Ep: 232 - Boys Club Live from a16z: San Francisco Tech Culture, Hudson River Trading's intern class, OpenClaw in China with guests Jimmy Lai (Vercel) on Next.js and AI Agents, and Seun Omonije (x-Google Quantum AI) talks quantum computing and bitcoin

[00:29] Here's this week's show. [00:31] you [00:33] Hi. [00:34] How's it going? We're live. We're live. [00:38] Thank you for being here. [00:41] So weirdly formal. Okay, this was live. We are coming at you from the A16Z offices in Selma. [00:50] The extremely gilded offices. Very gilded. Gilded Age. Here. It's a theme. You can see it. We are not in theme for. And I'm here with my good friend, Julie Rosenberg, who is co-hosting with me today, which is so fun. And we've got a really fun show. Before we get into all the things we're going to talk about today, we are... [01:11] going to say thank you to a few of the people that make the show possible. Um, [01:16] first off kate in tampa off screen but living living her best life for sure which everyone could see what's happening there glasses around there are palm trees in the background she's living large she's in like a muscle shirt of some kind um but polygon octant and versell make this show possible we're going to talk about all of them today uh and give you a little bit [01:46] Great guests were going to come. Jimmy Lai from Vercel and Siun. Who is working on a bunch of stuff. A legitimate quant, which I can't wait to tell him. We officially have a quant. We officially have a quant, which is huge for us. Truly. I have been making a joke for years that

[02:06] You're my quant. Yeah. And today we're going to find out what that has meant. Right. We're going to get down into the deets of the quant of it all. [02:17] Okay, but before we do all that, and then I have a little story and you have a little story that we're going to brief each other on and talk about. [02:25] Before we do that, though, I'm going to talk about Polygon. So every company moving money globally hits the same walls. Legacy rails are slow, expensive, and built for a pre-internet world. Stablecoin rails are fast, but they're fragmented across vendors that don't talk to each other. The plumbing to deliver on the promise of stablecoins hasn't existed yet. Polygon's open money stack is fixing that. One unified stack that puts global money movement onto proven rails that moves at [02:55] internet and why Polygon because they've been moving money, which is why their new brand campaign reminds audiences that it's not their first trillion. Check out the tweet that we're dropping in the chat and watch their new brand video. We love Polygon. [03:12] Heart hands. They're so great. The campaign work that the marketing team is doing there is really, really fun. And I love what they're up to. [03:22] Love. So that's why not. Okay. Do you [03:27] I feel like first of all, we should just talk a little bit about San Francisco. [03:30] Let's talk about San Francisco. Welcome to my city. For those who don't know, this is my city now. You're it. I own San Francisco. Daniel Lurie.

[03:38] I've been here for about a week. Yep. And not my first time in San Francisco, obviously. No, I've lured you in here before. And every time I come, I'm your passenger princess. I just sit in that passenger seat and you just take me around. But it is really interesting. There's so much. I was talking to Danny yesterday and there's so much. [04:01] You can like feel [04:03] what's happening in the air [04:05] Everybody's talking about obviously AI. Everybody's talking about websites. I'm just like, okay, the meme is real. Like, I think that that is, that's my main takeaway is like the meme of San Francisco is actually real. [04:17] a reality and i was talking to a friend who used to live in new york [04:20] and now lives here. [04:23] Uh, she works in Metro Capital and [04:26] I was sort of-- we were just talking about that, like the meme coming to life and how it feels being here. And she was like, yeah, it's really interesting. [04:35] you're sort of in the land of promise, but also not exactly reality. I feel like in New York, [04:41] Obviously, there's so much capital. [04:43] So, [04:45] much venture activity that's happening, but it's also, there's a lot that is, that does feel rooted in [04:51] like the reality of businesses. And of course that's happening here, but there's also this like layer of [04:57] secondaries and valuations. It's really interesting. Yeah. And I'm curious what your perspective is. [05:04] I think it's interesting because... [05:06] I do get a sense of when I'm here and you start to just hear the eavesdropping and the whispers like,

[05:13] Like AI. You step into a coffee shop and like there's just like this [05:17] Oh, we led the round. I've heard that so many times. No, it's everywhere. And I do think like an immediate reaction I've had is like, oof, like anxiety, like everyone is behind the times. Like this is the most pressing thing. Yeah, there's a sense of urgency around everything because it's like, so the discourse here. Whereas in New York, [05:36] There's just like, that's like a side plot. Like there's so many other things that are happening. Right, right, right. And it creates like a balance around the tech environment in New York, whereas here... [05:45] were basically just on an IV drip of-- [05:48] Mostly. Totally tech. Totally. And I think it's a good my conversation with this other friend. It was a good reminder of like [05:55] Of course, a lot of this is very real and very present and actually happening. [06:00] And there is legitimate. [06:01] changes that we're seeing [06:03] that [06:04] in tech that are legitimately going to be like a before and after in the way that people work and live and all these things. But there's also a lot that is not and like a lot that will go to zero and [06:16] It's sort of just a nice smelling salts on the urgency of it all. Yeah, I think that's a good... [06:24] gut check on the urgency. Like everything feels like it's going to happen tomorrow here. Whereas like the AI doomsday feels like it's very imminent here. [06:33] And I think that that is like [06:34] there's an inflated sense around everything that is discussed as a result. For sure. Yeah. I also will say [06:41] It's incredible that people are...

[06:44] so into tech when it's so beautiful outside like the weather and the the nature i know what the is going on ali from versell did a good tweet the other day that was like it's beautiful day sit inside on slack oh that actually became like a whole meme um um okay so that's the san francisco update um everybody's talking about ai and peptides as much as they say they are you do have a we will [07:14] - For San Francisco. - Oh, I have a special place in my heart for San Francisco. Everybody here wants me to move here. Everybody's like, "So when are you moving here?" And I'm just like, that is literally, [07:23] Like no shade, but that is not going to happen for me personally, but I do love it. I think it's beautiful. Yeah. I, the food is amazing. The people are super smart. Um, [07:33] I get to hang out with my wife. Great. Most importantly. I'm no fool. I don't even try to convince you because I just know that's not on the card. Okay, tell me realistically. Thank you. I appreciate that. Okay, tell me what you want. Tell me what you want to talk to me about. Well, speaking of very intense groups of people. [07:51] people. [07:51] There's been this viral paragraph from a Colossus article [07:55] going around on Twitter recently about Hudson River Trading's first ever intern class. Okay. And there's basically... [08:05] 10 people. It was their first ever. First ever. First, what is Hudson River Trading? Hudson River Trading is a high frequency trading firm. [08:13] Jane Street. Jane Street. Okay. They are doing high-speed bath. They're in New York. They are in New York. Okay. Okay.

[08:24] four of the 10 of this first cohort, first intern cohort have gone on to build [08:31] multi-billion dollar companies. Okay, name names. Okay, we're naming names. We have [08:36] Jeff from Hyperliquid. This is where the article... [08:39] where this information originally circus was. Okay. Colossus did an article on Jeff and Hyperlimit. Oh, I saw this. And their growth. And what his journey has even been. 11 employees. [08:52] 900 million dollars in the last year. Yes. No venture funding. No venture funding. Unbelievable. Insane. Insane. The highest of high margins. Really great stuff. Yeah. Incredible stuff. That's the goalpost. Proud of him. Yeah. [09:06] Which I feel like is a whole other tangent to go on. Oh, this came from that article. Yes, this came from that article. It was like a little subtext of the article. Oh, Jeff's classmates at Heads of a Retreating were Alexander Wang from Scale AI. Whoa, okay. Scott Wu from Cognition, Jesse Zhang from Decagon. These ginormous names that are in mostly AI right now. [09:31] And they were sort of like sub-treating each other, being like, "Oh, that was so funny. Remember the good old days?" Like joking about it, but it's just like a crazy... [09:41] concept that there was like this insanely high talent density. Yeah. And [09:45] I think it's just interesting to sort of think of like the new graduating classes. [09:50] Thank you. [09:51] I think very much the past version of that has been from schools.

