speaker-0: Hello, welcome to the Robotics Stack Podcast. Today we have a very, very special guest. His name is Randall ⁓ Briggs, and he's working on an open source robot platform that just recently came out of stealth. He also has a very interesting background in academia as well as a couple of practical projects under his belt. So we're gonna talk about it. This is episode 16. My name is Oscar. Hey my name is Leo. Welcome, Randall. We are super happy to have you here today. speaker-1: Really appreciate it. Thank you for having me on. speaker-0: Well we thank you for being available and ⁓ we also thank Bryce Scancioto who was the first guest who introduced us to each other. I guess we can start with a little bit about your background. It looks like you have been in robotics for a while, ever since you started studying. So maybe you can tell us a little bit about that and what attracted you to robots in the first place. speaker-1: Sure. Yeah, yeah. So I ⁓ I, you know, tr I chose mechanical engineering as my major. ⁓ I was at MIT both for undergrad and and grad school. While I was at MIT, I actually had to take two years off because my grades had fallen for a while. And ⁓ they have a pretty strict policy where if they see a significant drop for more than one semester, they ask for you to take some time away. That ended up being really good for me. ⁓ and for two years, then in the middle of my undergrad, I worked at a high speed centrifuge company. So obviously that's not robots. But the beauty of it was that it was such a small company that even as an intern, ⁓ I was able to get hands on, like understand the machine. And these are like five hundred thousand million dollar machines. These are not small machines, they're not cheap, and they're highly, highly engineered. You know, twenty thousand RPM, like twenty thousand G as well. I mean just very impressive things. But that that experience was really helpful. 'Cause then before that I think engineering to me was kind of theoretical. ⁓ and my I think that's part of what honestly, and that's why my grades were slipping, was I just felt like, you know, what's the point of this? Like these classes are so boring. I don't see how it's useful into building something interesting and cool. I I don't know, there just wasn't a practical handle. Whereas when I saw that machine and like I was like, ⁓ okay, now these equations that I learned, they actually apply to making this machine run properly and not explode, you know. It's very visceral when you're standing next to something that's spinning that fast. You know, so anyway, I did well, you know, during that time away and learned a lot there, took some classes at Harvard Extension, came back to MIT. The very first class that I took when I came back was the kind of famous undergrad robotics competition class. And we always use numbers, so we just call it two double seven at MIT. But the founder of that class, his name was Professor Woody Flowers. And you may know his name because he was also the founder of First Robotics. When he started this class at MIT, it was a the at at the time he took over it, it was very boring. We just learned how to like do engineering drawings and like textbook, you know, design work. And he was like, you know, we really should make this a competition or make this something interesting. Where you're building something real, you get a standard kit of materials, you have to compete. It's like a lot of fun at the end. And ⁓ you know, that model has turned out to be extremely smart. And so when he ⁓ then pivoted the class from like the first it was like mouse traps or simple things. He made it robots. Like that went really well and was very popular. And I think he loved it so much that he wanted to bring it to the world. So that's why he started first robotics. And then and now we see that it is worldwide. And it's such a big deal with educating the next generation. ⁓ but anyway, yeah. So that was the first class I took when I came back. And obviously I felt, you could say a little bit ashamed, you know, that I had been away from MIT and I wasn't a hundred percent sure that I deserved to be there. So I I think partly because of that and also just because I love building, I worked really hard on that class and on my robot. And ⁓ I took a different approach. Most people started building their robots right away, just building some prototypes. I just thought about it actually for a long time. Like I studied the strategy, I looked at old videos, I considered the different point method, you know, s scoring methods, and then eventually you know switched into like okay, high level design, like here's the theory, what I could do. ⁓ and then designed in CAD and actually it wasn't until two weeks before the competition ⁓ when my design was complete, you know, digitally. that I moved it over to the real thing and ended up getting first place. And that that was a big deal. ⁓ both for me and for my dad who was there. He, you know, was very encouraged by that. But the other reason it was a big deal was that my instructor for the course was Professor Sambekin. And at the time he was a brand new assistant professor, you know, still quite unknown, I would say, except for his the work he had done at Stanford, which was kind of famous with the robot get go. But ⁓ he had just gotten a contract through DARPA, or I guess a grant, you know, to work on the MIT Cheetah robot. along with Boston Dynamics. And anyway, h his work at that time ended up being foundational. Like he realized that if you want to build an electrically powered robot that is dynamic, that actually, you know, is is locomoting in an unstructured environment where you don't understand exactly where the ground surface is and you can't calculate it ahead of time, you need backdrive ability. And this is in contrast to like an early robot like Ozimo. You know, Asimo was very impressive, you know, really well controlled, you know, like a lot of incredible engineering that went into it. But ⁓ the fact of the matter is Ozimo could only work in a very controlled environment, like only on super flat ground, super precise stairs with just the right amount of ⁓ rise for each one. And you can see that whenever Ozimo would fall down, he would be completely rigid, like nothing nothing would move. You know, he was using very high gear ratio, very non backdrivable. And so it didn't have that allowance, you know, for variability in the environment. And so what Songbei realized, and this was somewhat based on the work of Mark Rabert, you know, who came before, but I would say the big difference though was Sunbei realized that electric control was better. Mark Graver actually believed for a long time that the only way to actuate dynamic robots was hydraulics. It was Songbei that actually changed the the game on that, that electrical actuators would be better. Because when you have that backdrive ability, one, you can handle unexpected terrain and collisions and things like that. And it doesn't hurt. You know, just like us, we're very very I always tell people we're very floppy. You know, like if we bump into something, it's not a big deal. I mean we have Course, we have direct drive, you know, from our muscle ⁓ to different things. And so Sangbei realized, like, yeah, you want to get as close to direct drive as you can. To be practical, it makes sense to have a small gear ratio, but with something like planetary gears, which is what he is, you can add extremely high mechanical efficiency. So it's not like you're losing much anyway. So the the benefit of that transparency is the ability to deal with unexpected terrain, unexpected collisions. And then the other big one ⁓ is because you have that transparency and the gearbox is not that inefficient. You are measuring through the current, you know, you're measuring how much fork is being applied. And and of course, with applying kinematics equations, you're you're measuring how much force is being applied to the environment. And that makes a big difference. And course, now that has become the foundation of all the reinforcement learning techniques, the simulation techniques that would not be possible if you tried to use a really high V ratio actuator. I see. speaker-0: Just before we move forward, maybe we have to go back to a few definitions. So first about act prototypes for the ones who are discovering what an actuator is, I really recommend you check out our episode with Bryce Sianciotto, who we mentioned already, where we go over the different types of act creators that exist. But you also mention two different things that I think we should talk a little bit more about. One of them is back drivability. What is that and then why is it necessary? And maybe tell a little bit more about Azimau, about the context in which speaker-1: build. Sure. So I'll make I'll start with Ozimo because I remember even even when I started working with Song Big Kim back then, Azimo was one of the first real humanoids, you know, that kind of took