The Exchanges

Q And so normally we're used to natural and this kind of begs the question, um, that we had talked about over dinner, which is where does knowledge come from? Yeah.

A So the knowledge that we human beings have that makes us so intelligent, uh, comes from a number of different sources. The first one, which people often don't realize is just evolution, right? We actually have a lot of knowledge encoded in our DNA that makes us what we are. Uh, That is the result of a very long process of weeding out the things that don't work and, you know, building on the things that do work. And then there's knowledge that just comes from experience. Uh, that's the knowledge that you and I acquire by living in the world, and that's encoded in our neurons. And then, um, equally important, there's the knowledge that, the kind of knowledge that only human beings have, which is the knowledge that comes from culture, from talking with other people, from reading books, and, and, and so on. So these are the sources of knowledge in natural intelligence. The thing that's exciting today is that there's actually a new source of knowledge on the planet, and that's computers. Computers discovering knowledge from data. And I think this emergence of computers as a source of knowledge is going to be every bit as momentous as the previous three were. And also notice that each one of these sources of knowledge produces far greater quantities of knowledge far faster than all the previous ones. So for example, you learn a lot faster from experience than you do from evolution an…

AI assessment note: “comes from a number of different sources. The first one... is just evolution”

Q And then the master algorithm is one view of where they all come together?

A Exactly. So each of these schools has its own master algorithm in, for example, the master algorithm, the connection, this is called backpropagation, because it's based on propagating errors from the output back to the input. And the Bayesian's is called, um, Uh, probabilistic inference. The evolutionaries have genetic programming. The, um, the symbolists have inverse deduction, and the analogizers have what are called kernel machines. Uh, the master algorithm would actually be a single algorithm that unifies all of these into one. Again, think of the analogy with physics. You know, so Maxwell unified electricity and magnetism and light into one set of equations. And now the standard model has actually unified those with, you know, the strong and weak nuclear forces. So the idea here is we should be able to have a single machine learning algorithm that can actually do what each of these five can.

AI assessment note: “The master algorithm would actually be a single algorithm that unifies all of these”

Q It seems like everybody's getting into artificial intelligence, artificial intelligence, machine learning from Facebook and IBM to Amazon and Google. Do you see, do you envision a world, like, 10 years out, I have a bit of a mischievous mind, so, where people are trying to feed other people's algorithms false signals to change the machine learning, or is that just crazy?

A Oh, this is already happening, and it's going to happen even more in the future, right? So what happens whenever you deploy a machine learning system is that the people who are being modeled Change their behavior in response to the system. Sometimes in benign ways, but sometimes in adversarial ways. A classic example of this is spam filters. The first spam filters were extremely successful. They were, you know, 99% accurate. They were very good at tagging an email as being spam or, or, or being a legitimate email. But then guess what? Once those spam filters were, were deployed, the spammers figured out ways around them. They figured out how to exploit the weaknesses of the spam filters and do, do things that would get through. And there's been this, you know, ongoing arms race ever since then, where the spammers come up with new tricks, the machine learning, you know, algorithms together with the data scientists come up with ways to defeat those tricks. And this just keeps going. And I think the same thing is going to be true in many other areas. In fact, two other areas where you can already see things like this very much happening. One of them is actually the stock market, right? The stock market is largely a bunch of algorithms trading against each other. And in fact, what these algorithms are doing Whether or not they know it is modeling each other. And what typically happ…

AI assessment note: “Oh, this is already happening, and it's going to happen even more in the future”

Q And you're talking about the Good Judgment Project. Can you maybe introduce us to that a little?

A Sure. Well, the Good Judgment Project is a research program that my wife, Barbara Mellors, and I Uh, started, uh, several years ago. Uh, it was supported by a branch, research and development branch of the U.S. intelligence community, known as IARPA, Intelligence Advanced Research Projects Activity, which models itself after DARPA in the Defense Department, and their mandate is to support research that has the potential to revolutionize intelligence analysis. So working from that mandate, they decided in 2010 to support, uh, a series of forecasting tournaments in which Major universities would, um, uh, compete. Researchers at major universities would compete to, uh, generate accurate probability estimates of possible futures of national security, uh, relevance. And, uh, we were one of the five teams selected for the competition in 2010. The tournaments ran from 2011 to 2015. They ended in June, uh, of this year. And, uh, the Good Judgment Project, uh, I am proud to say, was the winner of those forecasting tournaments. Uh, and I can explain more about what winning a forecasting tournament means later if you want.