[09:56] But I feel like in the age of AI, it's increasingly moving towards like these extreme high agency environments. Um... [10:03] And instead of thinking about like school, job, career, it's becoming like join the highest agency environment and then – [10:10] learn how to operate so you can just build your own thing. It's so easy to build things now. Yeah. What would be an example in your mind of a high agency environment? [10:19] That's a good question. I think... [10:22] joining like really small and mighty teams like this. Like early stage. Yeah. Super early stage. Joining Hyperliquid. Being one of the 11 and Hyperliquid. Join Hyperliquid. It's not hard. Join Hyperliquid before Hyperliquid becomes Hyperliquid. So simple. [10:37] I think that... [10:40] It's about just finding like really small groups of high agency environments. I think like the blueprint even 10 years ago was like graduate college, [10:48] get a job at Google, Meta, Amazon, like big tech, [10:54] Big IB, big. That's like 100% not the... [10:59] Not the case anymore. Anymore. Yeah. Yeah. Which is so interesting. Like, I'm very curious what the new, like, graduating classes are going to be entering into and what that model is going to look like. Yeah. And what… [11:10] environments people are going to be seeking out. [11:14] It actually reminds me of our friend, Kieran. [11:19] - Okay. - Kieran, to shout out. POV Kieran, shout out POV Kieran. - He's so happy to hear this. We'll make this cut down. He is a friend of ours from New York. He's a crypto founder and

[11:29] he has moved to China for the next year to [11:33] basically create content around [11:36] what is happening in the tech scene there, what innovation looks like, what early stage founders and builders and creative folks as well are doing and what tools they're using and how and when the government is involved in that. And it's, [11:50] He had told me before that he was going he was like I'm going to China for a year like kind of vibes. And I was just like what? And then he has first few videos on YouTube are so good. They're amazing. Amazing. And [12:06] one of the videos is about an open claw meetup that was happening I think in Shanghai and [12:14] was [12:15] One of the parallels that he's sort of making, or one of the things he's talking about is like a lot of these meetups are being funded by the government and they're actually like giving out credits to students and to early grads and... [12:26] really encouraging the use of these tools and innovation. And [12:32] that's interesting for a number of reasons but one of the things that has really stuck with me is he talked about 12 million [12:37] new [12:38] 12 million graduates every single year and like where are they going and what workforce are they entering into? And [12:45] that [12:46] the concept of 12 million people every single year going out and looking for jobs and um [12:53] that gives me like a bit of like anxiety just thinking about but it's it's interesting um but anyway i one recommend watching

[13:03] His video and just kind of like seeing what's happening over there and two I think I [13:08] To your point, it's like, where are the environments? Where is like the smartest... [13:13] environments for young ambitious people to spend their time right now and also [13:18] What does that mean for... Yeah, like what is the... [13:21] graduating class of what institution is going to look like this in 15 years um and yeah i think it [13:28] Jane Street used to be [13:30] these training firms or Big Tech or Deloitte or whatever it was. And that feels like totally not the case anymore. I know. I'm very curious to see like how sought after those opportunities will continue to be. Or like if they exist also. Yeah, totally. I think that's even a big part of this question too, is like the size of companies is just like increasingly shrinking. [13:49] Everyone is moving towards a more efficient model. Like what you just said with Hyperliquid, it's like an 11 person team. [13:55] Yeah. Like each employee is worth hundreds of millions of dollars in revenue. Like that's an insane, insane metric. Yeah. But like that's the age we live in now. [14:05] And so what will... [14:07] Yeah, like that. [14:10] hello come on down you can come join us no yeah come on why not he's like wait wait [14:18] No hormone. [14:19] Or you can hang out, whatever you want. Come on in. Come in whenever you feel like I. Thank you, Sherry. Jimmy! How's it going? I'm Natasha. [14:31] Nice to meet you. Nice to meet you. So nice to meet you too. Okay.

[14:35] Jimmy, you lie lie. Okay. You are at Vercel and we love Vercel. We're huge fans. We're doing a lot of fun stuff with you guys. We're actually going to talk about some of the things that we're doing with Vercel later in the show. But you are an engineering director. You're [14:51] mainly focused on Next.js. Is that correct? Yeah. Okay. Um, and so we're going to talk about that. We're going to talk about like open source frameworks in general, but, um, [15:03] I would love for you to just give a very quick overview on what Next.js is for the listener. Yeah. So Next.js is- Let's jump right into it. I'm sorry. We can do some other questions. Allie did have a really good question here. She was like, "You're an engineer." Wait, her wording is self-reviewed. "You're an engineer, but you have to explain." Which is maybe a softer question to start with. Yeah. It's the hardest one. Oh, really? Okay. You can think about that. So Next.js to start. I mean, I can start with that, I think. [15:33] The drip. [15:34] I think I say, so I'm French. And so I think there's like a little part of that. [15:42] But I probably blame my Asian mom mostly. Okay. She's a fun fool. Yeah. She obviously speaking, it's just not going to be... She's not going to know that. Yeah. [15:55] I wish you'd like it. [15:55] incredibly judgmental person and so i think every time i dread for you at home would kind of roast me that's powerful yeah and so i think that that was like [16:05] pretty

[16:06] pretty sort of like important in my childhood. Yeah. A little bit. So did you grow up in France? Yeah. Okay. We're in France. [16:12] Oh, in Paris. Okay, we're in Paris. Sort of like in the suburbs of like just basically right outside the... [16:18] 30th. Okay. Oh, nice. Okay. Have you been? Yes, I have. I've spent the last few, a month every year for the past few years. Yeah, it's really fun in June. You've been more in France than me. Probably. But yeah, it's, I feel like growing up there, everybody I know who has grown up in. [16:36] Paris has pretty much impeccable taste. So it really just kind of shape you. So maybe that's the answer your mom and [16:42] the environment. I think so, yeah. And every time you leave the house, it's now [16:46] What judgment would mom have on this? Yeah, exactly. That's a great name. That's important. I love that. Um, okay. Nice. Well, congrats on the drip and, um, that being part of your persona. Um, okay. Next.js, tell us what's up. Yeah. Um, so Next.js is basically, uh, a tool that like people use to basically build websites end to end from like the server to the client. Okay. Um, it's open source. It's been here for like. [17:12] nine years at this point and it's [17:15] I think it's like one of the most popular tool that people use to build. [17:19] Sorry, the websites that you use every day. [17:22] Basically. [17:23] Yeah. [17:24] Powerful tool. [17:25] Yeah. And it's open source. It's open source. Yeah. Okay. [17:29] Yeah. Okay. Tell us what that means. [17:31] Yes. [17:32] Open sources, I think at the [17:35] The lowest level, it just means that the work that we do is available for everyone. People can see how the tool is being built when we release it. It's available for free too. Okay. Those are...

[17:49] the two big tenants, like you can basically [17:53] you know, regardless of whether or not you're a reconversal customer or someone else, you can basically just download it and use it. Okay. Yeah. Pretty simple. [18:01] I think a lot of the ways that I think about open source is that like a lot of people are contributing to it. [18:08] Is that true of... [18:10] all open source projects, [18:12] I don't think all-- I think it really depends on which. OK. For next SEO, we get a really sort of healthy amount of people that-- because the-- [18:21] the way the tool is being built. [18:23] is available for everyone. There's a lot of people, you know, that chime in and they ask requests and sometimes they're just, you know, they're just like, well, I fixed it myself. Like, why don't you like, [18:31] take it and distribute it for everyone else. [18:34] which is really nice. I think [18:36] Some other tools like [18:39] some pretty popular tools like [18:40] you know, VSCode for example, [18:43] the [18:44] probably don't do as much because [18:48] even though the work is public sometimes people just don't really [18:51] Like, you know, they don't want to. [18:53] go and like go through the volume of like people sending them like sort of suggestions. Okay. They just feel very strongly about what they do. I don't, I actually don't think it's the case for VS Code, but it does, it does happen. Yeah. Okay. Are you [19:07] So are you like... [19:08] reviewing [19:09] submissions that contributors, is that a big part of your job? Not my job, but my team's job. Okay. Yeah. Yeah. Okay. We try to take a look at it often.

