Shopify’s Lütke Builds AI Search Engine to Counter Big Tech E-Commerce Platforms

Shopify的Lütke打造AI搜索引擎以抗衡科技巨头电商平台 电商格局正向大语言模型驱动的智能体电商转变，促使Shopify构建专有的产品世界模型，将产品质量置于庞大的营销分发预算之上。 在这场深入的对话中，Tobi Lütke回顾了Shopify二十年的发展历程，详细阐述了像“GSD”这样的专有内部软件如何塑造公司的文化与运营速度。他解释了电商从早期充满技术摩擦演变为如今高转化率独立网店格局的历程、Shop Pay的强大力量，以及针对智能体电商和稳定币的战略准备。此外，Lütke还探讨了他专注于垂直领域而非横向扩张的理念、在Coinbase董事会任职期间学到的经验、加拿大的文化与经济格局，以及他对赛车运动的技术痴迷。 找到一个美妙且无法彻底解决的问题是人生最大的馈赠之一，因为它能赋予人一份持久终身的追求。 (0:00) 优质产品是解构消费主义的终极良药，而消费主义主要是由于人们厌恶自己拥有的低质劣品而驱动的。 (0:00) 公司是一种未被充分研究的社会技术形态，它赋予人类为共同使命而协作的努力以合法性并加以协调。 (2:07) 传统的财务报表和商业指标难以有效衡量软件工程与研发成果，因为它们仍然是针对实体工厂设计的。 (4:31) Shopify依靠一套名为GSD（“把事搞定”）的定制内部操作系统运转，该系统利用交互式数字画布保持业务运作清晰透明。 (7:27)

Shopify founder Tobi Lütke contends that enduring technology companies succeed by obsessing over deeply specialized, unsolvable problems and shaping internal environments through custom software rather than traditional corporate playbooks. Under this operating philosophy, modern infrastructure primitives have inverted e-commerce, allowing independent merchants to technically outperform legacy retail conglomerates. Custom Tooling as an Organizational Engine (14:20) Lütke views companies as an underappreciated social technology designed to legitimize collective dedication to a shared mission. Because traditional factory-centric metrics—such as counting pull requests per engineer—fail to measure software R&D productivity, Shopify structured its workflow around an internal operating system called GSD ("Getting Shit Done"). Reviewed every eight weeks on interactive digital canvases, GSD establishes a centralized source of truth, bypassing off-the-shelf enterprise software like Workday whose embedded assumptions can subtly distort an organization's compensation and operating philosophy. Invoking Marshall McLuhan's maxim that "first we make the tools, and then they shape us," Lütke argues that digital environments dictate organizational ambition and velocity far more effectively than executive policies. Software adjustments deploy instantly, circumventing bureaucratic town halls. By building custom tools—ranging from an internal GeoCities clone to bespoke compensation software—Shopify deliberately designs opinionated environments intended to push both its staff and merchants beyond their initial operational goals. The E-Commerce Inversion and Peak-Load Architecture (19:47) A structural inversion has reshaped global commerce: while storied retail brands frequently rely on brittle, homegrown software frozen by legacy technical "cement," small businesses on modern platforms now achieve superior site performance, lower latency, and higher checkout conversion. Proprietary tools like Shop Pay deliver a reported 16% conversion increase by eliminating manual address entry on mobile devices. Paired with Instagram's visual discovery engine, this ecosystem enables niche direct-to-consumer merchants to acquire dedicated audiences profitably, turning Kevin Kelly's "1,000 True Fans" concept into high-resolution "4K commerce." Supporting this retail shift required solving extreme transaction spikes. Beginning in 2010 when viral t-shirt sales from theCHIVE repeatedly crashed upstream network routers, Shopify treated peak demand as a technical training ground rather than offloading challenging merchants. This engineering focus prepared the infrastructure for high-velocity flash sales from brands like Supreme and Kylie Cosmetics, relying on inventory pooling and specialized payment infrastructure to overcome database lock contention during concurrent checkout transactions. Vertical Craftsmanship Versus Horizontal Expansion (62:02) Lütke rejects conventional MBA doctrine that pushes mature companies to expand horizontally into unrelated markets once they achieve an equilibrium. He highlights NVIDIA's decade-long investment in GPUs and CUDA prior to the artificial intelligence boom as evidence of the compounding value of vertical hill-climbing. In 2008, venture capitalists broadly rejected Shopify's pure e-commerce software pitch because only 40,000 online stores existed globally, misjudging platform friction as a natural market ceiling rather than a latent opportunity waiting to be unlocked. In Lütke's assessment, true enterprise defensibility originates from specialized domain craftsmanship rather than interchangeable corporate administration. He points to software developer Erich Gamma, who built VS Code as his fourth programming editor across a multi-decade career, to illustrate the power of deep focus. Similarly, direct-to-consumer businesses that document authentic product development often see explosive organic traction—sometimes scaling to 100,000 or one million orders within 18 months—by establishing direct consumer trust that polished corporate public relations cannot replicate. Agentic Discovery and Merit-Based Retail (214:43) Online shopping discovery is shifting from traditional keyword searches and ad-saturated portals toward agentic commerce powered by large language models. Rather than navigating web forms, consumers are increasingly consulting conversational AI engines like ChatGPT, Claude, and Perplexity for targeted recommendations, completing purchases through integrated one-click checkouts. Shopify is facilitating this shift by connecting the Model Context Protocol (MCP) to merchant catalogs and assembling search engineering teams to construct aesthetic "product world models" that interpret taste, quality, and semantic similarity rather than simple text keywords. Lütke argues that this AI-driven discovery levels the retail playing field, allowing high-quality niche products to surface based on intrinsic merit rather than marketing spend. By helping buyers identify durable, well-made goods, this model addresses the root cause of consumer waste: discarding cheaply made items people dislike. Concurrently, Shopify's cautious integration of stablecoin payments settled into fiat currency via partners like Bridge provides internet-native financial utility, mirroring how QR codes gradually became ubiquitous after solving real-world friction. While Lütke remains confident that specialized infrastructure, founder-led agility, and agentic AI will continue to elevate independent commerce, unresolved questions persist around how legacy enterprises will adapt as their internal systems fall further behind unified platforms, and whether risk-averse institutional cultures can resist the urge to govern away the friction necessary for genuine craftsmanship.

Shopify 联合创始人兼首席执行官 Tobi Lütke 提出，科技企业应专注于不可完全解决的“美丽问题”，通过自主构建软件环境与深入技术底层来驱动组织进化。他指出，随着高性能基础设施与智能体商业的演进，注重品质的垂直小品牌正在技术效率与发现机制上逆袭传统巨头，而企业成功的关键在于抵制盲目的横向扩张，深耕专业能力。 软件塑造组织：内部工具与研发度量 (2:07) Shopify 将公司本身视作一种被低估的“社会技术”，旨在凝聚大量团队成员为共同使命协作。针对研发投入占总预算约一半却缺乏物理工厂式衡量指标的难题，Shopify 摒弃了源自 Bethlehem Steel 泰勒制秒表逻辑的传统度量方法（例如将每人每日 Pull Request 数量作为生产力代理指标），转而自主开发名为 GSD（“Getting Shit Done”）的内部操作系统。该系统结合 tldraw 数字画布进行每八周一次的项目全景评审，将业务真实状态显性化。 Lütke 援引麦克卢汉关于“人造工具、工具反过来塑造人”的观点，坚持自主研发 HR、薪酬乃至内部静态站点托管等核心工具，避免因引入 Workday 等外部企业软件而被动吸纳第三方厂商内置的世界观。目前 Shopify 核心代码量约 2,000 万行，为全球最大的 Ruby 应用，其 TypeScript 管理界面亦包含 800 万至 1,000 万行代码。在他看来，通过代码部署而非繁复的政策宣讲来调整日常工作环境，是驱动公司文化与业务野心进化的最高效手段。 零售格局倒置：高并发淬炼与结账基建 (19:47) 在传统商业模式中，小型零售商常因技术能力不足而在移动端体验上落后于巨头，但 Shopify 促成了一种“零售大逆转”：依托平台支持，中小品牌在网页加载延迟与全链路转化率等技术指标上，已经显著超越了许多系统笨重的老牌零售企业。这种高并发能力的形成始于 2010 年为 theCHIVE 处理比尔·默瑞（Bill Murray）T恤闪购的经历；Shopify 没有因服务器资源消耗过大而解雇该客户，反而将其视作“健身房”进行系统极限施压，从而为后来应对 Supreme 和 Kylie Jenner 等极端流量洪峰（Drops）奠定了底层架构能力。 针对数据库层面的“锁争用”（Lock Contention）难题——即数百万用户在毫秒级内同时尝试扣减同一个库存整数，Shopify 与 Stripe 展开深度协作，通过产品池与快速补货机制替代教科书式的严格串行事务。此外，前身为“Remember Me”的 Shop Pay 解决了移动端手动输入地址的高摩擦痛点，仅凭功能存在便带来约 16% 的转化率提升。Lütke 指出，电商领域 80% 的底层功能（结账、托管、支付）高度通用，大企业不应盲目固守由离职人员编写且难以维护的自建老旧系统。 坚守垂直问题：反思横向扩张与生态协同 (59:24) 硅谷商业文化常过度推崇寻找类似 AWS 的“第二曲线”，而传统商学院与风投模型往往驱使企业横向扩张至 CRM 等领域。在 2008 年全球仅有约 4 万家网店时，风投机构因市场规模预期过小而拒绝投资 Shopify，忽视了高摩擦本身才是限制开店数量的核心瓶颈。Lütke 认为，NVIDIA 长期垂直深耕 GPU 与 CUDA 最终迎来爆发的路径证明，在 AI 大模型逐步普及、通用管理能力逐步商品化的时代，真正不可替代的价值源于对垂直问题的专注与工艺精神。 这种理念延续到了长期合作策略中。Shopify 与 Stripe 维持了超过 13 年（自 2012 年起）的共生关系，双方严格聚焦各自的核心任务并避免基建重复建设，摆脱了科技行业常见的“囚徒困境”。在获客层面，Meta 旗下 Instagram 的发现场景与 Shopify 形成了高效协同，不仅助推了 DTC 品牌的爆发，更使凯文·凯利（Kevin Kelly）的“一千位铁杆粉丝”理论具象化为高清晰度的细分零售市场。此外，Shopify 正在审慎推进与 Bridge 合作的稳定币结算功能，使商家在无需承受加密货币价格波动的前提下接受新型支付。 智能体商业前瞻与品味驱动的搜索重构 (214:43) 随着消费者逐步转向使用 ChatGPT、Perplexity 等对话式 AI 工具进行商品调研，电商正步入“智能体商业”（Agentic Commerce）阶段。Shopify 正在为各大主流大语言模型构建专属底层接口，包括将模型上下文协议（MCP）接入每个独立店铺，使 AI 充当了解个人偏好的专属私人导购。这种交互将免除繁琐的网页表单填写，重塑产品发现链条。 伴随 AI 聚合能力的兴起，商业竞争的重心正由渠道营销力量转向产品本身的实用价值与品质。Lütke 指出，学术界主导的传统关键词搜索本质上是为论文检索设计，无法应对复杂的审美与相似度需求。Shopify 已建立专门搜索团队，探索嵌入向量（Embeddings）与“产品世界模型”，以捕捉品味与质感。他认为，消费主义泛滥的根源并非过度消费，而是人们厌弃并丢弃低质粗造的产品；帮助消费者精准发现寿命长久、极具工艺价值的高品质物品，才是解决这一问题的根本途径。 “不可治理”的领导哲学与组织防滞 (76:08) 在组织管理哲学上，Lütke 从出任 Coinbase 董事以及与 Brian Armstrong 的合作中提炼出“优秀创业者应保持不可治理（ungovernable）”的准则。Shopify 倾向采用职能型架构而非传统自上而下的科层制，公司领导层主动跨层级“直达”业务与技术前线，以防止脱离一线实际。对于技术领导者而言，迅速掌握 AI 等新技术，也是识别内部具备敏锐技术嗅觉骨干的重要标尺。 为了防范组织陷入停滞并甄别出究竟是“热爱问题”还是“热爱特定解决方案”的员工，频繁主动的组织架构调整被视为一种健康的干预手段。架构调整所带来的协同收益立竿见影，而其负面效应则滞后显现，这促使不适应变化的人员自主流动。Lütke 强调，研究企业失败案例远比研究成功案例更有价值，因为衰亡往往不是由于突发灾难，而是因与一线脱节导致的微小错误逐渐累积而成。 Lütke 将创业与商业视作动态变化的“热情项目博物馆”，坚信技术环境的设计权直接决定了团队与生态的抱负上限。然而，随着智能体商业将搜索发现权重新中心化，独立商家如何持续保持对算法中介的议价能力，以及高度定制的内部软件生态在面对极端规模化时能否长期避免维护陷阱，仍是 Shopify 在下一个二十年中面临的开放性命题。

[0:00] Tobi Lütke: The best gift in life is finding a beautiful problem that you can never solve... [14:20] Tobi Lütke: First we make the tools, and then they shape us. [19:47] Tobi Lütke: The rebel alliance is doing better than the large, established companies. [72:28] Tobi Lütke: great entrepreneurs are ungovernable [76:08] John Collison: If you're going to make mistakes, at least make new mistakes that haven't been discovered yet [85:07] Tobi Lütke: the world is a museum of passion projects [89:56] Tobi Lütke: everything's interesting [89:56] Tobi Lütke: eventually, you run out of other people's money [96:46] Tobi Lütke: report card of all your inadequacies as a driver [99:46] John Collison: the most zero-sum competitors you can imagine

