The Terminator "Coming Onchain" Is Officially in Production

Recorded at Solana Breakpoint 2025, day three.

Hodl Up: All right guys, welcome to yet another episode of Hodl Up at Breakpoint. This is day three, and this will be our final, closing episode. Today we have Amanda from BitRobot with us. I don't even want to get started on what they're building. I'm just going to give her the floor and let her do all the talking, because it's so cool that I'll get excited and just leave. So Amanda, the floor is yours.

Amanda: Awesome. Thank you so much for having me on. I'm the COO of BitRobot. We're building the world's open robotics lab, using crypto DePIN incentives on Solana to aggregate resources for robotics research and development: robots, operators who pilot the robots, compute, hardware. We aggregate these resources to produce outputs for robotics R&D.

Hodl Up: All right, but how did you get into crypto? Most people in this industry have a unique story about why, how and when they joined, so we really want to know your origin story.

Amanda: I worked in traditional finance for a couple of years, and then I wanted to get closer to technology and innovation, so I ended up at a venture firm, Collab+Currency. I was on the investment team there for a few years. I looked at BitRobot from the investment side and was super excited about what they were building, so I joined the team. That was in May of this year. I've now transitioned to venture partner at the firm, and the rest is history. I've been getting up to speed on robotics.

Hodl Up: I was going to ask why Solana, but it's a no-brainer: DePIN on Solana is uncontested in so many ways. I want to hear more about how the whole infrastructure works. How do you get the information, and how are participants rewarded?

Amanda: We're building a protocol on Solana. It's structured as an ecosystem of subnets. You can think of a subnet as a market to produce a particular output. For example, you can wear this cap. It's a wearable cap with cameras that collects egocentric video data. You buy the cap, wear it around, and you're rewarded for your contribution of video data. That's what we mean by a subnet. What Solana enables is attribution, who is doing what: different robots getting registered, different tasks getting completed. And then the incentives: what rewards are people getting? If I'm the subnet owner, there are rewards for me, there are rewards for the contributors, and there are rewards for validation. Through all of that we create verifiable robotic work and then do the payouts.

Hodl Up: All right. I really want to ask you about this big boy right here. What is it, and what can we use it for?

Amanda: This one is a sidewalk robot, but it's also a fun game, which is something a lot of the subnets have: they collect data, but it's fun. It's almost like a Mario Kart game where you can drive this robot around in the real world. You can pilot it in person, so I could pilot it around the Solana venue, but you can also pilot it remotely, so I could be controlling it from the US right now. It's quite fun. We actually had someone from the Philippines piloting it around the Breakpoint venue earlier. What it's doing is collecting urban navigation data. There's a lack of real-world data across the different robotics disciplines, and urban navigation is one of them. The founder actually started out building more of a gaming company. He was excited about the idea of playing games with robots, but found the data was super useful for researchers. Sergey Levine, the co-founder of Physical Intelligence, wrote a paper on it with his lab at Berkeley, co-authored with Toyota. Other researchers have engaged through the Earth Rover Challenge, which uses these robots to test their models on urban navigation. Most recently we announced the Grand Challenge, where we drove these robots from Stanford to Berkeley as a human benchmark, and now we're inviting researchers to complete that route autonomously, and in other cities once that benchmark is achieved.

Hodl Up: I was told outside, by the ring, to put a hat or something on this one, because it gives it a personality. I thought it was a learning machine model that basically becomes a pet for you, but I think I missed the point with that.

Amanda: I do have one of these in my apartment. I think it's useful and, again, fun. That's why a lot of people want to engage with robotics. This one is safe to use, simple and a couple of hundred dollars, so it's a great entry path.

Hodl Up: For sure. My other question: I always use the example of Apple, and how all of their products work well with one another. They have some of the best interoperability: you can pair your AirPods and MacBook and use them simultaneously. Does the same apply to your robot products? And if so, do I become Doctor Octopus from Spider-Man, with one arm for one thing and another for something else?

Amanda: Of course there are some complementary aspects. You could wear the robo cap while driving a sidewalk robot around, right? But what's quite exciting about what we're doing is that a big research trend right now is cross-embodiment: the idea of creating a model for general-purpose physical AI that works across all types of robots and tasks. That's what the big labs and robotics companies are working toward right now. There are three types of useful data: teleoperation, video and simulation, and our network has subnets tackling all three areas. So there are definitely synergies. Say I learn to control a robot arm and start doing teleoperation with arms, and then I learn about this other opportunity to wear the cap. Some of these things can be done at the same time, for sure.

Hodl Up: Now I want to talk about the humanoid robot. All of these gather data, that's basically what's happening when people use them, and that's how you incentivize users through the DePIN side of the project. What about the humanoid robot? Does it have to fight other humanoid robots? How does it collect data?

