Relentless
00:40:41 · Brian Armstrong
[00:40:41] S0: Well, think there was this initial panic or fear about AI generated content and like, we wouldn't know anything was real. Argument always seemed like a little funny to me because if you grew up on the Internet, you're very used to seeing things that are not true. Like, yeah. I mean, or just bad takes on Reddit or whatever. So now I think it's kind of been like a total nothing burger. Like, see AI things and sometimes they're pretty funny and sometimes you know, they're really dumb. And sometimes you see something and, is that real? And then there's actually a community note or a bunch of discussion people saying, no, it's not. And... But this would... This is the thing that is real behind it. So I actually think it's, like, not as... Not only not only is it totally fine, it's actually, like, great. I think the world is more interesting because we are getting... Like, the barrier to do creative content has gone down, and so we see more satire. We see more things that are funny, whatever. Like... So I'm not worried about AI slop. I think, you know, I think that it's it's fine for our democracy and everything like that. And then it's interesting your tie into prediction markets. I mean, prediction markets are a really good signal of what is true in the world or going to be true. It's probably... I think it's much higher signal than, say, like traditional polling, which can be biased or certainly like mainstream media. So I do think prediction markets are... Have a positive externality in the world of just helping people try to get truth about, you know, is the the Strait Of Hormuz or whatever going be open in a month? That's an important thing for a lot of companies. So prediction markets create a lot of value for that. I actually think that in a strange way, AI agents might be useful to resolve prediction market contracts because if you set up a contract today, you have to put the resolution condition as typically like, okay. Of these three traditional media, you know, companies, if they... If if two or three of them or whatever report that this person won the baseball game versus this one, then it resolves. And there's a human being typically behind the scenes, like checking to see if that CNN published this thing.
00:42:39 · Brian Armstrong, Ty Morse
To me, that's less reliable than, say, you could say an AI agent, you could say, okay, between Gemini, Grok, and OpenAI, I'm going to ask on this date, at this time, I'm going to ask all three of these agents who won the baseball game respond with a single answer. And if at least two of three agree, that is the answer. And so an AI agent can actually have a preponderance of evidence across the whole Internet and resolve that, I think, more truthfully than one media organization, let's just say, in an extreme case. So anyway, I think there is some intersection there, but I'm glad AI agents are I think overall on bet... On net, like, AI agents are gonna help us establish truth a lot more in the world and so will prediction markets. [00:43:17] S1: I think it's just like, the next evolution of the Internet, and the Internet was really good at expanding truth. [00:43:22] S0: People forget you get, like, biased or untrue answers from... When you ask humans a lot too. Right? I mean, this is that kind of like we were talking earlier about simulating that person's brain in your brain. But when you go ask the person that question, like, they're bringing all kinds of biases, misinformation. They're like, oh, I want to seem like I know the answer to this even if I'm not 100% sure. So like, we all... We already live in a world where you have to, be a little skeptical. And so I think AI agents so far are, like, exceeding the the average human bar. So I think it's net positive.
00:43:22 · Brian Armstrong, Ty Morse
[00:43:22] S0: People forget you get, like, biased or untrue answers from... When you ask humans a lot too. Right? I mean, this is that kind of like we were talking earlier about simulating that person's brain in your brain. But when you go ask the person that question, like, they're bringing all kinds of biases, misinformation. They're like, oh, I want to seem like I know the answer to this even if I'm not 100% sure. So like, we all... We already live in a world where you have to, be a little skeptical. And so I think AI agents so far are, like, exceeding the the average human bar. So I think it's net positive. [00:43:53] S1: I think it's really easy to hold other things to a way higher standard than you hold yourself. Yeah. If you had to look at the way that you operate Coinbase today and then you looked at your actions versus maybe a set of your peers like Sam Altman or Brian Chesky or Joe Gebbia, someone else, would run their company, what do you think they'd be surprised by?
00:44:14 · Brian Armstrong
[00:44:14] S0: Well, there is a common set of tactics that I think many of us use. Right? It's like we'll run an exec meeting, and we do an off-site every quarter, and we try to typically use OKRs. And... But I think there are some differences. I'm trying to think, like, at Coinbase, the one that might surprise them the most. I mean, the most surprising thing has been some of these decisions that we made, right, about being apolitical in the workplace, like suing the SEC. I think most of those CEOs you mentioned probably wouldn't have done that unless... Maybe there's a situation where they would have, but it probably wouldn't have done that. Underduress. Yeah. You know, and I think those are all great CEOs, by the way. I... I've learned a lot from all of them. Like, Chesky had a great thing with his founder mode, right, that he put out, which I think articulated a lot of his philosophies. One thing we've, done as a result of that is we've started hosting twice a year product showcases, like to really package up everything we worked on and tell it to the world. So, yeah, I've learned a lot from each of those folks. And most of the things they would be saying are good practice to run the company, I would probably do as well.
00:45:18 · Ty Morse
[00:45:18] S1: You mentioned that the thing that you probably fear the most or think about the most is trying to design Coinbase so that it's an environment where employees have the latitude to go pursue almost like many businesses inside of the main business instead of just going off on their own. And try to structure everything so that that can work. Mhmm. How'd you kinda come to that philosophy? And then, like, how do you actually implement it?
00:45:41 · Brian Armstrong
[00:45:41] S0: Yeah. So for folks who don't know, we we have something internally called next bets. And twice a year, we have a panel where anybody in the company can come in and pitch and say, I think we should be building this, and I have one or two people that wanna work on it. And you don't have to get a yes from everyone on the panel. That would be a committee where they're... Like, one no means nobody does it. You don't have to go to your your boss, your boss' boss, your boss' boss, all way to the CEO. That's also a committee in another form where one person's no kills it. And my fear was from reading things about, like, you know, Wozniak, Steve Wozniak at HP. Like, he went... Famously went to HP and said, I think we should build a personal computer. They said, no. He left to found Apple. And so your best people, even junior people on the on the team can really be on the frontier, like, coming with great ideas. Not all of them are gonna be great, but some of them are really great things. And so what we've done is we've created this next bets program twice a year that can in and pitch, but they only need to get one yes. So of the panel, there's different product group leaders who each have their own budget. They can decide to fund it out of their own budget. There's folks like myself, some young engineers like our president, CFO, etcetera. And so if you get any one of them to decide they want to fund it, you're greenlit, and it's kind of like coming in and pitching an internal set of venture capitalists. So that's what we modeled it after is if you go pitch 10 VCs and you get one term sheet, you can go move forward. So it lowers the bar for these ideas to be tried. Now we also want to make sure we don't lose focus at Coinbase. We can't just, like, fund everything willy nilly. Right? We we use what's called, like, seventy twenty ten resource allocation, so 70% of the of the resources are on, like, our core business, which has a certain definition of a scale and profitability. Make sure you keep the main thing the main thing. You know?