[0:00] The best gift in life [0:01] is finding a beautiful problem that you can never solve, [0:04] and even if you accidentally solve it, [0:06] if you're so unfortunate to solve it, hopefully, [0:09] it has, like, plenty of enlightened problem children. [0:12] Consumerism is like a thing that, like, [0:14] is being thrown around, but where does it come from? [0:16] People throw away things [0:17] because they hate the things they have. [0:21] The thing that solves consumerism is quality products. [0:24] A personalized ad is a wonderful thing. [0:26] I am scrolling past something that's monetizing [0:28] the free application that I appreciate using. [0:31] Yeah, yeah. You should tell the EU this. [0:32] Yes. Well, actually, I do, actually. [0:35] Some people fall in love with solutions. [0:37] Some people fall in love with problems [0:38] and I just, like, fundamentally, [0:40] I'm a person who appreciates [0:41] people who fall in love with problems, [0:43] but, like, it's very hard to know who is who, [0:45] because, well, if you don't change anything, [0:47] they look exactly the same. [0:50] What's the point of you only having half an Irish pub? [0:52] Oh, this set? [0:53] Well, this is our MVP, [0:56] when we weren't even sure that this would be a thing. [0:59] Tobi Lütke launched Shopify in 2006. [1:01] We've been working closely with them for over a decade, [1:04] and I always enjoy getting together with Tobi [1:06] and I always learn a lot. [1:07] Cheers. [1:08] Cheers. [1:11] Okay. Many places I could start. [1:12] What's the size of the code base? [1:13] In as much as, you know, [1:14] vendor stuff or whatever. Yeah. [1:15] I think the sort of core part of Shopify, [1:19] with all the, like, minus stuff like the identity system [1:22] and those kind of things, [1:23] it's like 20 million lines of code, [1:26] and then the TypeScript written admin interface [1:30] is like another, I think, 8, 9, 10. [1:33] It's apparently of the largest TypeScript applications [1:35] and definitely that largest Ruby application. [1:37] So pushed it into pretty uncharted territories [1:40] for our tech stacks. [1:41] It does so much. [1:42] Like, it's like a full FinTech business in there [1:45] with capital and these kind of things, [1:46] and all of what we are doing together, [1:48] and my company adopts the complexity, [1:52] the real-life messy complexities [1:54] of the commerce world into itself, [1:57] to then front it with approachable interfaces, [2:00] things that don't, you know, [2:02] just making all the pieces fit together, [2:03] and so that requires a lot of work. [2:07] So there's like thousands of projects going on [2:10] at any given time. [2:11] So, going through all of them is work, [2:14] but we have a good system. [2:15] We have a product operations team that prepares for it. [2:18] We have an internal system that's just for the reviews, [2:20] which is actually kind of cool. [2:21] What does the system do? [2:23] So I'm fascinated with companies. [2:25] I think companies are under-appreciated for the, like, [2:29] just we don't think of that generally. [2:31] It's like, you know, companies are technology [2:33] and tech themselves. [2:35] They are technology by which you create, [2:38] part of what they create is social acceptance for, you know, [2:42] people, maybe tens of people, hundreds of people, [2:44] thousands, tens of thousand of people, depending, [2:47] to spend all their day pursuing a mission together. [2:50] You know, it's really just like universities exist, [2:55] giving you time that you can use [2:57] for, like, hopefully intellectual inquiry into some topic, [3:01] and everyone's cool with that, [3:03] although if you try to do the same thing, you know, [3:05] just with your laptop on the internet all day long, [3:08] it has not nearly the same social acceptance. [3:10] It may also be valuable, [3:12] and then companies are ones that allow people, [3:14] a lot of people together to, you know, [3:16] try to make, like, pursue a mission or create something [3:18] hopefully world-class. [3:20] So I think companies are understudied, [3:23] and also this sort of incentive system behind companies [3:27] tends to be that [3:29] there's no absolute need for perfecting them. [3:33] There's just like, [3:34] you've got to be better than everyone else, [3:35] and at some point, in many industries, [3:37] like, if you get ahead, [3:39] it's really, really hard for others to catch up, [3:41] and so there's sort of a point equilibrium, [3:45] at which point you just— [3:46] Because you're good enough [3:47] and so there isn't pressure to get better beyond some point. [3:49] Yeah, I mean, I just, I mean, taking a sort of example, [3:52] I talked with some executives [3:53] in the gold mining space, right? [3:55] It's just, like, really sort of interesting, [3:56] because I'm like, what is your day like? [3:57] Yes. Like, the same questions. [3:59] Like, it's much more interesting. [4:00] And what is a gold mining executive's day like? [4:02] It's just like, I mean, my conclusion, [4:04] which wasn't maybe what he was trying to guide me to, was [4:07] that it's essentially [4:09] government relations or investor relations. [4:11] You know, like, if you are, you know, like, [4:13] you know how much is in the ground. [4:15] You know, sort of, it's a capital play, [4:17] but somehow, if the investors like you better, [4:20] then you get a higher multiple, [4:22] at which point it behooves you [4:23] to purchase all the other gold mines really quickly. [4:26] So I imagine it involves playing golf. [4:29] I think that's probably the technology there, so. [4:31] Isn't one of the conundrums for you, [4:33] as someone who thinks about [4:34] companies and organizations a lot, [4:36] the lack of measurability of R&D, [4:39] which is probably, you know, half of what Shopify, [4:42] you know, spends its money on is R&D, [4:44] and so if you're running a factory, [4:46] it's pretty easy to measure the inputs and outputs, [4:48] and, you know, how many widgets are we producing per hour [4:50] and what's the labor efficiency and everything like that. [4:52] It's very measurable, whereas when you're building software, [4:55] it's like, I don't know, we put a bunch of engineers on it, [4:57] and hopefully, in like six months, [4:59] they have a great product out there in the market, [5:01] but the intermediate measure is, yeah, [5:03] you can measure, like, [5:04] pull requests per engineer per day or whatever, [5:05] and everyone knows that's a very broken proxy. [5:08] I mean, lots of thoughts on this specifically. [5:09] I would love to actually [5:11] know how you think about this too. [5:13] So, traditionally, when you look back in business history [5:15] about when people got a little bit scientific about business [5:17] or actually started to improve them, [5:19] like, clearly, it all started sort of vibes-based, [5:21] and then, you know, Frederick Taylor came up and, you know, [5:24] brought the wonderful new technology of a stopwatch [5:27] to the production line and timed the various steps, [5:30] and, of course, produced, this is I think Bethlehem Steel, [5:34] and end of 1800s, and drove massive efficiency gains, [5:37] and so that's like the hero's journey [5:39] of business books, right? [5:40] because, like, and it's still [5:42] so much of the corporate culture today, right? [5:43] You know, it's how org charts tend to be structured. [5:45] It's how the financial statements are. [5:46] You know, the kind of business [5:48] that financial statements are optimized for [5:49] is kind of, you know, a factory. [5:51] In a world where almost no company [5:54] was terribly conscientious [5:55] and no company metabolized all the value available to them. [5:59] Yes. [6:00] Someone starting to hill-climb some source of, you know, [6:05] efficiency gains or just like becoming a better company. [6:09] It was like a breathtaking change, right? [6:11] Like, once someone starts hill-climbing, [6:13] everyone has to or, you know, gets left behind. [6:16] So that became the story, and it's remarkable to think that [6:21] that was probably good enough for like [6:23] probably 80 years or 90 years or, like, [6:25] from this point on, right? [6:26] Like, it's just like— [6:27] To lean, basically, has just become the 2.0 of that. [6:29] Exactly, so you said, [6:30] sort a car industry with exactly Toyota and so on. [6:35] There's some new ideas came on, [6:37] but they were all around the sort of drive for efficiency. [6:41] Mhm. [6:42] The downside of a drive to efficiency is that [6:45] it requires you to act on unquantifiables, right? [6:48] Like, you can only measure what you can measure, [6:51] like sort of a car industry. [6:52] So that's not the entire, like, [6:54] the space of how to make better companies. [6:57] Like, there's a huge amount of things [6:58] that are unquantifiable, [6:59] but taste, like quality, is hard to quantify. [7:03] Well, not only that, [7:04] because those are the obvious examples of unquantifiables, [7:06] but the big one is, I'm curious, like, [7:08] there's a team at Shopify [7:10] that's doing a great job of their product [7:12] and there's a team that's lost [7:13] and has kind of a woolly direction [7:15] and actually their dev tools suck, [7:16] and so it's very hard for them to make forward progress [7:18] and they're thrashing with lots of keep-the-lights-on work. [7:21] Just how do you distinguish [7:23] between the team that's struggling [7:25] and the team that's executing really well? [7:27] Yeah, I think that's, I mean, this is why [7:32] my answer is, I should go to, I don't know. [7:35] Here's what I've found works is [7:37] we have rituals by which we talk every, I mean, [7:40] at the latest, every eight weeks, right? [7:43] And so let the teams talk about [7:45] the progress against the goal. [7:46] It's like the new, like, [7:47] we have an internal system called GSD, getting shit done. [7:53] It's for central registry, clearing house. [7:58] It's our wiki. [7:59] Yeah. It's our, like, [8:01] feeds and stuff like this, [8:02] but it also has every project in it, [8:04] and so it's built around, you know, [8:06] teams updating everyone else, [8:08] registering, like, here's something we learned, [8:10] and so, like, that was step number one. [8:12] Let's get the actual state of a business [8:15] into a legible internal system [8:19] by which you can reason about it. [8:20] Where is every project? What are the deadlines? [8:23] What has changed? [8:24] And so this is what these reviews are about. [8:26] So we go, like, this is what we have on the screen. [8:28] They get a little tldraw [8:30] that they can drag images and mock-ups and everything into, [8:33] but that's their area, but, like, [8:35] all the metrics and all the things are, like, [8:37] around the edge on the screen right in front of everyone, [8:39] and we can talk about that [8:42] and how many people are on a team and this kind of stuff, [8:44] and giving everyone an opportunity to just talk to me, [8:49] even if it's only quickly, is incredibly valuable, [8:51] and I just learn tons of things about it, [8:53] and this is something that's in the calendar, [8:57] and I don't know. [8:58] It's exhausting and a lot of work, [9:00] but I don't know how to do it in any other way [9:02] because that is the thing by which people can say, [9:05] "Hey, it feels like progress is really well [9:08] and here's what we are building [9:09] and here's something we learned. [9:10] What should we do about it?" [9:11] Is it like, people who want to get into running, [9:13] the couch to 5K program. [9:14] Start with get off the couch. [9:16] Yes. Go walk down the block. [9:17] Go walk down two blocks, you know, and things like that. [9:19] It turns out that, from an organizational point of view, [9:22] having a centralized source where you track all the projects [9:24] and you list out your goals [9:26] and you post updates based on it, like, [9:27] that sounds too simple to work in the same way that, like, [9:30] getting off the couch is too simple [9:32] as a step to running a marathon, [9:33] but it turns out both