$39B founder says his company could 100x in 5 years

Key Takeaway

When a problem feels overwhelming, shrink the time horizon until the next action is obvious. Bret Adcock’s rule for difficult stretches is to make a to-do list, work one day at a time, and avoid mentally carrying the whole week—or the whole outcome. Today, identify one stalled project, write the next three concrete actions, complete the first one, and only then decide what comes next.

Episode Overview

Bret Adcock discusses his ambitions for Figure’s humanoid robots and Hark’s AI agents and hardware, arguing that AI is moving unusually quickly toward systems that can act in both digital and physical environments. He also shares how he hires for real builders, chooses difficult high-upside problems, manages intense work and family priorities, and persists through financial and entrepreneurial low points.

Full Transcript

Transcript of $39B founder says his company could 100x in 5 years from My First Million. Auto-generated from episode audio; may contain minor errors.

I think if you Google Bret Adcock's net worth, Fortune says it's $19 billion. So that's a pretty good amount. How do you feel about this ? I don't care at all. I don't care about it. So, Brett Adcock, the short version is that you grew up in rural Illinois. You founded Vettery, which we sold for over $100 million . Then you took Archer, a company that makes drones , I believe helicopters, public. And now you have a company called Figure, which is worth, I don't know how much, 40-something, 20- something, 50-something billion dollars.

You have another company, Cover, that fights mass shootings in schools. And now you have a new company, Hark, in which you have attracted billions of dollars in investments. And you look exhausted . I'm in a great mood. Feeling really busy, man. Is this your third time participating in this show ? It seems like you've been there every year for the last three years. You told a story about how, I think, when Figure started, you basically said, “ I had a fortune, I do n’t know how much, tens of millions of dollars.

I put almost all of it into Figure to get started, and at some point you said, "I have a mortgage on the house, and the rest of the money is in Figure, and some of the money is in Archer, and it's not doing very well there right now." And since then, if you Google Brett Adcock's net worth , according to Fortune , it's $19 billion. So this is quite a significant spread . How do you feel about this ? I don't care. At all. You don't care about it .

You are a very competitive guy. I think you said something like, " I just want to win." You said it many times last time we talked. I wanted to win for these reasons. I am very competitive. I want to give everyone my attention. I think it needs to be taken seriously, and in fact, I think it's really important that Figure becomes the biggest company in the world. You say that you definitely have this Napoleonic energy: I want to be the best, I want to conquer. If you think about me, I would characterize it this way: we, my companies, are just now reaching a turning point, and they are in a very early stage of development.

They can get really big. So if this works, it will bring in a hundred thousand dollars. So most of my energy is focused on how to make sure it works. There is no stagnation here. It's either going down or it's going up, right? This is a binary situation. Either robots are out of scale or they are not. So, in five years this will either be a very big event or a very bad one. And so all my energy is focused on making it a thousand or a million X from where we are now.

And so the pressure is very high, you really need to show results. Where are you now? What are your prospects for the next five years? It seems like three years ago, when you were visiting us, we said, I think I said that. I said then that your company would probably be valued in the $40 million to $50 million range, and I think that's the case now. But as for you, you still lack robots. You still need them to show up. Where will you be in 5 years?

What is your forecast? I think, broadly speaking , the AI ​​work we see now will be so large-scale that it will be 100 times larger than the internet. Everything just works so well. The system works well. Deep learning works. And everything is happening faster than I could have imagined. And frankly, after 15 years of working in software and the internet, it seemed to me that nothing happens as fast as it does in reality. And here, in the field of AI, everything happens exactly like that .

Can you give an example of something that struck you? We started with Hark, I have a new AI lab called Hark. About a year ago, I became very interested in the idea of creating a digital symbiosis of AI and humans. I think Figure will be the ultimate AI, and a version of it will exist in the digital world. This will be a human who will work in tandem with the AI. Eventually, it might have its own AI weights, its own memory, its own hardware. This will be very close to ideal.

The core of this concept was to understand how to make AI use computers for general purposes. You will never hire an assistant who doesn't know how to use a computer. Therefore, you need to be able to provide him with tasks that he can perform in the same way as you. Financial models, booking airline tickets, ordering food at home, everything you need. He must be able to do all this autonomously. But only one in a thousand websites has an API. Therefore, most computer use around the world takes place on the Internet, in a browser .

My main aspiration is that in two or three years you will have a system that you can communicate with and say, “Do this” or “ Do that,” and it will be able to, for example, go online and use it very well. Almost like a robot that can move a mouse and use a keyboard. Here's what needs to be done to solve the computer's universality problem: don't rely on APIs or MCPs. You need to learn to navigate the web the same way a person does. At Herk, we released our first model for preliminary testing last week.

It's very difficult for us to find something on the internet that we could ask her to do that she couldn't do . What did you do differently than others, because everyone is trying to use computers, right? For example, Elon Musk created a macro, and ChatGPT created its own model for using computers. Everyone does it . Do you feel like you've found something new? What did you do differently? Okay, there are a couple of things we did a little differently. First of all, everyone is trying to solve this problem using API and MCP.

For example, OpenClaw became so popular because it could do this, but it couldn't use the browser , it couldn't use DoorDash end-to- end because DoorDash doesn't have an API for consumers. So we tried to figure out how to use a virtual computer for each agent. So they don't need a MacBook or anything like that. You can simply run as many of these environments as you want in sandboxes. And, uh, you're going to need to give him the ability to, like, look at the screen and use, like, move the cursor and use the keyboard.

Yes, but I used Chat GPT on my computer and it did the same thing. I said, "Hey, book a massage," and he opened the browser and I saw the mouse moving and he was trying to type and scrolling through the results. It was bad. It worked poorly, but he didn't try to use the API or MCB. He tried to use the internet. Yeah, I don't know if it works well . That's the whole point. But if it gets there and starts acting up on the internet, then the whole point is that it won't go anywhere.

So what I mean is, what did you do to make it work well? Was this some kind of algorithmic breakthrough? This was in our post-training. So , it was in our reinforcement learning process, which we think maybe no one in the world has used yet. Well, let's put this in context. Beck... Okay, that's surprisingly easy to understand. Humanoid robots are a huge business if they work. If you can crack the code, I think you said the demand is unlimited. Hmm, I don't quite understand what this is .

