AI robotics startup Astrocade emerges from stealth with $71M seed and live robot interaction platform

Speaker 1: Let me tell you about Console. Console built AI agents that automate 70% IT, HR, and finance support, giving employees instant resolution for access requests and password resets.

Speaker 2: Up next, Jonathan Jacoby. Jacoby. Jacoby, coming in hot to the UltraDome.

Speaker 1: How you doing?

Speaker 2: Hey, guys. Thank you for having me. So, co founder Great to have you. Having this Great to have you. Let's get right into the gong worthy moment of this segment. Tell us about the news.

Speaker 10: Yes. So today, we launched the name out of Stealth. Was raising $71,000,000 seed round and announcing Million dollars for sanctions.

Speaker 2: It had been a while since I saw a 70 plus million dollar seed round, and I feel more at ease. Ribbit capital. Having having Thank you, Adam. So what

Speaker 1: yeah. Yeah. What what does the progress need to look like these days? How how mature is the product? How built out is the team to jump straight to that size of seed round?

Speaker 10: Yeah. So we we're a team of 26 people. Okay. So Actually, grown quick pretty quickly. And today, we also launched something pretty unique. We launched, like, the first AI robotic experience online so people can just browse to a website, to robots that online, and interact with real AI powered robots Yeah. Via our models and our interfaces. So this is kind of

Speaker 2: a double milestone for us, both, you know, emerging out of stealth and kind of putting our technology out there at the same time. Very cool. I saw the painter, which is a robot that you can control and, you know, paint in I the real was a little bit frustrated because I had I'd I'd written been writing down notes and talking with a few people about somebody should set up a robotic arm with that lets you paint on real canvases and then send it, like, send it to yourself. But you're effectively doing that. So you you you beat me to the punch. Yeah. So where where does the AI start? Where does the teleoperation

Speaker 1: end? Because I believe there's been demos of, you know, deterministic robot or, like, remote control, even like a remote controlled camera that you can pan, tilt, and zoom. That's existed for a long time. Where does the AI actually come into the process?

Speaker 10: Yeah. So, basically, we have robots. You can talk to them. So, basically, the idea is we put out different interfaces. One, in some of the interface, can as you can see here, you can talk to the robot and ask it things, and the robot just responds to you. We're gonna also throughout the week, we're gonna launch more and more different experiences, and we're gonna showcase different ways to interact with robots. So our company is a research heavy research lab, research heavy company that focuses on robotic foundation models, but we also think deeply and care deeply about the experience. We think novel user interfaces are going to be a huge enabler for robotics. So we kind of wanted to put out something that is very wide to learn how people want to interact with robots and take that learning and put it back to the research. So in those experiments you're saying, you're saying one robot that you can fight with, like a robotic sword fight. So here it is. So one of them is controlled manually, like a teleoperation, and this like, which is the user. And there is a robot that is an AI robot that responds to the user, it actually fights. So this is one way of interaction. Another one is you type to the robot something and you see what what it does. So the painter, you give it an association. It paints it plays like a Pictionary with you. And then after the experience, you can, like, ask it to do some pantomimes. You can imagine the the brush is anything, and it can be a wand. It can be, like, leaning the the like, as a broomstick, getting the floor. So this is where the AI comes in. This is where our models, our foundation models comes in comes into play. And we wanted to showcase both interfaces and the models.

Speaker 1: Very good. What's your opinion on robotics sort of or robotic control coming for free in generalized LLM development? I saw some demos of, I think it was Opus five controlling a robotic arm. It seems like a lot of these models, they don't always need to be trained for something specific, and then you wind up just getting some particular functionality for free or it's like a quick RL step on top of what the Frontier Labs are already doing. Is that relevant to your roadmap? Or do you think there's a convergence? Or do you think you'll go off and carve out a specific niche in robotics? I mean, obviously, it's a huge category, but how does that all fit together?

Speaker 10: It is a huge category. That's true. Yeah. So I I seen the demo you're talking about, I think, because it was Anthropic. Yeah. So we've seen those demos, actually we had from GPT 3.5, and before that we've had robots controlled by LLMs. I think the problem that we're seeing, you know, lab structures ourselves trying to tackle is the fact that this doesn't really either generalize very well or doesn't, you know, fit in with the requirements of, like, you know, dexterity or doesn't Yeah. It feels like cool demo, but not necessarily

Speaker 1: commercially Yeah. It also might just be like, it gets it's impressive to be like, wow. Language model controlled ARM. But if it's like, every time it moved, it was $20 in tokens, and it was 45% accurate. And if you just, like, did an actual model for the robot, it's like It's true. 2¢ and perfect.