[19:20] I would say my team is like 15 people working on framework every day. [19:25] every week or so someone takes a look at what's incoming and what are people [19:30] trying to add to the framework, then we sort of like [19:33] They review it and accept it and merge it. Okay. What is the current evolution of Next.js? What are your team of 10, 15 engineers working on right now? Yeah. Yeah. [19:44] So I think I've been managing for like four years at this point. And... You've been managing it for four years? Yeah. Oh, wow. Four years. Next.js through and through. Yeah. But it's nine years old at this point. Yeah. They created the project when... [19:59] when I started learning how to program. [20:02] So there's a lot of history there. I can't speak to the people that came before. But I think since I came in, the team has been somewhat stable. You find that for those kind of tools, [20:13] like you really don't want to have like you know 50 engineers working on this like it's it's kind of like a labor of love so you really want to like [20:21] just keep the core really small, have everybody understand well with your building. [20:28] Like, you know, like when the sites, when your apps kind of get too big, you can tell that like some other people have like a different hand in like, [20:36] you know, one page versus one another and the design becomes kind of messy, for example. - Too many cooks in the kitchen. - Exactly. Yeah. And for something so specific like a [20:49] Like a... [20:50] like a web framework. Basically, we try to really sort of like keep it well designed. Okay. I'd say we're very sort of nitpicky. Yeah. How we go on about this. I'm sure also everyone who uses Next.js is very opinionated as each new feature comes live or each new change comes live. I'm sure you

[21:07] spend a good amount of your time just fielding [21:10] uh, [21:11] complaints from very passionate developers. Yeah. There's no one that has as much standard as someone who's getting your project for free. Yes. That's so funny. The less you pay for it, the more you. Okay. Something really interesting. So this time last year, there was 9.6 million downloads per week. [21:31] And now there's 36 million recently spiked over to 40. And a huge part of, it seems why this is happening and sort of like how dominant the framework has become is that a lot of these AI agent applications are widely using Next.js, is that correct? Yeah. Okay. Can you talk a little bit about like what changes have happened that have made sort of this massive surge happen? Yeah. Um, I think there's two parts to it. I think like I brought it up earlier. The framework is just like nine years old at this point. [22:01] Thank you. [22:01] And [22:02] Over time, we just kept steadily growing our usage. So many other frameworks have gone and gone during that time. It's a bit of a thing in the JavaScript ecosystem where [22:16] Every week, people would create new tools. [22:19] And we're lucky that we survived up to the point that we've become sort of like [22:25] If you're looking to build a website like this, [22:27] very easy to stumble upon it. And so I think then there's like, you know, the learning cycle where the [22:32] The data sets just have a lot of information about Next. [22:36] And so they train on it. We work with the big labs to continue training the models on the newer version of Next2. So that's one part of the answer.

[22:47] which is just like [22:48] which is popular. And so, you know, it sort of like maybe compounds on itself. Like, okay. Yeah. Like you're already, XJS was an evolution of sort of building out a framework for [22:58] very clean UI building and so much of agentic [23:02] workflows, like developer workflows is like [23:05] creating very specific instruction sets. Next.js is already like the first instruction set that an agent is given. [23:12] Yeah, it's exactly that. [23:14] We're like one of the big ones. And so they, they learn through it, but [23:17] It's kind of easy, right? [23:21] we're like the biggest ones, but at the same time, like the competition is like very sort of like intense, because of the new frameworks coming in and they don't have all of this legacy behind them. So they're able to do new things, they're able to do it faster. [23:33] And so a lot of the work to like another part of like the answer is that we're [23:37] We just keep trying to improve all the time because we have 15 heads. So we're spending the full time on thinking about [23:45] Yeah. [23:46] What is the [23:47] how do people want to develop websites in the future? And how do we basically best guide our users, new features, new documentation, [24:00] Um, [24:00] Sort of new [24:01] patterns basically and so we spend a lot of time on that and I think [24:06] I think the agents can pick up on that. They know that it's not a dead project and that we're still actively... Yeah. These people are working on it. I think that brings up an interesting point. Something I've been thinking a lot about recently is just like,

[24:20] the [24:21] state of UIs and the importance of UIs. Like I think so much of our Excel story was obviously about like the ease of spinning up UIs. [24:28] Like, [24:29] you want to build a website, the easiest way to do it is through Vercell. It takes out like a huge amount of infrastructure pain to actually get to the point of like having a, [24:37] beautiful landing page or whatever that thing is. No. [24:40] But now, sort of the importance of UIs is changing. Like, we have all these headless bots that are running and scraping websites and pulling information and... [24:53] I mean, Vercel has done an amazing job in keeping up with [24:57] understanding [24:59] where demand is for [25:01] high-risk building and like very much a lot of your work is how to build Next.js for agents, not just for people now. So I guess I'm curious like how you're thinking about [25:10] just the power of interfaces and what that evolution is going to look like. [25:14] in this new [25:15] AI agent world that we're living in? Yeah, it's a great question. We think about this a lot because [25:22] You know, I don't. [25:23] I'm trying to think about my work in five years and like, am I going to be out of a job? What is important? - The question we're all asking. - Exactly. And the worst answer is like, you know... - Yes, the worst answer is yes. - We're about to communicate with everyone, it doesn't matter. - It's over. [25:43] Yeah. [25:44] But [25:45] I think about it and [25:48] Bye. [25:49] I think about the way I use like UI as myself and like the way I use like so like agents to

[25:54] potentially like should I create apps and like [25:56] what I care most is [25:58] So having access to information and being able to update it. And it doesn't matter the shape or form. [26:05] And so I think we're kind of orienting ourselves toward like, [26:09] Yeah, so like maybe you are less [26:12] so a ULS kind of word where [26:15] what matters more is like what you expose, right? And then [26:18] On top of that, let's say we jump [26:21] a few years ahead, and agents are able to kind of create like UI just in time. And they can basically create something really personalized to you. [26:29] Like, um, [26:31] When you say just in time, do you mean like in real time? Yeah. Okay. Like I'm thinking about, you know, what's the weather? Like in Paris today. Okay. Like as I prepare for flights. Yeah. And like, you know, like instantly I could have like a... [26:42] like a nice UI that shows exactly the kind of stuff I need. And that's going to be personal because like, [26:48] the information that I need [26:50] or that I really like to see is very different from the information from someone else. Yeah. You're like, how do I have the drip? [26:58] Exactly. What kind of, you know, what shops do I want to go to? What kind of food? [27:01] Should I try to visit according to my own taste? Yeah, type of personalization. Exactly. [27:07] And so I think [27:09] We're going to [27:10] After we're done with the Orca and Patchwork, we're going to try to think about this problem. And at the same time, I'm saying it's not about UI, but at the same time, I think in the same way people like, [27:21] nice clothes then and sort of like you know nice

[27:24] nicely designed, Macbooks, et cetera. I think not just because you can create any kind of UI you want. You still probably want to be able to consume [27:33] someone else's taste. It's not because everybody can cook that [27:40] you don't want to go to the restaurants anymore. Right. Yeah, totally. And so, [27:44] So I think that's what matters mostly. I think people are going to say [27:48] recipes about like those building blocks oh interesting cool um i remember [27:54] a few years ago, people, there was like this [27:56] phrase that cut [27:58] sort of coming along the timeline and people kept talking about like luxury software and like what that would look like and um i think it's like not exactly the same thing but similar to what you're speaking to of sort of this experience of so much of our lives are more and more online and so much more of the way that we interface with [28:15] with the world initially is like through this information that exists on some sort of screen. And, uh, [28:22] I think, like a positive view of where that could go, is that there is sort of like a new type of-- [28:28] like artistry and personalization and appreciation for different people's [28:33] like setups in a way that, um, [28:36] we don't have today just because so much of it has looked very similar for everybody. And actually, it would go into some of the Show Me Your Stack stuff that we're doing with Vercel that we'll talk about later. But yeah, it's really interesting to sort of see like how, how, [28:48] Like... [28:49] All of that end state comes from, with all of software, but comes from

