Since then, cost of hardware has come down. A lot of people are building.

There's an icon in your company you.

Can 3D print your own robot.

Yes, yes.

They're amazing guys.

And 8,000 bucks isn't that, you know, I mean, that sounds like a lot of money for an average household.

That's exactly.

But I could afford to put that on a credit card.

Absolutely right.

So it would cause an argument and I could spend it.

And. And the lower. Actually, the smaller one is pretty brilliant on those guys because you don't want that robot to fall on your cats and kids. Right. So lower to the ground is better in our case. What we realize is that, look, even if you have a humanoid, it's the brain that is missing. Because typical robot is a Christmas tree of sensors. Just because you have a 20 eyes doesn't mean you know actually how to navigate. So can you. And a lot of these demos, you will see that they would have accepted expensive depth sensors, expensive lidars, and yet they just keep bumping into things. And that's what we wanted to prevent that. Can we give it a brain? So the analogy we use that we are actually, you know, human child. What does human child first learn? They learn to crawl, they learn to walk, and they actually learn to navigate the space first.

Yeah.

Then we don't even give them a very sharp items. We give them soft toys initially. Then we give them things that are unbreakable. It's about eight or nine years before we trust them with scissors and knives. Right. So in the same exact way, can we build a robot that first just learns to navigate home and go from point A to point B precisely 10 out of 10 times, regardless of the changes in the environment, regardless of the changes in the structure of the home, whether you move table or whether you move couch, can it learn on its own and still get to the same place?

Did you have a cord to your Macintosh?

Don't. Yeah, don't pull the cord. Don't topple over cord. Avoid the bag or things. Kids learn that initially in the same exact way. Robot can learn that, and then you can start teaching it how to manipulate things, how to put things back together. So it's maybe, you know, kids are maybe around 8 before they can start putting their own toys back where they belong. So in the same exact way, robot is first learning to do it, but can we teach it to eventually give it a hands and teach it to say, okay, put all the things back together, put organize the home. That's where we are heading. So it's just evolution and we just think of it as a step, one of the three steps.

One reason I wanted to meet you because on your exit count, you put a 3D scan of the house.

Yes, yes.

Oh, they're building a slam map.

That's really cool. Yes, yes.

That's how you solve all my problems with the fringes and all that stuff. Right, right. But I was like, oh, that's cool. Because then, you know, then it sees it. Well, you could see five years from now when the humanoids really get cheap. You're gonna put this in. Are you gonna put a multimodal model in here?

Yes. So that you can deal with the camera. We are actually already doing it. So what we are doing.

The robot might do automatic shopping soon. It's like you're out of milk.

What? So. So, you know, they could be here.

Watching us go into the refrigerator.

That's right. That's exactly right. So we actually, you know, this is why privacy is so important and that's why doing it on devices is so important. So edge device computes where you can run it slowly and then latency. You saw all the issues with latency with some of the recent products. So even with the robot, even visual latency has to be instantaneous, which is if you just open the door, robot shouldn't be able to run into it right away, even if it's one feet. Right. Don't run into kids, don't run into pets, don't scare them. These are all the challenges. And VLMs are great because you can extract information out of it. So what we do is we already started using some of the open source VLMs, like Dyno and Clip and extracting some of this information for semantic understanding. And what we do is we build a map. So you saw that 3D scan of the map. What's interesting is that the way we do it is each voxel, so it's a voxel, it's like a cube, but pixel, but a cube for a three dimensional word. So each voxel knows that it belongs to a trash can, or it belongs to cabinet, or it belongs to floor or chair or wire. And that's how we actually avoid things because we use that information to give it a semantic understanding so it knows what is what, and that's how it actually understands the unity.

Who said every voxel needs a LLM?

That's right. That is exactly right.

You need to be able to talk to your refrigerator. Refrigerator's a million voxels.

Right, exactly. So be able to understand and then put it all together and say, okay, this is a shape. This is the thing. And now, you know, once you have a voxel understanding, it's very interesting because it does. It knows its shoes. It may not know its physics property, but you can potentially ask that, hey, is this breakable? Is this movable? That information does exist in reality.

There's a difference between these and these.