You can explain it like I'm an idiot, because Sean, you should have seen... I got a presentation, and there was your voice on the screen for about an hour , and then there was a team list of about a hundred guys who had just moved here from China and who seemed to have amazing bios. And it seems like you were just collecting money because the team was great, and that's all that was in the presentation. You just said it in the video. Romeo, that's all we had when we started.

So, okay, what is Hark? I think the best way to succeed is to look at how other people have done it, whether you copy them or just use it as inspiration, because then you will know what is possible. So, from the age of 24, I did this tirelessly and very methodically. I created a spreadsheet where I tracked about 50 people who had achieved incredible success. I took into account the year they were born, the year they started their education, the year they started their first business that led them to success, and finally, the year of their breakthrough.

I collected all this data, along with stories about what they did to become disciples and what they ultimately accomplished, and compiled it all into a database. And HubSpot found this thing that I honestly had forgotten about, but it changed my life. They found it, improved it, and made it available for free download right now. So if you click the link in the description or the QR code, you'll see this database I created when I was 24 and it changed my life. So if you want to be successful or are already successful and just want to get more inspiration, check it out.

I firmly believe that artificial intelligence will develop in two directions. And then , at some point, perhaps, maybe even, we will come together. First, we will have AI in the physical world that will do everything in the environment for you . For example, laundry, dishwashing, cooking, managing the supply chain from start to finish, working in healthcare. A humanoid robot will be used as a basis for this . It's just a human form. And he'll just go out and do whatever he wants. It's just one piece of hardware that can do it all.

You build a smart AI into it and it will do anything in the world. This is exactly what Figure is working on . More subtly, but gradually, a very close digital symbiosis of AI and humans will form . You will have something special that you can talk to, that will go with you everywhere, that will know everything about you, have access to all your memories, all your accounts and systems, and will be able to do things like a superhuman assistant. It will be something like Jarvis from Iron Man.

And he will be able to do almost anything, he will be a superman . He will know everything about your life. You will be able to access it whenever you need it. It will work in the background , helping you constantly. If you are flying by plane, for example, with a long or short layover, and you are late, he will already have backup plans, he will already help you sort everything out. He will be with you everywhere. We don't have that. We have very good programmers.

We have very good chatbots. But we don't have something that could suddenly become my Jarvis. To achieve this, we need to work on the model. It just has to be better than text chat. The system must be able to work with computers, have a near- perfect memory, be able to communicate with you just like a human, and, of course, have a vision system. She must see the world and understand what you see. And secondly, the AI ​​interface needs to be fixed. You have an AI and a human, and in between them is an outdated hardware system, like a MacBook or an iPhone .

They were developed 20 years ago. They are completely unsuitable for AI. This is not the right interface. So we started developing what we believe will be the iPhone's AI successor. It's a kind of renewal cycle. We see this all the time in startups. You see it too, right? We are in a cycle of updating computers and phones. They won't just disappear. New ones will appear. All computers, phones and systems will be equipped with AI. And they will be magnificent. They will work in real time. You will always be able to access them at any time.

They will always understand what is going on. They will always be able to refer to what is happening . You will be able to abstract yourself from most applications. You probably won't have an app store. You will probably have an operating system with artificial intelligence. This will be perfect for you. Eventually, you will have your own devices that you can own and carry with you everywhere. This will be really cool. And we hired a reputable team . The team is probably 80 or 90 people. The guy who leads hardware development and previously developed the last few generations of iPhone, MacBook, MacBook Pro.

He is simply a genius. He is magnificent . So we're developing what we believe will be the next generation of AI devices that will replace phones and computers. And then we develop the next generation of AI models. Models need to become much more multimodal. They should become much more expressive. For example, text encoding is not enough to make us truly feel like a real AI giant. So we are working on it. We have the first preliminary research sample of the computer agent , which we released last week.

I think we were among the leaders in some of the world's leading browser-based desktop performance tests . And they will continue to improve. This will get better and better. Every month we will make it better, smarter and faster. We are working on several other technologies internally in the AI ​​space. And then, in about a month, we'll launch Hark in traditional browsers, iPhone, and Android. So you can start using it, and then we'll have the hardware. We are working on it now. We have a device in our lab right now that we're testing, and it's something else entirely.

It's like something out of a science fiction movie. What do you think these devices look like? People are speculating because Jony Ive's store was acquired by OpenAI, and you saw the video with the puck , and then this little puck and earring. I don't know if it's real or fake. A Super Bowl commercial has been leaked . Again, is this real or fake? What is the story behind this video? And what do you think these devices will look like ? Is it a watch, glasses or something else?

I think my opinion on this has changed a lot over the last year or so, but we have a strong belief within the company. We believe that devices in the middle could reach a billion units per year worldwide . Currently, the only things in the world that even come close to this level are computers and telephones. They are called mega-devices. And around them revolve auxiliary devices like AirPods, Apple Watch, and other similar gadgets that don't sell billions of units a year. They account for about 3% of Apple's revenue and help the ecosystem as a platform.

At Hark, we aim to solve the problem within this large middle ground . To do this, you need to give up your computer and phone. There is no way to avoid this. Therefore, it is necessary to create a new computer or phone that will be better and will replace the existing systems from start to finish. And around this are devices that, like we have at Hark, help develop a family of devices that, of course, don't sell billions of units a year, but are important to the ecosystem.

As I understand it, you're saying that the next device might be a phone, but it might just be a phone with artificial intelligence, right? You won't try to change the form factor. No, that's not what I'm saying at all . You will need to radically rethink everything. The first version of the hardware that we now have in our lab is unlike anything I've ever seen in my life. Fine. What is outside our lab, such as glasses, pendants, wearable devices and the like, is not the main thing.

In fact, Meta glasses are probably one of the worst products I have ever purchased. They are just awful. Horrible. I ca n't even figure out how to use them. They don't have their own network. They use the iPhone network. This means your app needs to be open on your phone and pairing takes a long time. Overall, it doesn't work well. I can't think of any reason why I would need to have this thing attached to my head for 14 hours a day. It's just not the right device.

This is wrong. So, the end result is a brain- computer interface (BCI) in the brain? And before that, we will have AI-powered language processing devices within the next 10 years. And this is , in fact, the way. And these are not glasses. Glasses...I do n't even know if they will make it into our top 10 most popular devices . When you and your team brainstorm, do you have a concept that allows you to think outside the box? Because when you talk about something like that, I literally can't imagine what you're talking about.