are true. [9:35] I think, honestly, it's not a complex idea, [9:38] and it's extremely valuable for 15 other reasons [9:42] other than it powering these reviews, [9:47] but, yeah, it's not terribly complex and it works. [9:52] No, we took inspiration from GSD, your system, [9:54] as we built out ours similarly, and having a centralized [9:59] internal source of truth for projects that are going on [10:02] is surprisingly helpful, [10:05] and it feels like it shouldn't be, but it is. [10:07] Yeah. [10:08] You know, it's funny because, so, again, [10:10] it's an internal system. [10:11] Shopify has a culture of building internal software. [10:15] I mean, I tend to point out [10:17] Shopify itself started out as internal software [10:20] to power my snowboard store, [10:23] so it'll be funny to not do this, [10:26] and so I think fundamentally, like, [10:29] companies are all pretty bad. [10:34] Like, all the companies of today are pretty bad [10:36] compared to the companies we will have in the future, [10:38] and I don't know how I would possibly run Shopify [10:43] without having GSD, [10:44] and I think there's going to be some productizations of GSD [10:47] that'll be available to everyone [10:49] that's probably going to be renamed 15 times or whatever, [10:51] become a category of software that's going to exist, [10:54] and then everyone's going to be wondering, [10:56] how the hell did we build software before we had it? [10:59] Yes. [11:00] And I try to have Shopify live in that particular state [11:04] a little bit earlier, so. [11:06] Has your proclivity for developing internal software [11:08] ever led in just like a really hilariously [11:11] overbuilding internal software direction? [11:12] Oh, yeah, absolutely. All the time. [11:14] Like, what's the most ridiculous [11:15] internally developed piece of software? [11:17] We have, like, a static [11:19] sort of our internal GeoCities, [11:20] It's like static site hosting, [11:22] which is like, it's the most incredible— [11:25] Oh, GeoCities. [11:26] Like GeoCities, yeah. [11:28] It's like '90s web, with all the pros and cons. [11:32] Yeah, like, I mean, it sent us down wrong paths. [11:35] Like, it's very often, I say internal tools culture. [11:39] This is actually a little bit of a dangerous statement. [11:41] I shouldn't make it like as such. [11:42] We have a very strong appreciation of good software, [11:47] and good in the context of we have to use it. [11:49] Like, some software is really good. [11:51] Like, Shopify is really, really, really good software [11:54] if you are running, like, a D2C business [11:57] or a proof of product. [11:58] Some people try to stretch it [12:00] to subscription-based grocery stores or something. [12:05] Maybe that even works, [12:06] but I clearly didn't build it with that in mind, [12:08] and I actually, although I appreciate all my customers, [12:11] I'd rather not anyone use the software if it's bad for them. [12:14] Yes. Because that's like, [12:15] I built Shopify to avoid [12:19] people having to use shitty software, [12:20] not to add to the pile, right? [12:22] So, sometimes this means, [12:23] given our scale and our minimum quality bar, [12:27] that we have to build it ourselves, [12:28] and in those cases, we do that. [12:30] Sometimes, we build it where we have no business doing it [12:33] and that's where things go sometimes wrong. [12:35] Well, what I think is interesting is [12:36] you think it's important to build your own HR software, [12:39] and you would recommend people [12:40] at least think seriously about that. [12:41] Maybe you can share the worldview there. [12:43] Not sure, I exactly recommend it, [12:46] but we have ideas about how to do HR [12:48] in terms of compensation and so on, [12:50] that they're just different [12:51] from what's normally implemented. [12:53] This is most true in project management software, [12:56] like, again, GSD. [12:58] If you use software by others, [13:00] you have to buy into their vision. [13:02] Like, some vendors [13:05] tell you the software can do absolutely everything, [13:07] and that should be— [13:08] Softwares have a worldview, [13:09] so you're adopting Workday's worldview. [13:11] That's right. [13:11] When it comes to your HR, which— [13:12] That's right. [13:13] May or may not be what you want to do. [13:14] That's right, and I think that's important to realize [13:17] and it's important to use to your advantage, [13:19] and I think this is why it's really, really important [13:22] to buy software, like the people who buy the software, [13:24] other people should use the software, [13:26] and then because they have a better view [13:27] of what needs to be had. [13:31] So, in our case in HR, [13:33] we had some things that were not doable [13:37] with Workday and all these systems, [13:39] and so we did it ourselves. [13:40] I mean, this led to a journey [13:44] that might have gone a little bit too far on, [13:47] how should this be done? [13:49] This is one of our areas where, like, [13:50] probably Shopify's the weirdest bit of software ever made, [13:53] but, like— [13:53] But is it something like the internal systems you use [13:58] will subtly affect the decisions that you make? [14:00] Everyone's probably been in meetings where it's like, [14:03] "Oh, well, we can't do that [14:04] because the system doesn't support it." [14:06] And so again, do you want to have [14:08] a Workday-designed compensation system? [14:10] Probably not. [14:11] Like, you wouldn't frame it that way, [14:13] which is why, again, it's important that people think about, [14:16] yeah, what is the vision they want [14:17] and does the software support— [14:19] To make a... true true. [14:20] There were the kits in the LAN parties of the '90s [14:23] that set up internet sharing for everyone, right? [14:26] Like, when that was still difficult. [14:27] I'm just like, I'm a tool-maker, [14:30] infrastructure-thinker in my entire life, [14:32] and I deeply believe in environments that cause people [14:38] to accomplish bigger and better things [14:40] than what they even imagined they could've done. [14:42] Marshall McLuhan says, like, [14:44] first, we make the tools, and then they they shape us, [14:47] or something along those lines. [14:48] I have a terrible memory for exact quotes, [14:51] but the sentiment is right, [14:53] and so I think this is powerful [14:56] and it's something I want to channel at every level. [14:58] Again, I want the people who have a utilitarian problem of, [15:03] "Hey, I need an online store," [15:05] to accidentally catch Shopify [15:07] because this will go to software, [15:09] but then Shopify to inspire them [15:11] to build much bigger and better companies [15:13] than they thought they did, [15:14] and, like, elevate their own ambition [15:16] in what they are building, just like my ambition increased [15:19] from building snowboard stores [15:21] to helping millions of businesses be built, [15:24] and so it's the environment [15:27] that I think we have a lot more power over, [15:31] and it's the sort of missing ingredient people talk about. [15:35] You know, people talk about incentives shaping companies [15:38] and then policies shaping companies. [15:40] So both tend to be the two tools, [15:43] and I think then sort of generally, [15:46] like, if people are really insightful, [15:47] they talk about culture as well, [15:50] but it's the environment [15:51] that's even more powerful than all those things, [15:54] and it tends to be, as a software company, [15:56] especially if you feel like you have agency over the tools, [15:58] but you actually have more power over the tools [16:00] and therefore the environment than you have over policies. [16:04] Like, because those can be moving forward [16:07] by the speed of a deploy. [16:09] Like, you know, a new change to the GSD software [16:12] is immediately part of the environment, [16:14] where a rollout of a new policy [16:15] is going to be a consultative town hall, [16:17] and, you know, a lot of convincing. [16:20] So, in funny ways, it actually acts as [16:23] a fast and the most effective way [16:26] of evolving a company forward. [16:28] Having opinionated software [16:30] that you use to tweak the environment. [16:32] That's right. [16:33] So, the obvious question is, if software shapes us [16:38] and kind of shapes the actions of organizations, [16:41] and you should think about [16:42] what the vision embedded in software is, [16:45] what is the vision of Shopify for your merchants? [16:48] I mean, you said one part, [16:49] which is that they should be more ambitious [16:51] for their business's success, [16:52] and they can probably be more successful [16:54] than they might realize. [16:55] What else? Yeah. [16:56] What I tend to visualize is, like, [16:58] what are the things, you know, [17:01] if you're sitting with a friend [17:04] over drinks in an Irish pub, and— [17:07] I'm with you so far. [17:10] And the question is like, hey, [17:12] you've done an online store. [17:13] You've done an e-commerce business. [17:15] You've done a retail business for Shopify. [17:17] Would you do this again? [17:19] That the answer to the question is, like, [17:21] hopefully "hell yes, because," [17:23] and then there comes a list of things [17:24] that we would like to do. [17:27] Like, amongst those is, like, it just makes it easy. [17:29] Like, it allows me to build the business, [17:32] not feeling like I have, like, you know, [17:35] when you try to paint a painting and you have mittens on. [17:38] Like, it's just like this is sort of a, you know, [17:40] where there's perfect vision in your mind [17:42] but you can't get it out because the tools or the— [17:44] We've all used software like that. [17:45] Yes, exactly. [17:46] So, like, another thing is that [17:50] it keeps me cutting-edge, right? [17:51] Like, we are going through an enormous platform shift. [17:54] Like, Shopify predates smartphones, right? [17:56] Like, as software. [17:57] So we have seen multiple platform shifts, [17:59] and usually what happens when you study [18:00] any kind of industry upheaval [18:04] or, like, a recession or anything, [18:07] it's hardest on, like, the small businesses, [18:11] but, really, big businesses get bailed out anyway, [18:13] and everyone in between has probably [18:15] usually the line of credits or whatever capital [18:17] to make it through. [18:19] You're talking about an economic recession. [18:20] Economic recession. [18:20] So, like, something happens in the world. [18:23] Small businesses are the ones who are wiped out. [18:25] The failure rates during recessions is enormous, [18:29] and that's a huge drag on economies, [18:31] because like 60% to 80% of all people [18:34] work for small businesses. [18:36] I think people felt this [18:37] very viscerally in COVID with restaurants, [18:41] which were obviously particularly hard hit by COVID, [18:43] but I think people could see how little, [18:45] how kind of shallow the balance sheets were, [18:47] because