The Deep View: Conversations

If you apply super in front of that intelligence, ultimately what we're trying to do is create a product, create intelligence that goes beyond what we've ever done as a human species in every domain possible all at once. Our goal is democratizing super intelligence, super intelligence for all. It's not just this thing that's going to sit in some lab and only be used by special folks. And I don't think that's anyone's intention today, but because that word is so undefined, in artificial intelligence, artificial general intelligence, so I'm fine, it can feel scary to the sort of common person. And so if Lambda can build enough compute to ultimately democratize super intelligence across the economy, that is ultimately the mission that we want to get behind. In this episode, I talked to Robert Brooks IV, the chief commercial officer of Lambda.

Robert and I discussed the company's mission to build supercomputers for super intelligence, its vision of democratizing AI, and why it's beating the drum on the idea that compute is not a commodity. We also talked about why Lambda invests in research and what the research team does. We got into why all of the AI infrastructure companies are talking about co-design right now. And Robert emphasized why the company's experience and expertise in building out physical infrastructure is such an important aspect of what the company can offer. We also talked about Robert's journey in the tech industry, his $40,000 robot experiment, and his advice for maximizing your time in the age of AI. So here it is, our conversation with Robert Brooks IV of Lambda. So Robert, for those who aren't familiar in our audience, talk a little bit about what Lambda does and what your role is with the company.

Of course, so very simply Lambda builds supercomputers for super intelligence. We are the compute backbone to the AI economy. And we have this thesis around, you know, getting one GPU for one person in the entire planet. So we want to build billions and billions of GPUs out there. It's an ambitious goal, which is the right thing to have. In terms of my role, I'm the chief commercial officer, so I oversee sales, marketing, and customer success, and have been part of the company since really as part of the founding team. So I like to say that I showed up on day two, not day one, like our founders. And we've had this sort of front row seat to the birth of the transformer, the neural network architecture, that ultimately led to language models and then large language models, which back then were not as large as they are today, like billions and trillions of parameters.

And we serve that economy. And we're AI researchers. I'm one of the only non-technical AI folks on staff. And we actually build compute clusters for our own AI research as well. That's how the company started. Amazing. How long has Lambda been around? Technically, since 2012. So our founder and CEO, Stephen Balaban, is a machine learning engineer and built a couple of really bespoke machine learning and AI products in 2013, 2014, and 2015. These products actually look kind of like what we use today, but they were using extremely different models. So convolutional neural networks, a lot of things on like GAMs and rendering. And we were creating products that essentially could do this like style transfer with a picture of you and a painting from Monet, put those two together and create this sort of surrealistic image. This was really, really cool and fascinating work in 2012.

It feels like it's just like a prompt away on the Claude or ChatGPT app today. Of course. Of course. You know, it's funny. So I remember Lambda back in those days because I was working at a very enterprise site, Tech Republic. And I remember Lambda coming up with some of the early work that you all were doing. And so, you know, what a journey. And you know, they say there's a lot of things that you can't control in your journey as a startup. And one of them is timing. And you know, what a, what an exciting, you know, run to sort of be working on that at the time when sort of this, this huge wave now has come and sort of the work that you've set yourselves to the world needs a lot more of at this moment. Yeah. Yeah. I like to tell people that in 2018 and 2019, when I showed up to customer meetings, I had to try and convince them to use GPUs for their workloads.

That statement alone is insane in 2026. The dynamics have reversed. Tell me a little bit about this like one GPU for every person in the world vision. That's the first time I've heard that. And yeah, what would that mean? What does that mean? And yeah, how does that vision, how do you see the vision playing out? Yes. You'll see this in all of our press releases. You'll see this in a lot of our videos and content. We'll talk about one person, one GPU and making computers ubiquitous as electricity. These are grand visions that will be achieved over decades or potentially centuries. But that's sort of our foot in the game. Ultimately, it comes back to why I believe, what I believe our mission to be today, which is maximizing breakthroughs. So whether it's drug discovery or something related to robotics, it's going to have a