[28:54] like the infrastructure layer that you're building and speaking to. OK, so you guys recently had a big announcement [29:03] in partnership with a bunch of Vercel competitors. And I want to talk about is open source just a bunch of frenemies. [29:13] Yeah. [29:14] So, yeah, the big announcement that we had is that we created a working group with Gladfair, Netlify, Google, [29:22] about how we could work better together to support Biscuit and digest on their platform. Okay. Um, [29:28] because the part I didn't cover about like nexjess is that like nexjess is you know backed by Bercel and so [29:34] with where [29:35] MR PALLADINO: No. [29:36] But we're selling is like, you know, sort of like the framework that you can use yourself, but also if you want to use it to the best of its abilities, you probably, you know, [29:45] like can use a recel to deploy it and have the best performance. So are you guys like a-- is it like a freemium model? Like you get Next.js for free, and then like-- but what is going to make them-- [29:59] what is like most easily [30:01] plugged in is other Vercell tools. Is that kind of true or? No, I think it kind of implies like you can't [30:08] Sorry, access it with a historic painting. [30:11] But you can't, right? It's all open source and we sort of expose the right hook so that you can [30:18] If you wanted to, you could create your own persona. Right. It's more cloud agnostic. Yeah. Okay. [30:23] and but we're selling you know what we would sell you is like

[30:27] the time to set that up. I see. Okay. [30:30] In the same way that we would [30:32] in the future we won't sell like you know pre-made UIs because yeah because you don't want to waste your time doing it like it's the same thing we're selling you like [30:39] I don't know. [30:40] tasteful, sort of primitives so that you can sort of like [30:43] deployed. [30:46] And so other companies like Cloudflare and Natalify, they basically allow you to host Next.js in the same ways. Okay. But because, you know, there are like [30:54] Well, they're not like [30:55] the company that pays us. [30:58] It was a bit more difficult, probably, to kind of get access to the same level of details about, you know, how do you actually plug NextGest in the right way with your infrastructure. [31:07] Mm-hmm. [31:08] And so [31:09] They were pretty unhappy about this. OK. They made a bunch of super controversial blog posts. OK. Our CEOs started finding Weaveshutter on Twitter. OK, when was this? Like a year ago or recently? Maybe a few months ago. Oh, really recently. OK. It's been going on for a while. OK. It's been going on for a long time. OK. But during that time, we basically-- [31:29] we reached out to the engineers working on Next.js at their company. And we're like, [31:36] we're gonna wanna help you guys because like, [31:38] what matters most for me is just that that connection works well. Yeah. Everywhere. A peace offering. Yeah. [31:45] Because also, if it doesn't work well, people blame me. [31:50] And so it's just like, yeah, it's just a matter of like, I'm just blanking myself here. And so we started working on it together and like,

[32:00] Yeah, a month ago. [32:02] I think at this point, yeah, we announced like officially like a working group where we're just going to... [32:07] publish your findings, like meet with people and sort of like, you know, have a bit more of a process around how we support them. Okay. A beautiful CYA campaign. I love it. That's amazing. Yeah. [32:21] Okay. [32:22] Why should an everyday internet user [32:25] care about open source. [32:27] at all. [32:30] I think... [32:31] Open source is like [32:33] I think everyday users don't realize it, but a lot of the tools that we use, a lot of the [32:38] the [32:39] the laptops that we, or hardware that we use kind of runs on like open source software. Um, [32:45] And open source software, the way I feel about it is that it's [32:48] It's basically the closest thing to [32:51] Art. [32:51] for nerds maybe that you can do. - Okay. - In a sense, like, thinking about the reason why people do art is because they have like [32:57] they have an opinion, like a very strong opinion on how to do something or how they visualize something. [33:03] Most of the time, they don't even care about making money from it. They can just, you know, sing out there at a bar and it just feels good. I think open source software is like the equivalent for [33:14] for software engineers, where [33:17] people just feel so strongly about how you should [33:20] be able to build websites or how you should be able to like, um, [33:24] run your computer, that they go out there and they kind of do it like [33:29] for free to distribute it in the same way.

[33:32] And I think... [33:34] I think it's really important that we keep this open. [33:37] Yeah. [33:37] because yeah it's it's all were you um [33:41] Thank you. [33:42] in your career before you started working at 4SEL, were you contributing to open source projects? Like what was your relationship to [33:47] Well, I was always a big user, right? Because a lot of like most of the tools that we use to build websites are all open source. Like no one pays for like a license to like download the tool and then like do something with it. Okay. [34:01] It's all like insanely accessible, which is also why I think like [34:04] website. [34:05] building is one of the most accessible way to-- it's one of the most popular software engineering jobs, right? Because a lot of people start when they're 14, and they're able to just hack on websites for free. [34:17] Um, [34:18] But I never really contributed to it. [34:21] before joining Vercel. I was working at like [34:24] -Mena. -Okay. [34:26] But I was... [34:28] Quite the opposite of open source. Well, funnily, no, because Meta actually has a big open source. Oh, really? Oh, wow. Sort of like a DNA. A lot of the tools that we use, like React. Oh, wow, wow. And they don't make money out of it. They don't really care for it either. [34:46] they let it happen because [34:48] Wow. Okay. The final result of their website. And it's like, you know, a really well-proven technology at this point. Yeah. Yeah. And so I used to work on React Native, which was like the over-the-end four for mobile. Okay. [35:00] So I was kind of contributing to open source. That's definitely you contributing to an open source project. I was paid. I was very much paid for it. Yeah. Yeah. So it's a bit different. Yeah. I think, um,

[35:09] I have a lot of respect for the people that have sort of regular jobs and are able to spend their evenings working on those tools. Even though it's unrelated. A true artist. Like really for the love of the game. Yeah, I'm just a fake. I'm just a fake. We're here to out you. I am curious if you [35:30] in this art form of development. Like if you pursue any... [35:34] hacky side projects to even also just like test out what you're building with Next.js. Like, how are you sort of [35:40] feeling like you're keeping in touch with the current state of website building or product building. [35:45] Well, yeah, the big reason why I went into web development is that I just [35:52] I'm a very sort of visual person, and so I was kind of building my own tools because I like building my own sort of UIs. [35:58] Like I was saying, I really like [36:00] I hate going in on a website and [36:02] having to go through 10 pages of ads or in addition content. [36:07] um and so to this day i'm still doing it a lot where um if i want to like watch a movie i have like my own sort of like movie app calendar um oh really yeah okay whoa and i have like many tools that basically are using next yes and um i think that's the way i'm most useful to the team too yeah so you're already seeding these hyper personalized experiences to you yeah you're introducing the drip all around you exactly like um i love like um i think someone called it like [36:35] personal software, where this idea of like [36:38] you're just like doing like

[36:40] custom stuff for yourself. And now AI has made it super available. And I have a lot of my own personal websites. [36:48] that I don't care to share. They're just like... [36:51] They're just like nice and naturally it's kind of [36:53] really allows me to do this like really quickly. Yeah. So that's definitely an observation I've had where a show that we're working on together with voice club is a series called show me your stock where we're interviewing techies or I'm interviewing techies as the creator and host, um, and getting a screen level tour of what their tech stack is. [37:11] And I'm talking to a lot of very technical people, developers, people who are building out IDEs, people who are building out very complex AI infra, all these little pieces of the stack. And I think the thing that I have found most interesting is [37:27] every single person I interview uses a different IDE, a different way in which they're interacting with [37:33] the thing that they [37:34] is most important to them. And the ones that have been most popular, the ones that are emerging beyond just the classic VS code and things like that, are these hyper-personalized IDEs. For example, like Ghostly has become super popular. [37:54] based off of the customization that you can have in the layout of when you're writing code. [37:59] Which I find really interesting is that we're like hyper personalizing our own workflow rather than just like the end state product that we're working on. Yeah. [38:08] Like it, [38:09] Another expression of this, I think, is like,

