BUILD SCALE

“The input is electron, the output is tokens. That is in the middle, Nvidia. And our job is to do as much as necessary, as little as possible to enable that transformation to be done at incredible capabilities. Making that token, it's like making one molecule more valuable than another molecule. Making one token more valuable than another. The amount of artistry, engineering, science, invention that goes into making that token valuable, obviously we're watching it happening in real time.”

“Instead of having four humans that has, maybe a task might take you three weeks or so, like a cold email campaign might take you two, three weeks. Maybe it's three people helping you. Now you have one human that can probably do it in a day or two, right? And that's being generous. Probably takes a day. You maybe need to have some reviews done, but then you're often sending the emails the next day. Just the human timelines don't work anymore.”

“Toyota is one of our most significant investors. And we chose Toyota early on, one, because the Toyota family has dreamt about building aircraft for daily transportation, dating all the way back to the 1930s. Two, Toyota is known around the world for the quality and the reliability with which they build.”

“Circle’s Arc Blockchain, an economic 'operating system' designed for a world where machines drive the real economy.”

“I'm actually personally bullish right now on um, like one-person agencies. Like if if if I had to make my first freaking 10k a month, I'd just be one guy with AI, armed with AI, probably like a Claude Max. And then just try to figure it out from there. Um, and I'd post every day on socials, like one shorts one short a day, and just go that that route. It's probably cuz I know that route. Like I know that method pretty well, so I feel like I'm I could run it up. Like I don't know, gun to my head type of thing.”

“We had a mission, a vision of saying, we think that we can build human level AI, make it be something positive for the world, make the benefits be something that are distributed broadly, but how? And how do you get people to actually leave their jobs to come and join this thing? Initially, the set of people that I narrowed down to were actually Ilya, Dario, Amadai, Chrysola, and myself.”

“We are making these human robots to enhance our own manufacturing capability and efficiency. We haven't launched any 2C robots. So all the robots that we are developing, all the humanoid robots we are developing right now are used in our own manufacturing scenario. I think the video already showed that there were two robots that are working consecutively for three hours with a very low margin of error.”

“Big companies are looking for systems and assets that don't depend on one person's social media profile.”

“People often ask if you can do this with just cameras, but our stance is that Lidar provides a level of redundancy and precision that you simply cannot ignore. When you are operating a multi-ton vehicle in a complex urban environment, having that direct depth perception and 360-degree awareness is the difference between a research project and a reliable commercial service.”

“Now he’s back with a $1 billion bet that AI can make kids learn ten times faster, and that school as we know it isn’t just inefficient, it’s broken. The fix isn’t more money or better teachers; it’s rebuilding from scratch around mastery, motivation, and AI.”

“I think one of the things that was certainly true at Apple then is just this insistence on their leaders being exceptional engineers themselves. I remember a conversation that made a big impression on me. I was talking to one of our advisors who was a key senior executive at Apple at the time of like, how do you think about prioritizing sort of like management and leadership skills versus technical ability? And he just said, you need both. Like, you can't compromise. The best people are able to do both.”

“The rest of the day is devoted to what Liemandt calls life skills: leadership, entrepreneurship, teamwork, and real projects that kids actually care about.”

“You could use like a like a drop shipping model to test out an idea first. So, I think the best way to test an idea to see if it works and to see if you have the right unit economics, it's like yeah, drop ship it first, run some ads to it, see what the customer acquisition cost is going to look like. Um, and scale it up. You could you could basically launch a site within like like under a week. You probably depends on how cracked you are. With AI for sure. Like website, you can AI with it. You can get a bunch of like fast creatives, static using AI or or UGC.”

“Our new custom-built vehicle is a total departure from traditional car design because it doesn't need a steering wheel or pedals. We've designed it from the ground up to feel like a mobile living room, focusing entirely on the passenger experience, comfort, and utility rather than the mechanics of driving.”