AI assessment note: “the Good Judgment Project is a research program that my wife, Barbara Mellors, and I”

Q Excellent. Thank you. Uh, let's, uh, wind up with three questions that, I'm gonna try to ask everybody, at least for the start, so we'll see how this goes, but what book influenced you the most in your life, and why?

A Oh, that's hard. It's too hard to answer one, but I, I can mention probably a handful of them that, um, that were very influential. I mean, for, for, from a business point of view, certainly the book written by my mentor, Al Rappaport called Creating Shareholder Value was enormously influential. He's a, a dear, dear friend. We are collaborating on a project now, so that, that would be one. In terms of thinking, probably, uh, three books for me. One is, um, Dan Dennett's book, Darwin's Dangerous Idea. And I'm, uh, just a huge Charles Darwin fan, but this is this idea of how evolutionary thinking should permeate basically, um, almost, almost everything you think about. So, uh, Dennett, I think it did, it's a, it's a fascinating book. Another one I would mention, uh, Is E.L. Wilson's book, Consilience, right? Consilience really means the unification of knowledge, but the plea that Wilson makes in this book, to which I'm deeply sympathetic, is that many of the problems that we face as individuals and societally are problems that are going to be at the intersections of disciplines, and it's simply not enough now for us to, to use one discipline to try to solve problems. We need to bring people together to think across, uh, across those, uh, intellectual barriers. And the third one, which has also been a very big part of my life, is Mitch Waldrop's book, Complexity. And this really d…

AI assessment note: “I can mention probably a handful of them that, um, that were very influential.”

Q Aside from systems thinking, what are some of the mental models that you use to think through problems?

A I think one mental model that we've had in thinking through problems is kind of an irreverence, in a way, to the problem. You know, one of the decisions I think back to the early founding of Roblox, and I would call this a set of an irreverence, is the very earliest version of Roblox had a, what we would call a classic Roblox character We were saying, okay, we're building a user-generated gaming platform. We're gonna have an avatar. The very first Roblox avatar was, it was a very fast programmer art avatar, um, to get something working. Its proportions are ridiculously simple. You know, the legs are one by one by two. The body's two by two by one. The face of the early Roblox avatar was just programmer art that we put together. Universally, all of the gaming industry people we would talk to, whoa, whoa, whoa, do not ship that. Like, you need a full avatar, you know, you need a meshed avatar. It's got to be high res. Your objects need to look better. So we got that universally. I would say even within our team, we got that, but we decided it's better to ship that early and start getting feedback on it. I thought that was a great decision, and we found out, actually, that a lot of the input we had been getting, thank God we didn't listen to that, basically.

AI assessment note: “one mental model that we've had in thinking through problems is kind of an irreverence”

Q You mentioned a couple of false starts. What are some of those false starts where you guys went down the wrong path, and how did you realize you were going down the wrong path?

A Along the way to building an economy on the platform, we started with a very different economic model. It was a membership model. You got a few extra places to build on Roblox. In retrospect, that was a really bad idea because building should be unlimited. But along the way, that economic model started to fail, actually, and we had a situation where our user growth was going up, up, up, up, up, up. But less people wanted to buy this membership, and so this is really interesting, right? Our users are growing, it's all great, but money is flat to going down, and so what we did on that, I would say, is a very typical tactical response. Ok, something's up. Let's fix the current system. Let's make a list of 50 things that might be broken, 50 small tune-ups that we can do towards this system. Ok, let's do that. So as you would imagine, we made that list of 50 things, did like 30 of them. Oh, we still have the problem. Now what? But in the back of our head, we had been thinking about the, the economy, about virtual currency, about how this was like a whole outside of the box game changer. This wasn't just tuning up something that was arguably broken. This was like, this could be a hundred times bigger than what we're doing. And so then all of a sudden, like maybe in the, in a disparaging moment, like just, oh my gosh, you know, all of us in the company said, we just have to do this, l…

AI assessment note: “we started with a very different economic model. It was a membership model.”