entire, you know, streets [18:50] turned over with new storefronts. [18:51] Like, it was a very visual representation for people [18:54] of how fragile a lot of the small businesses are. [18:56] Yeah, and, like, I mean, [18:57] many, many places never recovered. [18:59] It's like, it's a very organic process by which [19:02] an area becomes, like, [19:04] sort of have the critical mass of interesting stores [19:06] that then have has people come there, [19:07] and that just, like, this builds up [19:09] over very, very long periods of time. [19:10] Entire city centers can shift in these times, [19:14] and so it's a very precarious balance [19:18] and they have very tight margins. [19:21] They often, like, [19:22] there's so much that makes them fragile, [19:25] like systematically, [19:27] that leads to all their sales disappearing, [19:32] because suddenly most people [19:34] purchase from their mobile phones. [19:36] that their website doesn't work on... [19:40] It's not their job [19:41] to stay current with these kind of things. [19:42] So, like, just like inoculating [19:44] from all these kind of shifts [19:45] is a really important part of a business. [19:47] So you're talking about [19:48] some of the disadvantages that small businesses have. [19:51] I was realizing recently, it actually really hit me [19:53] when I went to buy something from [19:57] a very large, famous brand, [19:59] name redacted to protect the guilty, [20:01] but very big, very storied, [20:04] and I was going through their e-commerce experience [20:06] and it was horrendous. [20:07] It was so bad, [20:09] and I realized that, thanks to Shopify, we are now, [20:12] I don't think it's hyperbole to say [20:12] we're living in an inverted world [20:15] where the very large retailers [20:16] are worse off in their e-commerce experience. [20:19] They have, you know, worse systems [20:20] and lower-converting websites than small businesses, [20:24] which have these, like, amazing, super snappy, [20:27] and they're just like more technically performant. [20:28] If you, you measure the latency, if you measure the, [20:29] you know, full funnel conversion and everything like that, [20:32] on any metric by which, you know, if you were, [20:34] with your stopwatch, and you went from Bethlehem Steel [20:36] to, you know, the internet, [20:37] and we're kind of measuring things with your stopwatch, [20:39] the small businesses are better off. [20:40] So, one, I thought that was just interesting [20:42] that, in this case, we have this inversion, [20:45] where, on a very tangible technical level, [20:47] the rebel alliance is doing better [20:49] than the large, established companies, [20:51] but then it made me wonder. [20:53] This is kind of a question for Shopify, [20:54] where you have succeeded on the small business side, [20:57] and I know that's probably anathema to you. [20:58] It's like, "Oh, no, we're just getting started." [20:59] Blah, blah, blah, blah, blah. Right. [21:00] But you have a very significant fraction [21:02] of the small business e-commerce market [21:06] with a very successful product there, [21:07] and so how do you think about Shopify expansion from here, [21:10] where you can go international, [21:11] you can go up market to large companies, [21:13] you can go into new modalities like agentic commerce, [21:15] and, like, there's many vectors of expansion, [21:17] but it does feel like there's a take-stock moment, [21:20] because step one is [21:22] deliver a great experience for small businesses. [21:24] Yeah, thank you, by the way. Me too. [21:26] I also have purchased from Nike, and so... [21:30] It wasn't Nike, but yeah. Those kinds of people. [21:34] Yes. [21:35] You know, that, I love this inversion, [21:38] because this inversion is super, [21:40] this is one of the very few spaces where this is true. [21:43] Yes, exactly. [21:44] And that is exactly what we set out to do. [21:47] Exactly, and access to the capital market [21:49] and all these other areas. Yes, yes. [21:50] Our fundamental belief is, like, [21:51] people should use amazing software [21:53] that really, really, like, [21:55] is like perfectly fit to the problem. [21:59] You don't bloat. [22:00] Exactly, but we can go deeper and deeper [22:02] and deeper and deeper into the space, right? [22:04] So, it's just really, really quick. [22:07] It's really easy to underestimate the size of, anyway. [22:09] Everyone underestimates the size of the internet [22:11] and everyone underestimates the size of retail. [22:13] We sit in an intersection of both those things, right? [22:15] Like, it's my favorite status that the companies, [22:20] the industry supplying wooden pallets to warehouses [22:24] is a $60 billion a year industry, right? [22:27] Like, that gives you sort of an idea [22:30] for how large retail actually is, at least directionally, [22:35] and so what we have done from a company perspective is [22:40] we were very, very clear that our mission is [22:42] to make entrepreneurship simpler, [22:44] and that's where, you know, our heart is, [22:48] and many of our customers have gotten quite big. [22:51] Like, we ended up raising one of our customers [22:53] to a billion dollars of revenue, [22:55] and so some of our customers have, like, [22:58] in the top 10 of largest customers, [23:00] many of them have started on Shopify. [23:02] The reason why they're still around is because [23:03] we actually, again, we take this so seriously [23:05] that we're actually building with them while they grow. [23:08] Yeah. [23:09] And, you know, it turns out, again, like, [23:12] all these stories of softwares [23:15] for a certain segment of a market are skills issues, [23:17] and that's exactly what they sounded like [23:19] when I first heard about it. [23:20] It's convenience to the company making the software [23:24] to say we are focused on one segment, and by the way, [23:26] they're all focused on the richest segment, funnily enough, [23:30] and we just said no. [23:31] We are starting with SMB, and then some of them grow, [23:34] and hopefully we build good software [23:36] and then we invite others over. [23:38] The most common thing that people said [23:39] with Shopify in the early days was always that [23:41] this will never work in real life, [23:43] because, like, in the real world, [23:45] retailers are messy, and you know, [23:48] just like have all these business rules [23:49] that we didn't implement, [23:52] and our opinion there was always like, [23:54] "Well, the real world sounds like a terrible place. [23:56] We like ours better." [23:57] And we just invited everyone over and— [23:59] Well, isn't there also something where, like, [24:01] the real world is messy, [24:02] but we find this with Stripe Billing, where people say, [24:05] "Oh, everyone's billing system is different," [24:06] and like the dimensions of their billing [24:10] that they want to encode and things like that. [24:12] Therefore, you know, there's a certain size [24:15] beyond which, you know, Stripe Billing will cap out [24:17] and people need a custom system, [24:18] and our view is, absolutely, billing is really messy [24:22] and it's also hard to test [24:23] because it involves time travel, right? [24:24] It's like, you know, what happens if people do [24:26] a mid-period adjustment [24:27] that needs to be prorated or whatever? [24:29] Therefore, do you want to be [24:31] using your homegrown version of this that was developed by [24:34] some guy who had never built [24:35] an e-commerce or billing system before [24:37] and has since left the company, and you know, [24:39] now it's kind of poorly maintained, [24:41] or do you want to buy it from people [24:42] who have been thinking about this problem space [24:44] and building a generally abstracted frame? [24:47] And, you know, our Stripe Billing now [24:48] has a time travel engine [24:50] where you can just move time around [24:51] to test all these kind of different cases, [24:53] because, obviously, you need that, and in Shopify's case, [24:56] similarly, yeah, we've been thinking about [24:57] inventory management for a very long time, [24:59] and, you know, how you model inventory [25:01] and all these kind of things, [25:02] and so I feel like it's funny that people sometimes say, [25:04] "Oh, this will never work at the large end [25:06] because our needs are super complex. [25:08] Therefore, we should be using a piece of software [25:11] that only has one user [25:12] and is poorly maintained by, you know, [25:15] one tiny corner of the company." [25:16] Yeah. That's right. [25:17] I think, look, [25:19] when I first started programming in my apprenticeship, [25:24] my meister taught me that, in the software world, [25:28] you have about two years' time, [25:33] any time you start a important piece of software, [25:35] to try to nail it, to try to nail the problem domain, [25:37] because, afterwards, it's just magically like [25:39] someone put cement into the code base [25:40] and you're never going to change a thing. [25:41] Yes. [25:42] And that's just like, that's the state of it, [25:43] like, software in the '90s, [25:45] and frankly just the way that software works. [25:48] Like, everything that got to scale [25:50] was finished in the '90s with cement in it. [25:52] Yes. [25:53] And then saw it in the 2000s, [25:54] and in a lot of decision-makers' life, [25:57] that was the time when software adoption really started, [26:00] and their priors are loaded up with software [26:02] that corresponds to roughly [26:03] what Jurgen back then was telling me. [26:06] Now, we've gotten a lot smarter about building software. [26:08] We have, you know, CI and automated testing in general, [26:11] and, like, we just like, [26:12] we know how to build software in a way. [26:14] Yes. [26:14] We are by no means perfect at it. [26:16] We don't even know how to measure it, [26:17] as we talked about earlier, [26:18] but the state of art has increased significantly [26:19] and we make better software, [26:20] and that's a completely different thing [26:22] than what people are comparing again, [26:25] but it's hard to tell them, right? [26:26] Like, you know, the first 80% of every e-commerce businesses [26:30] are the same. [26:31] It's like you need a check-out. [26:32] You need to take credit cards, of course, [26:33] and you're taking credit cards for us, [26:35] and, you know, like, you need the hosting [26:38] and a check-out and all those graphics, [26:39] and an admin interface. [26:40] So it's the last 20% that are actually quite unique, [26:43] and they exist in a particular, [26:45] very predictable problem domain. [26:46] Yes. [26:47] Like, what does the website look like? [26:49] It's like what my brand looks like and also business rules. [26:52] Like we have, over time, [26:54] just like your time traveling back test system, [26:58] figured out where do we allow programming [27:02] or programming-enabled extensibility in the system [27:04] so that these last 20% [27:08] can uniquely be created for