[38:12] You've probably heard of Open2L, right? Right. So people like [38:15] you know using it to just let it do anything they want and like [38:19] What's amazing about this is that you don't even need to code anymore. You don't need to go through like your config or just write your own like [38:25] scripts that you can just say, hey, check every week if there's spots available at the DMV in San Francisco or something like this. I have this running right now every two minutes, for example. And it saves lives. [38:40] you know, yourself like an incredible amount of time. Yeah. Yeah. [38:44] like, [38:44] you use it as like a personal assistant and [38:47] I think [38:49] I think that's the real sort of game changer. It's going to happen. [38:54] you know, the new products that the AI labs are going to come up with. How do you sort of, so what we're sort of describing is like that in the current age, people are building out these hyper-personalized systems around them, like personal workflows. And so much of what Next.js is and originally was built around was building these frameworks around [39:12] specifically UI developments and the ability to get to a front end. How do you think about frameworks that we're developing to [39:20] uh, [39:21] build out the systems around our own workflows. For example, I think Vercel does a lot of really interesting skill work and helping [39:29] you get very specific instructions that's for their agents but how do you generally think about like the agentic [39:35] frameworks that are being built for people developing out their personalized workflows. Yeah. [39:41] Well,

[39:42] We're, I think as a company, we're like, you know, investing heavily into this because we think that it's going to be, you know, the, [39:48] the next next yes, in a way. [39:51] which is sad for me, I think. But we're really banking on being able to provide those primitives so that you can then deploy your agent. Because it's not only about running it for yourself on your own computer. [40:06] is we're thinking more about the business, in this case, like the business agents, basically, where internally, for example, we have something called [40:15] I don't know if I'm supposed to say actually, but like we have something called like D0, for example, which is [40:22] this type of agents that is uniquely personalized to [40:26] be able to go [40:27] and reason about all the data that we have. OK. So instead of having a data analyst, having to call them, like being like, hey, can you send those 20 reports? It's just literally one message away. [40:40] And you say, "How many Next-JS Donuts do we have?" Like, "Who's the best customer?" "Who's the worst customer?" That kind of things. [40:50] As we built this, we were like, well, this [40:53] like we're gonna need this like everything because that just translates to like customer support like um software engineers too uh like [41:01] which is why we're going to be out of jobs here in Bedlake. Basically, anything like this. And I think they need [41:08] people do need like sort of like a framework list. [41:11] And a lot of companies, probably upsells too, are starting to like, you know, investigate like what that looks like. Yeah. What is like the final thorax shape through which you're going to expose those capacities.

[41:24] For the record, I don't think you're going to be on a spot anytime soon. But genuinely, I think because so much of what you are doing is creating hype. [41:32] like very hyper-opinionated [41:33] guardrails. [41:35] to [41:36] like framework development and even what you just described with like building out these [41:40] uh, [41:41] sort of opinionated agentic flows. I think that will continue to be something that is heavily reliant on people and building out these very specific opinions and instruction sets for [41:51] agents to follow. [41:52] And I think that will require continuous [41:54] hand-holding or continuous [41:57] Care. [41:59] throughout the experience of using these tools. Yeah, I'm not that worried about it. I talk about it a lot because I'm thinking about retirement and all. I'm excited about the next things. [42:11] But yeah, I think essentially a lot of people are just going to be judged for [42:15] purely their case, right? Like what's their... Yeah. Drek. Exactly. Okay. So you're using agents in your workflow around data. Is there any other ways that you are personally using [42:26] um i'm uh [42:28] in work specifically yeah i don't know if i should say this too but like basically i have like something that runs every hours okay and that just pulls data from like [42:35] every sort of like information that I have. Like Slack, Notion, Vanola. Okay. And I try to compile this into like sort of like a brain, like just like a knowledge base so that I can like later say like, hey, [42:49] I have a meeting with those people, like, what is the context, right? What should I remember about? Because the one thing we've seen with, especially with

[42:57] The raise of AI is just like... [42:59] there's just a lot more stuff happening. Yeah. A lot of information. Yeah. And honestly, I can't even keep up anymore, like, more slack. Like I have so many sort of channels left and read. And so [43:10] I have this agent that pulls the relevant things for me to look at. And then what's the interface where you're reviewing that data? Pure chat. I haven't gone as far as doing personalized sort-of reports or anything like this. [43:25] Um, [43:26] But I would love to. Yeah. It's like next weekend's project. There's a little bit of danger there because you can like spend so much time trying to automate it. Right. Instead of like just reading the. [43:36] yay for sure yeah um but uh so i don't want to spend too much time on it but it's already [43:42] I think it's been like probably close to like 100 hours. Well, okay. Well, it's not a solve problem yet. Like memory is like [43:50] I'm like watching very closely like what the [43:52] what the companies in the space are doing because [43:55] If chat GPT, Claude could really remember accurately everything that you've been up to, [44:03] I think it's amazing. It'd be super scary, of course, but I think there's a great unlock there to be. Yeah, totally. Yeah. Something we were talking about before you came in was just [44:12] how so much of this innovation is like [44:16] truly valuable will change the world. But there is a lot of like FUD and there's a lot that's just, [44:23] Um, [44:24] Just like the land of promise and like just kind of, I don't know, will go to zero. Stuff that will definitely go to zero and that we collectively as an industry in tech have sort of

[44:36] put some eggs in a basket that maybe is not exactly what we thought it was going to be. [44:41] And maybe you don't have an answer for this, but is there something that you use or have seen or even just like a Twitter... [44:51] rabbit hole where you're like that legitimately is [44:55] Like, everybody's talking about this thing and like, it's not going to work. Rose, it's a strategy. You don't have to name names either. But like, it can even be a concept. [45:05] Where do you think there's like the most five, I guess is the question. [45:08] oh um what do you mean like in terms of like this thing is like valueless or this it's like it could be a tool it could be like i mean sort of similar to what you were talking about of like [45:18] we can spend so much time over engineering [45:21] the [45:22] the like brain or the summarization or the process for like getting information in and then getting like an output that is supposedly supposed to save you time and energy but then at the end of the day if you really look at it like you could have just read all the stuff that's one example but like is there something else that you're seeing that you're like oh i just see that and i think [45:42] that actually is not going to be the way that we're giving it in [45:45] two years or five years or ten years. [45:47] Oh, that's a great question. [45:49] Um, [45:50] Like, what are we going to [45:51] What are we wasting our time? Yeah, yeah. I think [45:56] I think... [45:57] I think we're not wasting time anywhere. I say this because I think it's like slightly personally, because I think a lot of the work we do, for example, is... [46:05] like sort of like research and development on websites and like how people want to build websites and like you might just say well websites were working like

[46:12] 10 years ago, why are you trying to reinvent the wheel? [46:16] at all times. [46:17] And so I'm... [46:20] Every time I see something that's like, you know, [46:23] It's not as great as it should be. I think it's still like great that like people are exploring it so that like, you know, at least we have like a record of it, like not working. You're like stuff begets more stuff. Like there will be, it might not be that thing, but like that will give you some data and information. Okay. That makes sense. Yeah. There's a big bubble right happening at the moment. Like this insane amount of money being funneled. And then like, we're just like, [46:42] trying stuff out, we tried stuff out in Germany, right? [46:45] Mm-hmm. [46:46] OpenAI definitely tried a bunch of stuff out. I think they have a browser. There's no one really uses right now. : Or Sora or like... : RIP. : Exactly. [46:55] But now we've tried it. Now we know it's [46:57] you know, it kind of sucks. So at least we can kind of move on and learn from it. And that's why it's kind of exciting to be here in SF. Yeah, totally. [47:07] Yeah, you're really close to it. Probably half a billion dollar. [47:11] Yeah. Fun experience. Failed experience. That's true. Yeah. This is so fun. What tools were you thinking about? I don't know that I have a specific tool in mind necessarily, but yeah, I think that [47:25] people love to talk about their stack. They love to talk about the optimizations that they're, um, [47:32] doing any given day. And yeah, I think I'm just like a little bit of [47:37] I'm just skeptical that we're all just drinking the same Kool-Aid and we'll eventually feel like, whoa, that was crazy that we were all trying to do that.