Q I talked to somebody inside the turnaround who reports to you now, and he said, uh, you've killed more things in nine months than most CEOs do in a decade. What did you kill?

A In my first four weeks, I killed two entire business lines. Opendoor used to have, essentially, we provided general contractor services to other companies, which was, it was a profitable business for us, but it was not the mission of our company. Our job isn't to become the world's best contractor, so we killed that. We had a business where we were essentially called OD Select, which we essentially were, like, Quasi builders. Like we would take essentially homes that were like unlivable and just build them from scratch up. Profitable business side for us. Killed it off. Because again, that's not our job. They're our builders in the world. We're not one of them. We're a market maker. It's a different job. If you think of my time at Opendoor, I think it's reasonable to say I've only started three products. And we've shut down a few dozen.

AI assessment note: “In my first four weeks, I killed two entire business lines.”

Q Do you think it's a turnaround or like a refounding?

A I think it's a new company because AI has made this a different company. Because look, what, what about Opendoor's business is difficult? The operational complexity of the physical asset. And what do you need to make that operational complexity go away? You need very low cost, fast decision making at scale with data. We have built the AI exoskeleton around every human being at Opendoor, such that they can be three, four X more efficient than everyone else. To give you a sense, the last time Opendoor bought as many homes as we bought last quarter, Opendoor's OPEX was more than twice as high. In a world where prices have gone up for human beings, we've become twice as efficient.

AI assessment note: “I think it's a new company because AI has made this a different company.”

Q What are the most powerful mental models that you find yourself coming back to over and over again?

A Actually, you, you know me a bit, so you know this is true. Uh, I, I have these written down. First, friction is underestimated. Two, uh, map is not to train. Three, truth over feelings. Those are like that. Three. So let me give you the first one. Everyone has seen, uh, the supply and demand graph in their life, and, but people underestimate The, uh, invisible hand of friction on where those two things cross. The total market size of Walmart would have been massively underestimated. Same with Amazon, same with Google, because before these things existed, everyone thought the market was smaller because there was so much friction in the market. When you reduce friction in a system, you just get a lot more of whatever you reduce friction on.

AI assessment note: “First, friction is underestimated. Two, uh, map is not to train. Three, truth over feelings.”

Q So how does Opendoor make money then? Like, how do you become insanely profitable doing that?

A We make our money in two ways. First, we make money in each transaction, relatively thin margins of each transaction. But we make our real money from the first derivative of the business, which is a mortgage, Insurance, title and escrow, all the things that go with the home. So our job is to buy and sell the home at fair prices as fast as we can, and make our money on the services you get. This, by the way, is how we built Shopify. If you look at Shopify, you can buy Shopify, the software, for one dollar. Like, and it's one dollar because you can't possibly be less. Like, credit cards won't let you charge less than a dollar. And Shopify makes most of its money by you succeeding on Shopify, by its services it provides you. And this has always been true. People misunderstand how businesses work. It has always been true that some of the biggest companies in the world have been built on the first derivative of the core business. Google does not make money from you searching for things. Google makes money from showing you ads while you're searching for things.

AI assessment note: “We make our money in two ways. First, we make money in each transaction”

Q Are there any other questions that come to mind that you think are particularly revealing about people?

A What is the biggest obstacle you've ever had to overcome in your life? The follow-up there is, what resources Did you lean on overcome that obstacle? In other words, how did you overcome it? So people stop at what obstacle did you overcome? Where it becomes interesting psychologically is how did you overcome them? So going back to a really important interview question is, well, tell me the hardest thing you've ever overcome. And then what did you enlist? Did you go to other people? Do you have a network? What beliefs did you engage? It's really important to understand the method by which people overcome their obstacles. They have to rely on themselves, and this is what Tiger Woods was great at for a period of time. It's like, hey, here's, here's what his self-talk was. You got yourself into the, into this mess, now get yourself out of it. And that's the best version of Tiger. Obviously, we've seen some other versions of Tiger, but the best version of Tiger is Tiger telling himself, you got yourself into this mess, now get yourself out of it.