off. Especially one that could walk on its own. Like I think others were just more like animatronics, you know, it's just kind of a a stable robot that could move like an actor, but Azimo could actually walk. And if you watch their demos, like Azimo will even do a little jog and he will run a little bit, he will lean even in the turn. It looks fairly dynamic. But the the the clue that something is not quite right is when you see him fall down. And like I said, the main thing is that you just see that the body stays completely rigid, you know, in the in a fixed position, even when it does fail. And what that indicates is it's not back-drivable. What I mean by that, and and everybody has experienced this, I'm sure that everybody has grabbed onto like an RC car, right? Or some toy. And ⁓ you try to turn the wheels emanally and you can hear the motor go woo, you know, and spin, you know, because there's some gear ratio there. And ⁓ sometimes you can spin it. You turn a little bit of force and you can hear the motor turning at a higher revolutions per per minute. Other times you can't. Other times it's such a high gear ratio. When you try to move it, it doesn't work. That's what I mean by backdrivability. And you you've all felt this. Like if you grab onto an electric grille and try to turn it when it's powered off or something like that. And it turns out, you know, that in robotics, that's a really big deal. The way to achieve a h backdrive ability, well, there's one there's two ways, right? One is direct drive. You could just have the motor directly attached to the limb or to whatever is controlling the control surface or the control linkage. You know, that's perfect backdrive ability in the sense. You're feeling anything except for maybe some cogging torque in the motor, something like that. But the problem is if you do a direct drive like that, you have to use a pretty big electric motor to have enough torque. So typically you want at least a small gear ratio, like five, maybe somewhere between five and up to twenty. Even sometimes a little higher if you have a really good gearbox that's very, very well designed and very smooth, you can get away with a higher gear ratio than that and still have good back grabability. Sometimes it depends on which limb you're doing or which actuator. So like this one in the shoulder, for example, this can typically be a higher gear ratio than the elbow because it it's controlling a huge, you know, huge object on the entire arm. So it's okay if it's a little if there's a little bit more reflected inertia. That that means the the inertia of the rotor, the spinning part in the motor, a times the gear ratio squared. Anyway, you don't have to worry too much about the math, but I think we all have some intuition for this. And yeah, so anyway, that ends up being really important when it comes to controlling the robots. And we're now we're seeing it. is extremely important ⁓ when it comes to simulation and to the reinforcement. speaker-0: techniques. Yeah, I mean maybe that's a good point to dive into your project. Steelbot an open robot platform and tell us more about it, of course. speaker-1: Yeah. So yeah, exactly. Yeah, the company's called Steelbot. That's correct. Largely chose I thought it was a cool name and the domain was available. That's honestly why I chose it. Steel.bot as the website. The basic idea, I always want to give credit where credit is due. And to be quite frank, I was inspired by what the Chinese companies were doing, special Unitry, with selling affordable robots that were relatively open to program. And the reason I was so impressed by that was you could see the great variety of work that was being done. A lot of robotics researchers were buying them and doing really interesting things. Even companies like Amazon was doing some very cool like internal research with them. ⁓ and of course others were doing research as well without publishing it. And then the other side was that you were beginning to see startups that were buying these robots and building a business with it. Maybe one of the most famous ones is Rec Robots, the Fighting League, and San Francisco. ⁓ and at least at least initially that was all unitary robots. And ⁓ and there's others too, like you know, some of them I'm I'm now working with. But ⁓ the main thing that I think that they got right was I would say one was affordability, that these robots were actually at a price that companies and researchers could afford even if they broke. Before Unitree and these robots, typically humanoid platforms were anywhere from two hundred thousand to a million dollars or even two million dollars each in some cases that I've heard of. And that's scary. Like when you have a two million dollar robot that's doing, you know, and you're trying to do some experiments with backflips or things like that, it's a big deal when it breaks. So All of a sudden you had robot, you know, industries, the G one EDU versions about fifty thousand dollars. And it's like, wow, okay, this is a big difference in price. It's still expensive. I mean, it's still a BMW or car, you know, but it's much, much, much cheaper than what was there before. They were smart in that they made it a little bit shorter too. It's about four foot four. So ⁓ I'm not sure what that is in centimeters. I don't know, like a twelve year old child probably or something like that. And so that's a little bit less scary. To deal with too. It's and it's also a little bit less likely to destroy itself because it's closer to the ground. So it was a really good development platform. And I felt like they took the right approach, but I felt like there has to be an American alternative or at least Western alternative for a few reasons. You know, one is that during the recent trade war with China, they had proven that they were willing to cut off critical supplies, you know, to the United States or even to the West if they wanted to, like in order to intimidate or use a negotiating tactic. They had cut off the supply of rare earth minerals for a couple of weeks, I remember. And that was kind of a wake-up call to a lot of us. This is not friendly global trade, like this is a true rivalry. And unfortunately, they are willing to use these tools against us, use their supply chain against us. So partly for that reason, partly for the reason of spyware. The other thing is that research security researchers had proven that the Chinese robots were taking data, collecting data, specifically as audio. Video ⁓ 3D LIDAR stands of their environment and GPS coordinates. And they were sending that back in payloads of a few megabytes every five minutes or so to the UniTree servers. And of course in China, and this is not Unitries' fault necessarily, but in China the C C P has direct access by law to every company's server. So that basically means that each of these robots is a walking spy machine. That's obviously not acceptable. For a lot of situations, you know, especially if you're gonna use this in a sensitive manufacturing environment or for security at an airport, or you know, I mean, obviously military base is even crazier, right? So I just knew that there had to be an American alternative. And specifically, I needed an open American alternative. Because I felt like the other companies like Figure and One X, ⁓ as impressive as they are, the approach that they were taking was to deliver a final product that is polished and sealed, kind of like Tesla, right? Like when you buy a Tesla card. The FSD's great. You know, there's a lot of wonderful features in there, over the air updates. You cannot tinker with it. You cannot change the programming. You cannot customize it for your purpose. It is a closed environment. That's kind of like the way that the robots were taking in America, the robotic companies in America. And I felt like we needed one that was just open, open as a development platform, open to program, open to even modify physically if you wanted to, or add different sensors to the network, you in the n the robots network. And so that's the approach that I'm taking with Steelbot. And in so far, like the feedback I've gotten is outstanding. Like people So many companies, so many researchers, everyone today that's working with a unit tree or a Chinese robot, they really want this to exist. They are actually frustrated with the limits of what they can do and with the Chinese robots because they're open to some degree, but they're not fully open. And whereas mine is gonna be like root level, root access level open firmware fully open source. software completely open source. My goal is that Steelbot is so open that you can ⁓ with very low risk, you can build your business on top of it and just get reliable hardware and you don't have to worry that Steelbot's gonna tell you what you can do or what you can't do with your robot. speaker-0: That's super cool. And also the parallels to the car industry as always. ⁓ I was just imagining like an open source car. How would that look