Let's get this straight at a basic level. What changed first? What's changed is that we have a new type of computer, and I think of AI as a new type of computer. A new type of automation is here. Automation can do several things that we see as 10x better. And this is where we want to develop a system based on key principles that will enable us to achieve a 10-fold improvement. If it's one or two times better than your phone or computer, you won't use it.

It will be literally 10 times better. What is it about deep learning that makes a system 10 times better? There are several such advantages. First, AI can now essentially think and use computers and systems for you just like a human. He can talk to you. He can see. He has visual understanding. It can perform speech- to-speech conversion in real time. It can use computers and systems for you as quickly as a human. Over time, he will become as good as a human, and even faster in terms of success.

So you have a system that is almost human-like in its capabilities. It can also have memory, meaning you can put memory into it. It will not forget anything. Over time, it will become almost perfect. So you have a system that is almost like a man in a box, with all the same capabilities as a human. And it's almost as if you could carry a little person with a computer on your shoulder with you everywhere. That would be madness. But that would only be for Sam. Only Sam could see him, only Sam could talk to you, and he was only there to help Sam.

And this would be your life, and it would become smarter and better as it developed, it would have a perfect memory, it would be able to use computers, talk to you and see. You'd think, 'Damn, this thing could do everything you do on a computer.' Okay, your first step with your team is to simply get rid of any limitations. What would be the coolest and most magical thing if we had a little person sitting on our shoulder, an all-knowing artificial intelligence that could see and hear everything we see and hear, and give us advice?

What would that bring that would fundamentally change all of this? Fine. And based on this, we must design the system around this. The competitive advantage here is that it is human-like in its capabilities and has a near- perfect memory, allowing it to return to old data and use it in the future. My phone can't do this . For example, I added a contact to my phone last week and found out I was busy when I entered the number, and then a day later someone asks, " Hey, have you called this person?" I think, “I don’t even know the name.

Forgot. I can't even ask for the phone number. This is just stupid. The whole system is like this. And then I go in there and order DoorDash, like a monkey, pressing something every day. I don't do anything like that with Hark . He does everything from start to finish for me on my way to work. I just say, "Order me a coffee," and it's done. It does everything for me in the background. I don't need to touch anything. Everything is abstracted. And if you had this little man with you everywhere, you would just say, " He probably would have predicted, Brett, that you wanted coffee today," and I would say, "Oh, yes, I do.

Let's order. You know what? " Make a double batch today." And, you know, forward the order to the Hark office instead of sorting it out. I can just place an order. Everything is ready. Let's take care of it . I don't sit like a monkey on my phone for three minutes trying to place an order on DoorDash. A phone is almost like a tool, like a hammer, right? If you want a hammer to do something useful, you have to pick it up and swing it. But the next generation is essentially like having a jack of all trades by your side.

And you just say to him, "Hey, can you fix this window?" He'll just go and fix the window. There is no need to pick up a hammer and figure out how to use it. Start with this. And then you need to quickly create prototypes. So when we arrive, we have already developed everything that can be thought of. We printed this on a 3D printer. Which of our developments didn’t work, but were quite interesting ? Which of them, despite the failure, were quite interesting? The thing is that we are now creating a lot of different devices covering a fairly wide range.

We've been designing some pretty crazy stuff, so it's not like you look at it and say, 'It looks like this, it looks like this,' and then it looks like this. It's not as simple as drawing parallels. This is quite radical. But we are quickly creating prototypes of all this . We have a production workshop that does this . We have a whole design studio where we work on this. I want you to use these things over the coming weeks and months. I would say… "Either carry it with you or wear it, whichever you prefer, in the end we choose, and the choice narrows.

We had one of the biggest telecom executives in the world here, who helped Steve Jobs work on the iPhone 1. He was here two weeks ago and just returned from a meeting with Tim Cook. Tim Cook is going to leave Apple, but he was just at Apple, came here, saw our products and said, "Holy [ __ ] , man! This is the first time I see someone who can push the big players out of the market.” "Well, would it be fair to say that with the Archer and Hark figures, the difficult problem seems to be whether you can just mass- produce it?

The difficult problem is not this. We believe that the most important constraint that really needs to be addressed now is to create a truly intelligent robotic system that can power the entire world. There are many robots available for purchase today . You can buy a few in China and they will turn out to be complete junk. They can't do anything . You can control them using a joystick. That's all you can do. You press a button and it waves. He has no arms. It has some protrusions .

And you think, “What do I do with this? This is a toy." It's similar to what I used to do when I bought a DJI drone a few years ago, and I was just playing with it, and then the next day I was like, "What am I going to do with this?" "Yeah. Uh, it was a bit tricky to set up. It was n't working very well, well, not very well . I just crashed into a bunch of trees. It didn't work. I was like, ' What am I even doing with this thing?

" Robots are just like that right now. We can produce them in huge quantities, but unless they're very smart, they're not going to be of much use. We're trying to achieve true human-level intelligence. We really want to solve the problem of how to make it so that a robot can stand in any home and do any job that needs to be done. That's what we're working on. We think that's the biggest problem, the biggest gap in the schedule that we need to solve. That's the thing.

Also, people sometimes confuse consumer electronics manufacturing with automobiles. There's not a single major company in the world that's afraid to produce consumer electronics in large volumes with such high demand. It's simply possible. We produce billions of phones, almost by hand, in the world ." And with some automation. But cars are a whole different story. Companies like yours would die trying to make cars. There are a lot of companies like yours, and I haven't seen anything like it, you know, I was a commercial customer of BMW.

I haven't worked for BMW and several other groups, it's very difficult. The reason why making cars is so difficult is because is that you can't hold the part in your hand . With phones, you can always hold it in your hand , change something or something else, move it and hold it, but with cars, you can't. Physically, you can't. You need robots that literally hand the part off to other robots that install the parts on the chassis. And if any of them breaks, when you have thousands or 800 robots working, then your entire line will stop.

And it's like you're building a huge giant robot that builds a car. And with Figure, you can hold any part in your hand. So , I think if we 're between cars and consumer electronics, we're in a more challenging situation. Here. Closer to that level, you know, we're at about 40% here using mobile phones. We just built our 1,000th EVT robot for Figure 3 last week or the week before. When you say you 've built a thousand, that means a thousand that are going out to customers like BMW , or you're building prototypes internally companies.