[47:47] I can't have to use. Totally, for sure. We can't say that word on the show anymore. Jimmy, thank you so much for coming on. It was so nice to chat with you. You're welcome anytime. Truly. Thank you so much. Nice to meet you too. Okay, you got it. You got your way out. Okay. A great chat. Jimmy. [48:07] um okay i just love the everyone for brucelle that i meet just the coolest the best truly the best people and a huge shout out to brucelle they're also sponsoring show me your stack yes series that we're working on i'm so excited for the world to see it wait a long um oh oh no okay our second guest is here let me see if he's gotten in the building oh great um i think while it's happening it might be [48:32] good time to talk about [48:34] Vercel a little more. [48:36] Let's do it. Okay, great. So if you're a startup building with agents, look no further than Verzal. Their agentic infrastructure gives you and your agents everything you need to ship. And if you're supported by one of their hundreds of VC or accelerator partners, you have access to exclusive discounts and benefits. Deploy faster, scale globally, and ship world-class AI products. [49:05] Yo. [49:07] There's some rumors there at versell.com slash startups and see the live link in chat. [49:14] Great job. Great job reading. Thank you so much. Try to give it a little character. Oh, I love it.

[49:20] What a great, really interesting conversation. I know. I really appreciate that he is truly bringing Dirk into every piece of the staff. It really came up. And I-- It's a methodology. OK, I think that our next guest is-- [49:36] in inbound and i think that have you made any hyper personalized [49:44] setups for yourself over there. [49:46] No, no. [49:47] We're working on it though. [49:49] Yeah. Sure. I feel like, I don't know, with a lot of, I mean, obviously I use cloud every single day. I use it for a bunch of stuff. [49:58] I [50:00] Um, [50:02] have [50:03] I talked about this with my therapist where I feel like I have some- Let's bring therapy in the room. Let's do every episode. [50:09] Where I feel like there is a gun to my head. Yeah. And people are like- [50:13] use these tools and then I'm just like whoa I now never want to like there's something that and um something to analyze. [50:22] I guess. There's still just a lot of overhead involved in it. Like I think whenever I try to do these things I realize how early they still are and I'm like oh this just requires so much hand holding. Like I feel like I'm [50:34] My [50:35] digital companion is like an eighth grader. 100%. Yeah. Which like, love them. They have so much energy and it's [50:41] So sweet. They're so sweet. But just like, they're so eager, but they have so much to learn still. And I think like, [50:48] what I have less patience for as a non-technical person is just like

[50:53] doing the reps on a lot of these tools. For sure. Which I think someone like Jimmy has so much – has such depth of experience doing. Yeah. In – [51:00] He's run like a million versions of React and Next.js. Totally. That have just gone nowhere because [51:07] It was too early or not right. Yeah, exactly. And I think there's a lot more tolerance for [51:12] the fumbles that happen along the way. Yeah, for sure. I think that I have maybe like a negative [51:19] perspective that it's like [51:21] He'll figure it out and then I'll learn from him. Totally. Hey! How's it going? Hello! Come on down. Thank you, Sherry. Come on in. [51:29] - Connor! - Hello! I'm Natasha. Nice to meet you. Nice to meet you. Long time no see. Welcome. You're gonna have to scooch in tight on your heel. [51:40] You're live. We're live. We're like, yeah. We're in the building. Okay, it's she, you. She, you. She, you. She, you. She, you. Okay, super excited to talk to you. We, um, [51:53] have made this joke already, but you weren't here for it, so I'm going to make it again. No, let's run back the joke. Oh, let's run it back. I... [52:01] I've made the joke for years that a variety of people are my quant about all sorts of different things. I am trying on an outfit and the shoes aren't quite right. I text my friend. [52:13] We're good. All good. I text Kate. I'm like, [52:17] Kate, you're my client. What shoe should I wear? Perfect. Really in... [52:21] accurate way of using it. And that's part of the bit. I mean, I'm a joke, but you are a researcher, ex-engineer at Google Quantum AI. You are legitimately a quant.

[52:39] Kind of. Okay. Okay. [52:43] I have experience with quantum physics. Okay. Quant is like finance, bro. That's a traitor. It's a traitor. Yeah, yeah, yeah. Okay. I'm like- Similar mouth under the hood though. [52:53] Kind of. Yeah. For this talk, kind of. But my skills are actually like building computers. [53:00] When a coin trader is like... [53:02] If I'm trying to like short the price of something later, I'm looking at like satellite imagery in some random country to see like how many cars in the Target parking lot. Right? Hearing out like how things are going. It's the industry meme. Yeah. Yeah. Yeah. Or like, what is it like, Wall Street? I think this is my quant. Right? I'm more like, [53:19] a lot of physicists, like kind of nerds in the corner who are just exploring, like, how does the universe fundamentally work? Which at its like smallest level is like quantum physics. Quantum meaning just like small things. Yeah. So that's kind of the difference there. Okay. [53:35] There are a lot of applications in the financial industry, which, of course, crypto is off-shoots of that. Yeah. Yeah. Okay. And it still stands to a certain extent. This is my point. This is my point. Okay. So you're working on something new. We're not going to talk a lot about that today, but as much as you want to talk about it, we will. And we are going to talk about, though, a piece that you wrote. [53:58] about quantum computing [54:01] breaking bitcoin for sure and um there there it is it's great it's um heady it's very a good read but you got to sit down and really think about it um or put it in clod and the mouth of some words are in there for sure um and i want to definitely get into the piece but the first sentence that

[54:22] of of that article you say crypto quantum computing and ai are victims of social science of a social science concept that i call telephone entropy can you explain what that is so i'll give you more background just on why i wrote the article i started it kind of like tongue-in-cheek right because a lot of people are mixing these words like ai is a big buzzword right crypto is a big buzzword quantum computing is a big buzzword [54:46] What I was basically trying to do in like a funny way is make a meme title, right? [54:52] how a quantum computer can break Bitcoin using AI. That's going to, people are going to look at that and say like scam, right? Okay. But the thing is, and what telephone entropy, to my point is it's going to be long term. [55:03] is you basically have these data dense concepts, right? Like cryptography has a lot of stuff under the hood that is really legit. And there's like a lot of really good computer science, a lot of really good like tech there. AI, of course, as we know, is really, really complex under the hood. There's a lot of like social implications on top of that. And now you have this new thing like quantum computing, which [55:23] has had its like good and bad times. There's been times where people were really, really excited about it and there was a lot of like buzz. [55:29] There were times where people think it literally will not work. Even businesses sometimes think it will not work. [55:34] And my point of like making this a thing or saying telecon entropy, [55:39] is similar to how like when we grew up and we were playing like the game of telephone where i said the word like cool and you pass it along like a circle and almost never on the first time did that word like make it all the way around without getting jumbled right my point there is for these concepts that are like way more complex than just like saying the word cool down a line a lot of the meaning and the truth behind these things gets really jumbled up yeah and so what happens is people have opinions people you know they make money they lose money and they basically

[56:09] like feelings about how these things are going. And a lot of the real like meat of what is possible here gets kind of lost. [56:16] In physics, we talk a lot about entropy, which is the randomness or chaos in a system. [56:21] And so because I wanted to make this seem very like it, [56:24] researchy or more like legit, I use that word to basically say like, a lot of people are saying a bunch of stuff. But that's my whole like telephone entropy. And I also wanted to protect myself a bit, right? Because people are gonna do that link and be like, Oh, this is gonna be something which is not not correct. A disclaimer, like I'm putting map out there. [56:51] I want people to look at it and find a flaw in it. Yeah. But I think the whole point of why I wrote this is we should be looking at these three things from like, [57:00] an information theoretical perspective meaning like how are these problems being represented on the ship and how can we like build upstream [57:08] Thank you. [57:09] computers, algorithms, systems on top of that to figure out signal from noise. Yeah. I love that. Before we even get into the article, I feel like that is such a good concept right now. There's so much just like slop. Obviously, there's all this AI slop, the Hulu's nations. And I think it's so easy to just get lost in the sauce of what we're even talking about. Yeah, for sure. In the AI world right now. It even gets to the question of what is slop? [57:35] Right? Because people are purchasing things that feel good. Like even Claude or Codex, right? Or Chad B2 or whatever.