everyone who wants them, [27:11] and I think that's what it means, [27:14] and I think that's a place you can only get to [27:17] if people who work on the problem really care, [27:20] because you kind of have to fall in love with a problem [27:23] to want to look at it from so many sides, [27:25] but at some point, you understand it well enough [27:27] to not solve the problems that people state, [27:31] but build infrastructure in the form of primitives [27:34] that can be composed to solve all the problems, [27:37] and that's a completely different level, [27:38] but it's what I deeply appreciate. [27:40] I massively admire it every single time any category [27:45] gets to this quality of decision making and infrastructure. [27:51] Another thing I enjoy on the software quality side [27:55] that both Shopify and Stripe sell on [27:59] that's just a funny domain to pursue is peak load, [28:05] where Stripe sells on just [28:08] you can put a lot of transactions [28:10] through the system at once, and obviously Shopify similarly, [28:12] it feels to me that [28:12] how you displace other e-commerce systems is, [28:15] if you are getting very spiky sales, [28:17] you can get, you know, put a thousand requests per second [28:20] through the system and that's totally fine, [28:21] and most systems just, at a technical level, can't do that. [28:25] I guess I find it funny [28:26] because I think we kind of grew up together, [28:28] kind of expanding the, you know, [28:31] requests per second thresholds, [28:33] and you guys pushed us pretty well there, but I'm curious. [28:36] Like, where did this come from in e-commerce culture? [28:39] Because, like, the marketers are saying [28:40] everyone will come to the website [28:42] and click buy at the same time, [28:43] and the engineers are like, "No, please do not." [28:44] Stop encouraging that, [28:46] but it's become like a real part of e-commerce culture, [28:48] like the drops and things like that. [28:49] Maybe you can fill me in on the drop history. [28:51] So it's funny you— [28:52] You seem to have started it. [28:53] Yeah. Like, it's really funny. [28:56] I'm just as amused as you are because, like— [28:59] No engineer would ever suggest this. [29:00] Because it's a very, very strange thing because, [29:04] but it was this fantastic business built around it. [29:09] The one that did it for us first, [29:11] it was somewhere in 2010, [29:14] a business called theCHIVE, [29:17] which was like some, I mean, it's a website. [29:20] It was like some community. [29:21] Yeah. [29:22] And it had like some T-shirt with Bill Murray on it, [29:25] and that thing, every single time it went on sale, [29:28] it took Shopify down. [29:28] It was just, like, unbelievable. [29:30] We actually couldn't, we didn't know the demand level [29:33] because we just couldn't observe it. [29:35] It was like, upstream, [29:38] the routers could not send enough traffic [29:40] and it's like, first of all, who was theCHIVE? [29:42] We had to, like, catch up very quickly, [29:46] and so it was one of those things where [29:48] there was a very specific decision moment where we realized, [29:52] I mean, this is obviously [29:54] Shopify down is the worst possible thing. [29:56] This is like, the entire company exists [29:58] to never make it so that anyone's down, [30:00] no less the entire thing. [30:02] So, we were like, well, [30:06] the traditional business thing is, like, [30:08] this is super expensive from our reputation, from money, [30:11] and we can't even, in our business model, [30:12] ever make money back from theCHIVE, [30:15] given the amount of resource they consume, [30:17] so we clearly needed to fire the customer, [30:19] and we decided, no, absolutely not. [30:20] We are using this as a gym, [30:22] and we are going to use them like, [30:24] we are going to build the system in such a way [30:25] that we can deal with this [30:26] because this is exactly what good engineering looks like, [30:30] and that was fortuitous because [30:35] I think this sort of pattern ended up really sticking. [30:38] It was like Supreme, which really drove us further. [30:41] It was the streetwear world. [30:42] I remember Supreme and I remember Kylie Jenner. [30:45] Yeah, lipsticks. Those lip kits. [30:47] That's embedded in my memory, [30:49] from kind of 2013, 2014 Stripe-ish, [30:52] that that was definitely, like, [30:54] that was a board-level topic. [30:55] Yes, and we had board-level discussions [30:57] on our side too, right? [30:59] It's actually really funny to know that [31:01] we probably sweated exactly the same amounts on both sides [31:05] because, obviously, we have an incredibly long-standing, [31:08] and I think, at this point, storied partnership. [31:11] It's tremendously difficult to synthetically test [31:18] all the pieces that go into commerce, [31:20] because it's just, like, so complex with so many things. [31:23] There's so many APIs. [31:24] There's so much we have to, like, [31:26] the shipping to, of course, payments, [31:28] and payments is like, depending on the mix of bins, [31:31] it's like a different profile on-lot, right? [31:35] Kylie Jenner, like the store, [31:36] had always a lot more debit cards, [31:38] and that just kind of exercised [31:39] different parts of infrastructure, [31:41] and so, anyway, these kind of things you find out, [31:45] and on our side, we just like, yeah, like, [31:49] no sin was forgiven. [31:51] Like, every single sin was revealed [31:55] and found us to be wanting very publicly, so. [31:58] Is there anything particularly technically hard to scale? [32:01] Well, I mean, so if you open a book on databases, [32:05] like sort of page, probably chapter one, [32:07] is like an example of how a transaction works is like, [32:12] I mean, either moving money on a ledger [32:15] or moving an item out of inventory and into an order, right? [32:19] That's the textbook definition of transactionality. [32:22] Like two textbook definitions in a row, [32:25] so you fundamentally have to serialize at those points. [32:30] This is like the lock contention moment. [32:32] There's architectures which we employ that, like, [32:35] you can get around this. [32:36] There's like pools of products. [32:38] Like, we take our products out of an integer [32:40] into like Stripe's thing, [32:44] and, like, there's all sorts of things that we do, [32:46] and then try to refill this as fast as possible, [32:49] but the point is that [32:50] you really deal with lock contention, [32:51] and lock contention is difficult. [32:54] It's very, very hard to scale [32:57] if everyone at the same time [32:59] wants to decrement an integer in the database actually, [33:02] then also the transaction involves [33:04] having to go to interchange and move money. [33:06] I was reading some histories of Oracle recently, [33:08] but I think it's kind of interesting because, you know, [33:12] proprietary databases were the leading databases. [33:15] Like, the LAMP stack only really took over the 2000s, [33:17] but in the 1990s, if you were building a thing, [33:19] you actually needed to go buy your database, [33:21] and there was these proprietary database wars in the '70s, [33:24] and I was wondering, are there lessons for us [33:25] for the kind of current open-source, [33:27] or, sorry, current AI wars and model wars [33:30] from the database history? [33:31] But it was striking, [33:32] talking about kind of the early feature wars of Oracle [33:36] versus its competitors, [33:37] and locking was such a big differentiator [33:40] in those early versions, [33:42] and, like, row-level locking. [33:43] Row-level locking was a big thing. [33:44] But it came in like Oracle 6 or something. [33:46] You know what I mean? [33:47] You would think it's like an Oracle 2 feature, [33:50] but, no, it took them a few revs to get there, [33:53] and row-level locking was like the bee's knees [33:55] when they released it. Yeah, yeah, [33:58] and I'm very glad they did. [34:01] So, I mean, this is dot-com predates Shopify a little bit, [34:04] but, like, yeah, dot-com tech company creation was like, [34:10] day one, you send a million dollars to Larry. [34:13] Yes. [34:13] To get your Oracle database, [34:14] and I think, you know, day, like 180, you went public [34:18] or something, like a very different time from today. [34:21] Yes. [34:22] I mean, I guess we're kind of getting back to that [34:23] a little bit in AI where, you know, [34:26] step one, send $300 billion to Oracle for a data center. [34:30] You're a couple days past 180 [34:32] on your journey to becoming a public company, [34:34] I have to say, but like— [34:36] Yes, yes. [34:38] Anyway, moving on, I want to talk about agentic commerce. [34:43] I had a really nice shopping experience recently [34:45] where I wanted to buy [34:47] an integrated travel adapter and power brick, [34:51] and so I did some product research in ChatGPT [34:54] and it was really nice. [34:56] There was no ads getting in the way or anything like that. [35:00] I actually found a product [35:01] that I hadn't come across from Momax. [35:02] It was really good. I think I was telling you. [35:03] Yeah, yeah. I have two. [35:04] Okay, yeah, and what's the verdict? [35:06] It's very good. [35:07] It's very good. [35:08] Okay, yeah, we'll put a link. Perfect. [35:08] And it kind of linked me out and I purchased. [35:11] Obviously, it was super easy, [35:11] one click with Shop Pay and everything like that, [35:14] and it felt very futuristic to me [35:16] both on the product discovery [35:17] and the completion side of things, but it feels like [35:23] filling out web forms [35:25] is not a value-add activity for people. [35:29] Like, when you travel, you probably don't [35:30] directly book the hotel yourself on the website. [35:32] You know, you have an assistant who does it [35:33] or something like that, [35:34] because choosing the hotel might be desirable for you, [35:37] but actually filling out the web form [35:38] is not a value-add activity for you, [35:40] and so what is your vision for [35:42] how agentic commerce happens specifically? [35:46] Because it seems relevant to your interests. [35:47] It's relevant to my interests. [35:49] I mean, it will be extremely valuable, [35:51] and it will be something that will be done a lot [35:55] and might even be, like, a majority of commerce. [35:58] Like, on the internet, I don't know, [36:01] but, like, I think my role in this is going to be, [36:04] like, largely infrastructure providing, right? [36:07] Like, this is the, our job is to keep our, [36:11] like, all the millions of business on Shopify [36:13] extremely current, and so what we want to do is [36:18] make sure, like, everyone's plugged into [36:20] the various chat systems [36:22] and there's an MCP connected to every Shopify store, [36:26] and, you know, a global catalog [36:28] that's really nicely presented. [36:29] So, we are building software specifically for [36:34] OpenAI and Claude [36:36] and the Perplexity to make it really, really