massive impact into the world and in our quality of life and our ability to potentially live longer. These are these issues that are largely going to be achieved with the adoption of these neural network architectures and scaling compute on top of it. So Lambda's position in this concept is providing as much compute as possible to those geniuses that will do their work on that substrate. And so that is our part to play. Now we come at it from a little bit of a different angle with our founders sort of being machine learning engineers and researchers. We understand what it takes to produce and get to that scientific outcome because we've done it ourselves. But to the point of where we actually add value back to the ecosystem, it's in, okay, we know how to build the clusters for ourselves. Therefore, we know how to build them for the largest enterprises and hyperscalers in the

AI labs in the world. And we've proved that thesis over the last eight years. We've been a GPU infrastructure company. And today, ultimately, if we can help maximize AI breakthroughs or breakthroughs in general, then that's our position within the economy. You know what, it's funny, you mentioned the supercomputer part of the mission at the beginning. And that is kind of how I think of Lambda today. You know, when I think of Lambda, I think of that part. But you mentioned this research mission as part of what you do as well. Tell me a little bit more about that. Like how many researches do you have working for us? What are they working on? You know, how does that play in to sort of the economy, you know, the economics of the company, all of that? Yeah, of course. So it's ultimately just part of our identity and why we exist and how we feel that we can give back.

So ultimately, you know, when the transformer paper came out in 2017, you know, there were a few sort of infrastructure players that were looking at that and going, like, wow, we're going to be able to really scale up neural network architectures. And essentially, it's almost boundless in terms of the compute that you can provide to scale these neural networks up. There's a few clouds that we're thinking in that way, maybe none, maybe only Lambda. Okay. And I think that perspective is just very different in terms of why we exist and what we're building towards. So yes, we had six papers published at NeurIPS in December 2025, which I don't think any other cloud did. You know, of course, like I'm sure scientists and hyperscalers that are on different teams than what those core businesses are have papers published there.

But like we have the exact cloud engineers that you're potentially working with. Now, when you're buying compute from Lambda, actually publishing at these conferences as well. And when we can go super deep into the stack, not only at this sort of like technical driver and framework and layer, but also into the sort of optimizing the actual architecture of the model. This is how we give back to our end users. It's not just the dollar per GPU per hour and the compute that you're thinking about. It's the value out on top of it. And that's where Lambda's special sauce is. Tell me a little bit more about that. What are some of the things that your researchers are working on? What are the things that they, you know, provide? What are some of the more like tangible benefits that you've seen from sort of investing in research?

Because it's not, I think writ large, in many cases, we're seeing, you know, that kind of sort of investment and focus on sort of pure research is a little less, you know, interesting than it used to be. Unless you're Google DeepMind, right? And you have a couple hundred people and, you know, they don't have to provide any economic value or whatever the case may be in there, you know, for a company that large. So of course, there's things that we show off, like from time to time. We don't, we don't show everything. But one of those last year was a concept called neural OS, where essentially an operating system could be run by neural networks. And essentially you could create any software just by prompting it into existence within your operating system. And you can connect those pieces of software with other people that have the same neural OS.

This is a vision for the future that might not manifest itself tomorrow, but it's in a similar vein of what, you know, big AI labs are trying to build towards. So the factory to compute provider has a similar thesis and has already gone down a little bit of the research rabbit hole to understand it. Probably build the trust in terms of, you know, being able to work together and supply this really important substrate to you. So that's one example. We have a lot of like, Palantir-style engagements with customers where we're actually, again, not just providing the end service, but sort of customizing it to them. It goes beyond the sort of infrastructure level all the way to that metal level. This is helpful because researchers cost a lot of money and they're in high demand and that might approach them for a billion dollars.