Yeah, exactly. Exactly. And if it. And so you need to.

If you're a robot, you know, you can treat this a little rougher than you can treat.

That's precisely right. That's exactly right. So that's where we. Instead of asking questions, because. So the best way to do it is even though we are building this 3D scan of the world that allows it to navigate precisely using slam, as you mentioned. But if you just have slam, people.

Who don't know the term SLAM is simultaneous location and mapping. That was designed for the Mars Rover.

That's right. Exactly.

Because we couldn't drive it from Earth, Right?

Yes.

The latency is, what, a minute or something? Right.

Actually, so it has to.

If it's headed toward a cliff, it has to know.

Okay, that's right.

That way I gotta turn.

That's right. That's exactly right. And actually, a lot of that Roomba team is Mars Rover team. Ex Mars Rover Team. They're the guys who actually brought those robots. So it's amazing how the technology works. But. Yeah. So slam. SLAM is needed. But SLAM alone is like giving. Like, dogs and cats can navigate. Perfect. They don't bump into our furniture. They don't get tangled up in wires. Right. Yeah. But if you tell it fast. And they could do it fast.

I have a crazy relative with a crazy dive who runs around, like, in circles really fast. I haven't seen a robot do that. Not in the house. With jumping over people. No, no, no, no. You know, Boston Dynamics ain't quite there yet.

Boston Dynamics. Yeah. Yeah. The analogy I was trying to make is, you know, your dog can navigate, but you can't tell your dog that, hey, go sit on the couch. Yeah. They don't understand that information or say, go to the bathroom. Get me something not that good.

Yeah.

Or go to the front dog, give.

Me a Diet Coke. No, no, no. That's why I need a little robot.

So, so that's where. That's where VLMs and some of the advances in AI is amazing because now we can give it a semantic understanding. We can give it a thing and then it can do all kinds of of stuff and say, hey Matic, go clean kitchen. Hey, Matic, go clean the kids near the kids toys. And it understands all this information. Hey Matic, push that shoes near the shoe rack. And you can do all that information even with the robot that we have built today. And that's the idea. And eventually we expect robots to have a very similar chat interface or natural language instructions. And we've built it too. But then it goes to the next layer, which is having a vision means that it can understand gestures.

Oh, so you can gesture to it. Hey, go away. Or hey, hit that spot over in the corner.

That's exactly right. That clean this. And then this was one of my use case because gestures. Yeah, yeah, this was one of my use case because I used to have a golden retriever. Golden retriever would have this hair, which are very hard for sometimes us to see because it's golden, blends into the floor. But if I see it, why can't I tell robot that hey, just clean here. Right here. This is where the stain is or the chair is. Instead of just cleaning the whole area and wasting 20 minutes of time. So from quietness to cleaning efficacy, to just building end user interactions to make it very good, simple for users to use. All these are part of the product we've built today.

That's awesome.

Super excited to show it to you.

Yes, I want to try it.

All right, so then you can just, if you do me an honor and lift the. Lift the robot. Excellent. Maybe come around this side. Yep, perfect.

Yeah.

And then come do me a favor, please come on to this side and press the power button for me. All right. Here for a second. Just the power button right here.

This one.

That's it. Just simple press. All right, There we go.

That's a new kind of unboxing.

Yeah. You know, if it's a robot, it has to roll out of the box. How can you build a robot that you would pick it out? It defeats the purpose. So fully autonomous means that it knows when to come out of, of your home. It says, hello. Feels like, okay, it's a friendly being belonging to home. Yeah. And. And that's what really try to do. So this is a, this is a Full and onboarding experience that we have here. And I will just put this dock for now. Let's just do it here for a second.

Okay.

Okay, perfect. So it will know and so it.

Figures out where the dock. Is it automatically around the corner?

Yes, yes. And that's actually automatically when you're using.

Does it have a Bluetooth beacon on it?

There's nothing there. It's pure visual.

It'll just figure it out.

It will just figure it out. And you can move wherever in your you want. It just figures it out based on, based on the, the way it looks and the visual recognition. The same exact thing.

Cool.