What does that mean? We have two big clients. We have an engineering organization and an AI research organization that need robots, like, every engineer needs a robot. We need robots in every lab. We need to do tons of testing. We just need to do a lot of work internally . We might call it an engineering park, where we need to go, and the second thing is getting robots to customers. So we need to do both right now. This week, we actually shipped robots to our third client.

What do they do with customers? What, what, what can the robot do? What can it maybe not do at this stage? We do a lot... Right now, we do logistics, like packaging. We have experience in other industries, mostly in manufacturing and logistics. But we're also talking to companies in other industries. And so, when the product gets to the customer and, I don't know what you said, like packaging, sorting, tearing, or whatever , the robot just did a live stream on YouTube, and There were hundreds of thousands, maybe millions of views of people watching a robot sort packages on a conveyor belt.

Yeah, I've seen that. So , is that the type of work? It's kind of ... Give me an example of one of those jobs. It's just... Yeah, that's an example of one of the most challenging jobs we do. The client says, "Oh, that's great, because I can't find labor to do that. It's too expensive for people. This is much cheaper." Or is it just, " Look, it's not necessarily faster, cheaper, or better today, but it's an investment in the future, where in two years this cost curve will work out, and it'll be faster, cheaper, and so on." No, no, no.

The point is , they come to us and say, " We have a catastrophic labor shortage . We have a very high turnover rate. In some areas, it exceeds 100% per year. Finding talented employees is very expensive. We have a huge labor shortage. Talent is very expensive. Wages are rising, and we can't solve this problem. We can't automate all this work. And we need your help. We have the opportunity to earn good money on our contracts, and our clients get a very good return on investment from this ." Imagine a robot that can work multiple shifts a day , 7 days a week.

We will have a lot of uptime . The task you saw on the logistics line that we broadcast live was a real use case for one of our clients. This must be done within 3 seconds per parcel. And this needs to be done 5 hours a day, 5 days a week. We did this for 200 hours straight at a rate of 2.9 seconds per parcel. So we are already at the level of human speed. We're already doing it here. Now they are already getting a return on their investment.

And we are now in the early stages of implementing these solutions for clients and scaling. Over time, this will generate billions of dollars for these groups. Can you help me with the topic "truth is above all fiction " because one of the strange things is that the enthusiast or the layman who is excited about this future cannot understand it. It's very expensive or difficult to test, right? So I see a Chinese robot for 20 thousand dollars if I want to buy it. I have no idea what he's capable of.

I see Elon Musk coming out and saying, 'We're going to build a million of these robots next year. We are going to ship them out." And then you get a 1X robot and they show their hand and say, " Look at our hand. Look at this, this is the best hand you've ever seen." And then there's a service in San Francisco that sends a robot to clean your apartment, and they say, "Yeah, that'll work today." So, can you help me separate fact from fiction? It seems very complicated compared to most categories where I can just quickly try products online or buy them and test them out.

Firstly, the amount of information noise in the market, as you mentioned, is simply off the charts. There is so much crap on the market. It's very difficult to understand what's going on at all. So let me briefly outline what I think is most important and work backwards . I think the most important thing is the ability to autonomously deliver robots on a large scale into useful work environments. For example, so that they can cook dinner, wash dishes, make the bed, manage the supply chain from start to finish, work in healthcare, construct buildings, do logistics, and so on.

This requires built-in AI. You can control the robot and perform autonomous work. This cannot be solved with code . This needs to be done autonomously. This needs to be done over a long period of time. And you'll probably need to move around using your hands and move objects around the world. Do you know what I mean? In fact, everything needs to be done economically. Electrons need to be moved . So, I think at a high level, we're not interested in, for example, the best robot that can do a backflip, run the fastest mile, dance , participate in a parade, or run in the woods.

We are not interested in this. Man, I can't wait to see the figurine, like, taking a smoke break at the BMW factory. Yes. I could have been great in high school, but I messed it up. Now I work at the BMW plant. I used to joke with you, saying that what you started with Veteri was just a recruitment drive. And now you're working on projects that change the world. And you said: “Veteri Veteri really changes the world, and here’s why.” And you presented your project. He was very good.

You are very good at presenting projects . You are very good at attracting money. You are very good at being charismatic and convincing people. When you prepare a presentation to attract and convince people to change their lives, move from one place to another , trust you and start a company, how do you do it? And what was that presentation like for some of your companies? I mean, most of them are on the internet. For example, Figure's master plan is posted online on its website. Archers has also been available for a long time .

I wrote about this thinking that deep down I really want to find people who really care about this and are obsessed with it. And I think I spend most of my time not learning the details of the presentation. Most of the time I try to find such people. I found that even in the San Francisco Bay Area, which probably has the most people in AI and engineering, 90% of them are not fit for the job. How to determine who is suitable and who is not? I appreciate them all.

Yes. To do this, do you have to be technically as good or even better than them in order to rate someone? I need to know certain guidelines . For example, I need to know whether you did the work yourself or watched someone else do it. If you've done the work, it's like a scar you 'll carry with you . It was as if he had dug into you. You know all the details. You can talk about it freely. You don't even have to think. You will understand how to, so to speak, reconstruct everything you have done and discuss it.

Those who have n't done it can't do it. They just can't get through one stage, for example, and they constantly have problems. They can't talk about it . They don't know why. Out of 100 candidates who sound good, how many of them, whose resume looks good, do the recruiter consider suitable? Out of 100 candidates, how many do you think actually meet the required criteria? Let me give you an example. Here at A Figure, we have a very challenging recruitment process for Mechanical Engineer positions. You need to be able to create actuators from scratch.

There are bearings, motors, gearboxes, and other sensors inside the system. It is very compact. You know, this is a very difficult task. And the requirements are very strict. We've been doing 10 cases a week for 6 months and have n't hired anyone. This is madness. It's just crazy. But when you find a qualified candidate, and their offers from companies that are larger or more liquid than you are, they're in the tens of millions of dollars a year, right? The situation with AI is roughly the same.