“People often ask if you can do this with just cameras, but our stance is that Lidar provides a level of redundancy and precision that you simply cannot ignore. When you are operating a multi-ton vehicle in a complex urban environment, having that direct depth perception and 360-degree awareness is the difference between a research project and a reliable commercial service.”

“I think they've already increased capacity by 50% this year, and then they want to double it, and then I think it's just going to continue doubling probably twice a year. So it's just going to keep getting better and better and better and cheaper and faster.”

“At Alpha School, students spend two hours a day on AI-driven instruction and score in the top 1% on standardized tests. The rest of the day is devoted to what Liemandt calls life skills.”

“Not only that, it generated, we're talking about a lead from a multi-billion dollar company, actually two multi-billion dollar companies, is interesting, right? Because you're not just talking about generating views, you're actually talking about generating pipeline, and the agent is helping you do that. I'm not saying you should just let your agent YOLO and do whatever it wants all the time, but you should let it get you to the point where a human needs a review, and then you're okay to publish it.”

“The number one constraint we're facing right now is building more drones faster. The demand for these products has really just exploded over the last couple of years. We a couple of weeks ago announced a $50 million, 3,000 drone order from the US Army. That size of contract is actually not that much of an outlier for us these days. So it's a good problem to have, but we will be tripling production over the course of this year.”

“I think the number of agents are going to grow exponentially. The number of tool users are going to grow exponentially. And it's very likely that the number of instances of all these tools are going to skyrocket. Today we're limited by the number of engineers. Tomorrow, those engineers are going to be supported by a bunch of agents. And we're going to be exploring out the design space like you've never seen explored before.”

“IRL events are the next big creator play for high-engagement media brands.”

“From a product perspective, I think the course that we're charting is towards our drones being like flying agentic AI, just like you have an agent that you interact with on your computer or in the cloud, this thing is an agent that can move and do more for you in the physical world and you should interact with it in similar ways. Like it should have the intelligence and domain expertise to be useful to you in that way.”

“I think marketers are primed to win the biggest in this space. Cuz developers are obviously clapped. Um, to give you guys some insight, we recently stumbled upon a tech walk in Vancouver. And half these guys are looking for a job. The other half is saying they like to think with their brain um, rather than use AI tokens. And I think I think Tony said something like if you're thinking, you're not using tokens, which is like NGMI, like automatically, right? Um, Yeah, I mean, there's a there's a thinking machine for us. Like why use that type of like deep thought, deep thinking?”

“You have to figure out how to eliminate key-face risk so the brand lives beyond you.”

“The reality is that although the visuals do look fantastic, those visuals actually aren't accompanied by an understanding of the 3D world, understanding how objects can move, what the consequences of different actions are, and that's what's really needed for spatial intelligence. So, I mean, a term we sometimes use is that you need action conditioned world models, that you only actually have a world model if you can predict, given some action is taken, what is going to change in the world because of it.”

“The cases that the SEC and the Department of Justice brought against Nader have been dismissed. The DOJ dismissed in March 2025, and the SEC dismissed with prejudice — meaning it can’t be refiled — in March 2026. This comes after the founder was charged with defrauding investors of $3 million and running an unregistered securities offering.”

“How many hoops do you have to hop through to do that? Send them your driver's license and all this information, and there's a whole KYC process, and they limit you on how much you can purchase per day until you pass these different verifications levels and all this other stuff. It's like, what are we doing here?”

“You have cron jobs that are firing at 2 a.m., which is what we have, that will ingest all of my content from my podcast, my YouTube channels. We have Cold Outbound launching on Saturday, sending thousands of emails before Monday, right? You can do that. We have Deal Resurrection crons that will find the right time to reach out to people that maybe we lost the deal 60 days plus in the last couple years or so.”

“I think a default expectation in five years is if there's an emergency, you call 911, a drone shows up in a few seconds, and that's going to be everywhere in the US, hopefully everywhere in the world. And that's going to change the way policing works. It's going to get better outcomes, you're going to have fewer officer involved shootings, faster response times. And I think you can also do that while protecting privacy and transparency. Like these things are essentially flying body cameras.”