And that was actually one of the critical part that a lot of these indoor robots, when they build a. When they build a map, they build a map starting from where the dock is. Those maps are relative. They only robot only knows where it is from where the dock is. That's like coming to your home and saying, I can only navigate my home if I come through front door, but if I came through side door or back door, I don't know. Versus what we do is build a full blown visual absolute map. That means you can move around your dog, you can move around everything. And it just figures it out.

So what do we need to do now? Get my iPhone.

So first let me just quickly give you a quick understanding on the, on the, on the industrial design piece of the puzzle. So a lot of people ask us, hey, you know, why is your robot tall? Or why did you decide this way? And one of the first thing is we said is that let's do a quick thought experiment. Imagine sending a Roomba to 60s, teleporting it to 60s and then let me know if you're still recording.

Yeah, I am.

Okay. So imagine sending it to 60s, sending Roombar to 60s and asking people what is it? Yeah. Most likely they wouldn't be able to guess that it's a robot or a vacuum.

Yeah.

So we wanted to build a robot that looked like a robot. Then we give it a cleaning head. Just like in upright vacuums. It can clean sides and corners very well. And it's clear that this is a cleaning robot. This is not a robot built for. It's not a Lexa on wheels too, to be honest. And we completely redesigned the entire mobility system because roombas were built for a roll 25 years ago, was wall to wall carpet. Now we have thick pile rugs and wires and thresholds. So we had this big wheels. It's very much like an SUV that allows robot to climb over about a 1 inch of threshold shag rugs, all kinds of surfaces. Then we build a cleaning head that goes up and down. It actuates so there is a clearance, 1 inch of clearance here. So it can go over various strains. But this also means that cleaning head actually automatically adjusts to the thickness of the rug. So if you have a really thick pile of rugs with things, it's not going to dig into it and start sucking up everything. It actually automatically detects what the rough thickness is and adjust the brush roll speed, suction speed, all kinds of information to just do right amount of cleaning. Then what we did is you were, we were earlier talking about the robot, you know, having a lot of hair stuck. So this front part of the vacuum is the, is the brush roll. This is first of its kind, tangle free brush roll. We've completely redesigned it from scratch. It does a really good job of cleaning on the hard surfaces, really good job of cleaning the pet hair rugs. And it's very simple to take out clean. You can wash it if you want. But the most important part is no hair ever gets tangled up in it. There isn't. You're never gonna have to take it out and cut it. So we've designed it so. And the way it works is the simple part is that it just tapers off a little bit. This, it just slightly slides. All the hair actually just slides to the either side of it.

Here, let me get, let me get a close up of that.

Yeah, so people can see.

Yeah, yeah.

That's really well designed and it's very easy to do.

You're a brilliant dude, man. I mean, how many people worked on this?

We're about a 60 people team at the moment. 60, 60, which is for the entire robot. And a perception system we've built is actually relatively tiny. I think other companies are usually around 200, 300 people if they're doing hardware and software system. So it's relatively tiny, but we spend quite a bit of time. Then the back part, you'll notice here that this cleaning, it actually tapers up a little bit. So the reason is because when it vacuums, it doesn't mop. And this backside is literally first of its kind self cleaning mopping system with a microfiber brush roll. Okay. And the way it works is that we have a water tank in here. So if you open up the lid, here's the water tank. In this tank we can put in water, we can put in vinegar, we can put in bona soap, any sort of non foamy soap. Goes in there, you put it, you fill it up. It tells you when the water tank is empty. You fill it up, you put it back in, press it in. Now when you start mopping now, do.

Most people just use water or people just use water?

We've had people who use water. We have people who put a little bit of vinegar in it, where people who use some sort of a soap. Soap combination with the, as I said, bona soap. So we have all kinds of people. It just depends on preference. But water itself is fine as well.

Yeah, this floor, again, is clean with water.