Amanda: The exhibition we have here is UFB, Ultimate Fighting Bots. It's a project we spun off, and a separate team now runs it. It's essentially a media and entertainment franchise. They've been covered by the Washington Post, the New York Times and Ashlee Vance, and they've sold out venues in San Francisco and LA. So there's a whole entertainment and media opportunity, which can then finance the robots, and the robots can be used on a BitRobot subnet for something like humanoid navigation. We also have another subnet focused on data collection with humanoids. So the connection is that entertainment and gaming are the use case right now, and the question is how you use that to finance and activate data collection.

Hodl Up: I think there was a vote on one of the fights. Is that kind of data used for...

Amanda: It feels like there's a lot of product-market fit for betting, I'll say. Essentially, we're seeing a new entertainment and media franchise form, and people love to sports bet. I don't know how useful that data is for robotics R&D, though. But what's awesome is that people are really getting excited to engage with robots, and I think that's the first step on the path to general-purpose physical AI.

Hodl Up: You're in this sector, so I have to ask you, and we need a prediction market on this, guys: what are the chances we get taken over by robots in the next 10 years? These look dangerous. Have you seen the humanoid robots? They dance better than any dancer I've seen.

Amanda: They are dangerous.

Hodl Up: Wait, I was expecting a no.

Amanda: No, they for sure are. The Unitree G1s weigh about 80 pounds, so if one falls on your foot, that can be a problem. We're seeing robots start to enter people's homes, which I think will happen in the next year through companies like 1X and Sunday Robotics. These robots won't be fully functioning. They'll be teleoperated, getting feedback while data collection happens. I was filling out one of the surveys to be involved, and they don't want you to have any elderly people or kids in the home. So there's definitely an important safety element. The weight of the robot and how you make it safer are important. I think it'll take some time, given where we're at. You can probably see the robots around here: they're a bit dangerous and not super productive yet. Mostly they're being controlled. Even in the demos you might see online, they probably took a lot of takes. If you change the lighting or the location, they may not achieve the same goal. So it's pretty early, but 10 years out, it could be good.

Hodl Up: Okay, perfect. That's peace of mind for me. I feel safer. We've just seen too many dystopian movies, so you get where that comes from. On day one at Breakpoint, I saw the humanoid dancing, and I felt like it was looking at me, so I waved. It waved back, and I was creeped out because I didn't know it was being controlled. Then I was smoking a cigarette, it came close to me and put its hand out like this, and I didn't know what to think at that point.

Amanda: The humanoids we have here at Breakpoint have pre-programmed moves. One of them can do a Michael Jackson dance, others have fighting moves, others wave or shake your hand. But they're just being piloted right now. Maybe you didn't see the controller, but someone was controlling it.

Hodl Up: All right. Now I want to talk about the role blockchain plays in what you're building. I can think of multiple ways blockchain could come in, but I want to hear how you're approaching it specifically, and why a traditional approach might not work in this case.

Amanda: Think about data collection and what's required for general-purpose physical AI. In the world of language and vision models, you can scrape the internet and get what you need. But in the physical world you need actual real-world interaction. I need someone to drive this robot around in many different cities and environments. I need a lot of robot arm data collected in different environments. Crypto is perfect for this: how do you crowdsource different interactions around the world and create in-the-wild data collection? That's our initial focus, because there's a huge opportunity there. Most centralized players may be collecting data, but they usually do it in a lab or a controlled setting. So how do you know the model works on every doorknob? Collecting that data, and also evaluation, testing the model around the world in different environments, is what our network is uniquely able to facilitate.

Hodl Up: When you think about it, taking into account the technology behind crypto and blockchain and how easy it is to build these incentive layers, I think it's the best way to incentivize people to actually do the work and help you get this data. DePIN is literally the best approach for gathering data.

Amanda: 100%. Of course you could hire data collectors. We have a couple of subnets focused on egocentric video. There's the one where you wear this cap, and there's another that Virtuals launched, called Seesaw, where you upload egocentric video from your iPhone. You could do that the traditional way: Figure recently raised a billion dollars and did a partnership with Brookfield, the real estate company, to collect this type of data. I'm sure they're paying a lot of people to collect this data in different environments. But it's way more capital-intensive, whereas we can do the same thing just by using crypto incentives and our infrastructure.

Hodl Up: Exactly. To wrap things up, I have one last question. What should we expect in 2026? Or is that too early to ask?

Amanda: Well, we're in December.

Hodl Up: Yeah, I know. For BitRobot, and for robotics.

Amanda: For BitRobot, the exciting things are, first, you'll see many more subnets launching. We'll be launching our ecosystem-wide points program and gearing up for testnet and our broader mainnet launch. In terms of robotics, I think next year will be the year of more in-home pilots and commercial pilots. A lot of robotics companies plan to ship their initial humanoids to people's homes next year. These humanoids will very much be extreme beta, but it's exciting to see robotics start to enter everyday people's lives.

Hodl Up: All right. I'm kind of scared, but I'm putting my trust in builders like you, so let's see what happens in 2026. Amanda, thank you very much for being with us today. We look forward to you delivering even more, and maybe having you on the podcast again in the future, virtually. So thank you again, and until next time, see you guys.

Amanda: Awesome.