“We are now at the point where we are doing nearly 500,000 paid rides every single week across 10 different cities. This isn't just a science experiment anymore; it's a massive global scaling operation where we've moved from the lab to the streets in a way that is repeatable and safe across diverse urban environments.”

“They also discuss the tokenization of real-world assets, the move toward full-reserve banking.”

“Move over, context engineering. Now it’s time for Harness engineering and the age of the token billionaires.”

“A lot of the money is moving out of software into fiscal goods. So, I'm like pretty bullish on physical products right now. Unlike software, there's like software really has no moat right now. I've seen Jackie Vibecode like absolute trash. Um, Yo, chill. If and if Jackie can do it and if I can do it, uh, there's going to be like a like 10 million other degenerates that are going to code up the same [ __ ] So, I think like at least for venture and like private equity and growth equity firms to invest in software is going to be very challenging moving forward. It's an unknown space and money's going to money's going to rotate to the fiscal.”

“It became clear to me that bank-backed stablecoins were going to win because they were what we referred to at the time as geopolitically aligned. This basically means they're not going to be shut down because they funnel foreign money into US treasuries. We realized that was going to happen in 2018 when Coinbase and Circle launched USDC.”

“When I was actually working with Nvidia on the Synthetic Data Foundation Model Training Project, we were actually generating a lot of these synthetic data and showing that these synthetic data are actually as useful as real-world data when it comes to multimodal pre-training. But then, there's a lot of dollars being paid out to external vendors or other folks to manually curate these types of data.”

“We've agreed with the FAA on every detail of the design, not just testing. The FAA designated engineering representatives have signed off on each of those drawings before we start making them. Then the full aircraft comes together and we now begin flying it. First step is Joby pilots fly that aircraft... [Commercial rides begin] as soon as the end of this year.”

“The philosophy for the company is that can we do a smartphone that is much better quality and we can offer to users at much more affordable prices and that's something that you see perminated in the history of our products, right? We try to offer, you know, very high quality solutions to our users at affordable prices.”

“optimize an entire codebase, workflow, and organization around agent legibility instead of human habit.”

“The local partners are willing to work with you developing products that are more customized to what we want and as a result it's not just faster but it's also our ability to offer much more customized solutions that fits better into our smartphone, that fits better into what we want and as a result, you know, we go to market we are able to offer, you know, products to our customers at some offer something that they really want.”

“There are many moments along the way where you feel like it's real now, it's going to really happen. The economy is going to transform into this compute powered world. And I think that those moments are not yet at the end. I think that we have many more breakthrough moments where you realize that the next stage is possible.”

“We are now at the point where we are doing nearly 500,000 paid rides every single week across 10 different cities. This isn't just a science experiment anymore; it's a massive global scaling operation where we've moved from the lab to the streets in a way that is repeatable and safe across diverse urban environments.”

“If our next several years is a trillion dollars in scale, we have the supply chain to do it. Without our reach, the velocity of our business, just as there's cash flow, there's supply chain flow, there are turns. Nobody is going to build a supply chain for an architecture if the architecture, the business turns is low. Our ability to sustain the scale is only because our downstream demand is so great.”

“As you want to scale from hundreds in a community to thousands and then tens of thousands, you really want to do that with autonomy. I think this is another area where the administration and the FAA, the DOT are really leaning in to look at how do we make our airspace safer and how do we increase the capacity of it.”

“There are no lectures, and kids don’t move forward until they master the material. He argues the traditional classroom was designed for a narrow slice of students and wastes everyone else’s time.”

“If you're simply, you know, trying to predict the next video frame, that's not so difficult. But what you actually want to do is understand the consequences, likely consequences of actions minutes into the future. And to do that, you actually need much more of an abstracted semantic model of the world.”