Water, exactly. So water is fine. You don't need to use just simple stuff. So, so, and. And some people have a really, like a real word. So sometimes they use vehicle soap. So that's because of water might damage the thing. So that's where we built a system where you can put all kinds of stuff into it. And then what it does is robot, when you tell it to mop, robot takes water out of the tank, drips it onto this, makes this mop roll that. Then this mop roll spins. And actually. So instead of throwing a water onto the floor and trying to kind of drag the mop, or it only drips it on here, then as it cleans. The best part is that as it cleans it as it spins, we have a ringer in there that takes all the dirt and dirty water out, squeezes back in, and then puts it. Puts it up into the. This. This HEPA bag. Yeah, and this HEPA bag, you'll see it actually has a dirt because we've been cleaning and demoing today, said this dirt, it has water in it. And then we also have this version of our diaper salt, this white powder, diaper salt, as well as the. As well as the antimicrobial powder that turns into a gel. And you see that some of the water is already dried up along those lines. And when the bag is full, it just tells you where the bag is full. And when the bag is heavy, it also automatically locks. So you don't. It does. The content doesn't spill out. And this top part is the HEPA filter. So every single time you replace, the bag, filter gets replaced as well. And you as a user never have to worry about doing any of those work. And you'll see that on the bag, it says bag is missing. So on the display right here, it says bag is missing. Right. As I put it back in, it will say lid open. So as a user, you never have to worry about it. Robot just tells you exactly what's going on.

It knows what it's on.

And it also has already figured out that this is a hard floor that we're sitting on a hard floor. Right here on the display, it's actually telling you. And then, by the way, when it vacuums, when it vacuums, it has a cleaning head like this, but when it mops, it actually tilts cleaning head this way. So when it mops, the vacuum doesn't touch and when it vacuums, mop roll doesn't touch. So you can go through all kinds of stuff. Now that's the cleaning system part of it. Then we also have a. This black border. What we refer to internally is a black crown, the way it works. And. And this is the entire sensor stack now there because it's about one feet high. Cameras. Yep. So there are two cameras in the front. Same exact. Two cameras in the back. Just like a human eye. There's a camera here, there's a camera at the top which sees at the ceiling. And that's how it's able to see people, pets, anything that's taller. That's how it's able to see our gestures as well. So that's what we use. We also have a 4 omnidirectional microphone, so you can give it instructions using voice. We'll do that. Array microphone, actually, just so we do. We do array ourselves. So their only direction. But we do it ourselves on the device as well. Yeah. And then these are the IR eliminators. So this is if you have a NEST camera or security camera, sometimes you see people in a gray. So that's the same exact thing that we don't see robots exist. So there's one here, there's one in the back and one in the front that illuminates the thing so it can cleans at night. And it's the quietest robot in the world. So now you can enable it to do clean at night. And if you see you're in your bedroom, you won't even hear it. And that's really the point of view of the robot that, hey, even at night time you can just turn off the lights and say, hey, robot, clean it and make sure the floor is clean before I wake up. That's really the type of thing they're building. So with that, let's actually jump onto the front part of the robot, which is the, which is the, the, the clean. The intelligence and how it builds the map. Yeah, yeah. So I'll just quickly. All right. So as the robot comes out, we'll just bring it up here. So as the robot comes out, you know, we already see our legs here. We're standing behind it has the front camera, back camera view. And like any human being, we go into new space and explain explore. To build a map on the fly. So in the same exact way I can just tell you to explore. And it starts moving. And as it moves, it literally builds a full blown 3D view. Wow. Full blown. And can zoom in.

That's the slam map.

That's exactly right. It knows precisely where it's going. It's going to move around and says, okay, I'm going to go and try to navigate the area now if I move around it. So your leg was already here and it's getting erased. If you kind of see it, this part of the area, it's getting erased. So it's dynamically capturing all the information. It will say, hey, I'm gonna go and try to figure it out. So and, and the way it works is literally it says, okay, until I see, until I've covered all the areas where there is a there where I can go, I will keep exploring. So first it comes into your home. Yeah. Builds a map. It takes about 15, 20 minutes to explore the entire home. And once it builds a map, it actually knows exactly what's going on. As you may notice, it hasn't bumped into your shoes yet. It hasn't gone over the tassels yet. And it's getting this chair, legs and all kinds of information pretty accurately. Not only that, but it actually has a semantic understanding. So it already knows that we were on hard surface, that there is a carpet right by your front door and what are the rugs? So because it knows all this information, if it's cleaning, it knows that, hey, I can never mop a rug. So as it moves on to the rug, it already figures all that information out right here. As it moves on to the rug, it knows that, hey, there is a rug here.