AI assessment note: “What is the biggest obstacle you've ever had to overcome in your life?”

Q What does it mean to know the bedrock of the industry? We live in a world where people scam, they want the gist of things, they want, give me the summary, give me the executive summary.

A I'm gonna tell you a story. So, um, my partner at Benchmark, Alex Balkansky, would go to this charity auction that I think Andre Agassi would run. In, in Vegas. And one year he bought a dinner with John Lasseter, the creative genius behind Pixar. And we go to John's house and he serves us in his movie studio. He serves us in his viewing room, a 10 course meal. And each piece of the meal is tied to a classic cartoon that he believed was, was super important to understanding animation. And he would show it and he would talk through it and explain it. And you see that and you're like, holy crap. Like he knows more about the history, you know, and, and then here's another data point that I just love. There's a, you know, world chess tournament and they take a break and run a trivia contest and Magnus Carlson wins the trivia contest. And it's all about the history of chess. We do live in a world where information is really cut up, but we also live in a world where you can have access to more information than you ever could, and that's even more true now with LLMs. I mean, you could just sit there, you could have an hour drive, and you could sit there and talk to OpenAI and learn about anything you want to, and I think more people would benefit by studying the history of whatever field they're in. There, there's another one that we mentioned is, uh, Picasso Was a wildly successful re…

AI assessment note: “more people would benefit by studying the history of whatever field they're in”

Q You had a falling out with your dad, and that seemed to be one of the pivotal early moments in your life. What happened?

A My dad, we all kind of played a role in his movie, and he loved being in a fraternity in college, and was like, defining for him, and he wanted me to do that. I wasn't into fraternity, and they weren't into me either. It was like, mutual. And, and I initially went to this Big 10 school, University of Michigan, and I was just not in the right place in any way. And I just was not, On lots of fronts, becoming the kind of man that my dad wanted me to be. He wanted me to be just like him. This got more and more tense, and our family was on a sailboat, uh, in the Caribbean, in the Virgin Islands, and we were bare boating, so it was just us crewing, and my, we had grown up sailing, and my dad was the worst sailboat captain ever. I mean, just famously, like, epically bad captain. Like, we would get stuck On a sandbar, because he didn't read the tides right, you know. We were in this harbor in, uh, this island, Virgin Gorda, and the keel got stuck, and we always had something go wrong, and the boat was going in circles and headed eventually for this, like, rock barrier, and I got in the dinghy, and I turned on the engine and grabbed the rope, and I pulled the boat Into the slip and save the day. And my dad was furious, and he said, you could have killed us all, and there could only be one captain, and, you know, I think he was also a little humiliated, and that escalated to this bigger …

AI assessment note: “I've decided to take you out of college to finish raising you, and I said”

Q I know you're all thinking electrolytes are for athletes, but you don't have to be an athlete to benefit from it, and it tastes great. Stay sharp in the afternoon and grab a free eight-count sample pack with any purchase at drinkelement.com slash TKP. That's drinkelement.com slash TKP. You had a near-death experience with Facebook and Zynga. Can you walk me through what happened there and how you navigated that?

A The entire Facebook experience for Zynga was a near-death experience, and not just us, for everyone in their app ecosystem, it was the least stable app ecosystem ever imagined or invented, and many companies did die in it. Most. Um, there's only two companies that ever actually survived out of it, it was Spotify and Zynga. Um, and I'd say my experience in building Zynga, and the reason I kept raising money, even though we were profitable, was I felt like I was on a five-story high unicycle, and I kept adding another story. And it was like, whoa, it's gonna be really far if we fall now. And we didn't have any stable, there was not a stable platform. We didn't have a stable agreement between the companies. They could and would Change their platform all the time, and some low-level product manager who wants to promote events would deprecate the whole left rail of their homepage where all the apps were. Like, they'd be behind a more button, and you're like, oh my god, nobody can find us anymore, you know, or it would just always be moving. They had to move fast and break things. Most of what they broke was all of their app ecosystem. I would walk their hallways Every week, trying to convince them that games and apps were this great business for them, and they didn't believe it. They didn't think that was what their platform was for, and then eventually they did, and it wasn't even …

AI assessment note: “The entire Facebook experience for Zynga was a near-death experience”

Q And so what happened next? You get this, you get the response from the lawyers and you're like, holy shit. And then what?