like? But that's a be a different topic. How far along are you right now with like the bot itself? speaker-1: Yeah, I would say that the design of the robot is fairly mature. Like so I've been working quite a bit on the actuator design, on the the communication protocol, like all the different parts of the robot, how they work together, both on the electronic side as well as mechanical. Definitely have made a lot of progress on that. Today I'm actually gonna go and look for more industrial spaces where I can set up the robot factory here in San Diego County. And so the plan is to start manufacturing as quickly as possible. And it's still gonna be a few months, I'm sure, to get to that point. My goal is to have the robot shipping by Q one of twenty twenty seven. So hopefully in the first few months of 2027. That may be a little aggressive, but I think I think it's possible. And I also think it's important, especially if it's okay if I pivot. You might have heard about the FCC ban in the United States on the Chinese humanoids. So I'll just bring that up now because it's very relevant. Towards the end of July, the FCC in America added Oran robots. They gave it a designation or added it to a list called the covered list. This is similar to what they did with drones. It basically means that going forward, the models that have been approved already for an FCC license for their electronics basically They will remain approved, but no new models will be allowed. The government in the United States was trying to take a balance, right? Between they want to close the door to the Chinese robots, but they don't want to slam it shut. They don't want to cut off the supply immediately. They want to make it gradual. You could argue both ways. I you could say, well, that's kind of anti-competitive, right? Like now the American companies are protected from foreign competition. Sometimes that results in worse products. That's true sometimes, you know. That is a real danger to to this. On the other side, you could say, well, China also doesn't play fair. So Whenever they want to dominate a certain industry, they typically subsidize it like crazy. And they ⁓ like I said, they put spyware in there too. So they even try to collect data while they're dominating. You know, there's a tricky balance there between like ⁓ free trade, which is generally good in theory, and unfortunately a malicious adversary, you know, who doesn't play fair, who doesn't play by the rules. And so I understand where the government's coming from. Obviously, it helps me, so I'm biased. You know, I mean that closing the door to the Chinese robots leaves a very clear opening. For an American alternative or Western, you could say, although that rule does apply to every country, unfortunately, you know, for better or for worse, all foreign robots now are banned in the United States. So anyway, that just happened recently. To me, that felt like the starting gun. Like, okay, now the race is on. Like who's gonna be the American company that's gonna come in and fill that gap? And so I'm trying to move as quickly as I can to be that company. speaker-0: There is one component of robots and of act traders that China has almost a monopoly on, which is the permanent magnet, right? So how do you get around that, for example, in your case? speaker-1: Yeah, well the good news is thankfully they don't have a monopoly anymore. When they when they did that ban, it kind of caused a massive scramble here, ⁓ both from the government as well as from investors. And there was already a few companies that were starting to produce domestic ⁓ permanent magnets and they all got supercharged. You know, they got massive cash flow from investors as well as from government grants. And so that domestic supply chain is actually spinning up quite fast. The mining is another side, if I remember right, I think there might only be one mine that that actually ⁓ mines the rare earth elements that you need. and refines them for that process. But that's limited. So you're right, there's both sides to it. On the one hand, yes, we are spinning up a domestic alternative pretty quickly. We are planning on using the domestic materials actually in Steelbot, which on the one hand causes us to go a little bit slower right now, but I think it's worth it because we want to establish a safe supply chain. On the other hand, today if you want to buy magnets for your robot, it's a little bit hard because these domestic companies are pretty much committed. to supplying the magnets for like the Fe Department of Defense, you know, critical industries, power generation is another big one, right? For all these data centers. So yes, right now we're in kind of a difficult middle period where there's still a shortage. And if you want to get magnets today, you pretty much have to get Chinese magnets. And you can right now. So there are companies that are importing them and building them. They're building US made actuators like for drone motors or for robots and they are using Chinese magnets. At least right now China's allowing it. But it's it's risky 'cause as they've proven they're willing to turn off that faucet ⁓ whenever they want to. speaker-0: I mean that's pretty good insights. Kinda positive that we see so much movement everywhere. But maybe now back to the specs of the core product that you're building and what you can tell about it, of course. speaker-2: Yeah, speaker-1: been relatively open with it, especially when I'm talking of course when I'm talking to potential customers or even even online of like published certain things. And I'm happy to say some parts of it and one approach that I'm taking is relatively high degrees of freedom. Some robots, like Asimov, not saying it's a bad one, but I'm just saying they have taken a different approach of kind of limiting the number of actuators in the robot. I understand why. You know, it's to reduce the cost, it's to maintain simplicity. ⁓ in some ways it makes it easier as a development platform when there's fewer parts that can break. I'm kind of taking the opposite approach of adding pretty much as many degrees of freedom as possible within reason. Like I'm not I'm not going crazy with that, but I just have this belief that if you want the robot to be truly general purpose and work for all kinds of applications. You actually need a lot of those and I'll focus on one in particular. ⁓ with Asimov, if I remember right, is that it only has one degree of freedom at the waist, which is we call it the Z or you know, or the rotation. Mine will have all three, which is typical. Like ⁓ actually the G one has three, right? You so that with the G one it can turn, it can lean forward, and it can lean to the side. Those are the three ways degrees of freedom. I believe those are actually important. For a number of reasons. I'll just give one small example which I like to talk about is, you know, we have these different impressive demonstrations that robots have done with like backflips and karate and all that kind of thing. And I definitely acknowledge that those are super impressive and it would have been impossible to do those things even a few years ago. It's only because of the AI techniques and the reinforcement learning and the simulation training that we can actually do that now with such a high degree of freedom robots. But I would argue a little bit that that is not the most interesting thing to look at. You it a humanoid that you want to see tasks that are a bit closer to like what what is something that's economically useful, both if the robot's working in a home or in a company. And just one that I like is like just imagining a robot carrying a tray ⁓ of champagne glasses, like a waiter would, right? You know, you see these people, these waiters will walk around, they have a tray full of food and glasses, and they can go around people, go behind chairs, you know, do it very gracefully without spilling a drop. And I actually think that's really hard for a robot to do something like that. It c I obviously it can, or I should say. Obviously it will be able to. But I think that even for that, if you think about it, you kind of need the waste degree of freedom. You need a little bit of flexibility to be able to move the upper body relative to the legs if you want to maintain balance of something. ⁓ as you're like making maneuvers and going around a person like if you want this robot to be a waiter or to be even to be a busboy in like New York City or some city like that, you know, it's a very cramped environment. A wheel-based robot's not gonna work. There's not enough room, clearly. So you need bipedal or at least some sort of legs. Or something like that, some sort of narrow platform. And then you because of that, I would argue that you need higher degrees of freedom to do a graceful motion like that. And so yeah, that's that's one approach that I'm taking in the robot structure. Another one is modularity, kind of like I mentioned earlier. That is manifested in two ways. One