And it, it said, oh, there's a tassel here. I'm going to study that a little extra.

That's exactly right.

So did it really know that there's a tassel there?

It knows it. Rug and then there is a tassel and then there is also the other. It also started. So all the blue color things you see are wires actually. So it already knows that this is a bunch of wires on the floor over there and capturing that information. So that's how we figure things out. So tassels, wires, all kinds of stuff. It does it, you know, I'M not.

Gonna let you leave with this.

Right? Yeah. And you'll see like it kind of gets your pattern of the rug as well. And then the idea is that now you can just tell it to clean this rug if you really wanted to do it, that something just spilled here. You can, you can do that now. What's interesting is I'll stop it for a second just to kind of do the demo for you, but I'll stop it for a second. And this is still the beta version, so. So this particular release doesn't have a room detection yet. So for what it will do is it will turn various areas into sort of sections. Yeah. In lieu of segments. And now you can prioritize based on 1, 2, 3, that hey, clean this one first, clean this one second. Or you can say, hey, clean this one first and second and this third. So it gives you full control of when it cleans. But what's amazing and around this is now this is where we were standing. I can just tell it to say, go here. So I just long press on the map, it points the location, puts a target and then I press the button, go here. And now watch. It will actually create a Google Maps like a directions. Yeah. And we're standing in the middle, so let's see what it does.

Oh, so you told us, told it to run over our feet.

Yeah, we said, hey, there's here. So. So you're standing in the middle, there's a tassel. So it stopped. And if you kind of see it, it will start capturing all this information. So all of a sudden your legs came and it found a little space around it. Look at that.

Wow. It's like an autonomous car.

That's exactly right. So that's really what we've built, which is Tesla FSD for the indoor world. And. And now if I were to tell it to explore again, it will say, hey, I'm going to explore that area.

What kind of computers on board?

So we are actually putting Nvidia Jetson Nano GPU on it. It's one of those. Yeah.

So now I'm gonna have two of them. Yeah, Nvidia's getting rich off me.

Nvidia, among other things, they have incredible software, they have incredible support, and by the way, just recognize the dock. I don't know if you saw it or not. Yeah, yeah. So you just figured out that, hey, that's where the dock is, all this information. So. And again, it knows where the rug is and it's a full blown 3D map. That's the Whole point that it actually knows where the location is, all those things. And this one we're only building about 1ft off of the map. The reason is because our systems are still not optimized and it's a function of a compute before the production version. We're going to build three feet of the map of your home. So it's not algorithmic restricted. If we can put a thousand dollar computer on it, we can build a full blown 3D of your home inside it. So. So it's really just a function of how much compute we want on the device. And that's where it is. Yeah. So as you can see, you gotta make this affordable. Yes.

Right.

So now it's kind of exploring all your.

The retail price of this right now is what, nineteen hundred dollars for like sixteen hundred.

That's how we started. So we started collecting pre orders at. Around, around. We priced it at 795 and started collecting pre orders for 1500. But we're actually bringing the price down. So over this last six months we received a feedback from users that hey, that's really un portable. So we actually are bringing worked hard to bring the price down to 995. So starting on May 7, you'll be able to purchase and reserve this thing just for nine.

I'm keeping this one. I don't know about you.

So here's an example of it. Robots navigating. Now another beautiful part is that I'm gonna stop it again for a second and let's say I'll bring it here. We're standing on this location. So let's say, you know, killed kids spill something right here. Now what I can do is I can tap on that pencil icon and just say, hey robot, I just want you to clean this area. And I'll hit vacuum. And again it will find its way. So it says, okay, I know how to get here.

Is that this place over here?

Yeah, it's right here. It's coming right here.

So yeah, it's a little where, where the heck is it? There.

Yeah.

So it's like, oh yeah, I gotta go over and clean up the cereal.

Yep. And then it's figuring out that hey, I'm gonna do it. And now you'll see.

Okay.

And we'll lower the cleaning head.

Much quieter than my friends.