AI is evolving, and it's all driven largely by Meta. Like in Hark, I've never seen anything like it. I thought maybe Meta had paid these people about a year ago and everything would have gone quiet. But they did n't stop. What's that crazy story you heard? We offered a job to a very experienced employee who came from XDI. XDI has completely disintegrated. Everyone just left about 6 months ago. It's just ...It's as if Macrohard has completely disintegrated. In general, a lot of things happened . We interviewed a pretty experienced guy from the AI ​​infrastructure department.

That was great. I think I gave him a very good Series A stake in Hark. It was, I don't know, 15-20 million dollars in shares. In 4 years? Over a 4-year period, we do five deals for my companies at early stages, and then move on to four a little later. So we're still at five. And you know, I thought, "I think we can grow Hark's market cap 10x pretty quickly." And I said, “ Okay, you have 15-20 million, I think 10 times that— several hundred million dollars.

If we have more time, you will have several billion dollars ." And I think we can do it. I think we need to...obviously it's going to be difficult, but I think we can do it. And he was offered to move to Meta for $36 million for four years. And he said, " It's practically guaranteed money. You know, I go there and I have to weigh it like this: $200 million in Hark, or $20 million , or maybe $36 million in Meta ." And he said, "I'll go to Meta." And they do it, every candidate we talked to makes up some absurd things.

They just don't stop. They've been doing this for about a year or a year and a half. They buy talent. They are making their way into the AI ​​race. What do you think about this strategy? You know, even if you hate her, do you respect her? Or do you just think it's a pointless exercise? What do you think about this? I really like this. I think in AI, I've found people who really understand how to do pre-training, mid-training, and post-training of language models, especially pre- training, and also understand supercomputers and data and evaluation and all the things that go into that, and the number of people who really understand the right recipes for what makes transformers work well, and, you know, in the MOE space or whatever, I think that's very hard to find.

It's actually very difficult to find people who really know their stuff. I think my rough calculations show that in California there are probably 20 to 30 people who can build really good AI models. Wait, does this apply to later employees too? You said there is an experienced specialist, but do even those who are not so experienced, 20-year-olds, young 30- year-olds, receive eight-figure sums per year? No, young professionals, guys 20 years old and a little older, earn several million in total. That is, the basic salary is 200-250 million.

They get another million or so a year in restricted stock units (RSUs). So they pay between a million and 750-2 million a year. And that, uh, was achieved thanks to Meta, but all the other labs followed suit. When I asked you what you thought about it , you said, " I like it." Were you making a sarcastic joke or saying, "No, that's actually smart, considering how hard it is to find such talented people"? I think it was really smart and I would have done the same thing if I were Mark.

I would buy myself a seat in the race. And I think that's what he's doing now. I do n't think I would do that. I want to understand this, and I want to find the right people first who are really deeply interested in this, rather than hiring mercenaries. Yeah. And he hired a bunch of mercenaries. They are driven solely by money . They came there. They know no one wants to go to Meta. They simply said that they were going there because they were guaranteed a block of shares with guaranteed rights to sell, just sitting around doing nothing.

And the point is, you don't need a thousand, 500, or 300 people to develop AI models. You can have a really good team of 20, 30 or 40 people. And you can achieve your goal without resorting to this. And those people are likely to be much more interested in the mission and where you are, and more committed to the cause than just throwing money at solving a problem. But I think if I said, "I think this was a really good strategy and it's working," I'd be like, "Hats off to you ." Great execution , their recruiting efforts and the way they structure it all, and I think it's paying off for them." It's still unclear whether they'll actually be able to release real products.

I think the problem with these bands is that they're traditionally not good at doing new things. I think Facebook will probably go down in history as one of the greatest acquisitions of all time, thanks to Instagram and WhatsApp and the different ways they've broken into those spaces. But, you know, if you look at Ray-Ban and everything they do, it's just not a great job. So I think the question is, how do you really do great work? I think we even use the Hark system now and it's so good.

It's far better than anything I use today. You must send it to us. Yes, can we use it? Well, yes . We can show it to you. Yes, sure. Preview of the study. There are about 500 PhDs there, and then there's me and Sam. Yes, exactly. No, well, just like on any other platform, it's Hark. What's the weather like outside ? I can answer that . Yes, no problem. But I guess what I'm trying to say is that every week there are five or ten AI startups or something like that popping up.

They're just not very good. This whole area has reached the point where nothing really good appears. I think programming is probably really great right now, but everything else is just not so great. On January 1st of this year, you made four forecasts for the year. I would like to know how you think they will be implemented. First. Number one: humanoid robots will perform multi-day tasks without human intervention in homes they have never seen before, controlled solely by neural networks. Long-term horizons, moving from pixels to torques.

How are we doing with this? On the right track, lost, or done? On the right track. You have 4 months. Yes, I think I see every day what we do. I think we are on the right track. The tricky part here is that we're already converting pixels into torque. This simply means that we take the video feed from the camera and get data on where to place the motor so that it fits precisely in the right place or joints. In general, we do it. The challenge is to get the robot into a new home where it has never worked before.

We are working on it. I work on this every day. I spend about 3-4 hours a day on this, 7 days a week. So if a robot action figure showed up at my house, what would be its problem? What would be her bottleneck now ? The fact is that he would n't know what to do , where to go, he wouldn't be able to accurately tune in and cope with my dishes. How would he bother me? We might be folding laundry, for example, but then moving to another location where we're folding laundry in a different place, with different lighting, maybe a different table height, different types of laundry, and different scenarios that the model has never encountered before, it's like she's stepping out of her comfort zone.

She does n't know what to do. It's like if you took away all the information about pyramids from the LLM pre-training, she wouldn't know how to talk about pyramids. Right. And we simply don't have enough of this kind of data. They are not available on the internet, so you have to collect them yourself. So we need to know how much of this data we need to collect from around the world to train a model to, for example, fold laundry at your home. Hey, stupid question. Why do all the robot companies care about folding laundry and doing the laundry?

Wouldn't it be more commercially viable to just say, "Hey, we're going to create a better warehouse worker"? There are already 20 million of them in the world, and this costs billions. And, of course, this will give us the opportunity to complete the development of the folding robot. But why do they even care today? Why not just do industrial work that people don't want to do, but that companies need done and are willing to pay for, and that's not like my house where there are all these other sensitivities?