is that every part of the robot is really easy to repair and to replace. If you lose the rope the elbow actuator, for example, you pretty much just have to unbolt, remove a few bolts here and remove a few bolts here. Take out the actuator, put a new one in, bolt it back in, and it's ready to go. Very quick swap. That is not the case for most robots, even including the Unitries. The other thing is that if you want to modify it, I have friends, for example, that for their application, they don't need the wrist degrees of freedom. You know, so for them, actually, they would love to just get rid of that. You know, basically to just put like a solid piece of, you know, of material or just a solid fixed hand for whatever the job is that they're doing. And then with unit, they actually can't because. The software gets mad if you try to remove one of the actuators. So they're trying to like figure out how to spoof it or do get around that that limitation somehow. Whereas with Steelbot, though it's super modular, super flexible. You know, remove it, take out the bolts, remove that part, attach whatever you want. In the software, you have full control. So you just remove those actuators and then same in the simulation. Or change the lengths, right? Like let's say for some reason you want to keep the wrist actuators, you just want to make the forearm really long, whatever it might be. You know, so you add a you 3D print a new structure, you put it in there, same thing. You can just change all the parameters, you can train a new policy. ⁓ you have that full flexibility with steel bot. speaker-0: The Azimov you were talking about, you mean the open source one from Enlo? Yes. Yeah. That's right. But I think they are not shipping with hands, for example. Are you planning to do hands as well? What is your take on that? speaker-1: My belief is that there's already a lot of interesting hands on the market that are available for purchase. So I'm also not shipping with hands. What I'm actually gonna ship with is a standard connector at the wrist. You know, and that connector is both mechanical and electrical. ⁓ so when you plug into it, like you are connected to the robot network and you can communicate with the computer immediately and it's easily removable, easily swappable if you break a hand or if you want to change. Sometimes you might want a five-finger hand. And then maybe for a different job, the robot just removes the hand and it puts on like a drill, you know, or some power tool or something like that. So yeah, I'm not doing a hand myself in-house. I'm only doing the robot body and I'm just providing a standard connector there. I am standing on open sourcing that connector with no strings attached, no licensing fees or anything like that. So if the industry wants to use it, they are welcome to adopt it. I think that it's important that we do have a standard connector so that we can have a full ecosystem of ⁓ accessories developed. Kind of like we do for computers, right? There's all kinds of webcams you can buy, all kinds of mice. And joysticks and things like that and they all plug right into USB. So we have that standard connector already with computers. speaker-0: I have a general question about the company. How many people are you working with? Have you raised capital? Because there is very little information about steel bots online. I'm guessing this is intentional, so I don't know what you can say. speaker-1: Yeah, actually it's funny. I don't know how intentional it is, but we have finished raising a pre seed, so I'm thankful for that. That was earlier this year. And then right now we're just kind of fully focused on delivering the first robots. My goal is to use the pre-seed funding to get to low volume production. And at that point, once the robot is the design is working, you know, most of the bugs are figured out and we have some good feedback from customers. ⁓ we'll be limited on how many robots we can make per month at that point. ⁓ because of the approach I'm taking. Most of it's being done in-house, like I'm actually buying the machines to make most of these things. At that point, once the robot's working, if there's room and demand and we're actually running into supply constraints, then I would probably go out and raise the seed round at that point. ⁓ as far as the team, right now it's basically just two people full time. It's just me, me and my designer. He's the one that helped with the website, with the logo, with kind of the whole branding for the robot. He's also helping on the industrial design for the robot, things like that. And then I have a couple of other folks that look like they want to come on board. So we're kind of in the middle of talks there. speaker-0: I also have another question. So you mentioned okay, I want to make a totally open robot. I want to give you root access to the robot you buy pretty much with ⁓ no overhead, no bloatware, no nothing. And you say that unit tree is kind of open but not totally open. You also mentioned spyware. For example, if I get a unitree G one robot today, what can I do with it and what what am I prevented to do, basically? speaker-1: Sure. So if you buy a Unitri G one today, like the E D U versions, like if you buy the cheap one, you can't do anything. If you buy the thirteen thousand dollar one, all you can do is walk around you use the control light and just walk around and can waves. And that's about it. Yeah. speaker-0: Just to give more context, it's just a demo robot, it does not have sensor, it does not have embedded compute, it does not have a GPU basically. So it's just a remote control or scripted demo robot. speaker-1: Exactly. It's based in it, I would just call it like a very expensive toy in a sense, you know. So it's kind of interesting for the novelty. I think some YouTubers have bought them, you know, just for fun, just to like show the robot walking around, showing people's reactions on the streets. Obviously robots are brand new right now, so everyone's kind of excited when they see one in public. Yeah. ⁓ it it I laugh 'cause it makes me think about the early days of video, you know, when people would pay a lot of money to go into like a little theater and watch like a two minute or even one minute clip. ⁓ you know of like a train on video and and there's even stories of people getting scared right that it is gonna come out of the screen or something so it's you know things like this always have a certain novelty factor and it makes sense to actually take advantage of that at the moment. And that's why there's a lot of applications in it in entertainment ⁓ because it's so interesting and so new. But ⁓ yeah I think at some point that will wear off. I'm not sure when it'll probably take a few years. But anyway to go back to your question, if you buy the G one, the EDU version, so this is the one that you can program It has two computers on board. There's a locomotion computer, and that's kind of the one that's that's locked down that has cryptographic locking, you know, cryptographically locked. That's the one that sends out the commands to the actuators on the RS four eighty five. That computer basically handles the low level stuff. And then there's another computer, which is typically a ⁓ Jetson Orin. And that one is available to the customer. So that on that one you can put your custom policy or whatever you want to do with it. Maybe you're doing a manipulation policy like a VLA. And in that in that case, you might even run it over Ethernet or network. Or maybe you're doing locomotion policy and you want to do karate or jumping or backflips and you put that on that computer. And that computer, the user-facing computer that communicates with the locomotion computer sends the commands, basically sends the joint position commands. ⁓ and then that implements it. And if I remember right, I think the refresh rate's around 200 hertz, something like that. But the problem is you don't have access to that locomotion computer. So if you want to train change like the PID tuning or like change different timing things or stuff like that, you don't have access to it. ⁓ like I said, you even have trouble like if you want to disable an actuator, say you don't need all of them, so you just want to turn it off, you can't do that. The robot will respond in a negative way. So even the friends that I have that are trying to do something like that, they have to create a custom board that just spoofs the actuator, you know and they put it in place. So it's ⁓ that's the current architecture for the G1s today. That's that's speaker-0: Super interesting. This is really something I did not know. Because for example, I mean I know Figure AI when they explained their system, they mention system one and system two. ⁓ One is slow, one is fast. So the fast one is the one handling locomotion, is basically the one that cannot stop because if it stops, basically the robot will just fall down over itself. And system two is like the slow one that will output, you