Yeah, yeah. And we are having conversation even though it's really just about a 1ft away from us, which is almost impossible for any sort of thing. And then you'll see as it cleans, it actually tells you which Area. It's cleaning. And it even gives you an ETA that, hey, it will take up me about four minutes. And to figure this, clean this properly. That's cool.

That is so cool. All right, you sold me, man. I'll pay a thousand bucks.

There you go. And we'll do it. So that's. That's the thing. The other really incredible part. So I can tell. Just off for a second. I can also say that, you know what? Just the same exact area. Let's say I'll put this as a clean mode. Pick this area, and let's just tell it to mop and say.

Okay, if you're vacuum first and then mop.

That's right. So if you tell it to do both, there is an option to select both. Then it will vacuum first and then mop so that all your floors are clean. So we'll just do it here for a second until. Until this happens, just to kind of show how it works. Yeah. And in the case of vacuuming, as I said, it first gets the water onto the brush roll, makes it wet. We call it a saturation process. Then it will lower the cleaning head just the way it did right now. And then you'll see that it actually will tilt the head and move backwards as it does. And in a second, we'll do it. Yeah.

And it shows that it's mopping.

Yep. It told you that there's mopping.

And it's not wet behind it.

Oh, you. You'll see it in a second. Now.

Now I see it.

Yeah.

Okay.

So it just hadn't adjusted before. It was trying to get into the right position.

Got it.

And now it will see. You'll see now how it tilted. And now it will. Behind the things. You'll see. Yeah, red patches right away.

Yeah.

And we put just enough water on it so that it actually. It actually does it. So give me a second. You know what it's doing. I know. I'll do it one more time. Here we go. All right, let's do clean. Mop. Okay, now it'll be slightly better. So it was trying to do the edges first. Edges. It does it far slowly than the normal version of it.

Got it.

That's where. That's where I just told it to stop doing edges and just clean them.

So is it vacuuming and mop at the same time?

So it just vacuums first, then it mops in a sequential manner.

Got it.

The reason is because if you try to do it, and let's say you had a spill, it will vacuum up Spill as well. And you don't want that as a user. You don't want to put a. Put a liquid on it and then do it. So we can kind of do the same exact thing. So. So now it will do it very, very well. Okay, so let's. We'll get into position and start in a second. There we go. Oh, and you can see that perfect patch. Yeah. Okay, what we can do, if you allow me, maybe I can even grab a.

Now I can see it.

You can see it.

Yeah.

Maybe we can put some. Can I get up a glass?

And maybe the glasses are over here.

Oh, glasses are there. Maybe I can put some water on it and kind of just show you how it works.

Sure.

So here we go. And. And whether it's a, you know, any sort of a spill or anything like that, I would just do it. So. So it's gonna take a turn onto the right here. There we go. So I'll just put some water in it. Yeah. As you can see, it's perfectly gone.

Yeah, that's cool.

Coffee spills, anything like that, and we just put enough water on it. So instead of spraying water, we do enough water so that it actually. And as you'll see, it won't bump into things, but it actually dries up within 30 seconds to 60 seconds, because we didn't want kids or pets to sleep slip on.

You know what this looks like? It looks like the Zamboni ice rink. Yeah, it does the same thing. It's just a lot bigger. It's like a little Zamboni from my kitchen. And it's laying down, you know. Nice, Nice.

Actually, we were at CES and this reporter from Virgil was there, and she literally brought. Poured a glass of milk onto the floor, and it mopped it out. And. And that was. That was really a tougher test than even just doing it. So that's the mopping piece of the puzzle. Now, another really cool feature that I want to show you. So now that it's mopped, I'm gonna tell it to, let's say, go even further here for second. Let's see. Yeah. So I'm gonna tell it to come here so I can say it'll go there. So it's going to come near you. All right? And I'm gonna move around, and you can record me and the robot both at the same time, hopefully. So let's do this. Let me just set this up. Voice command. There we go. All right. So I can say, hey, Matic, look at me. Clean this. And I'm pointing my hand and a finger and say, clean this. And as you'll see, it actually understood the command. It's planning now and it literally sees my. Me as a person. My hand, my finger triages where my face and eyes are looking at and then it's coming over there and starting to clean right here. It says, okay, you told me to clean over there and it's right there.