[57:45] there is like RL post training on these algorithms to make them [57:49] say what you want to hear. Right. And you can market that as [57:54] oh, we're making users have a better experience. [57:58] We're making it feel safer. We're making it feel more friendly. But at the end of the day, it's propagating like human psychology of what they really like. [58:06] And so if you have these things that are just kind of like feeding you [58:11] what our brains neurologically want to hear. It's like what even in that case is like a good or a bad outcome. We're getting like super meta here really early. But it's like the financial point, right? Like if people will pay for something and there's like a dollar exchange of information, [58:25] you now have this like quantifiable metric of this is like good versus bad. And so if the AI is putting out slop, but people are paying for it, and GDP is growing and everyone's happy, then where do you kind of draw the line on, hey, I'm kind of going to pull this back or I'm going to get back to the basics of what's happening? Um, [58:42] And I think that's like a dichotomy that you see a lot in just like research and industry. Yeah. Research is fundamentally... [58:47] figuring out how things work, right? Like, what is this table? What is the table made out of, right? How do the legs work? [58:53] How does all the air in the room interact with the table and figure that you're not thinking about money, right? You're thinking about like, [58:58] universe structure. - Yeah. [59:01] I would argue that industry is the opposite, right? Industry is how do I sell something to somebody or say something to somebody [59:08] to make them feel comfortable. And a lot of the winnings of that game are [59:14] is financial value. Yeah. So, yeah. Yeah, it's interesting. Very interesting. But slop versus like giving people what they want, I think

[59:24] yet to see yeah i also think that um we cheated i also think that like there is [59:33] I don't know, there's like immediate [59:38] value, sort of what you're speaking to, like immediate commerce or, um, [59:43] an exchange of money that might happen that in the moment, [59:47] is valuable because... [59:49] that exchange happens but in the long term is cheap and like doesn't sustain and [59:55] I am curious about in the conversation around this lab, like something I have thought about is like, oh, there's these things that feel-- [1:00:02] um that are instant gratification that feel like immediately good but [1:00:06] I believe will not like stand the test of time. Like a really stupid example is like, did you see the, um, the AI fruit, um, [1:00:13] love island stuff yes okay and like that is coco melon for adults [1:00:21] um yeah there's like an immediate like something fires in your brain when you see that and like you kind of can't look away but that it's not like nourishing to you and your brain and your soul and i think eventually that we will see like the the results of that over the course of the long tail not actually having value and not producing like positive sum in the world i would argue that this [1:00:43] world already works. I think AI is just accelerating and making it way more obvious. Yeah, I think like there were always like, [1:00:51] downstream effects to software technology, like building things up and

[1:00:56] at the end of the day, whenever you want to, [1:00:58] like build big things, you kind of have to, again, [1:01:01] go against what humans inherently desire. That's the easiest way to make money. [1:01:04] Now, because it's so quick to give people exactly what they want, I just think we're kind of going to [1:01:09] increasing exponentially increasing. Yeah. That makes 10 sense. Okay, I want to get into discussing [1:01:18] quantum computing and cryptography. And there's been this conversation for a while around how the future of quantum computing will [1:01:25] essentially break different cryptography, Bitcoin being like the example that is talked about a lot. There's a lot of like fear around that. And [1:01:34] And in the last month, there's been [1:01:36] there were some research articles that came out that basically, [1:01:41] talked about that that might be sooner than we think that it might be easier than we think and then the timeline blew up and everybody's talking about it and everybody's arguing whether that's true or false and then there's like a combination of um some fear mongering and then some i think really like cogent takes can you just give like an eli 5 explain it like in 5 on what that concept is and sort of [1:02:03] how that would be possible. Perfect. So [1:02:06] I'll go back to my article when I wrote it in February. [1:02:09] The industry standard by Google was around 768,000 qubits to break an ECC curve. Okay. ECC curves are... [1:02:19] based like elliptic curve photography, if I'm saying it correctly. [1:02:22] But they are for like ZK snarks on like layer twos. Okay. For Ethereum and Bitcoin are like the Pygagonic protocol that is very widely used. Okay. Shor's algorithm on a quantum computer is

[1:02:34] very good at breaking that. Okay. A lot of people, the way that they approach explaining this is [1:02:41] I have some very genius algorithm that is like figuring out the solution way faster than a classical computer. Okay. Which is theoretically correct. That's fundamentally correct. [1:02:50] Um, [1:02:51] a better way to look at it in the way that I posted and what I wrote [1:02:54] is there's also an information theoretic way to think about this problem where [1:03:01] We are currently... [1:03:03] encoding data like cryptocurrency on CMOS transistors. [1:03:09] A CMOS transistor is basically an advanced wire where electricity is going in and out of that wire. Okay. Yes or no, one or zero, binary. All of computing is based on electricity going through a wire. Okay. A quantum computer is like a fundamentally different way of expressing information. Okay. So it is not just a wire where data is going in and out. [1:03:33] in some cases. So... [1:03:35] we're getting a little bit technical. If the purpose is this discussion, it's a way bigger search space or way different like piece of matter that's being used to do computing. [1:03:45] And with that different piece of matter, you're able to represent [1:03:49] the curve problem different. Okay, so wait, let me just make sure I'm sort of tracking. Yeah. [1:03:54] Basically there are [1:03:57] the way that we have... [1:03:58] Thank you. [1:03:59] There's computing that exists today and then there's quantum computing. And other forms of computing, not just quantum.

[1:04:05] People might think that quantum computing is just like a layer up from the way that we've been doing compute. Yes. But what it actually is, is like one is an apple and one is a dinosaur. Yes. Okay. Yes. Okay. It's like very different. But there are like... [1:04:21] It's like different directions, right? So everything we've built has been like apple tree. [1:04:25] But maybe you plant an orange tree. Maybe there's a strawberry... Whatever, I don't know how strawberries grow. [1:04:31] I don't know. Strawberry apparatus, like pear tree. And maybe by exploring how to represent the same information, like loading this onto a different root, the fruit that comes out of it will be... [1:04:44] way better way easier easier to harvest whatever but the thing is because [1:04:49] we haven't had AI and it's been so hard to do these other things that made no sense. Right. Okay. Because it took way too much time to explore and theorize about this stuff. Yeah. You got to wait for strawberry to grow. A hundred percent. It's just too much. Okay. Okay. Now we have like, [1:05:02] I mean, you guys know Minecraft? Like, you know, bone meal where you basically just like put it on the root... Not that familiar with Minecraft. There's like one thing where if you want to grow a crop, you just like kill a... [1:05:14] uh, [1:05:14] enemy and then you get bones and it like grows the wheat really, really quickly. I see. So in that moment, we're in that moment where like AI allows us now to explore all these different trees, which we were able to do before. Okay. And now it's like, [1:05:27] both fundamentally viable and economically viable to explore because AI is eating everything else up. Okay, cool. Wow, great job. Really good explaining it. Sharon gives the best explainers. But to succinctly answer your cryptography question, though,

[1:05:41] The problem with quantum computers versus classical is with a wire, the wire is like blinds the problem. Okay. So if I have this like really big number that I want to decrypt, the wire does not know anything about that number. I see. It's like guessing and checking. [1:05:57] to see can I find a solution to that number? What a quantum computer can do [1:06:02] is not just... [1:06:04] guess better. [1:06:05] but it can actually load in the entire number where it's not a guess and check. It's finding needle in a haystack. [1:06:13] Okay. So the solution is already existing like on the physical chip. Like it can be represented on this chip. Okay. And the algorithm is like finding some way to orient the chip and get something out. Okay. Rather than I'm just like guessing and checking randomly to see if I get a solution. Okay. No, you go. Can I just see if I'm tracking? So you're saying one version of it is [1:06:36] like saying, go and look, but you don't know what you're looking for. And the other saying you're looking for this very specific thing, you still have to like search around and try to figure it out. But like you're looking for the needle, I would say both don't really know what they're looking for to start. Okay, but it's kind of like, [1:06:53] If I have like a box of toys, right? [1:06:55] You can basically tell somebody and maybe I'm reading what you said. [1:07:00] you could tell somebody the toy is in the box versus the toy might not be in the box. Okay. And so the search space [1:07:07] from your perspective is way more defined because you know to look in this area. Okay. Whereas if you're outside of it, then you don't know if the solution is in there. It may be somewhere else. Okay. That's actually that's not exactly what's happening. No problem. We're getting close. Okay, so