So when you need to think about how you're going to build your business, whether that's a AI lab or an enterprise. You need to think about the type of partners that you're choosing along the way and their specialties. And we'll get to this concept about computers, not a commodity. This is a big part of it, right? Just handing you a GPU here is one thing, being able to add value on top of that is another. Now it's time for a word from this week's sponsor Airia. Bringing AI into your company is exciting, but can also feel risky. Between data privacy concerns, security threats and constantly evolving compliance requirements, it's easy to hesitate. That's where Airia comes in. Airia provides a unified security layer with advanced threat detection and strong governance tools designed to protect your AI ecosystem from day one.

You can safeguard sensitive data, monitor activity and stay aligned with your compliance standards, all while continuing to innovate at speed. So instead of slowing down because of uncertainty, you can move forward with clear guardrails and real confidence. You don't have to choose between innovation and security. Airia helps you do both at the same time. If risk has been the thing holding you back from going bigger with AI, this is your green light. Visit airia.com to learn more or start a free trial. Go to airia.com. That's airia.com. We thank Airia for their support of the Deep View. Now back to the show. Does that plug into one of my buzzwords that I keep hearing a lot in 2026, you know, throughout the AI ecosystem is co-design. You know, it's like, you know, working with chip makers, working with infrastructure providers to really optimize, to optimize the compute, right?

Because compute is very expensive. It's getting more expensive because of the commodity, you know, not the commodity, but the components are getting so, so expensive, getting more expensive. Up in some cases, 600%, 400% year over year. It's crazy. So people were, one, were very compute constrained and then two, it's so expensive. And it's also one of those things that we know that when there's a lot of competition, things like response time and all of that matter a lot, right? So, so my understanding of co-design is that, like, if you have a, you're a software provider, you want to provide something. And then you're an infrastructure chip provider and you've got sort of the, you know, the machines to run it on. Working together is what makes sort of that work really well. Is that something that, that you all are coming across and that are, I'm assuming, pretty involved in?

Yeah, of course. So for companies that need tens of thousands of GPUs within a large supercomputer, you sort of have to think about ultimately where their needs are, where their orchestration stacks exist. And how they want to, you know, take advantage of such a large fabric of GPUs. And it's not just by creating a product architecture, handing it over to them and saying, good luck, right? Or, you know, we'll help you a little bit with some forward deployment engineers. It's meeting them where they need to be. And so Lambda has done a really incredible job of that. So we do run an insanely big public cloud where anyone with a credit card can sign up right now and get access to enterprise-grade NVIDIA GPU instances or clusters. And that's sort of a set model where you sort of come in, you pay by the hour, you pay by the month, and you're good to go.

It's a really technically challenging platform to just keep alive and up, and online. And not a lot of our sort of like, neocloud competitors have sort of gone after that market. That's where Lambda has started because that's kind of where AWS started, sort of building a platform first. Well, it turns out that companies need to go way beyond what a public cloud can offer them today, especially the big AI labs and hyperscalers. And this is where that co-designing customization comes in. And, you know, whether it's sort of a bare metal that's managed offering, or whether it's a bare metal instances offering from Lambda, we can find certain ways to meet you exactly where you want to be and still get the sort of telemetry and healthy metrics that we need to run a healthy cluster with a high amount of uptime. And that takes a lot of co-engineering co-design to your point.

Look, Lambda has the tiger by the tail right now, right? Like the amount of demand, I'm sure demand is just falling out of the sky for you all. What does the next sort of stage of Lambda look like? Knowing that we're in this incredible build out period, that there are new startups coming, we've seen all the money, essentially like 50% of all the VC is being, by some statistics, is being put into AI. AI is going to need essentially what you have. So you've got demand as far as the eye can see. But what does, when you think about the future of Lambda, when your team thinks about the future of Lambda, what does it look like? What's your all's vision for where all of this goes and the part you can play? And what is probably the largest build out that we've ever seen in tech? And obviously you have a role to play. Yeah, so we talked a lot about super intelligence for all.