A We, so we got all of those outputs and, uh, the thing that was happening kind of like simultaneously is Gabe had some friends, um, that were thinking about, uh, joining this company called OpenAI. We basically said, hmm, we're using GPT-III to do this. Let's just cold email Sam, um, Sam Altman and, uh, the general counsel at the time, whose name is Jason Kwan. I think he's the chief strategy officer now. And we emailed them the results and we basically said the, I mean, the body of the email, Was basically, did you know that the models were this good at legal? And we emailed them that they got back to us like pretty fast. We met with Jason first and just kind of like went through what we did in the process with Reddit. We showed them like our chain of thought prompts and things like that. And then we met with the rest of the leadership team on actually the fourth of July and we pitched them our idea for the company and we raised money after that.

AI assessment note: “Let's just cold email Sam, um, Sam Altman and, uh, the general counsel”

Q Or do you capture it or do you just let it go? Like, what is that?

A When I get really bad at it, I ask my chief of staff to force me to write a paragraph about why I'm going to take a meeting, full paragraph about why I should take a meeting. And it's so easy if you do it that way, because for 99% of meetings or events or whatever, You start writing the first sentence and you're like, I don't want to do this. And if you don't want to do this and you're like, this is a waste of time, probably the meeting or the event is also a waste of time for the things that I think are really important when I'm like, I got to write that paragraph. I could write 20 pages.

AI assessment note: “I ask my chief of staff to force me to write a paragraph”

Q What was the darkest moment as you were building this when everything felt like it was just caving in and you were about to give up or quit?

A One of them was super early on. We thought, we, we basically thought that there was gonna be a way to, like, one-shot ourselves into, like, building the company instead of doing it the hard way, and the way we, this was, like, um, early twenty-twenty-four, so about, like, a year and a half into the company, a little bit less. The idea was basically we were gonna buy this company that ended up, it was 10 times bigger than us, people-wise, um, and about the same, like, we were gonna, trying to buy the company For the same, a little bit higher valuation than we are valued at. We tried to buy it. We kind of like signed actually the deal, um, before we had the money. So like what old, uh, private equity funds used to do, uh, right? Like KKR used to do this all the time, where they'd basically like sign the deal for the leveraged buyout, and then they'd go raise the money. Um, it was a very common thing they used to do. And that's kind of what we did. And so we got the term sheet. We like locked them up for a certain amount of time, and then we went to go, uh, get the money. And of the money we were short, uh, uh, we got like a lot of it. We were trying to raise, um, like around like seven hundred million or something like that. And we got, I think like 500 in clean equity. And then the deadline was there. And we had an option. We could have taken like a, a loan basically with like p…

AI assessment note: “One of them was super early on. We thought, we, we basically thought”

Q Oh, is that why you guys stopped showing reasoning?

A That is part of it. So there's two reasons. One is to think about distillation, but the second, in some ways more important, is that we had this insight when we first developed the reasoning paradigm that it gives us a interpretability mechanism we had not been anticipating, because you can really read the model's thoughts. You can see exactly how it got to an answer, so you can interpret How, like, what was actually motivating that answer? Now, the problem is, if you train the model to have a chain of thought that looks good, then you lose all the faithfulness, right? It's just going to be like, the model knows that part of the answer that is desired is for the chain of thought to look a certain way, and so it may not be representative of how it actually arrived at that answer anymore. And so we were, we made an early decision to say we want to avoid any temptation to train these chain of thoughts to look at Favorable to look like something you could present to a user, and so that really made us lean out for multiple reasons, for competitive reasons, for safety reasons, from the idea of showing these intermediate thoughts.

AI assessment note: “That is part of it. So there's two reasons.”

Q At what point did you realize that like this nonprofit thing just wasn't going to work?