know, the command and the high level action. So basically in this case You have zero control over system wide, which is the one actually controlling the robots. Not that open in the end, definitely. speaker-1: No no. I mean i you know, ironically it's open enough, like where people have already done some really cool stuff with it. Yeah, you could say it's open enough. It's open enough that progress has been made. But people are running into the limitations of it. Yeah. speaker-0: Yeah. I also saw that Asimov for example, they have something called Asimov OS, which is like their operating system that they're planning to ship with their hardware. Is there something like comparable that you are going to ship, like some basic policies or something like that? Or are you just like, Okay, I'm just mostly doing hardware? speaker-1: Yes. So even though my focus is hardware, you know, I mean I'm I'm a mechanical engineer, like that's my specialty, and that's kind of the main thing that we're delivering with SteelBot, I would say that software is just as important. The one thing I would say though is that what we're trying to ship on the software side is just again, everything is open source. So like all the software that we ship, I think we'll almost certainly have the Apache license, which means you can modify it in whatever way you want, you know, for commercial purposes or otherwise. It's meant to be a suite of development tools. I don't know if I would go so far as to call it an OS right now. Like it's pretty much just giving you access to joint level controls, you know, or access to the sensors on board, the microphone, the stereo cameras, the speaker, that kind of thing. So it's more like a low level API or you could say something like that. But as time goes on, for sure, I want the software that comes with it to be more and more developed. I always tell people that in the long term, like let's say seven years from now, I would love it if you could go to Costco or some big box store, buy a steel bot, you know, put it in your car, take it home. Turn it on and you just basically tell it what to do verbally, you know, like you would if it was a human helper. So long term, yes, I actually do want to be very, very refined, very, very polished software that comes with it. I just believe that the way to get there is not to aim directly for that future, but actually the way to get there is by just developing the ecosystem as a whole. And one thing I would say too is I want Steelbot to be such an open platform that even I have to compete with the other providers. If another company is making a really great robotics operating system and most people, when they buy a Steelbot, they like delete, you know, the Steelbot software and they install the other one. That's okay. Like I I think that's better for the customer. You know, and then the classic example is like how Microsoft always comes with the edge browser, right? Or their browser, which most people hate, and then they immediately delete it and it's like sound on Chrome. And so ⁓ I'm okay with that. I hope that we can make the best software. But again, even if we do, it's open source. So anyone can fork it, you know, just like Microsoft's coding environment, right? DS Code and Cursor famously forked that and built their whole company on top of it. So yeah, that that's what I want. I actually think it's a good thing for the customers. I'll give you another example while we're here. Honestly, I I'm sad to say this, but I I kind of hate Apple. I used to love them. I used to think they were such a cool company, but they have gradually become more and more close, more and more unfriendly to developers. more and more unfriendly to the customers too. They just protect the margins. They protect themselves. It feels like a yucky monopoly to me these days. And one example I'll give is AirPods, right? Like on the one hand, that's a a great headphone. Obviously I'm using it right now, but it's kind of sad to me that they give AirPods special access, a low level access to the iPhone, right? So that It connects very easily. You can swap between devices, between my Macintosh or my iPhone without any trouble. And things like that. So they give themselves an unfair advantage because they own both, right? They they make the AirPods and they make the computer. So they just like they make it seamless. And anybody like Bose or other companies that want to compete, they're always limited in what they can do. Another example is the Apple Watch, right? Like that has certain privileges that no other smartwatch is allowed to have. So like Pebble, the Pebble Two smartwatch came out super cool. very modular, very open source too. ⁓ people are doing super cool things with it. But unfortunately you can't even send a text message from your Pebble watch using your iPhone because of the limitations that Apple has in place. Yeah. With Steelbot, one, I'm not gonna do that. Two, even if I did do that, it's all open source. So you could just like literally just change the code yourself. And I would also take the opportunity to say that open source is important for a couple of reasons. One is that I think is modularity, flexibility, and adaptability. So you can use the robot however you want to. You can use whatever accessory you want to, whatever software you want to. I think the other big reason that why open source is so, so important for humanoids is security and trust. That when I get my Tesla and I drive around, obviously there's a camera, right, and a microphone in the cabin. And I have no idea what the code is doing, you know, with that video feed and with that microphone. I have to trust Elon Musk. I have to trust that Tesla is not spying on me and not using it for some malicious purpose. But I don't know. I don't have access to the code. I was honestly just trusting a company. I think with humanoids, there has never been a product that is gonna be more personal and more intimate in a sense. Like this thing is in your home, you know, it is walking around your house, it it is hearing everything you say, ⁓ it is seeing many things that it probably you wouldn't want the public to see. or a company to see for that matter. So I don't believe it's even gonna be possible ⁓ for a closed source robot to compete with an open source robot, both for like industrial use if you're using this in a sense at a factory or even for the home. I really believe that in the long term, people will realize this has to be open source. I have to know where that video is going. I have to know if there's a remote control channel that's open. I have to know if my robot can hurt me or or damage my house or something. If it gets taken over, it has to be like Linux, where it is so scrutinized from security researchers and scrutinized by everyone. ⁓ and of course now scrutinized by AI agents themselves to to look for any holes or to look for any vulnerabilities. I believe that in the long term a closed source robot won't even work. speaker-0: That's strong conviction. And I think it's interesting looking at the parallels to the AI space where we had the foundation models mostly on the US side or Western side. And then we had like the open source movement coming pretty much from China, you know, as a sort of a catch up. And now, since China has a really strong hardware background and manufacturing background, and now it's kind of almost like an inverse ⁓ reaction like with with the West again from the hardware side. So it will be historically extremely interesting to follow this. But yeah, maybe some applications where you see it or partnerships that you have been going. speaker-1: For yeah, no, this is a good question. Actually, it's interesting. Like, even if you look at Apple and how close they've been, how controlling they've been, it's ironic because if you look at Apple's history and the very first products they released, I mean the Apple One board, like Wozniak, was literally giving away the schematics, you at the homebrew computer club. He actually didn't even want to build it. He thought I would just give away the plans for free and let people like, you know, get you know get popularity that way. ⁓ and then when the Apple II came out, which you know, okay, finally Steve Jobs decided, okay, we we should make a polished computer that we sell that's pre-assembled. You don't have to solder it yourself. ⁓ it comes with a case. It's actually kind of nice, you know. That's so they made the Apple II. But the funny thing is, I think when you look at the Apple II, it actually came with no software at all. And there was literally no not a single program that it shipped with. It only came with a basic programming environment, which Steve Wozniak wrote himself. And it came with a demo CD that had like a few little games and things, but these were just demos. They're not really like a full software package. So the funny thing is, even when Apple started, they had no idea like On the one hand, they knew this is the general purpose