“He explains how Timeback works under the hood, why he’s comfortable streaming student screens to AI in real time, and how he plans to scale it for a billion kids.”

“This is an agent that helps us on sales across the board. It will actually create all the cold email sequences, it'll come up with all the leads, it'll scrub the leads, it'll de-duplicate the leads as well, and it will send it on a sequence, and it constantly iterates over time. It will constantly self-improve. And so this agent we have, that's in our stack, we have a stack called Single Brain, which is all of our revenue agents.”

“I really think the most significant fact in this whole thing is how small it is. We're in the very rare and harder position amongst robotics and AI companies of actually having rapidly declining capital needs. And it's a testament, I think, to the strength of the core business, the demand for these products, having a really elite team that's capable of operating extremely efficiently, that we actually don't need that money to keep scaling and even make more and more aggressive bets in building new products.”

“The acquisition of Beacon following the acquisition of Kodiak, which is followed now by the acquisition of TopBuild, takes us from eleven months ago where we had no building products revenue, let alone EBITDA, to the second largest publicly traded building products distributor in North America with more than $18,000,000,000 in combined company revenue and more than $2,000,000,000 of combined adjusted EBITDA. It's a big deal in the industry and a big deal in the market as a whole.”

“And we just kept scaling PPO and we exceeded the performance of the best humans. And that itself was the finding, right, that actually massive compute with simple algorithms, right, that that is something where we cannot just—it doesn't just work in theory, it works in practice. We can really make it happen and in this incredibly messy environment where you cannot program it, you just need this almost human like intuition.”

“I think it's really important to talk about what a game changer AI is. You take one of the greatest aerodynamic minds on the planet and you enable him with something that makes him 10x as productive. The benefits compound in a crazy way. This is the most profound technology, I think, in the history of humanity.”

“Jeremy’s predictions for double-digit GDP growth as AI and blockchain reach their 'broadband moment.'”

“In 2017, we started to think very hard about, first of all, how do we really achieve the mission? What will that look like? And we start to do the math on compute and you start to realize that it's gonna take a big computer. Elon, Sam, Ilya, and I all agreed that the only path forward for OpenAI, the only path to achieve the mission was to create a for profit entity associated with OpenAI of some form.”

“Over the past six to nine months on Solana, the markets and the activity have just gotten so good. There's a lot of smart engineers working on exchanges and things like that. They're just so good that it's actually much cheaper to buy Solana on Solana.”

“borderline “negligent” if you aren’t using >1B tokens a day”

“instead of prompting it better or to “try harder,” the team would look at “what capability, context, or structure is missing?””

“The old economy is still here. It's not generative AI, but it's arguably maybe it's even more important. You couldn't have the new economy without cooling and heating. The actual business of making things and moving things, heating and cooling, is very much enmeshed with the new economy, which is all about generative AI.”

“There is this misconception that if you just keep improving driver-assist systems, they will eventually wake up as fully autonomous robotaxis. That's not how it works; the architectural requirements for a system that requires no human fallback are fundamentally different from one that relies on a human to pay attention. You can't just bridge that gap with more data; you need a different foundation.”

“I think one of our chairman saying many years ago which resonate a lot with me was 99% of the problems there was already a solution out there. You just need to find it and maybe with AI you can find it fast enough.”

“We will get a lot of business from data centers, but not just on the installation. Data centers need roofs too. Data centers need waterproofing, very much so. Data centers often need lumber related products. So so data centers are big consumers of building products. Now TopBuild itself has single digit percentage exposure to data centers, but it's very fast growing.”

“no human REVIEWED code before merge.”

“Our new custom-built vehicle is a total departure from traditional car design because it doesn't need a steering wheel or pedals. We've designed it from the ground up to feel like a mobile living room, focusing entirely on the passenger experience, comfort, and utility rather than the mechanics of driving.”