easy [36:40] for these products to show up [36:41] and reason about them and so on, [36:43] because we think it's really important [36:45] that they can show up. [36:46] So, like, there's an infrastructure angle in it, [36:48] but, like, how is being used? [36:49] So I think the best way to predict the future, [36:52] in most cases, is just like, frankly, [36:54] look at what rich people buy for themselves, right? [36:57] Like, you know, Uber is a wonderful company. [37:00] It's also extremely predictable [37:01] because rich people had car service, [37:03] and, you know, so it's figuring out a way to scale that [37:06] in such a way from cost basis so that everyone can use it. [37:09] It's just like, yes, that works. [37:12] So, personal shoppers, right? [37:13] These are places where, you know, [37:16] memory and understanding people and personalization [37:20] are just, like, pay so many dividends [37:23] and are so clearly also good. [37:25] It's like people make so much out of data and advertising. [37:29] You know, a personalized ad is a wonderful thing, right? [37:33] Like, if a personal ad that tells, like, you know, [37:35] I am scrolling past something [37:36] that's monetizing the free application [37:38] that I appreciate using, and the fact that— [37:41] You should tell the EU this. [37:42] Yes. Well, actually, I do, actually. [37:45] And the fact that it's highly personalized to me [37:48] means, like, at this moment where this platform monetizes [37:51] because they're giving me a gift here that I get to use. [37:55] It's win-win. [37:56] Yeah. It's win-win. [37:57] Using this moment also that it's now not just like [38:00] a fucking, like, Ford 150 truck, [38:03] or I don't even know if that's how you say that, [38:04] but, like, I'm not in the market for trucks, right? [38:07] So, instead, I see a travel adapter like what you send me, [38:11] you know, just like a— [38:13] I am your personal shopper in this case. [38:14] But the funny thing is you send me the link, [38:17] and if I would put that link in ChatGPT, [38:19] which has my sort of, you know, lots of my history, and ask, [38:24] is that the kind of product that I would appreciate? [38:26] Like, it would just return, "Yes," right? [38:28] Because it solves the problem in its entirety. [38:30] It's like a travel adapter, which is like for any country [38:32] to any plug to any other, all sorts of stuff. [38:37] Good product, right? I want it. [38:38] Okay, so we're getting down the road here. [38:40] I mean, it feels like step one [38:43] that's going to happen with agentic commerce is, [38:45] I feel like existing aggregators, [38:47] if just aggregation is their raison d'etre, [38:52] then they're going to have a lot of questions [38:55] that they need to answer, because ChatGPT and the AI apps [38:59] end up as a new aggregation point, [39:02] and so it seems to very much benefit the tail. [39:05] It benefits all the Shopify merchants, [39:06] just like, you know, we had a new set of content creators [39:09] emerge on YouTube [39:10] that effectively compete with kind of TV. [39:13] They're both just entertainment, [39:15] and the tail becomes much more powerful [39:17] when you have over-the-top distribution through YouTube. [39:19] Similarly, it feels like, [39:20] once you have over-the-top distribution through AI apps, [39:25] you have a much more powerful position as a tail merchant [39:28] because, you know, Momax is a tiny brand, [39:30] but they can just get recommended as the best product. [39:33] So that's step one that I think happens. [39:36] I think a slight modification I would just make to this, [39:39] I don't think the tail part matters. [39:41] I think merit matters. [39:42] Yeah. It's just like, [39:44] it's kind of an important point, [39:45] because, like, you know, often, [39:47] when you talk a lot about commerce and products, [39:50] at some point, just because of prior discussions, [39:53] people are like, [39:54] "Yeah, but aren't you just feeding consumerism?" [39:56] Right? And so this is sort of like the negative take, right? [39:59] And, again, consumerism is a thing that, like, [40:02] is being thrown around, but where does it come from? [40:04] It comes from people throwing away a lot of things [40:06] that they probably shouldn't, but they throw away things [40:09] because they hate the things they have, right? [40:13] These things don't work. [40:15] They fall apart. They don't do the thing. [40:16] They're, like, cheap, plastic-y, this kind of thing. [40:20] The thing that solves consumerism is quality of products [40:25] which you want to keep using, right? [40:26] Like, you want to be excited about this, [40:28] and, like, you want this travel adaptor to last forever, [40:30] and so you never have to think of, [40:33] you literally never have to Google [40:35] what plug is being used in Singapore. [40:37] Right. [40:38] Like, you just know that the thing that you have, [40:39] you can deal with everything to everything, [40:42] so problem solved, [40:43] and I think that's an important modification on the point. [40:50] Yeah, I guess implicit in my view is that [40:53] there's good stuff in the tail, [40:55] whereas the head has distributional advantages [40:57] and so it is more meritocratic [40:59] to bubble more stuff up from the tail. [41:00] That is, I think, [41:01] the reconciliation of those two world views. [41:03] Okay, so that's kind of step one that happens. [41:05] I think you're getting to step two, which is [41:08] it feels like there's a lot of changes potentially [41:10] that can happen around personalization, [41:13] about agents actually acting for you, [41:16] and maybe one thing I'm curious if you have thoughts on it, [41:17] it just seems like a lot of search in e-commerce [41:19] is really bad, where it's keyword-based search still [41:23] and you can't do things like, [41:25] "Let me know if this becomes available in gray," [41:29] or, "What other products do you have like this?" [41:32] And you want some kind of embedding space [41:34] that thinks about similarity between the products, [41:36] and no one can do this. [41:37] So when are we getting better personalizing and search? [41:39] I just want really good search of the products [41:42] that are out there. Totally. [41:42] I am obsessing about this now, and like, [41:44] we have, like, we should have been— [41:46] Tell me the secret Shopify roadmap. [41:49] Shopify should have solved this earlier and hasn't, [41:51] and I'm pretty mopey about this. [41:55] I, like, we— [41:56] You guys are okay. [41:57] We will, yeah. [41:58] We will, I think we will solve this now. [42:01] So, like, the thing is that [42:04] it's really, really fascinating. [42:06] I really deeply appreciate it as a field. [42:08] It's so different. It has its own lore. [42:12] Like, everything's interesting is my fundamental opinion, [42:14] and search has been a wonderful area [42:17] to, like, actually learn a lot more about the history now, [42:20] and, like, the fundamental thing is, like, [42:22] there is a generic bias in search [42:28] and text is king, [42:29] and, like, frankly, the people who work in search [42:32] are, like, much more interested, [42:34] we're fundamentally more interested in searching [42:36] through papers than through products. [42:39] Yeah. And so [42:40] there's very few of the best people in search in the world. [42:43] The people who know most have actually ended up [42:45] working specifically on product search, [42:47] which really should be seen [42:50] as its own completely different domain [42:52] that just sort of looks like it, [42:53] but it's really clearly deeper [42:55] with lots and lots of complex features and so on. [42:57] So, like, I got to, like, build up over the last year [43:01] like, a, you know, sort of really, really, [43:04] really, really great search team, [43:05] and exactly look at the, you know, [43:07] embeddings and these kind of things, [43:09] and it's just, the amount of improvements [43:12] that's just, like, unexplored is just staggering there. [43:14] Shouldn't you do search across, I mean, [43:17] I find it interesting when Shopify does things [43:20] broader than the Shopify universe, [43:22] because, like, the Shopify app can track packages [43:26] not bought from Shopify merchants, right? [43:28] Mhm. [43:29] Yeah, and so shouldn't you do the search thing [43:31] across the whole web and not just Shopify? [43:33] Oh, should we? [43:34] I think so because the thing that feels like [43:37] the big opportunity to me is [43:40] clustering was always so hard to do [43:42] because all the features were developed manually, [43:44] and so you had to, like, go say that the dress is red [43:46] and everything like that, and it feels like, [43:47] have you ever used an LLM [43:48] as just a recommender engine where, like, [43:50] I like feeding in, [43:51] "These are all the books that I've read recently. [43:53] Give me a narrative nonfiction book [43:55] that someone who's read all these books might like." [43:57] And it does it extremely well, [43:58] but you're kind of using the LLM to access its world model, [44:02] and it feels like now you can have [44:05] an interesting kind of product world model and access it, [44:08] but people, like, for the people using it, [44:12] whether a merchant's on Shopify or not [44:13] is probably an incidental detail that they don't care about. [44:15] Yeah, I totally agree, and you're talking about how, [44:17] but, like what is the really important thing? [44:19] I just want the LLM to be proactive and just tell me, [44:22] "Hey, this outfit would look good on you [44:24] and here's a preview." Yeah. [44:25] Right, like, and just like, [44:27] here's the full cost, and do you want it, right? [44:28] Like, it's the personal shopper outreach kind of component, [44:31] which I think is just like really valuable, [44:33] and this is not going to be like [44:34] some kind of, like, you'll get this from everyone [44:37] because it's really easy to put together. [44:40] It's hard to get the data and the embeddings [44:42] and the sort of understanding about, you know, [44:45] like, we have to figure out... [44:48] It's funny. [44:49] The most important thing again in the AI space [44:52] is also about unquantifiables, [44:53] It's the vibes, people call them. [44:56] It seems like Midjourney specifically [44:58] is one company that managed to actually scale taste because— [45:01] They have a very distinct aesthetic [45:03] in the images they produce. [45:04] Yeah, and maybe this is because [45:05] they drove towards a particular aesthetic, [45:07] but they just sort of kind of fed it [45:08] what they had ownership over, [45:10] but it's one that's clearly appreciated. [45:12] But David is also, like, deep founder mode type, [45:15] how he operates. Yeah, yeah, exactly, [45:16] and so like, I mean, when something outperforms, [45:21] it's usually a product of a place [45:24] that doesn't make decisions by consensus, right? [45:26] As you