That's crazy.

That's cool. Yeah.

The magic of ar magical AI.

Now it will just start cleaning that area. So that's. That's one of the. One of the things then. So. Yeah, so it just kind of picked up that area where I pointed. Draw a little square around it and say, hey, I'm gonna start cleaning this thing.

Yeah, yeah, there we go.

Yeah. And you can even, you know, now is it this in your. In its path?

Yeah. So if I needed it to get underneath there, I just.

Too much. Yeah.

Or I can just draw a boundary around it.

No, no, actually it goes underneath it automatically once we start enabling the edge cleaning. So it knows exactly that. Hey, I will go it and doing it. I was just trying. I didn't tell it to do edge cleaning yet, so I was just telling it to just come and get it.

That's cool.

So we'll just come and get it in a second. So let's just do it now it's gonna move and then it will get the dirt. I just wanted to sure that it actually picks up everything.

It's interesting that it. It doesn't go straight like a box. It goes in various. Is that on purpose?

It's. It's based on the area it has drawn, you know, which is. That's a good question. So once we start. Actually, it's just a function. So now it captured everything. All the dirt is gone. Perfect. Okay. So it's just a function of an area that it has drawn. So right now it's drawing it based on this border and it thinks it's to going back and forth based on this water because I told it. But if it's a normal cleaning square. Yeah, normal square. Then. Or if I just clean it the whole area, then it takes into the account the walls and everything and then does it. So it's a function of that.

I know this is your feature.

Yeah, yeah. My feet are right here. It actually captures this. I can even show you one other cool trick actually. Hold on. So. So let's see if I can find this. Nope, not here. One second here. Okay, here we go. So. All right. I'm Going to do this now. I'll stop it for a second. And I'm going to try to have it clean this area, which is right in front of the refrigerator. Y. All right. And I'll draw a box. I'll draw square around it and say, clean this area. Yeah. Okay. It goes vacuum. And before it goes there, I'm also gonna throw a wire right in the middle.

Yeah.

And it hasn't seen that wire yet. And it already started picking up the blue part. Wire is already appearing.

So it knows not to go near the wire.

Yeah, it won't go ever. The wire is. It actually already figured it out. And the. The. I'll show you. So we'll start planning around it. There we go. This is. Okay, I'm gonna go straight over there. It will flip now turn straight. That's it. So I'm going, going. And as it plans around it. So we'll plan around that a little bit. And as it plans around here, it actually. So the white line is basically a next path is going to take. Yeah. So it's always planning next. As it comes near the wire, it will avoid it.

That's cool.

So in a second, we'll go. There we go.

This runs ON Android or iOS?

No. So this. This is. So this particular thing is just iOS at the moment. The app itself, we are going to give it an Android support very soon as well. So as you can see, it plans around it. But now the beautiful part is, let's go pick it up. And as it sees that area, you will see that slowly, slowly, wires will disappear. So ours are disappearing. And it will actually come back and plan that area as well and say, you know what? I didn't clean this area. Ours are gone. Let me actually. So ours are completely gone now. And all of a sudden it said, oh, I'm gonna go clean. Exactly at the area where wires are. Wire was there about 30, 60 seconds ago.

Got it.

And that's the point that it dynamically adjusts everything because it can really see and see the world and how to do it. So that's the cool part about it. And this is running. You're right. This is on the iPad. So it is running iOS at the moment. It tells you that it has three minutes left. Constantly plans. And I'll stop it for a second. And then I really want to show you the part we talked about where it is an absolute map. So if it's an absolute map, one of the biggest challenges, and this is a classic computer vision problem and research problem, is that if you kidnap a robot, what happens? Okay, so what I will do here is, and you know, I will also ask you to hold this iPad in your hand. And what I'm gonna do is I'm gonna cover this robot up, all the cameras with a, with a lid. So it's blindfolded now, right?

Yeah.

As I blindfold it, pick it up and I'm gonna take it to the room or your other room where it mark. Okay.

Yep.

And as I, as I lift it, watch what happens on the map and onto the robot.

It moved.

It moved. It relocated itself.

It figured out where it was.