A In 2017, we started to think very hard about, first of all, how do we really achieve the mission? How do we actually build an AGI? What will that look like? And we started to do the math on compute, and you start to realize that it's gonna take a big computer, and we came across a company called Cerebris, which was building a unique piece of computing hardware, and the kind of computer that they were promising, we realized was going to be far advanced of where our compute calculations looked. As you start to realize if we could buy a lot of those computers, we can actually probably succeed at building an AGI. If we could get exclusive access to Cerebris, that could give us an overwhelming advantage. If we could buy very large data centers, that could be something unique as well. And the thing about nonprofit fundraising is I think that there is essentially a cap to what is possible there. And so 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. And so we were committed to that direction, and that is something that we knew was the only way to achieve the mission.

AI assessment note: “In 2017, we started to think very hard about”

Q And if I was an employee and I'm working there and I'm on the dock and I want to see my specific performance or a driver, can I do that?

A A hundred percent. So the way we make it work and we implemented those tools as well. And you know, there's one thing, everybody goes to work every day wanting to do a good job. And when you get feedback and positive reinforcement, obviously you're gonna do it in a more effective way. But every time one of our drivers or dock workers Scans a pallet on the dock, they automatically see two dials on their handheld. The first one tells them what their productivity is versus their peers, so they can see effectively if you had 10 people working the dock, are they number one or are they number five? And if you're number one, there's a lot of bragging rights that go along with that, but that's more on the productivity side. And then the second dial shows them how many damages they could have caused by either not strapping the pallets or not. So every time you scan a pallet, To move it across our docks from one trailer to another trailer, we actually show folks what their numbers would look like. And similarly, the supervisors, they have a system as well, very modern platform, all proprietary, our own engineers build, build these solutions where effectively they can see then the performance across the entire workforce or a, or a certain shift and how they can actually manage that more effectively.

AI assessment note: “A hundred percent. So the way we make it work and we implemented those tools”

Q On the notion of risk taking, how did you navigate the yellow bankruptcy? Somebody reached out to me, a mutual friend of ours, and told me it was pure genius. Can you tell us that story?

A Yeah, so one of our large competitors went bankrupt in twenty-twenty-three. Very sad, obviously, for the company and for, for the folks, for the folks involved. But at the same time, as a less than truckload network, one of the most important forms of capacity to grow and service the customer is the terminal network that you have. So how large are your buildings? And some of these buildings, Jane, are, are a mile long dock, effectively. So very, very large facilities. And having enough capacity in key market is incredibly important. And it's very hard to be able to permit and build a large trucking terminal. So effectively when, when Yellow went bankrupt, all of that real estate came up for sale and we had to figure out, well, how much we want to buy and in one markets we want to buy. So in that particular case, and I was a new CEO, Going in there, the question is, well, how much do you buy, and what markets do you buy them in? How do you justify the return you're gonna get on the capital that you are deploying? One of the things I learned from Brad very early on, as a CEO, one of your most important part of the job is to make sure that you are allocating capital in a way that can create the most amount of shareholder value over time. Now, when yellow went bankrupt, we started looking at, well, okay, if we look, what is the data telling us? Based on different markets, where are…

AI assessment note: “when Yellow went bankrupt, all of that real estate came up for sale”

Q I want to come back to the family part there for a second. So you have a wife, you have three kids who are all young. How do you run a business and have time to be a part of the family?

A So I mentioned earlier on that one of the, the most important, um, things that as you see, as a CEO you have is where you spend time, where you spend your personal time and then where you have your organization spend their time. And I think I once get, got a very good piece of advice that when you're trying to do a lot of things, naturally you have less time, but you also have to simplify the goals and the targets and the things that you are driving to what is the most important. So when it comes to family, you know, obviously when you have, when you're doing a lot of things at work, you have a limited amount of time, but I try to make it the most amount of meaningful time that I can. So when I'm with my kids, for example, I don't use my phone. I put it away. I do meaningful activities with them as a engineer. Last summer, we actually built a robotic arm that we programmed and we actually tried to make it solve a Ruby cube. It didn't, didn't work as expected, but we kind of give it a fighting shot. There, or whether it's taking my kids fishing and spending time with them, actually doing activities that are outdoors. So I think when you are limited on time, what you can focus on is making the quality of that time be the highest possible. And also figuring out where you want to spend your time. So for example, I don't watch TV. I don't play golf. I don't, you know, have for me, I…

AI assessment note: “I try to make it the most amount of meaningful time that I can.”