know, so. [45:27] Yeah. [45:29] So, you know, how do we, like, [45:31] we need to figure out taste in this product sense. [45:35] Yes. And we'll figure this out. [45:39] Someone's going to figure it out. [45:40] Maybe everyone's going to figure it out. [45:42] Maybe it's already figured out in the neurons [45:43] and we just don't give enough information [45:45] for it to reason around taste, and so putting, like, [45:50] I think it will be very, very vague with personal shopper. [45:52] The thing that I think the aggregators still, [45:54] to just like get back to your point on the aggregation side, [45:58] there's many reasons why people buy things. [46:01] Just, like, lots of things are impulse purchases. [46:03] Like, it's like the things [46:07] next to the checkout counter at the supermarket. [46:09] I don't think people are going to use ChatGPT [46:10] to buy Mars bars [46:13] or something like a candy or something that's just like— [46:15] Well, they just don't do so online. [46:17] Right. Exactly. [46:18] But, like, is there an online impulse purchase [46:20] that is not suited to happening in an AI app? [46:25] Interesting question. Nothing comes to mind immediately. [46:27] Yeah, I mean, I guess online impulse purchases are like [46:30] things where you're scrolling through Instagram [46:32] and it's like, you know, all the COVID stuff. [46:33] It's the sharpest knife ever developed, you know? [46:36] I wonder if that's a truly— [46:37] An umbrella that changes color in the rain. [46:39] I wonder if that's a truly impulse purchase. [46:41] I mean, they sort of feel like impulse purchases, [46:42] but they tend to be, like, [46:44] very few people part with money unless they are like, [46:46] "Holy shit, this the thing I always wanted." [46:48] Like, this is actually— [46:49] It's just tapping into demand [46:51] that's been there for a long time. [46:51] It's sometimes a little bit, [46:53] the opinion, when you look at log files, [46:55] you're going to get a very incorrect view of the world. [46:58] It's very often the longest consideration phases. [47:02] Like, for years, I wanted a travel adapter [47:04] that can deal with everything through everything, [47:06] and, like, it goes in a chat from you or in an Instagram ad. [47:12] Doesn't matter. I'll insta-purchase at this point. [47:14] Mhm, okay. You were just ready to do that. [47:16] My other purchase I made recently that I really liked is [47:19] I was going kind of on a long day hike, [47:22] and if you read the accident reports, [47:23] it's often people day-hiking who get into trouble [47:25] because people who are going out on a multi-day hike, [47:27] they have their tent with them and they're prepared. [47:29] You know what I mean? Oh, yeah. [47:30] And then people go— [47:31] That's a problem with bicycle helmets too, right? [47:32] Yeah, yeah. It's like the same issue. [47:33] It's the short journey, exactly, [47:35] just to the store that gets people in trouble, yeah. [47:36] So, day-hikers, they get in trouble, [47:38] and so because they think they don't need any preparedness [47:40] and then they end up out after dark, [47:41] and, you know, a rescue party and everything like that, [47:43] and so it was like a, it's called a Bivy. [47:45] You know, like a little sleeping bag for emergency purposes, [47:49] and it's made out of the reflective foil, [47:52] and so, again, if you end up stuck, [47:55] an emergency foil sleeping bag [47:58] just so you don't die of exposure, [47:59] but it folds up to, it's like that size, [48:01] and so you can just stick it in a day pack, [48:03] and now you are somewhat more prepared for the elements [48:05] than you were before. [48:07] I don't know. Do you have a favorite recent purchase? [48:10] I just really like these... [48:14] I mean, someone going just like to ridiculously further [48:18] and to just like almost, [48:21] like products that just celebrate craftsmanship in a way, [48:23] even in places where it's just like no one else does. [48:26] Like it's like, a real example, [48:29] but, like, I, you know, [48:31] take all these vitamins in the morning, [48:32] and, like, I've bought these, like, [48:34] CVS vitamin pill cases at some point just for travel, [48:38] and then I found on Shopify, like, [48:41] this Japanese, perfectly CNC-built, [48:44] like unbelievably beautiful pill cases. [48:47] Like, the entire website is just like a love letter to— [48:49] The most beautiful pill cases, yeah. [48:51] To, you know, precision-engineered, [48:53] like tolerances of sub-million— [48:55] What's it made of? [48:57] Like, metal, which is like absolute, [48:58] like it feels like, it feels so... [49:02] Anyway, whatever. I'm really excited about it. [49:07] Let's put something. We can put a link in. [49:08] Yeah. Put a link in. [49:09] I forgot the name sadly, [49:11] and now I gift those things because they're just beautiful. [49:14] Yeah, yeah, yeah. No, that's cool. [49:18] How is Shop Pay doing? We referenced it there. [49:21] Shop Pay? Shop Pay is awesome. [49:25] You know, that started out its life as [49:27] a feature called Remember Me, [49:28] which, like, is literally the name, [49:30] what was on the chat box. [49:31] We also had a thing that was called Remember Me. [49:32] Actually? That was like [49:33] the original name for Link. [49:35] No way. [49:36] Yeah. [49:37] That's trippy. [49:38] That's very funny. [49:39] So, which, you know, is like, I mean— [49:42] It makes sense. They're kind of the same thing. [49:43] I mean, Shop Pay is much broader but they're similar things. [49:46] Yeah. Actually, Shop Pay feels more narrow. [49:48] It's in commerce specifically. [49:50] But you guys do all the shipping and tracking. [49:52] Yeah, that's right. Yeah. [49:53] And so, yeah. [49:54] So, yeah, Shop started, [49:56] like, I mean, Shop is Shopify's brand for buyers, right? [50:02] Like, so it's, you know, Shopify is what you use [50:06] to be on Shop. [50:08] Yeah. [50:09] And, at least, that's how we think about it. [50:11] So, you know, there's a Shop app and, you know, Shop Pay. [50:15] So, I mean, again, it started, [50:18] because it's a ridiculous thing that we asked, [50:21] anyone asked anyone to fill out their address [50:25] with their thumbs when you purchase something. [50:26] That is insane. [50:28] This was meant to never be a thing, because, you know, like, [50:32] I don't know when they came up with HTTP, [50:34] but there's the 402 error code payments. [50:37] Payment required. [50:38] Payment required, right? [50:40] It's like built into the HTTP standard [50:42] and no one ever implemented it, [50:43] and so moving money around was supposed to be [50:45] built into the platform and it wasn't, [50:48] and then, you know, [50:50] credit card forms were hilariously insecure over normal web, [50:54] and then we came up with HTTPS [50:55] and made it reasonable, I suppose, [50:57] and then it became good enough. [51:00] There was always efforts. Apple Pay. [51:02] Like, I think actually PayPal was around, [51:04] but then sort of didn't do anything. [51:06] You know, so there's so many people doing this, [51:08] and it was just like, it was just like, [51:10] okay, well, clearly this is going to be solved soon, [51:12] but, like, might as well [51:14] just put a Remember Me thing in there [51:16] so we can solve this now [51:18] as almost a poorly fill for the future of the internet, [51:20] and then no one did. [51:21] I mean, very few people did, [51:23] and it just sort of didn't happen, [51:24] and then now Shop Pay is, like, incredible. [51:27] Super glad to have it. [51:28] It's like— [51:28] But what's interesting to me— [51:29] The most converting thing we've ever done. [51:30] Yeah, what seems interesting to me about it is [51:32] I think it's a significant part. [51:34] It's not the only part, [51:35] but it's a significant part of the inversion where, again, [51:37] now the small stores are significantly better off [51:41] than the large stores online, [51:43] and that's just a striking fact. [51:45] It's striking and it's a 16% conversion increase. [51:48] Yeah, yeah. By it being around, right? [51:52] So it's a remarkable product now, [51:56] partly because, again, it's part of branding. [51:58] It's purple always and people learn to trust it [52:01] and associate it with a store [52:02] that's, like, actually recent and cared for, [52:04] and it doesn't, you know, even know they don't have to, [52:08] before they get to checkout, they know they will not have to [52:10] type in their address with their thumbs [52:11] like some kind of heathen from the '90s. [52:14] Yes. [52:15] Which just, like— [52:16] You mentioned personalized advertising [52:18] and being a big fan of it. [52:19] It seems to me that [52:20] there's a real symbiotic relationship between Instagram, [52:24] where people are kind of scrolling in an aesthetic mood, [52:27] looking at nice things, [52:30] and then Shopify stores, where they need customers, [52:34] and so, especially during the pandemic, [52:36] there was this huge growth of direct-to-consumer brands [52:40] who really cracked the Instagram CAC to LTV math, [52:44] and they were able to grow in a big way there. [52:47] Then, of course, the Apple ATT changes happened [52:49] where, for a brief period, it got harder. [52:51] It seems like Meta has mostly solved that, [52:53] but, anyway, I'm just curious what's the current state of, [52:56] like, is this worldview I have true [52:58] that there's a lot of Shopify merchants [53:00] that have been able to really scale up [53:02] thanks to personalized, you know, direct advertising? [53:06] Yeah, 100%. [53:07] It's a main growth channel, is advertising on a platform. [53:13] The Meta platform specifically [53:14] is just so incredible for this. [53:16] Like, I think the Meta-Shopify alley-oop [53:20] has created more businesses than, like, you know, [53:24] any government policy in history has. [53:27] So it's like, I mean, if you just compare it to television [53:30] and consumer packaged goods before. [53:34] The television channel required broadcast, [53:36] and so the only things you could really advertise [53:39] were, like, mass-market, high-margin products [53:41] that could afford the advertising. [53:43] Washing powder [53:44] and all that stuff. Exactly. [53:44] So it's like, you know, [53:45] laundry detergent and toothpaste. [53:47] Yeah, yeah. [53:48] And so the Instagram-Shopify [53:51] is the precise inversion of this. [53:52] It's like highly merit-ful products. [53:57] It's the first time you can advertise niche products. [53:58] Exactly, and the niche products are actually doing better. [54:00] That's the crazy thing. [54:02] Like, advertising directly snowboards [54:05] is, like, impossible there because— [54:08] You can't take out a billboard for snowboards. [54:10] The brand halo of Burton means you can't compete with that. [54:14] There's intersections. [54:15] Like, snowboards only for powder snow [54:18] on the, you know, like, US West coast [54:21] and, like, in certain