platform. Okay, you can do anything. Okay. And then you just asked them, okay, what are you gonna do? And they're like, I don't know. Like, maybe you'll save recipes on it. Like they they basically didn't even know what you're gonna use it for. ⁓ and and so I thought that was funny. And what happened, of course, was that one of the only times that Steve Wozniak actually fought with Steve Jobs and put his foot down was when it came to designing the Apple tube. Steve Jobs only wanted to put two expansion slots, so these are physical plugs basically on the motherboard. Then have access to the data bus and the memory bus. And Steve Jobs was like, All you're gonna need is a monitor as well as maybe a printer. So he's like, let's just put two. That's all you need. And Steve Wozniak was like, no, we need more. And and you know, Jobs would ask like, well, why? Why do you need more? And he's like, I don't know, I don't know exactly, but I just have a feeling that people are gonna make really cool accessories for it and we have to make more. That proved to be extremely pressing ⁓ because all kinds of cool things got invented. There was like scanners, disk drives, there was monitors, there was different types of printers, there was joysticks. There's all this this rich accessory ecosystem that got built around those expansion ports. So Steve Wozniak was 100% right. And that ended up being so important. And then two years after they released the computer, the big killer app came out for the Apple II, which was VisiCalc. And so that was the first spreadsheet program. ⁓ before that, all accounting was done basically on paper, right? Like you just wrote down all the transactions, you add it up manually at the end. Now all of a sudden you have a spreadsheet where you can change your value and everything calculates automatically. That was so powerful. I think in the l next year or two after VisiCalc came out, literally fifty percent of the Apple II sales were because of that software. Like companies ⁓ were buying a computer just to run that one application. So my conviction with robotics is that it will be the same. When you make a platform that is open, ⁓ that is programmable and even that is physically modifiable and expandable, I think that there are gonna be use cases that I could never predict, you know, that will be the important ones. And I find that already, even when I talk to friends and I say, Hey, I'm doing a humanoid company, like, ⁓ what if you used it for this? You know, they they have their application, right? And I'm like, I never thought of that. That's really cool. I have a friend that works in the pharmaceutical industry and he works in biologics. So they're using custom genetically modified cells, you know, to grow proteins, different things of interest. The problem is that if you get any contamination, like bacterial contamination, it ruins the whole bash. And currently they were losing something like twenty percent of the production yield to contamination. And the main reason is because there has to be a human in the sanitary environment right now. That human is changing valves, that human is pushing buttons, the human is monitoring things, like and they do their best, you know, to sanitize the human and to take precautions. But unfortunately, they're still getting contamination. And he was like, dude, if you made this robot, like I would program it, you know, for teleoperation or something like that, use it in the pharmaceutical industry. I have never thought of that before. And for them, a robot, like spending twenty five thousand, fifty thousand dollars on a robot is nothing. It's like literally disposable. For for for the pharmaceutical industry. So I have a feeling that there's gonna be so many applications that are beyond what I ever could have predicted. And that the most important thing I can do is just enable the ecosystem, is just provide our solid, well-designed hardware with root level access, some basic software that is a starting point that you can modify or adapt, you know. And I think from there the the market will take care of itself and then there will just be a lot of creativity that gets unlocked. speaker-0: Yeah, I have a question to expand on that idea of unpredictable applications and changes. So for example, you mentioned the creation of Apple and the first personal computer basically. As far as I remember, it was late 70s, early 80s, right? Something like that. And at the time it was a bunch of hobbyists, nerds pretty much building the first computers, and no one had any idea that one generation later, everybody who has even a remotely corporate ished job Would consist in being in front of a computer eight hours per day at least. Which was not the case one generation before. This is like one of the most tremendous change in society I think we have seen. Yeah, it was. I mean, on the similar scale as, you know, the industrial revolution where people were going from the fields to the factories, pretty much. And I think no one had any idea that at some point we are going to stop to be all farmers and we are all going to become factory workers. I think with AI and humanoid robots. The future is becoming a bit more obvious to people. Even people who don't know much about robots, if you show them a demo for humanoids and you tell them, okay, maybe in five years it's going to be able to be autonomous, they definitely understand that they can have one at home to make their tasks, to, you know, replace some people at some jobs. So I feel like the future is not there yet, but we definitely see it coming. Is there maybe like an undervalued or an under expected piece of the future that you see coming that maybe not everybody is seeing? speaker-1: You know, one thing that I posted actually on X the other day, I just said we're going to need millions of robot technicians. And I'm not exaggerating. Like I you know, it's so ironic. ⁓ on the one hand, you could say this is unexpected. On the other hand, every technological progress in history has actually resulted in more employment, right? Or more total jobs. ⁓ even if there are there are certain losses in the short term, no doubt. Some industries will be disrupted. It could be very painful for some people. But I mean, partly I'm an optimist, but also I think if you just look at the data and look at history, like Almost always when there's a technological change, new jobs. You just get more jobs, you know, that are different. And these robots ⁓ you know, some people will say, like, ⁓ the robots are gonna fix themselves and I'm like, Well, after a while, I mean, it it's I don't know, humans are underrated in a sense. And I think we're seeing that even with AI. Like, on the one hand, yes, AI is like superhuman at math, you know, or superhuman in other ways in just its thinking or whatever. But then you'll just see it mess up in the dumbest, silliest ways, right? That like any of us could have caught. And you know, or just the fact that our memory, that our context window is basically infinite. Like we have eighty years of context window or whatever, right? And we're kind of like thinking back on all that. Obviously not directly like that's being compressed. But my point is that I just think humans are underrated. And I think that even with robots, as as awesome as they're gonna be, it's gonna be gradual. These are not gonna be like iRobot or something day one. You know, they're not gonna be doing everything. That human can do. They won't have that dexterity. They won't have that intelligence. You know, for a long time, they're just gonna be used in certain kind of niche cases, or maybe or maybe less niche, you maybe like fulfillment will be a big one or something like that. But like look at Amazon, right? It's like on the one hand, they're the number of robots that they have deployed has grown exponentially. Like it's it's like hundreds of thousands, maybe even millions by now. But at the same time, at least until recently, I think now it's gone down a little bit, but at least until recently, their total employment was also just kept going up and up and up and up, even with all the robots. So if anything, they were just getting more productive. With that extra productivity, they actually needed more humans for different things. I really think it's gonna be the same with with humanoids that on the one hand, yes, there's gonna be so many cool applications, ⁓ so many cool ways of using them. It will expand the possibilities, it will lower the cost of certain labor tasks, you know, that are currently too expensive or currently too difficult for humans. Like even a good one to think about is like dog food fulfillment. I know this sounds funny, but like, you know, you think about a dog food bag. It's like fifty pounds, right? Like twenty kilos or, you know, something like that. ⁓ and if you're trying to carry that, you know, all day long for eight hours a day, you know, for years at a time, you're just gonna hurt yourself so badly. And that's why they have so