“Yeah, so we're very excited to be announcing our Series F. This was $110 million fundraise, $4.4 billion post-money valuation led by insider investors. And I really think the most significant fact in this whole thing is how small it is. We're in the very rare and harder position amongst robotics and AI companies of actually having rapidly declining capital needs. It's a testament to the strength of the core business, the demand for these products, having a really elite team that's capable of operating extremely efficiently, that we actually don't need that money to keep scaling.”

“TopBuild is the largest installer and distributor of insulation. And as you were just saying, everybody needs insulation. Every house needs insulation in the walls. Every office building needs insulation everywhere. It's a needed product, and it's not going anywhere.”

“The gap between being a creator and a founder is what actually determines if your business is scalable.”

“Jito is the largest liquid staking protocol in Solana and we're kind of view ourselves as the economic growth engine for Solana. So we have a liquid staking protocol, and then we also build a validator client, which basically tries to optimize the transaction ordering on the network.”

“At some level, the instantaneous demand is greater than the supply upstream and downstream in the world. And it could be at any instance, we could be limited by the number of plumbers, which actually happens. I actually went to the hardest one. Yeah, plumbers and electricians. And the reason for that is because, if we're too far apart, the industry swarms it.”

“A year and a half ago, we had the great honor of being sanctioned by the Chinese government. And it was a pretty aggressive action. They announced the sanctions, and then they showed up at the suppliers that we still had in China, shut them down, stop them from doing business with us, really tried to kill us. So that was a fun adventure for our supply chain team. They've done incredible work, and we've been able to maintain supply. And I think the good news piece of this is that we now have by far the most secure drone supply chain in the world, independent from China.”

“What creators avoid—but shouldn't—are the scaling pains required to make a business acquisition-ready.”

“We use a combination of Simulation and what we call 'Critic' models to train our AI drivers. This allows us to run billions of miles in a virtual environment where we can test the most extreme edge cases that you might only see once every hundred years in the real world, ensuring the system knows how to react before it ever hits the pavement.”

“A little bit over less than three years we launched our first car, we designed it and then we built a factory as well. So China speed. If you believe that a car a EV is going to be another piece of consumer electronics and that's something that we have a lot of experience in whether it is managing the software hardware integration whether it is managing the supply chain. I think those are the stuff that we have experienced working with so many consumer electronics products before.”

“This conversation covers his full arc from sleeping on the floor at Trilogy to being mentored by Jack Welch, to deciding that “kids must love school more than vacation” was a non-negotiable design principle.”

“We came up with what I would really say is almost the technical plan that we have pursued for the past ten years. Number one, solve reinforcement learning. Number two, solve unsupervised learning. And number three, was gradually learn more complicated things. After that off-site, I sent offers to everyone and said, hey, we wanna get started in the next two to three weeks.”

“In terms of the material we use for that rear floor we generated over we stimulated over 100 plus formulas and use AI to predict the performance of each of these formulas and ultimately we picked two to be the material for that. You don't use humans to do it anymore. You just use AI to take pictures very quickly. They can diagnose whether this piece is good or not. And then you can use AI in terms of material generations and a lot of those things.”

“DESO is short for decentralized social, and you can basically think of it as like Twitter or X, but on a blockchain. What that means is that when you make a post, it's actually stored on a fully decentralized blockchain and nobody can censor it. Just like nobody can take your Bitcoin, nobody can censor your content.”

“In 2012, when I was in college, I actually was able to mine 24 Bitcoin on free campus electricity. That's how easy it was to mine Bitcoin back then, you could just do it in your dorm. I remember I was so taken by it that I took a wad of cash to a CVS convenience store to send it to Mt. Gox just to buy more.”

“AI agents can already collaborate, but they lack a trustworthy medium in which to store value and execute contracts.”

“The fact that Nvidia's downstream supply chain and our downstream demand is so large, they're willing to make the investment upstream. And so, if you look at GTC and people are marveled by the scale of GTC, it's the entire universe of AI all in one place. I bring them together so that the downstream could see the upstream, the upstream could see the downstream, and all of them could see all the advances in AI.”