Why do you care now? In the past, we didn't care about this. When we first started, we decided that we would handle the commercial side to recoup the costs of the house in the long term. And that was our strategy. It made a lot of sense. On the commercial market we can charge much more. It's much easier. The entry threshold is lower there. We work on a small construction site. We can work around the clock. It's much easier. What I realized now is that today all problems are solvable.

We can choose not to deal with this problem and just sit here and work in the warehouse, but neither I nor any of my guys want to solve it. We want to create a robot that can operate in any environment using only language. We want to be the first to do this. This could probably be done with 100 robots and a team of 50 people . So this company will reach a trillion dollar market cap overnight. Sounds good. Let's do it. We do it. This is what we will be guided by.

We are going to solve this problem. I think we will be the first, we call it " Solving Common Robotics Problems." Don't they attack people in " I, Robot" ? I don't remember this movie very well. Yes, don't worry about it. Okay, not about that part of "I, Robot." Who can win? Who can win in a fight right now? Can man still win? Yes, man can still win. Fine. What other forecasts are there? Another forecast. Um, one of the things you brought up here is that the use of AI will change.

People will move from text input to multimodal voice agents with persistent memory, which will become commonplace. This will bring AI closer to the synthetic human intelligence we imagined in science fiction. We do this at Hark. We will release it in a month. And our first version will get better and better. Um, I think we're on the right track. Have labs , like ChatGPT or Claude, ever published data on this ? I use voice features a lot. Sam, do you use voice features often? Yes, I hardly type at all.

Yes, I'm interested, it's probably already a huge percentage of people, to the point that all offices need to change something. These open- air offices that are so popular in startups are not so popular now, because I want to talk in private. Yes. Yes, many engineers now have microphones through which they whisper. They just whisper into their computers . Yes, I spoke to someone last night . I asked, “Claude, why am I so indecisive?” And then my wife said, "Yes." She thought, "Oh, [ __ ], man.

She can now hear everything I talked about with Claude." Yes, I know. I talk all the time , but it's so awkward. Yes, even the speech is still terrible. She's still not very good. It's like you're setting it up, you have to go there. It needs to be turned on. She's not very good at remembering what you just talked to her about . She has trouble calling up tools and using the computer. It's just limited, you only have to use it during a certain session .

I don't know if we'll succeed this year, but by 2027 you'll definitely pass a full-fledged human Turing test with speech. You will be able to receive phone calls from the artificial intelligence system on your phone. And I can deceive you. I can make it so that you can get a call from both a human and a robot, and I don't think you'll be able to tell the difference . This event is in 2027. I'm sure of it. Okay, what about the third and fourth? The number of school shootings has increased tenfold over the past 10 years.

By 2026, the first full-scale scanning system capable of detecting weapons from 20 feet away will be built and tested in a K-12 school. We will begin development of our first full-scale system in October. And I think we will raise this issue before the end of the year. I don't know if this will be at school. So we might skip this project for a quarter. Do you have a separate CEO who is leading this project, or are you also the CEO of this company? I have a chief engineer from JPL and NASA, a very good specialist, and this is mainly a purely engineering project.

There's not much to do in terms of business. There are, of course, some supply chain issues and all that, but a lot of the work comes down to whether it's possible to create a system that detects weapons well. It's partly a hardware problem, partly an AI problem. This is, in general, a large-scale task that requires a deep technological and engineering approach. And my entire team consists of engineers. They are really good . We've made some pretty big changes to Cover. We would have entered the market a long time ago, and I rebuilt the entire technological system about a year ago.

The way we developed it was that essentially I found a way to do everything very cheaply on silicon crystals and microchips. Reduce the price by about 90%, make it much more scalable, improve the functionality, and we've refocused. The problem was that it took about a year to develop and release our own chips. So, we just got these chips a couple of months ago, we've been testing them, and they're amazing. Now we need to produce more and it will take another 6 months. So now we are dealing with very long lead times for silicon chips and very challenging deadlines.

At some point we will get out of this business, but these are not the kind of chips you can buy in a store. These are custom-designed chips that no one has designed before. Then we had a special manufacturer in Europe who needed to make them. And it took about a year. Hi, it looks like you're doing great professionally right now . And I would like to know what you pay for the life you are living now? You are very optimistic, you seem inspired, but what are all these compromises?

Yes, about 5 years ago I had children and there was a problem at work: I think of my life as three parts. I have a job that is very dear to me, my family. I have three children. They are still very small. And then there are, as they say, other things to do, like having friends in town, the annual golf trip, the bachelor party, a wedding in Europe, or anything else on that list. And I felt like I needed to make a decision that I didn't have to do just one of these, well, I kind of realized I couldn't do all three.

What I really wanted was a family and a business. I want to be at the top in these areas. And so, in essence, I stopped at the third point. I don't do anything here anymore. So I had a friend I met in college, my roommate freshman year , and he said, "I'm in town for 10 days in the Bay Area, I want to meet up. I have n't seen him for a long time. It would be great to have some coffee." And I replied: “ Honestly, I don’t have time.

I can't meet." He said, "I'll find time, come." And I replied, “I literally don’t have time.” Every minute I spend away from one of these two activities is a minute I spend either with my family or at work. And I can devote a virtually unlimited amount of time to this. Can I ask about your work process? You joked when you said you don't use Slack. If it's convenient for you, could you show me your phone now? What's on your home screen ? What? What? What applications do you have installed?

All notifications. Well, you need to open it. Oh, what's there? Oh my god, you have a ton of messages. Because it's all, I think, from Slack and messages. I use Slack, but I just can't get through all the messages throughout the day. I have Hark that looks at everything, and Hark sends me messages with links to all the important things I need to look at. So how is your work organized? How do you work every day? Do you use a laptop or just a phone? Laptop.

Yes, a laptop very often. Both a laptop and a phone. I would say I now use Hark for all my AI-related tasks, from start to finish. Even to track what I'm doing in engineering projects, in recruiting, all of that is tracked, it's in my email, it's in my Slack. What about your to- do lists? All this is in Hark. Hark runs it all. What was before Hark? My to-do list was created in Google Docs. I had a document I called " Replanning" and it was constantly updated every week.