[1:07:25] So... [1:07:26] Basically, so what happened in March? Yeah. [1:07:30] So Google dropped something to 168,000, around 100,000 qubits. [1:07:35] or atomic. [1:07:36] which is a Google e-startup. So it was actually founded by people from Google Quantum AI, who are very locked in at that group. Dropped it down to 10,000. Okay. How they did it was... [1:07:47] similar to the tree analogy that I gave where, so now we're looking at just the quantum tree. There are like different... [1:07:55] species of quantum trees. They basically said, forget the old way of building a quantum computer, we're going to build it in this way that's way more advantaged for this specific problem. I think it's more advantaged for just quantum computing in general, but they're saying there's a fundamental way to build the actual processor quantum computing. And it allowed for a 10x decrease in the number of qubits, like physical qubits, full stop to break encryption. Okay, well, [1:08:22] My point with the article... [1:08:24] is [1:08:25] I think that we're kind of approaching the encryption problem wrong. [1:08:29] because we don't need or I don't think we should think of this as like we need a big fault tolerant quantum computer to break it. We should think of it as like what is the best [1:08:38] like combination of species to piece together to like break Bitcoin or break. If I was an attacker, that's what I would do. Why would you do that though? Why would you do that? Okay, wait. So when you say, when you talk about like [1:08:55] the computer. Like, are you talking about hardware? It's a combination of hardware and software? Yes. Okay, can you talk a little bit about that? Yeah. So I'm of the opinion that

[1:09:04] And [1:09:05] through working with it, it is still very hard, right? I'm like grossly underestimating or like overshooting whatever, like the complexity here. [1:09:15] But in the long run, I really feel strongly that AI is going to do a lot of like the software piece of things. Okay. So the best way to look at like solving these problems is [1:09:26] designing new hardware modalities that make AI's job way easier. Okay. [1:09:31] And so I do mean like the entire stack of this is a computer that we're like, it's like, I'm touching it. I'm feeling it. Yeah. I'm going to load AI onto this new computer. I'm going to tell us to go to work. Okay. Okay. And I think that that is likely going to be. [1:09:47] What does it? [1:09:50] Thank you. [1:09:50] Is it financially viable to people? Who knows? But that would be my take off. Okay, so what's happening here? [1:09:57] On that map. Maybe if you're like Ramona, you get to something. I got all the max specs that you have no idea of what's going on this. My thing is I think the real problem is AI. [1:10:13] I think so many people say, "Why don't you do that?" And yes, it is. But the problem is once you build these things and you give it to AI, [1:10:22] it will then learn. So like that, those research papers that are out, I am [1:10:29] So I think I know for a fact [1:10:31] that people are already training models on this paper or like tuning existing models to try to beat

[1:10:37] What is already done. - I see, okay, okay. And they will very likely be successful in that. - You're betting against AI if you think that's not the case. That's my answer there. - Okay, the bar is like, [1:10:45] set and then that gives a bunch of information to then move the bar. 100% and AI will just continue to do that. Okay. And the thing is as you narrow onto the problem [1:10:55] AI gets more efficient because it's not looking for things that are wrong. So as it gets more and more correct and converges to some answer, because like the search space decreases, AI will become more efficient. And I think you'll only increasingly accelerate to the correct answer. OK, yeah, I this might be a dumb question, but OK, let's say. [1:11:17] People are doing, look, Ethereum, Bitcoin, whatever. Let's use Bitcoin as an example. [1:11:21] Is the motivation like, okay, I can break Bitcoin. [1:11:27] And [1:11:28] then I get all that Bitcoin and then I sell. Like, yeah, is it a financial motivation? Is it like a gigabrain person? Okay. [1:11:35] I just want to be the person that broke it. It's just cool. It's just cool. It's complete ego and I just want to do it. Okay. And probably the ego is on the table. I just want to do it and talk about it and put it open source and let's see what happens. Okay. Whoa. Whoa. Let's go burn. Hopefully don't get Harry or Strict. But like, so, you know, tongue in cheek, but to my point, like, [1:12:00] I also wrote an article... [1:12:02] cryptography is [1:12:05] So I'm not a liaison for the entire quantum computing community. But from what I've seen, cryptography is like a side effect of a quantum computer. There are like way cooler things that a quantum processor can use

[1:12:18] it can be used for than breaking [1:12:21] No one cares about that. [1:12:24] But... [1:12:25] people need to get made money people need to raise money people need to do things to get high it actually goes back to [1:12:30] - The thing you're changing. - Yeah, yeah, totally. [1:12:33] it is actually very, very clear. And the best quantum algorithm right now, the clearest one that Peter Shor created decades ago, still holds to this. [1:12:42] So they're gonna exploit this opportunity? Yeah. While the opportunity is here and then I think maybe there'll be bigger things like [1:12:48] high temperature semiconductors or like stuff on the line that are more impactful financially. Okay. And some people try to sell the anti, they try to sell like, oh, we're gonna protect against [1:12:56] crypto threats. [1:12:58] Maybe that's a viable business, but I'm not I don't like selling things to people who don't believe. Right. Like if I have to convince you to do something. Totally. I'm not wasting my time. I'll just take some of it and put it on a proxy and we'll know. Yeah. Yeah. Yeah. Yeah. Wow. Really interesting. Okay. So. [1:13:16] Should I sell? Should I sell? So this gets to my point at the very beginning where [1:13:22] If I could say one thing, I don't want to look at the camera. [1:13:35] AI is now in a place where if you have [1:13:39] motivation, if you have agency and you have like done some work to understand, you could probably figure out for yourself what the good answer is. I will say I hold crypto. Okay, I hold crypto. I like it. Okay.

[1:13:51] But I'm also confident in my abilities that when something happens, if something happens, I think it's going to be winning. [1:13:57] um [1:13:58] I will be able to like crawl back whenever I'll be fine. Yeah. Okay. I think if you're like, [1:14:04] listening to what people should say you should do, I think you've kind of already lost. Yeah. Um, [1:14:10] Because I think to the telephone entry point, incentives are so misaligned. Totally. Yeah. [1:14:15] it's especially now hard to really figure out what to do. Yeah, yeah, yeah. But I hold crypto. I will also put disclaimer, I know nothing about the cryptocurrency industry. So I mean, I know a bit, but in terms of like, [1:14:28] corporate professional like actually having like intelligent research knowledge to say yeah i have not spent nearly enough time to say like this is gonna happen they're gonna defend against it there are also like protocols that exist that are going to like beef up these encryption algorithms to be quantum proof right a really good example we'll take the snark for stark example like zk starks because i think they're [1:14:49] lattice-based or hash-based, [1:14:50] They're more robust against quantum attacks. [1:14:53] The only problem is if you have some like, [1:14:57] Like I'm not going to name the name, but there's some like L2 solutions that they use snarks for efficiency. Okay. So the problem is if you like make the encryption harder to break, you lose the speed. [1:15:08] yeah okay so there's kind of a trade-off there where you're kind of you've got to figure out a fast way to like make sure it stays [1:15:13] um again i'm not on the defense side of this so there's definitely going to be people that will have like good solutions it will only like

[1:15:22] get better over time. [1:15:24] My only concern with that is what we're seeing with like Claude Mythos, right, where [1:15:29] even if we find some way to protect now it is never going to be like completely 100 secure there's always going to be innovation that happens what is the timeline maybe it's not relevant to someone to make a financial decision right so i'm not [1:15:44] Again, I'm not speculating there. [1:15:46] But what I will say is I don't think that this should be something where you just feel like. [1:15:50] I'm completely safe in history. And somebody's going to take care of it. I would say do your own research. [1:15:56] And maybe their structure is... Or get a quote for it. [1:16:01] Start a podcast. [1:16:04] Nice. [1:16:06] Two thoughts. One, you're invited back anytime. Two, you should do more media. I should do more. I know. But like, I told you. Yeah, a star. A star. I just got a mic. If you, the more you get into media, I think I want, I like being, [1:16:20] incognito yeah for sure actually yeah it's a gift to not have your face on the internet i just i love that's yeah that's huge yeah honestly don't listen right oh we're gonna keep learning unfortunately well this was so fun to be here with both of you and at acpc and kate [1:16:36] I think you can close us out. [1:16:39] She's saying no? No she can't close us out? [1:16:42] Oh, oh, I have to close this out. [1:16:45] Great. Great, great, great. I'm ending the stream. [1:16:48] - Yeah.