Q Is that real time? So like when I take that photo, it reminds me of something, or is it sort of post when we aggregate the data the next day, you're like, Hey, this truck went out yesterday. You didn't see this, uh, just a reminder to strap it down better.

A Historically, we've done it where you are looking at it after the fact. So historically we met, we manage all that data through the shift. You are still getting it in the system in real time, but the person taking the photo was not getting immediate feedback. So let's say I'm a supervisor. As I close the trailer door of a given trailer, I would take a photo. And then I would actually rank the quality of loading of that particular photo. So in that case, it's not AI doing it. And then the next day we can tell which super, because then the next person in line in our, in our terminal network, when they open up your trailer, they take a photo and they rank your quality of loading. So you're effectively getting somebody, if you think about software development, it's almost like doing a code review, but you're doing a trailer review for the next person who's actually opening up Opening up that trailer. But now with AI, with the new versions we are launching, we actually are intended. It's not, it's not live yet. It's gonna go live, um, over the next few months, but effectively as a supervisor, whenever I take the photo, AI automatically tells you what it found wrong in your trailer. And now we can actually fix that before you actually send that trailer on the, over the road.

AI assessment note: “Historically, we've done it where you are looking at it after the fact.”

Q How do you teach the life skills? Are you doing anything different there than we would think?

A You would, you would think these great project-based workshops in the afternoon to teach these skills. So we're like, okay, how do we teach one percent better? How do we teach atomic average? Like, okay, let's have them run a, you know, five K. Uh, you know, we teach financial literacy. It really is like our fifth and sixth graders launching food trucks, running Airbnbs, right? And so you can teach them when you have all afternoon, you have all this time to teach them. Now, the biggest thing we do versus probably most others is we try to quantify everything. Like, a lot of life skills are very soft, right? And that, I, I can't run a scalable system if everything's, hey, you know, maybe it works. So like, we talk, grit. Everybody wants their kid to, you know, have grit, right? You can't read, like, Angela Duckworth's book and, ok, grit, right? That doesn't mean anything. So every third grader in our school can do a Rubik's cube. Every kindergartner can do a hundred piece puzzle. And so we, right, every eighth grader Has passed a teamwork grit test, which is they all run a Tough Mudder, if you know what the numbers are. Tough Mudder and pass cross the finish line at the same time.

AI assessment note: “the biggest thing we do versus probably most others is we try to quantify everything”

Q And then you ended up buying it back. I want to hear the story about that. And then what changed your mind about being public with Trilogy?

A Uh, the PCOrder.com. The story helped me realize I didn't want to be public. So PC order was, um, actually one of the trilogy co-founders, Christie. You know, had this idea of, okay, I'm going to build PC order, um, which was online selling computers online back in.com days and, you know, built it up super successful, had all the biggest resellers aboard. Um, and then she went public and then basically Dell put her customer base out of business. And so then she, she, her business went terrible. And so the stock went from 50 to five or something. I can't exactly remembers, but you know, down 90 90 plus percent. So it went public, uh, you know, ran up in all the hype and then imploded, uh, and then we bought it back and just, you know, said, okay. Um, and you know, for me, uh, so I've kept Trilogy private, you know, to this day, it's still private. And a lot of it is just, um, I prefer just to be a hundred percent in control.

AI assessment note: “The story helped me realize I didn't want to be public.”

Q You mentioned sort of getting to 90%, and that is, that's sort of the target you want to get to, and you want to do 10 deals. How do you think about de-risking deals? Do you isolate particular variables, or how does that work?