I usually came on Sundays and updated my plans for the week. And I updated them there. What about health ? Do you do anything health-related? Yeah, I've gotten access to some specialized doctors, for example, and now they basically send you for quarterly blood tests, full- body scans, cardiac CT scans—all that stuff . And this, frankly , is simply incredible. What's so incredible about this? The amount of data received and the significance of it all . For example, we know that if you get a CT scan of the heart for just $100, I think you can prevent heart attacks.

You will be able to do a full body MRI, and I think that will help detect cancer at an early stage. Many blood tests can detect abnormalities that can be corrected, improving your health. So, there are probably about a dozen such methods. Yes, but the solution to all these problems probably lies in something you are not willing to do. You may be willing to eat whole foods, but getting up, walking, and exercising may not be a priority. Yes, unfortunately, I didn’t have enough time for regular training, but, you know, eating right, like I do now.

Yes. Look, something has to give up. I can't sit all day and eat, but I need to go to work, which I love, I want to succeed in these companies. When you wrote in our pitch document, you said, "I put everything I had into my first three startups and basically hit rock bottom every year." Can you describe what you mean by "rock bottom " and how you deal with it? What conversation do you have with yourself or what is your entrepreneurial strategy when you hit such low numbers?

Yes. Basically, I experienced a constant shortage of money for almost 15 years. You know, in Battery we had a few twists and turns at the very beginning. In 2015, we raised approximately $500,000 in convertible bonds. At that time, I think I took out a loan for 50 or 100 thousand dollars. I did n't pay myself a salary because I lived in New York. I was completely broke. I was in the red. We attracted convertible bonds. It did n't look very good. And, I think, 6 months later we launched the marketplace.

It just took off, and a year later we sold it for 110 million. And I think the period from 2012 to 2017 was such that I was in debt. Things weren't going according to plan. And it's hard. What are you telling yourself inside? The internal monologue sounds like this: “This is terrible. It's very painful." I think at that point you just have to live day by day. You just have to live one day at a time. When things get really bad, you need to make a to-do list and just get through it .

The only way out is to go through it. So you need to make a to-do list. And you need to live one day at a time . You can't live for just one week. You can't live one day at a time without looking at Friday. You need to move forward, every day. All difficulties must be overcome. I was training for an ultramarathon and I hate running very long distances. I read a story about a guy who helped me, and he said, "All you have to do is pick something, it doesn't matter if it's 100 feet or half a mile, even if there are 49 miles left to go ." Just pick something half a mile away from you and tell yourself that once you get there, you'll think about giving up.

"And then you get there and you think, 'Okay, maybe I have one more thing.' And you choose something else, just 200 yards away. You think, "Okay, I'll think about leaving when I get there ." It was great, I think, that's exactly what I thought. But it was like, "Okay." Then I sold Veterinary and then I started working on Archer. I thought, "Wow, he made $110 million. We're 12 times bigger than all the venture capitalists ." And then I thought, " We're going to raise money. I'm going to raise money.

Everything will be fine." And everyone's like, " What are we doing? What are you doing?" For Archer? Yeah, everyone's like, "What are you doing? We're not funding this. What are you talking about?" How do you deal with that internal monologue when everyone's telling you you're stupid, wrong, and that it's pointless? Just go and do software. It comes from conviction. I know I'm right because I did it, I understand it. I'm on the edge. Yeah, but the odds are still against you, right? But that's it, it's a game.

When you play this game, it's like you sign up and 95% of everyone around you will fail. I remember when Vetergy and I started in the NYU incubator . I was so excited. We ended up in one of these incubators, and there were, I think, about 50 companies there. We started in Soho. That was great. We had a great time. I think if I look back, I think in about 5 years, me and another guy will be the only two people who made more than zero dollars.

Another team. 48" Companies collapsed. And I thought, "Damn it! " If you're in this business long enough, everyone dies. And it's everywhere. This has been going on for about 20 years now. I've been watching it. You see all these articles on TechCrunch and X about how people are raising money and all this stuff, and over time it all just disappears. And it is truly cruel. So, I had to buy a house and then invest all the remaining money into Archer. And I had blocked access to shares.

So even when I went to Figure, the shares were already unlocked. I financed Figure with Archer shares because I had no other money. The stocks were falling. Literally like a falling knife. I think they went from $10 to $2 and then went back up a lot. But I had to take action. I even need a second mortgage on my house to finance the Figure project. We asked you about one of your philosophical ideas, and you said, “I believe that doing complex things is in many ways easier than doing simpler things.” Can you explain?

Everyone is trying to do simple things. When you work on more challenging tasks, you tend to have less competition. A complex task probably means that it could bring in a very large market size and a very large return on investment if it works out . It's a kind of risk/ reward ratio. There are people who want to work on complex problems. Probably the best professionals in the world want to work in this field. In general, complex challenges have a binary benefit for investors. They really want to fund these projects because they can get 100x returns on the portfolio.

And I think there 's a non-linear scaling curve here. That is, I think many difficult problems are not 10x or more difficult. 100 times more difficult. I think that complex things are sometimes two, three or four times more complex. Maybe five times harder, but not 100 times harder . So you can get 100 times the return, but the cost may be three or four times higher. Let me give you an example from robotics. I think that making four-legged robots , like four-legged robot dogs, as opposed to humanoid robots, is probably three times more difficult.

Maybe at four. That's all. But actually, no, I don't think there's a real market for humanoid robots like these dogs. I think it's just a niche area. I don't think there is a real business for this. And I do n't know anyone who really wants to spend a lot of time on it at the moment. So if you're building humanoid robots, it's certainly three times harder, but the payoff is probably a million times higher. Probably a million times or a billion times higher return on investment.

What do I mean ? For investors, for people who want to work there, get shares, participate in the growth and all that. Why would anyone work with four-legged dogs? What economic value could a robot dog bring on a large scale? If you really understand, I think everyone is trying to do the easy job and it becomes very difficult. Look at all this AI crap, like the Open Claw systems that exist now. This is all nonsense. All this is bad. They will all disappear, and I do n't think any of them will survive in the long term.

There may be some consolidation, such as acquisitions, hiring, and so on, but this will continue until the end. Dude, you say so many absolutes. Has n't this ever caused you any problems? Do n't know. Mark my words. You're thinking, "Have you even used Open Claw since then?" No, I don't know how. How do you use Open Claw? I never...I never trusted Open Claw to set things up. I didn't intend to use it. I don't use it anymore. He's not very good. It was as if a wave had passed.