“Mr. Lee himself is a entrepreneur, serial entrepreneur, but he's also a product person, right? I think if you look at when we get into the car industry, he personally has personally driven 150 different models of car himself and take detailed notes so that he understand what's on offer in the market and how can he improve on it. He make us all take professional racing car licenses in order to know how to drive properly.”

“I think it's fair to say that, you know, vision understanding sort of stalled out; you got to object recognition and then progress just wasn't being made. There's really an interesting research question as to why that is, and at heart, the ideas behind Moonlake are an attempt to answer that, believing that there can be a really rich connection between a more symbolic layer of abstracted understanding of visual domains, which aren't in the mainstream vision models, which are still trying to operate on the surface level of pixels.”

“I think everyone was really focused on L2s and scaling through that method and some other methods, and Solana was just like, we think we can synchronize this entire state machine on one network versus many different networks. I started looking at the code and everything and just I was like, oh, this is a very cool network.”

“The most important thing is when you're setting up these agents, you want to set its own soul.md, so give it a personality, talk about what you like, what you don't like. Memory.md, so for persistence, obviously you want it to remember things as well. And lessons.md, too, you don't want self-correction after mistakes, right? So you don't want it to just, you make a mistake and you keep making the same mistakes over and over.”

“Now we're in a position to use that core technology to apply it to different form factors and use cases. And that's where the indoor drone comes into place, because a lot of dangerous work happens indoors. It's where the fixed wing drone that can cover much longer ranges comes into play. And it's a very exciting, fun time to basically take these mature technology building blocks to be able to pretty quickly assemble them into a fundamentally new capability.”

“At the core, this is an electric aircraft that can take off and land like a helicopter, but it has a wing. And so it can transition and fly on that wing that makes it more efficient and it makes it quieter. The acoustics are critical to what we're trying to build. We want a vertical takeoff and landing aircraft that you can land as close to the communities that you're looking to serve.”

“All of the evidence from neuroscience and psychology is that most of what comes into people's eyes is never processed. You're doing fairly fine-grained processing of exactly what you're focusing on, but as soon as it's away from that, you've sort of only processing top-down this very abstracted semantic description of the world around you. Human beings are working with semantic abstractions.”

“It depends like where you sit in the stack, right? Like um, now for Meta, it's like all it's all like creatives. Um, a lot of people are going to cycle into like AI creatives and then that's going to burn out. And then it'll be the next thing. Might be UGC creative uh, creatives again. Maybe affiliates. But it really like any good marketer uh, is going to figure out what the next arb is and they're going to maximize it. So, if you're like a real marketer, not just like one shot one shot wonder to understand Meta, then yeah, you're going to be able to you're going to be able to sell and you're going to be able to sell a lot faster in this new era.”

“When we complete the merger, we'll be number one in insulation, we'll be the second biggest in roofing, will be number one in waterproofing, and will hold number one or number two positions in certain geographies within lumber and building materials. So we'll have a huge addressable market, several $100,000,000,000.”

“There is this misconception that if you just keep improving driver-assist systems, they will eventually wake up as fully autonomous robotaxis. That's not how it works; the architectural requirements for a system that requires no human fallback are fundamentally different from one that relies on a human to pay attention. You can't just bridge that gap with more data; you need a different foundation.”

“I ended up returning the entire $140 million to investors; they got back like 94 cents on the dollar. I also didn't pay myself while doing it, so I literally had less money after than when I started. I had full control of the entity and over 75% of the equity, so if I wanted to grab cash, I certainly could have, but I decided to do the right thing instead.”

“Jeremy explains why traditional banking fails to support the needs of AI agents, and how stablecoins like USDC facilitate an internet-native economy.”

“We use a combination of Simulation and what we call 'Critic' models to train our AI drivers. This allows us to run billions of miles in a virtual environment where we can test the most extreme edge cases that you might only see once every hundred years in the real world, ensuring the system knows how to react before it ever hits the pavement.”