much turnover. So things like that, robots are gonna be great. ⁓ but for all of these applications where you could say, ⁓ the robot is taking the human's job. On the one hand, maybe that's true on the surface. I would argue that okay, well, for every ten robots that you have, you're probably gonna need one technician. Because the robots is so much cheaper than the humans, you probably are gonna have ten times as many as you do humans. So like it actually balances out. If anything, I think employment will go up with robots. So like I don't know, I I think the Doomerism is short sighted. I think it's dumb. I think it's kinda like, I don't know, I think people are just trying to get attention. Like news in general, right? Negative headlines always get more attention. Maybe it's boring to say like everything's gonna be fine. It's fun. I really believe that truly. And if anything, I am gonna steer this company in that direction. Actually, the thing I'm most excited about with Steelbot is I want Steelbot to be a platform on top of which millions of businesses can be built. I want someone to buy three Steelbots and start a window washing business, you know, and then they have some contracts and then maybe they even have an open source car, like you said, you know, it drives the robots around, it goes and washes the windows at those businesses. ⁓ I want someone to buy steel bots and do landscaping business. I want someone to buy them and do ⁓ event services like waiters. You know, you can imagine sometimes you need fifty robots to be waiters at a wedding or something like that. ⁓ and if anything, if it's a nice wedding, you're probably gonna get humans. If it's a low wedding, you probably get robots. I don't I don't think humans are going away. I think it's already been shown, right? Like that's why we still have Swiss watches today. That's why we still have gas cars. Like there's always gonna be a attraction to the old ways. And even I think humans are actually irreplaceable in a lot of ways and will always be useful. But my point is more that I actually think that this is an opportunity, just like AI, right? Like AI now, one person can do a business that they never could have done before. They can develop a mobile app even if they're not a programmer. They can get legal help, even if they're not a lawyer, or they don't they don't have money for a lawyer. They can get medical help if they need it. Like all kinds of specialist advice that would have been inaccessible before is now possible so that one person can start a little small business and do something that incredible, right? They can actually take care of their family and beyond if they're creative and hardworking. And I think it's gonna be the same with robots is I literally want there to be millions of small businesses that get built. And that'll be my focus. I love Walmart, you know, I love Amazon, but to be honest, like I don't really care about them as customers. I want the little guys. I want the little guys to have robots and that's my my focus with steel bot. speaker-0: That's really great vision. Maybe also connected to that because recently we have seen big companies, for example Tesla, you know, figure One X, but also I think yesterday or today, X Peng in China announced that they are ⁓ starting mass production with their humanoids. And one question that came up was, okay, they are actually starting to produce literally like thousands of robots per month, like per company, and we will quickly see like hundreds of thousands per month. probably being produced. What are your thoughts on that in terms of demand and what do you think is their strategy? Are they just so confident, okay, demand is gonna be there, we're just gonna produce a how we're right now, see when the software kicks in, or what is like the the game plan? speaker-1: that's part of it. A lot of what you're seeing in China is ⁓ the top down approach of the government, right? So they're actually artificially kind of increasing these numbers. So if a province wants to buy humanoids, the government right now will basically fully subsidize it. ⁓ and so the province to try to be modern, they'll just buy thousands of robots and they'll put them in a data collection center or something like that. ⁓ but I don't think it's natural. I think it's a skewed situation right now where it's just being subsidized. ⁓ I think if you look at the United States, which is much more natural right now, and look at the number of robots that are being purchased and deployed, it's in the thousands, right? It's not it's not a huge number. It's not something massive. I think that's the true situation. Like if you look at the robots that are actually economically viable and actually paying for themselves, it's not a big number of applications. You know, we talked about rec robots or entertainment, you know, and things like that. There's only a few things right now where the robot is truly making money. If you want to use a humanoid robot, and I'm talking about a mobile one for some sort of job and you want it to be fully autonomous, it just honestly it's not good enough yet. And I think it's gonna be a while. So I think if you look at the natural demand curve, yes, it's going up. Like it's grown a lot from last year and maybe even a hundred percent or something like that, but it's still not a huge number. I think what you're seeing in China, ⁓ and even with Optimus, right? It's like we haven't seen Optimus do anything useful yet. I'm very pro Tesla. Like I'm not I'm not against Tesla, but like I understand the reasoning, like you want to be ready. For when the when the software does unlock it, then you want to be ready to crank them out. But honestly, I think it's gonna go slower than than people predict. If you look at twenty fifteen or twenty sixteen, like go back, look at the headlines. You know, at that time there was a big hype cycle of self-driving cars, a lot of money, like billions of dollars were being invested. Like some of these companies, like Argo AI, I think had like two or three billion dollars of money raised. And if you look at the headlines, everybody, including the automakers, they were like, We're gonna have self driving cars in like three, four or five years. Like we're gonna have full of self driving cars. We gotta get ready for that feature. You know, like we gotta ramp up, we gotta, you buy cruise, like, you know, just like it was a lot of excitement. It turned out to take a lot longer than people expected. Obviously, now we're ten years later and I would say it's just starting to work. I think it is working, right? Like I think I think Wamo is working, I think the cyber cab is working. And now there's gonna be a big growth. But I'm pretty sure that there's gonna be a hype cycle with robots where right now it's kind of like a little bit overinvested, even maybe I'll say. And then there may be some disappointment. I could be wrong. Maybe because of AI, the software is gonna move so fast that actually the hardware will still be worth it. ⁓ that's possible. This is a different world that we live in compared to 2015 and 2016. So it's hard to predict, but I do try to comfort people, like not to worry too much as far as the displacement. I think we're gonna be okay. I think humans are underrated. I think that it's gonna be a while before these robots are really taking over many of the things we do. Again, similar to AI, right? Like AI is great. It's superhuman and it's definitely expanding us. It's definitely giving us leverage. But like it doesn't replace a human yet. You know, it doesn't replace the way a human behaves in the company and the way a human remembers things and the way a human has built up knowledge. It's it's getting there, but I I don't know. And anyway, that's just my feeling. speaker-0: Yeah, about self driving, we we tend to forget that one of the companies that was investing the most in self driving ten years ago, twenty fifteen, twenty sixteen, was Uber. And they had I think one of the largest team, maybe second to Tesla at the time, with stellar researchers are at the top. And they completely shut it down, I think around twenty twenty, because they were like, Okay, we've done a lot of work, it's not working and we just cannot keep going like this for another five years so we're just gonna stop it there. speaker-1: Exactly. And I so I do think there's gonna be s I think there's probably gonna be something like that. I could be wrong, but I I think there probably will be a period of disappointment. ⁓ where it's like, Okay, you know, maybe in China. I I think here because it's kind of a natural market force, it's growing slowly anyway. So maybe here we won't have as much of that downturn. I would worry a little bit about the big companies, right? The big humanoid American companies that have raised a lot of money and they're like, ⁓ we're gonna like automate your home tomorrow and I'm like, Ew, I don't think so. I I think it's g you know, ⁓ it's a little scary to actually raise