Figured out where it was. And that's the whole point, that if it actually is a absolute math. If I know your home, you blindfold me and take it to the other part of the home, I'll know where I am.

Is it reading my ex?

No, no, no.

That's what I want.

Now the other.

Hey, Matic, what's, what's the news?

And the other part is because it can. So you saw it relocated, so it actually remembers maps of a different floors. Ah, so if you move it, if you're a multi story floor and you move it to the second floor, you don't have to tell it that, hey, it's on the second floor, it will just figure out and relocate itself and reload the map, you bring it back.

So you can have three floors and.

You can have one robot, you can have four floors, it doesn't matter. It stores all kinds of maps on it. So as a user, you don't have to worry about it, you don't have to tell it. The other interesting part is what we can do is let's see which part of the home it hasn't. Maybe this will work. But I'm going to try and do this where I'm going to take it and put it in a part of a home.

Let's go in the.

Yeah. So it hasn't mapped.

So like down here in the bathroom.

Yeah, so. So I'm gonna put it here. It has a map now because it hasn't mapped, it actually doesn't know where it is. So now as we start exploring, I'll start here and say, hey, explore. And I'm gonna stand here. So robot comes on that side, but as it comes back to that area, it will actually be. So now it will literally do 360 turn to see, figure out where it is. It doesn't know where it is yet, but as it goes back to the same area that already mapped, it will relocate itself. So it's just gonna start exploring and building a new map.

Figured out it's on a hard floor.

Yep. And eventually it will reload a new map as well. So let's see what happens in 3, 4, 5 seconds. So first it says, yeah, I'm gonna build this map here.

It's like I'm lost. I don't know where I am.

That's exactly right.

And is it gonna figure it out? Should start figuring out soon.

It's like a kid that got lost. And it hasn't figured it out yet. Hasn't figured it out yet. It's just moving. And I think it might just have happened. There it is. Yep. So it just mapped and then. And also merged the map of the previous area that it had actually done it together. So that's. That's really the vision part of the system that makes it so amazing and unique.

This is cool. All right. We've gone an hour.

Yep.

I gotta let you go because it's Friday night. I'm keeping this.

Thank you so much for having us. This is great.

Thanks. I'll send you a thousand bucks.

That sounds great. And it knows that, Hey, I got a dock. So we'll start docking, figure it out and give all the information.

That's cool. Where do we learn more about it?

You can go to maticrobots.com we're going to start releasing a lot more information. And this is just the starting point. We're adding all kinds of things. The metaphor we want to do is we're telling it to detect dirt so it can do continuous cleaning. We're telling it to be able to push the shoes back in so it can tidy up. We're building a full blown map. It will remember where things are. So all kinds of amazing features are coming on top of it.

That's really, really brilliant product.

Thank you so much. Yeah.

You guys have a great team to build this. Thanks. It took on all my objections of these things, you know, And.

And by the way, now you'll see really quickly actually. So it's not. It. It will just back into it. It will. And try to dock itself. There we go. Yeah. And. Oh, here we go.

There we go. It's like me, I have to figure out how to back up this.

Yeah. And this is all done visually, right? It just. It just. There is no beacon or there is no information. It just does it. There we go. That's cool. Yeah.

Thank you.

Yeah. Thank you so much for having me.

Thanks for coming over and giving me a personal demo. This is amazing. Amazing technology.

We are hoping to. To finally, finally solve some real problems using AI and save people a lot of time and energy. And that's where maticomplay.

I got an AI oven.

Yes, I know founders of those. Yeah.

We love our gin.

Yeah.

For similar reasons.

Right.

You put food in there and it automatically cooks it and doesn't burn anything. It's really, really amazing. But here is another example of AI coming into our homes.

That's exactly right. AI finally being used. Used for problems that are real. Yeah. This is part which was surprised us. We started doing digging and it turns out average American family ends up spending about 45 to 60 hours a week doing these home chores. We just never think about it because it's not a time we spend in one shot. And if we can save, you know, 10, 12 hours of those.

Now you need a little short dude to do my dishes and do my laundry. I can see a whole family of these things coming over the next decade.

So great start.

This is really amazing technology.

Thank you, thank you.