Right now, super intelligence can feel like a scary word. It's somewhat undefined. I have my own personal definition of it, but it's not defined at the sort of macro level. And I think it's actually a way more powerful word than just- I want to hear your definition too, by the way. Yes. Okay, okay. It's a little bit like emotional in some sense, but I can- Okay, good. Let's go there. You know, for me, super intelligence, well, let me back up a little bit. The word computer, when I say that word today, you just think of a box. But that was actually a word to describe a human job at one point. Yeah, yeah. Right? Now, take a word intelligence. How do you think about that word? It's, for me, it's extremely abstract. I just think of like really smart people. Sure. In some sense. Right? Now, if you apply super in front of that intelligence,

ultimately what we're trying to do is create a product. But I guess that can be called a product in some sense, but create intelligence that goes beyond what we've ever done as a human species in every domain possible all at once. So, you know, was Edison necessarily smart in this domain that he didn't cover? No. Right? And if you can create some sort of tool, better word than a product, a tool, where we can leverage as humans. So, we take our intelligence, combined with the super intelligence of that tool, we're going to be able to get to drugs that we need as a species to cure illnesses that are affecting us right now. Or we're going to get towards tools or products or cars that help us get from A to B in a much more safer fashion. Or we're going to get to places within the economy that help us, you know, move a lot faster as well.

And so, we're going to solve human problems with that super intelligence. And I think that is ultimately for me the most important thing. So, intelligence will become a more human term in the sense of when we create the super intelligence tool. All right. Your take on super intelligence, AGI, however you want to frame it, this is really becoming a bit of a litmus test issue in the tech world, as I'm sure you're well aware. Like, there's the crowd that's like, everything we're doing is to build super intelligence, which is this like thing that's smarter at any human than, at a human than anything. And then the other crowd that's sort of like, no, these are tools and there's no single human that knows everything, right? So, it's not really possible. And it's the wrong goal to try to create one model that's good at everything.

What you need instead are more specific domain specific, task specific models that are going to do these specific things well. So, let me know. What's, how do you think about it? Yeah. I mean, I don't know how the future will look in terms of how the tool is built and spread. I can tell you that, well, there's a distillation of a large model into specific domains. Obviously, that seems to be the flavor du jour today. But to go back to your original question, like, our goal is democratizing super intelligence, super intelligence for all. It's not just this thing that's going to sit in some lab and only be used by special folks. And I don't think that's anyone's intention today, but because that word is so undefined, even artificial intelligence, artificial general intelligence is so undefined, it can feel scary to the sort of common person.

And so, if Lambda can build enough compute to ultimately democratize super intelligence across the economy, that is ultimately the mission that we want to get behind. The way that looks is if you can make the infrastructure layer affordable enough, that the downstream effect will be that AI is more affordable, more approachable for everyone on sort of any device on any, is that am I understanding that? In some nature, I think eventually, but right now there's still a lot to go in terms of the build out of super intelligence, right? And in terms of the chips that are needed to train the next model, you continue to reach sort of these ceilings in the capabilities. And I do believe in scaling compute is ultimately one of the solves in getting to that new function. And so, whether that's the Blackwell series that's out from NVIDIA or the next chips with

Vera Rubin or even beyond that with the Feynman series, it is our job as Lambda to make sure that those chips get deployed on time at a massive scale in the hands of the most important AI researchers in the world to ultimately affect the human space, she's positively with that work that they're doing, whether that's drug discovery or other parts of the economy. Okay, so that leads us nicely into this phrase that I've been hearing from Lambda lately, that I like to poke at a little bit, which is this like compute is not a commodity. And the interesting thing is like the temptation is to think of it as a commodity, right? And increasingly, the talk is like models themselves are becoming more commoditized, right? And then compute sits a layer beneath them, right? And so that it's essentially because what's when people say we don't want this to be a commodity,

what their meaning is because what happens when something's a commodity is it just competes on price. And so nobody ever wants to even be in a business where it's a race of the bottom and you have to just find the best way to get to the lowest price. And so what does that mean for you all for Lambda when you think about this? Because clearly what you're saying is not all compute is created equal. Yeah. So just based on what I've been saying, clearly we're bitter lesson pilled here at Lambda is one of the biggest solves and that's our contribution back. And so when I say computers not a commodity, well, in the simplest way you can think about if you've ever gone to a data center that's being built, these are massive infrastructure projects. Yeah. I mean, at one point in a particular large data center that we had, we had 3000 people concurrently working on it.