Speaker 10: That would be Exactly. Exactly. And and it's kind of one example that one of my friend told told me was was assume you could have had an LLM, like, do an ASCII art and, like, paint as if it's ASCII art, but you we still have image models. Yeah. Totally totally. Make sure that the, like, the natural language of Yeah. Of the way the LLMs work, and that's why it's a bit off. But I do think that, I think, onision models for robotics is such as what we're building and what, I you know, some other labs are doing as think it's going to take a huge part of how robotics is going to evolve. And we're also very lucky to be in this, you know, time of of great tailwind. Right? So there's a great tailwind of research. So we most of the team is doing research, we have a lot of table from, you know, LLM's research. We take inspiration from that or from video models or from world models, and that's pretty that's pretty that's pretty unique.

Speaker 2: Time to time to revenue. How are you how are you thinking about, like, first commercial application? Sounds like you're in the age of research, but when when do you see opportunity to start commercializing and kind of where?

Speaker 10: Yeah. So we have partners we're working on, working with right now across different kind of segments. So from logistics to entertainment, to health as well, some cool things on that that we might talk about on those three categories, and there's more. What we're seeing is we want to be able to get as much value to users and customers as fast as possible and make sure that when we are scaling it up, it's scaled up in the right kind of directionally right approach because you don't wanna be way off and then be dependent on kind of kind of be in a in a big tech debt, I would say, in product debt. But but I think I think we're at the point where we're also seeing we're start we're seeing value. We're seeing, you know, value from from from from partners, and and and it's actually a kind of

Speaker 1: showing a a sign that we didn't expect to to to occur that that fast. How do you think about vertical integration? It feels like there's robotics companies that are saying, we're gonna build the robot and the model. There's others that are like, we're gonna build a modular robot and the model. There's other companies that are saying, we'll just build the model and the robot builders will show up. I haven't seen actually that many new startups that are like, we're just building a dumb robotic arm that you can load whatever model you want on. We want nothing to do with the with the actual AI inside. We're just doing manufacturing. Maybe that's an opportunity. But where do you see your scope growing or staying focused?

Speaker 10: I think it's a great question because it really is a divergent kind of topic. Yeah. Lot of opinions. Yeah. We see ourselves as so we're software software, like, first company. Yeah. Right? So we think we work with our products, our kind of software stack from the models to the interfaces to to talk to them. We work very hard on making it generalized across embodiment. So Mhmm. We want to get to a point that even a new manufacturer comes out and says, alright. Here's a new robot. I would want I would want to be able to, within a week, have everything work on that robot as if it I had that robot for months. So we're building a lot of infrastructure to be able to cut like, this is part of our research, like how to be able to calibrate all our technology across robots easily. We think, and we see a lot of vendors, and we think doesn't have it doesn't necessarily have to be coupled. And we see a lot of, you know, great companies Yeah. Around the world that that sell the the hardware. So if we could be a good provider for the brain and for the interface of the brain, we think it's a it's a very viable kind of product offering. And I think it makes sense as well. Yeah. Yeah. That's cool. I mean, yeah, obviously, there are a ton of robotic

Speaker 1: production or supplier companies. There aren't enough startups in that world, I feel like. I feel like we need we we we need like, I'm a direct competitor to Kuka or something to and I'll make a big robotic arm, and I'll make it cheaper or whatever. I'm sure there are that there were probably What were you doing before you go? What were you doing before this?

Speaker 10: So I I worked I worked at a company I worked at a company acquired by Wizz, and then we got acquired by Google. Oh, cool. Congrats. That's awesome. And actually, me and Asaf, the CEO of Wizz, we we go way back. When I was when I was 17, he he was running Microsoft in Israel, and he brought me to join Microsoft. Cool. And so we have a long history. So I joined Wizz, and and, yeah, I think I think the the era right now is Long story short generational run.

Speaker 1: Amazing. Well, thank you so much for coming on the show, breaking it down. Congratulations on coming out of Stealth. Great demo. Thank you. Matt's a super way to launch. Huge. And we'll we'll talk to you soon, hopefully. I'm sure we will. Thank you very much. Goodbye.