A So much of what we do at Brookfield is, um, de-risking different business activities in such a way That we can turn the construction of a project or the operations of a project into a long-term inflation-linked stream of cash flows. We are very comfortable taking execution risk, operating risk, development risk. We don't like to take market risk, and we work very, very hard to structure our deals or execute in such a way that we're, we're not taking market risk. And I'll give an example of that. When you, uh, build a renewable power plant, let's just say a solar farm, there's really four key drivers, uh, of what your end return is going to be. It's your construction cost, it's your revenue offtake, your power purchase agreement, it's your EPC and your financing. We are very fortunate to have built one of the largest, um, renewable power operating and development platforms around the world, Whenever we build a new project, we do not like to put capital in the ground unless we lock in our CapEx contract, our off-take contract, our EPC contract, and our financing contract all at once. Because if you lock in those four things and you execute, it doesn't matter if interest rates go up or down, you've locked in long-term financing. It doesn't matter if power prices go up or down, you've locked in a long-term contracted revenue project. It doesn't matter if inflation goes up or down, …

AI assessment note: “we do not like to put capital in the ground unless we lock in”

Q How do you think about leverage? I look at what's going on in real estate today, and so much of that seems to be people got a little bit over their skis, which creates an opportunity for you, but how do you think about going into that with leverage, where you're sort of trading off a little bit of financial return for survivability over market variations?

A I like the way you said that. There's two things about how we finance our businesses around the world. One is just the approach we take. We focus on asset level, Non-recourse, long-term fixed rate financing. It's sometimes not the cheapest financing, uh, but it has some features that we really like. It takes away that market risk that we talk about, about interesting interest or financing cost changes over time. And the other thing is we like to do asset level non-recourse financing. That by choice is harder. You're doing a lot more individual financings rather than just grouping huge Portfolios of assets and putting a debt facility over the top of them, but what it really ensures for us is if you ever run into something unforeseen, and I say something unforeseen to the downside or equally something unforeseen to the upside, everything that you have to work through is done on an individual basis, and you're never, um, tainting, if you will, An entire portfolio with the dynamics of an individual asset. And obviously people will focus on if you have an asset goes bad, it's nice if that doesn't taint a broader, but it's the same on the upside. If you get an incredible bid for a single asset, but it's stuck in a debt facility that won't let you release it, that inflexibility is not helpful to running your business. So the first thing we do is focus across all of our platforms, non-…

AI assessment note: “We focus on asset level, Non-recourse, long-term fixed rate financing.”

Q I think the story is you were told to go to renewables, not necessarily asked. How did you feel about that?

A So this was in Concurrent with that move, I, I switched teams, and, uh, people always say, oh, did, did you want to join renewables? The, the honest answer is, no, I didn't have some weird desire to, or some strong, specific desire to go to renewables, but Bruce and, and Cyrus Madden, who built our private equity business, asked if I would, and I, of course, said yes, and if they'd asked me to go into infrastructure or real estate, I'd probably have a different business card today. I love the firm, and, I do whatever they asked me to. Again, I was very fortunate that I joined the renewables team, you know, in the early innings of what has been one of the largest and fastest growing industry builds in history.

AI assessment note: “I didn't have some weird desire... but Bruce and Cyrus Madden asked”

Q There's two paths I want to follow here. One was, you mentioned consensus, but the example that came to mind for me was Westinghouse, which seemed like a very non-consensus idea, at least from the outside. Looking at, now it's proved out to be very correct, but how do you think about taking bets that are maybe non-consensus?

A Westinghouse is probably a good example of, sometimes we get asked, what's an investment committee process, or what's that iterative process like? We focus a huge amount of time, the vast majority of the discussion will be focused on the downside. We like to believe that if you buy high quality businesses in good markets that have strong downside protection, If you underrate the worst case scenario really, really well, the base case or the expected case will end up being very attractive. And, and Westinghouse was a great example of that. You know, when we initially invested, it was not an in favor sector by any means, but it was a market leader. It was critical to the global supply chain of nuclear power, which at the time was not growing, but had a very, very Long life tale to it, of which Westinghouse was a critical supplier, and we felt it was an industrial operating business that could be run better using some of our operational expertise in other industrial businesses that could, could be brought to bear, and I can tell you we spent all of our time focused on the downside, and what was interesting is that proved out to be right, you know. Westinghouse is a market leader. It is absolutely critical to the supply chain. We were able to drive, drive significant operating efficiency within that business. All of that would have led to a, a very good outcome. And then we got the …

AI assessment note: “We focus a huge amount of time, the vast majority of the discussion will be focused on the downside.”