that much money expecting for this technology to be ready before your funding runs out. ⁓ I don't want to be negative. I don't want to like put down other companies, but I would be scared. And that's why I that's why I'm taking the open approach. Because then I'd feel like, okay, I'm taking an organic kind of natural path. I will let the market explore and figure out like what are the applications that make sense today. I will let the market kind of dynamics play out. And in a sense I get to be almost like an infrastructure company. I'm like selling hardware, you know, like they say picks and shovel or whatever. Like I'm just selling it to whoever needs it. Someone's gonna figure out how to use it. ⁓ that's that's nice for me. speaker-0: Yeah, that's cool. ⁓ maybe back to Steelboard a little bit. I asked that before, I think some partnerships or clients or milestones that you can speak about where you say like, Okay, this is the next step for us. speaker-1: Yeah, I don't think I can say much publicly about like who the customers are right now. I will say that we've got over a hundred pre-orders at the moment. So that's encouraging. That's plenty for me now for where I am. In fact, I don't want too many pre-orders because I know that the hardware's gonna take some time to perfect. So there's definitely interest. I would say basically all of the companies who are interested in Steelbot are companies that are using Unitree or other Chinese robots today. So it's almost always that they are running into the limits of the industry robots or they're worried because of the FCC ban and they need to find an American alternative. Or even before the FCC ban, a lot of them were just concerned about the spyware or about the limitations. So almost all of my customers are companies that are using the Chinese robots today. And they want an alternative. They want Steelbot because it's American or they want it because it's more open source than the others. speaker-0: Yeah, no, that absolutely makes sense, obviously. Challenges, something where you would say, Okay, maybe I could need some help or something where you seeking out speaker-1: Well, I'm always interested in people, you know, that want to join or kind of believe in this feature, either of a customer. If you're a customer and you want to talk more, totally just feel free to reach out to me. My DMs are open on X. My handle is Randall ⁓ Briggs, or you can find me on LinkedIn. Definitely feel free to reach out if you're an investor and you want to talk. Like we're not raising right now, but you know, happy to add you to the mailing list. Or if you're a customer, we would love to talk to you. We'd love to know about your application. Maybe you can even have some input in the way that we design this robot or in the way that we design the software. And if you want to work on robots and you're interested in enjoying Steelbot, we'd love to talk to you. I would say right now my focus is hiring people that are more experienced and more ⁓ independent. So less so on the new grads, but don't rule it out. I definitely want you to reach out if you're open and I may have something that you could work on. But ⁓ yeah, I think that's where I am. I'm really excited about You know, where things are going. I feel like I have a strong conviction that I'm on the right path. I would argue, you know, and the FCC ban was a big tailwind, no doubt. So I'm thankful for that. ⁓ I feel like Steelbot is taking the right approach. It's not easy. There's a lot of people that will tell you I even had some investors they did not invest, but I had some investors they said, like, can't you like make part of it closed source? Like, does it really does it really have to be fully open source? Like, can't you close something? You know, so people are locked in. And ⁓ I don't know. I mean, like I could, but I actually believe that being the most open will be the competitive advantage and that companies will choose Steelbot for that reason. So I don't know. I think I'm taking the right approach. I think it's one that does that does take conviction to follow. There's an easier story to tell investors if you are a closed source robot company, because you can say like we're gonna capture all the value. Whereas I'm saying, Yeah, we're gonna capture less of the value, but we're creating a gigantic pie. I just think that's the right way to do it. speaker-0: That's amazing. Maybe if you look at the robotics landscape ⁓ right now, is there something where you would wish people working more on speaker-2: this. speaker-1: Yeah, that's a good question. I do think one thing that's gonna happen is that there's gonna be a lot of companies that are kind of in the middle. Like let's say with AI right now, like with AI, there's a lot of foundation model companies, so they make the fundamental intelligence. And then there's companies that use AI services. So like if you're a law firm, you know, you want to use AI in your service, you either use the foundation model directly, or a lot of times what you do is you you reach out to a specialized company. I think Harvey is one of them for law, right? Harvey is not a model company, like they're just using the open AI anthropic models. But they're making really good application software on top of it and then selling that to the end user. I think that there's going to be a big opportunity in robotics. So I would say if you're in software and you want to get into robotics, I think some of the best opportunity is going to be you can call it the middle layer, the application layer, you could call it the deployment layer. I'll give you an example. Let's say you specialize in hotel housekeeping, you know, or something like that. I would argue that most of these hotels, they are not going to have the technical expertise to buy a steel bot and program it for their purposes. They want a company to come in that buys steel bot, that buys the Chinese robots, or maybe even other ones. That they want a company that buys open robots, a company that programs them, that makes like management software, you know, it's for scheduling or for, you know, programming it for that detailed purpose. And that they just ⁓ act as a contractor, you know, to all these hotels, basically. So I would argue there's huge, huge, huge opportunity ⁓ in that middle layer ⁓ to be like a deployment company, to be a specialist, to be a company that focuses on a certain industry and just nails it and does such a good job that. they you know, for them it makes perfect sense to hire that company and they're willing to pay a premium. They don't want to do the programming themselves. So that's I would say if you're in software and you want to get into robotics, I would look in that direction. speaker-0: Yeah, that's amazing. Really, really interesting thoughts and ⁓ really inspirational in terms of ⁓ the conviction that you bring to the table here ⁓ with this project because yeah, of course open source is not easy but I think you have a very good wave on your back right now. So yeah, we did not mention it, but I mean the hugging face acquisition by Nvidia is definitely an argument in your favor right now. speaker-1: Yeah, you're right. Because not only are they a platform for open source models and, you know, things like that, or open source tools, they also acquired pollen robotics, right? So they're making these little open source robots like the Ricci Mini and the and the Micro Duck now. Maybe maybe the last thing that I will say is a lot of people ask me, Are you gonna sell your actuators? And at first I would say no, because I was like, you know, I really want to build a good robot and I want you to buy the robot. I don't really want to deal with selling just actuators. But I have changed my mind on that and I've decided that I will sell the actuators too. The only caveat is I'm just gonna sell them exactly as they are in my robot. So because it's open source, because it's modular, you will be able to, if you want, to buy steel bot actuators and to build your own hardware. Let's say you work in a special environment that is where electrostatics really matter and you have to use a certain material, you know, or that's working in a bio lab or whatever. Like there's gonna be different applications where on the physical side, you need to build a custom robot that has certain characteristics. So just so you know, I am gonna sell the actuators as well. So that is another option for people to build on top of the steel bot platform. Those will be fully open source. You can use our software or not. You can use our firmware or not. These will it'll just gonna be high quality actuators at an affordable price. To empower the industry. speaker-0: Ending on a high note here. Thank you so much to Randall for this really insightful talk. This was the Robotics Stacked Podcast, episode 16. Check out Randall Briggs and all his social media link below. Thank you so much, Randall. My name is Oscar. And my name is Leo. speaker-1: Awesome. Thanks, guys.