I mean, can you the scale of that is actually really impressive. Sure. And if you ever get a chance to visit a data center during the build out, not after it's done, but during the build out, you're going to see thousands of different machines, thousands of different connections, thousands of like mechanical and electrical and plumbing componentry, all having to come together in order to make sure that that one GPU cluster stays with a high reliability and high uptime. And in order to build these things, you have to wear a hard hat. You have to wear steel toe boots. You have to get a little muddy. You have to get a little dusty. And this is different from, you know, just the concept of, you know, a polished software product that's sort of been built in a, you know, a glass ceiling building, an all glass building and, you know, hand it off to you,

you know, via an API, this is real physical industrial build out. And so you can't know what the way to do it, right? Like, this is the know what the way to do it with that. Yeah. Yeah. And so GPUs and the availability of them, you know, in 23 and 24, you know, was potentially treated as a commodity. Oh, we're going to have way too many, there's stories about, you know, AI being able to like catch up to this amount of build out, and we have concerns. And we've just continued to blow through that expectation over and over and over again. And so at the very simplest sense from a physical industrial sense, the computer is not a commodity, like I own steel toe boots. When I show up to a data center, I have to wear a hard hat. These are different parts of the economy that are all contributing back towards this build out. Secondly, the way that you set up these data

centers, the way that you set up these clusters, there's so many just incredible nuances and just a million different things that are happening concurrently to make these things go live. And they aren't expressed to the end user. They're only seen by the people that are actually building this out. Sure. And so there's, you know, within a given compute rack, there's tens of millions of things, components that are currently working all in tandem, the build out of the data center compounds that multiplies that even further. And the level of complexity just goes through the roof. But ultimately, it simplifies into you just being able to see that the GPUs are available, you know, in your command. And so if I could take every single ML researcher and everybody and just show them the amount of hard work and physical industrial sense that goes into this, I would.

And that would definitely expose how complex this type of system is and why it's not a commodity. I think the final thing is, we saw in the news a couple weeks ago that XAI's utilization was rumored to be at 11%. Well, there's different ways to measure utilization in clusters. Everybody took that mean fleet utilization. They're only using 11 out of 100 GPUs. That's not what I was talking about. I was talking about model flops utilization. And so ultimately, what that's measuring is the ability to make use of every GPU within the server for a particular training run or inference run, more so on the training side. And you do that through software optimizations, model architecture, understanding all the different sort of loss curves that go into what you're building, the data that you're selecting. And this is where it gets

closer to the bits rather than the atoms that I was just talking about. And so you have to combine the bits and atoms world and to, you know, all of these things currently to get to a point where compute is actually the opposite of commodity. It is one of the most scarce resources in our economy today. Very good. So what does that look like? So comparing if I have, you know, a training run that I want to do and I need infrastructure to do it, I'm trying to pick between Lambda and one of the neoclouds. What is your pitch to say, you know, hey, here's why you should do it with us? Here's what you're going to get, you know, if you do it with us versus, you know, somebody else? Yeah. So you would never, let's let's think of a good example. You would never buy a service that's abstracted from you without like understanding where it's currently located. So let's say,

let's say for example, you buy an airplane ticket, you don't just go on some random site and go, hey, we'll just let you know who the the airline is and what airport you're flying out of. And whether or not, you know, it's a good jet or a bad jet, you pretty much trust who the provider is, and you assume that they're flying like an Airbus or a Boeing airplane, and it's going to be up to a certain level of standard. You don't have to worry about that. It's been sort of abstracted away from you. The brand is you trust the brand. Well, in this world, when you're buying compute from different providers, you should be asking questions as to well, what data center is it located in? Is it a tier one, tier two or tier three data center, tier three being the better one, in terms of uptime and reliability? And are these of these data centers like former retrofits of

