The Exchanges

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 How do you tell what to keep, what to get rid of?

A So Open Door had gone fully remote. So at nine a.m. that Monday, I told the entire company, hey, Open Door is going to be back in the office next Monday. If you don't plan to be in one of these offices, we appreciate your help, but you're no longer with us. Then the second thing that happened was that Open Door had become a place where you succeeded by creating process, not outcomes. So a bunch of people chose not to be part of the company because we became so outcome-oriented, uh, because we demanded ownership. But the third thing that was impressive is, I'm just actually, that's been the best good surprise of Opendoor. I would, every once in a while, I find someone who I thought was exceptional. And I would honestly ask them in that first couple weeks, why are you still here? And it was a series of these people that had kept the company alive on their shoulders, and we've now rebuilt the company around these ICs.

AI assessment note: “a bunch of people chose not to be part of the company because we became so outcome-oriented”

Q Well, that sort of begs the question, how important are the words that we use to communicate ideas? Like, how do we make ideas more transferable in a company?

A Yeah, um, early on in my career, I used to ask everyone who worked on my team to read Orwell's, uh, essay, uh, Politics in the English Language. It's an incredibly good essay, because in it he talks about how English is among the most powerful tools ever invented by humanity in the English language, and I say as someone who was not, English is not my first language, but Orwell says English is in trouble because words have lost their meanings, and the word he uses is fascism. In the essay he says fascism doesn't mean anything anymore, because everyone's a fascist. Like, everyone gets called a fascist. Anyone I don't like is a fascist. Therefore, fascism doesn't mean anything anymore. I used to share this essay with people I work with because in order for language to be useful, we must hold it properly. And English is an incredibly powerful, powerful tool if you speak simply. And that is an incredibly important thing to learn inside a company. And it's incredible to guard against it because Most business talk is designed to confuse the listener, deliberately confuse the listener. I challenge anyone to find any sentence that includes the word leverage that means anything.

AI assessment note: “English is an incredibly powerful, powerful tool if you speak simply.”

Q How do you think about risk? A lot of people think, you know, for instance, that delaying a decision reduces risk.

A I think it almost always does the opposite. Like, there are very few decisions that gathering more facts Will help. I actually did this the other day. I was in a meeting, and some, one of her, um, colleagues said, hey, we're gonna run this A-B test, and then decide what to do. And I said, great. Guys, I'm from the future. I'm from six weeks from now. The A-B test was 51% positive. Tell me what you're gonna do. What most tech companies end up doing, AB tests are like the most, because people misunderstand, um, statistics and misunderstand, like, what mean average error means, and the compounding nature of mean average error, people misunderstand these things. What, for most tech companies, AB tests have done have been the safety blanket for management.

AI assessment note: “I think it almost always does the opposite.”

Q Why do you have to tell people to get shit done?

A So, let me tell you a story. Toby Lucas is among the most thoughtful people on the planet. There's a reason why people call him the founder's founder. If you walk the offices of Shopify, first of all, they're incredibly thoughtfully designed, but they all say on the walls, um, do things, tell people. And I always found that was odd. I actually haven't had a conversation with Toby about this, but, like, I realized that, like, in most companies, the order is reversed. Like, it's most companies is tell people, then do things. In most companies, this is what happens. You have an idea, you spend a lot of time writing memos about that idea, to get buy-in about that idea, before you can have a meeting, someone can approve that idea, And there's a whole process that takes many months before you can do the thing. There's a chain of command designed to minimize risk. And there's two ways to deal with that chain of command. The first is to do what Amazon said. Amazon became a default yes company. Most companies are default no. The chain of command is designed to reduce risk and say no to many things, right? That's what it does. Amazon said, no, no, no, we're a default yes company, which worked incredibly well for Amazon. But I think a better way to deal with this problem is to not have the chain of command Like, to minimize the number of people who can say no, and only have people that yo…

AI assessment note: “in most companies is tell people, then do things... chain of command designed to minimize risk”

Q super easy to schedule. I have it run every night while I'm in bed, and every morning I wake up to clean floors. Messes happen, and when they do, Matic rolls into action. Go to maticrobots.com today and put Matic to work for you. Do you think of AI as replacing people and labor, or do you think of it as amplifying the ability of a particular person I see?

A I think what AI does is two things. One, it lowers the, lowers the need for human management, right? If you look at any company, you end up with these very long tails of managers, managing managers. Before anyone does any actual work. And AI, because it takes the communication barrier down so much, we can share more information more broadly. It just collapses the company, um, chains of the company. The second thing it does, and what it has done, it opened door. Is AI fundamentally changes who wins and who loses. The people who use AI well will win, and there'll be more of them. Opendo has more engineers today than I did when I joined. Because we use AI well.

AI assessment note: “The people who use AI well will win, and there'll be more of them.”