Do n't know. I'm just trying to say that I think the most important thing you can do as a founder is to think carefully about what you're going to do. Because you will spend the next 10-15 years on it. And this will, as it were, pave the entire path, the probabilistic path. It is a probability-weighted decision, weighted by the probabilities of different potential outcomes. But you are talking about a very specific game. For example, as you said, you think: we will become a trillion-dollar company or we will go bankrupt.

It's binary. Most businesses are not binary. You know, you're playing a game where the binary Result is whatever you like. But for many people, things are not so . For a lot of people, if they can build a really cool business that's doing $ 10 million in annual revenue , that's a huge success. Is this true? If they were able to achieve that level of success, and you were looking back at 70 or 80 years old, would you ask that same person, "Hey, you built a really cool business that was doing $5 million or $10 million in revenue .

You've been doing this for 30 years . You didn't do anything else . You have n't tried anything else while you were doing this. You just enjoyed working in this business. Would you go back 30 years and try to make a bigger leap? “Would you have done things differently than Vetteri? Vetteri was like that . Vetteri is my bridge . I worked in Vetteri for about 7 years. We literally created a marketing automation tool for ourselves within the company. And then, a year later, I thought, ' Wow, look at this.

This is outreach.io." And it was a billion-dollar company . We created it internally a year or two before. And then, watching all this , I thought: “Damn, we did some of this work ourselves, internally . It's useless compared to some other groups." “This whole question of what to spend time on is incredibly important for startups, given that I think startups are pretty simple. Even those who make $10 million probably fail 90% of the time despite trying. So I think it's just difficult. And I think, man, respect to those who reach five or ten million dollars in business.

It's really difficult . Especially to do this, having, perhaps, little or no capital . Hey, let me quickly ask you about other things you saw. I'm sure that because you're doing really interesting work, you meet other founders who are doing interesting things in, shall we say, unrelated areas . Not humanoid robots, but no less cool and interesting look into the future. I think you've probably seen more of the future than we have, and certainly more than most." Among the listeners. Can you tell me about something you've seen, heard, or read about a company founder you've met who does something that you recognize as exactly like this ?

The future is going to look exactly like this, and we're just...well, it's not evenly distributed among everyone else yet. I enjoy thinking about the problem: what will the world look like in 30 years? Where is everything heading? I think we have an energy problem. Not with energy consumption , but with its production . So how do we...Well, I mean both , but ultimately, how do we as a species produce more energy? I think there is a long-term trend here that is worth exploiting and it will really help.

And I think a lot of work has been done to compare this with people's standard of living . So I think there's a lot of work here on ...What is the next generation? Solar energy? Wind energy? Nuclear energy? There are also many different areas here related to thermonuclear fusion, nuclear fission, and so on. I think it's a really exciting area. I think it will take a long time, but we need really good entrepreneurs who can solve these problems. I think AI will dominate many areas in the next 10-20 years for all of us.

I think it will be 100 times bigger than the Internet. I think it will be so huge. AI will consume the entire internet. It will swallow him whole. And I think this will be a huge trend both physically and digitally. What products or companies are you currently considering that have not yet become mainstream but that you think have potential? Any good examples of what you're talking about? I mean, we're working on it at Hover, we can figure it out. Um, that's unclear. We are still in a situation where it is really unclear who will succeed in this matter.

We 're in a foggy area on a lot of these things, there hasn't been a breakthrough here. Well, there have been some initial victories and some initial breakthroughs, but there is a next stage that we still have to go through. And we are already in it. I think we'll learn more over the next year or two about what this will actually look like. But, um, I mean, I've used pretty much every AI device there is. We weren't very...Um, I don't know if you've seen anything like this on the market, but I wasn't, you know, thrilled with the product.

I like little things. For example, I like Whisper Whisper Flow, it has changed things quite significantly. How I communicate. Yes. Um, that was pretty cool. I think this has been the highlight of the last 6 months for me. And what about the last question, what about people who inspire you? I really admire people who are completely dedicated to their work. You will watch a documentary about Michael Jordan. He just says, "I just want to be the best in the world at this." I think it's the same for us startups .

And I think first of all, you know, what I can read, I never met Steve Jobs, but my god, the stories I heard and everything, this guy was just an incredible product-focused leader and founder. Um, I also got to know Jeff Bezos pretty well. He invested in Figure, and he's been here many times, and I think Jeff has been a really good conversationalist for a lot of the things we've gone through. Um, I had Jensen here last week, we're pretty close, and I think Jensen is just an incredible professional.

He's very hands-on, he has a unique approach to running Nvidia and his organization over the last 30 years, and I think he's doing a lot of really good things. What advice did Jeff give you that was meaningful to you? Jeff said that the last time he was here, he said, 'Look, you 're in a very interesting period because you've somehow figured out how to do this.' "And in a year or two, they'll figure out how to break through and really make it work on a larger scale, otherwise you're not going to make it , and it's time for you to take action.

You just have to connect and figure out how to break through it, make it work and scale it. And you are at a really interesting moment. I don't know how you got here or why, but you're here and you need to figure out how to take the next step, you know , you're in a big field now and your next big push will determine success or failure. So I think he's pretty much right. I think we now have robots that do this autonomously using AI models, it's crazy.

I think 4 years ago you would have said, "No way. " No way." Dude, 4 years ago I came into your office and your knee was acting up and I was like, "Oh, that's..." The knee. This is great. Anyway, it was just a knee. There were five engineers, and you're like, 'This guy just finished building the Tesla X or the Cybertruck or something . This guy did something incredible. This guy is a [ __ ] cancer. Look at how the knee moves and how the ankle dorsiflexes ." And we just sat there and looked at this knee, and it was incredibly cool.

I know. And now we have an AI that is working on a humanoid robot. We receive data from cameras. He draws conclusions on board. It displays information about the movement of all joints, you see, it's incredible. And it's crazy that it works. And, you know, the next step is to make it work on a larger scale. So, I don't know. He was great. I think it's ...I don't know. I think these are really good people who are worth emulating, who really love what they do .

Do it deeply and with full dedication. Well, Brett, I think it's time for you to get back to work, my friend. Great. Thanks guys. It was nice to see you again. Thank you so much, man. Okay, that's all for now . The podcast is over .

Go deeper

• All My First Million episode guides