older data centers, are they retrofits of former crypto mines, who's actually operating the data center? And it's an important distinction in terms of like the end result that you get and what you're ultimately going to be able to do with it and your satisfaction with that service. And so it's really important that like lambdas chose them to grow a little bit slower than some of our competitors, because we only work with tier two and tier three data center providers. These are providers that have been in the business for decades. These are names that you trust. I would eat my dinner off the data center floor of these. I mean, they're so perfectly built and clean and well done. I mean, these are amazing structures. And as a cloud provider, we trust these people in terms of the environments and the actual physical infrastructure that they're building.

And that flows into everything that you do, the rack up time, the reliability, the ability to meet the service level agreements that we have with our customers. And over time, if you're doing a contract with us for three to five years, these things are going to become extremely important. If you're having downtime, if you're having reliability issues, if you're having throughput issues, latency issues, there's a million ways for this to be a bad experience. Right? So you booked that United flight to fly to New York, and it turns out that you're on a propeller plane, you know, and there's a lot of turbulence, you're going to have very different experience than someone that, you know, booked a boeing plane and got there in four hours. So it's hard to explain this nuance to customers. But because the build out has been so fast,

people have cut corners and go, All right, we'll just throw these GPUs into container and we'll blow a bunch of fans on top of it. And we'll use fresh air cooling. This is a term that I've never heard before, not just air cooling, pumped in HVAC system, fresh air cooling. Like if the wind's not blowing that day, your GPU temp is potentially higher, those corners that are being cut are not suitable for a large AI build out. And Lambda does not cut those corners. Robert, how did you get your start in the tech industry? I'd love to hear about one of the questions I'd love to ask people. You've obviously had a great run with Lambda, but how did you end up in this crazy wild ride of an industry? Yeah, I appreciate that question. I grew up in Cleveland, Ohio, really incredible place to be from. Definitely like strong identity with the

Rust Belt and how big their economy got and then contracted and the understanding of how that happened. I was ultimately a stockbroker out of college at Charles Schwab. There's sort of two routes that you can go with in Schwab. You can go to the technical analysis side, which definitely fascinated me, or you can go to the sales side. And I didn't know what sales was. I just thought like talking to people and building relationships was part of the gig and found myself in a position where I was in the 95th percentile for the metrics that they tracked in terms of lead conversion and bringing assets in. And I wasn't even trying. So it was inherently built in, probably credit to my dad, who was a similar, very jovial and very charismatic guy. And at Schwab, I was based in Denver, Colorado. A lot of Silicon Valley and New York offices had their second offices there for

engineering and sales. I was able to find Stack Overflow. I really wanted to get into learning how to write code. And so those two things meshed together. I got to live in the industry, but also learn from the industry. And for a little context, Stack Overflow is one of the prime examples of a company being disrupted by AI, which is sort of interesting. But the one thing that Stack Overflow took really seriously was machine learning. At that time, we more so talked about it in terms of data science. And so the next company I went to was Tempo Automation. That was a robotic company using computer vision to automate printed circuit board prototyping. And then ultimately, I went to a hardware hacker event, random Tuesday night in 2018, and got to meet the founding CEO of Lambda, Stephen Balaban. And they were five people,

they needed a salesperson for one of the first non-technical hires to jump on board the crazy rocket ship that was being built. It was not an obvious decision at all. I remember meeting my wife at that time, girlfriend, and bringing her there. And she thought I worked at some cool startup with all the perks and amenities. And it was just a factory floor with a bunch of computer parts all over the ground. And so I've been very lucky in some sense, but also intentionally always staying close to the engineering side and always wanting to explore what's next. Those are the key things that got me to where I am today. One of the questions that I love to also ask people is about, in this AI era, the promise was that we would get automation and so we'd work less. We could give stuff to the AI agents and they would do stuff for us. But what we found instead, at least what I found is most people

are working more. They're trying to keep up, they're trying to learn, they're trying these new things, the technology is changing so rapidly. So you're constantly in a learning cycle, a learning phase, a learning mode. And so time is more valuable than ever. You're also a leader. And so leaders talk about this in terms of leverage often, right? How do I leverage, which just makes, which means how do I choose what's the best things to spend my time on? As you think about those things, what's like one tip that you have for people who are trying to maximize their time in this age of AI? So I fundamentally believe that work is an important part of what it means to be human. You could give a lot of people $10 million and sure they could spend the next six months on a beach and surfing. But over time, they would want to get back to

contributing somehow. And today we manifest that in the form of a job. It could be manifested in any way, but ultimately it does look like something called work. And part of your identity is associated with that. And I think it's important, I remember growing up, like, why do these millionaires and billionaires and extremely successful people are still pouring in all this time to their companies and doing all that sort of thing? It turns out they enjoy it. It turns out that it's scratching at the surface of everything around you has been built by someone else. And so for me, the adoption of these AI tools, I think about it very simply. I don't have any sort of magic sauce. And it goes back to my definition of superintelligence. I cannot be an expert in every single domain that I currently act in. But I am one prompt away from knowing

what's going on. And that can be in meeting. That can be right before, that can be right after, that can be the research that I'm doing at night or on the weekend about that particular thing. And that is a very powerful thing to have at your fingertips. A very simple example, more so in the personal life, is if you're looking for a house, you can take the schematics in the floor plan and the square footage that they're advertising and run it past the three models that you choose. And all of them will calculate it and potentially they'll say different things. And you can go back to the realtor and say, Hey, this is in 1500 square feet. This is 1472. So that power at your fingertips, I think is really powerful for me in my professional life. I am just becoming an expert in real time by prompting. Now, of course, I can take that and

build that into certain documents and spreadsheets and presentations, but ultimately be able to get up to speed on the extreme vast amount of concepts that I deal with on a daily or weekly basis. That for me is how I'm leveraging AI. Very good. How about your time, the tools that you spend your time with? What's one tool right now that's making a big difference for you that you would recommend other people consider if they're not? That's a good question. I think I have a little bit of exposure to robotics at our companies. We bought a Unitree robot about a year and a half ago. I had some extra marketing budget and we decided to take that on and we took our machine learning team at Lambda and we started building models and playing with it and understanding how to actually program it and get it to do things we wanted to do. And I think not everyone has that

experience today. I'm very lucky that a company was able to fork over $40,000 for this robot and make the investment. Yeah, me too. Our marketing budget lasted way extra $40,000 and I also bought a robot. No, no, no. I'm pulling your leg. Yes, please keep going. Tell the rest of the story. But I believe that whether it's what 1X is doing with the NEO robot and the rumor price being around $20,000, that people are going to start experiencing that in their daily lives. Now, it's very, very simple tasks. But at the end of the day, I want to maximize my time focusing on the things that I love to do. And if this robot can just do a very simple thing that I don't want to do, whether that's loading the dishwasher, unloading the dishwasher, loading the dryer, unloading the dryer, folding the clothes, this is something these are tasks that I don't want to

spend my time on. I'd rather go hardcore into work or hardcore into leisure. I'd rather be on my mountain bike than folding that t-shirt. And so everybody talks about their exposure with LLMs and Claude Code and all this sort of stuff. And of course, I'm doing that today, but I've been able to glimpse a little bit of the future with this robot and having something personal do something for you. And I think that's a very special thing that's going to enter our world in a very short amount of time. And it's going to have a very interesting impact in terms of how we think about the tasks that are in our daily life. And if we have more free time, what could we be doing with that? Are we painting? Are we writing poetry? Are we writing on our mountain bike more? Are we doing more podcasts? Right? This is where stuff gets fascinating. Robert, thank you for your time.

Love the conversation. Appreciate you being here on the Deep View Conversations. Thank you so much, Jason. This was a lot of fun.