environments with, like, just like, [54:24] you know, like, this is not the best example, [54:26] but, usually, there's like [54:27] three different intersections of something, [54:29] and it goes back to, you know, [54:30] Kevin Kelly wrote about this in the early 2000s. [54:33] He's like, it's like, he wrote an essay of 1,000 true fans, [54:36] and the internet is about finding your 1,000 true fans [54:39] and everything else will happen, and restoring that idea [54:46] or bringing this idea into retail [54:48] has done a lot of good things, right? [54:50] Like, it just created better products. [54:52] It created this symbiosis, [54:54] and, like, Meta allowing you to discover one of those people [54:58] who has a potential to be one of those 1,000 true fans [55:01] is an incredibly powerful thing [55:03] that just, like, again, led to enormous good and employment [55:07] and all these kind of things. [55:08] So it's a very, very powerful combination. [55:11] There's all sorts of other ways [55:13] how people, like, grow, [55:15] but this is a very, very large part of everyone's strategy. [55:18] You're saying this enables 4K for commerce. [55:20] There's this, like, higher resolution, more specialization, [55:24] as opposed to the blurry generic products. [55:26] Yeah, exactly. [55:27] Find the people we're really interested in [55:29] who appreciate the products. [55:30] Hmm. I like that. [55:32] How is the stablecoin rollout going on Shopify? [55:35] So, you know, we're obviously doing lots of stablecoin stuff [55:37] with lots of people, including you guys, [55:39] but I feel like you guys are [55:41] one of the biggest bets on stablecoin pay-ins. [55:47] Everyone should just be able to, you know, [55:49] use stablecoins as a payment method that we've seen, [55:51] and so how's that going? [55:52] I'm obviously, just like you, [55:54] a huge fan of the concept of stablecoins. [55:57] I love the work you're doing there, [56:00] and congrats on the launch of Tempo, [56:02] and the internet just needs its own currencies. [56:05] I think it's important. [56:06] At least it needs its own infrastructure [56:08] to move things by internet speeds, [56:10] and I'm excited for it because, I mean, sort of, [56:13] I think a valid criticism of crypto in the past, [56:16] as exciting as it is, has been that [56:19] the utilitarian difference [56:22] between a real US dollar and a US dollar crypto coin, [56:29] it's like that the US dollar crypto coin is mostly like, [56:33] it's sort of like a gift certificate for scams. [56:36] It's like, there's not that many things [56:37] you can actually purchase with it [56:39] that are of high utility value. [56:41] It's like, obviously, this is I think why— [56:43] The economy was too circular. [56:44] Exactly. [56:45] It was like most of it was like you purchase other coins [56:47] that then you might have, I mean, [56:51] hopefully, you diamond-handed into some kind of upswing, [56:54] but it's like that's not usually what happened, [56:57] and I think that's valid criticism, [56:58] but I think the most important thing to happen there is [57:01] we need to expand what can be purchased with stablecoins [57:06] to the things people want to purchase. [57:09] Like, outside of just the sort of [57:11] insular crypto FinTech. [57:14] But, yeah, like, we are working hard on getting [57:17] all Shopify stores to just take stablecoins. [57:19] How has the merchant response been? [57:21] Very positive. [57:22] Again, because through the work that Bridge is doing, [57:25] we can settle into US dollars directly so the merchants can, [57:31] they've just gained a new ability [57:34] without having to make a strategic choice [57:36] to go into new industry. [57:37] Like, there's demand for being able to spend stablecoins, [57:41] and they can now sell to everyone [57:43] who wants to particularly do that, [57:45] and so it's like, it's totally transparent, [57:47] and that's a beautiful thing, amortizing R&D [57:51] over millions of businesses with us centrally [57:54] because we can just build up the system [57:56] and everyone gets the benefit of it, [57:58] and so, you know, obviously, it's in testing. [58:01] Like, the conversion rates have to [58:03] not be negatively affected, obviously, by its presence. [58:05] So there's some, like, [58:07] we have to make absolutely sure this is the case, [58:09] and, like, overabundance of caution [58:11] on everyone that goes into check-out [58:13] because it's like a probably, I mean, [58:16] it's in the busiest checkout on the internet, basically, [58:18] so we might not want to reduce conversion rates ever. [58:24] Yeah, I think it'll also be interesting [58:25] to watch the consumer training aspect of it, [58:27] where, like, QR codes is always my example of this [58:28] where, like, the tech has existed since, what, the '90s, [58:32] but it was really only during the pandemic when, like, [58:34] we installed the software in people's brains [58:36] to understand how QR codes work, but there was a moment [58:40] and now they're kind of everywhere for everything, [58:42] and obviously they were in Asia [58:43] before they were in the US and everything like that. [58:45] I think there'll be something similar with stablecoins, [58:47] where they'll be offered at the point of checkout [58:50] way before people become super familiar with them, [58:53] but we're probably only a small number of years away [58:56] from some mass-market familiarity. [58:59] I think it's going to be very, very natural for people, [59:03] but also in a way that they will just not think about it. [59:05] It's like it's a normal thing. [59:06] It's like it doesn't matter. [59:08] Like, it's going to be so well tooled up [59:10] to bridge between these things that it's, you know, just, [59:15] using crypto rails while using the internet, [59:18] and until you need it out of the internet, [59:21] you don't have to bridge anything, so. [59:23] Yes, yes. [59:24] You've referenced a few times [59:25] the kind of stuff Stripe and Shopify are doing together, [59:27] and, you know, we've caught up at various points on, [59:30] and you even talked at the meta level [59:32] about the Stripe-Shopify relationship, [59:34] where there's some, I know, is it like Apple TSMC [59:37] or, you know, something there [59:38] where it's like deep, intertwined building in a stable way, [59:44] but also kind of expanding over time, and no confusion as to [59:48] who's ultimately the one building the end product [59:49] versus who's kind of providing [59:51] some of the underlying infrastructure, [59:53] but I'm curious what you think has worked well, [59:56] because, yeah, a lot of corporate relationships do not last. [60:01] I mean, I guess if you date it to 2012, [60:02] which I think is accurate, maybe 2011. [60:05] 2012, like 13 years. [60:06] It's certainly more tenured than some. [60:08] Yeah, and I think it's almost more impressive [60:12] when you think about it as market cap expansion, right? [60:14] Like, I mean, we were both very, very small companies. [60:15] People get greedy. Yeah. [60:17] Well, also, I mean, we have had [60:19] a highly functioning, excellent symbiotic partnership, [60:24] and, probably, there's like, I mean, [60:26] how many more digits on the market cap above companies? [60:29] Like, we've expanded many, many, many times here, [60:32] many order of magnitude, and that's pretty rare, [60:35] because, like, again, partnerships, especially in tech. [60:39] Tech is extraordinarily bad at partnerships, [60:44] and, I mean, it's always treated as a prisoner's dilemma [60:49] that, at some point, someone in the company realizes [60:53] defecting gives you five points on the mouth, [60:55] and that sounds really, really good [60:57] with the short-term focus. [60:58] So everyone mistrusts each other, and so it's fairly rare, [61:01] and I think it's cool that we pulled it off, [61:03] and, I mean, I appreciate the partnership. [61:05] It's been really, really fantastic, [61:08] being able to just focus on my space [61:11] instead of having to, you know, [61:13] duplicate things that Stripe would always have done better. [61:17] I think a lot about main quest and side quests, [61:23] and I have such good, I guess, [61:25] founder-market fit with commerce. [61:26] I just, like, I really care about this [61:29] and I really, I can spend my entire, like, [61:32] Karl Popper calls it finding, like, [61:35] the best gift in life is finding a beautiful problem [61:38] that you can never solve, [61:40] and even if you accidentally solve it, [61:42] if you're so unfortunate to solve it, hopefully, [61:44] it has plenty of enlightened problem children [61:48] that you can then tackle, right? [61:50] Yes, yes. [61:51] And so finding one of those is good, [61:53] and not having to, like, [61:54] meander into some complete adjacency to support it [61:58] is actually wonderful. [61:59] Yeah, yeah, I think there's also something about, [62:02] when you talk about this idea of focus, [62:05] I think second acts are kind of a bit over-valorized [62:10] in Silicon Valley somehow, where everyone talks about, [62:13] oh, we want to do, like, [62:14] the new AWS is kind of the coolest thing ever [62:16] where you do a second thing [62:17] that is completely unrelated to the first thing, [62:19] versus NVIDIA is the largest company in the world [62:22] and they just started making GPUs in the '90s [62:24] and turns out a lot of people want GPUs, [62:27] and, I mean, they did obviously [62:28] lots of incremental expansion from there. [62:30] Like, calling them GPUs is doing a disservice, [62:33] but it's in a very hill-climbing expansion way, I think. [62:36] They went vertical, I think, instead of horizontal. [62:39] Business books tell you to go horizontal because, again, [62:43] the chief competency of most businesses ended up being [62:46] having a stopwatch. [62:48] Like, it was a management system world, [62:51] but we are actually in a different world now. [62:52] It's like this is commoditized. [62:54] It's easy to access and probably GPT-5 is better at it [62:56] than most managers, and it's now actually like, [63:00] well, what problem do you understand better [63:02] than what's in the training tokens for GPT-5, [63:04] and what do you care more? [63:06] Which is a human capacity specifically, [63:08] and so, like, Jensen just cared more about GPUs, [63:12] and, you know, saw more potential in that. [63:15] Isn't there a bit of MBA-ism here too? [63:18] Where if one was trying to, you know, [63:20] when Shopify was a thousand times smaller, [63:22] raise money for Shopify, it would just sound ludicrous [63:24] to say, "We're just going to do e-commerce software, [63:27] and it turns out that's a really big market, [63:29] and here's our current projection [63:30] and here's when it's a thousand times larger." [63:32] It's like, "Get out of here. Come on." [63:33] Whereas if you say, "Oh, we'll expand to do this [63:35] and we'll expand to do that. [63:36] We'll build a CRM or whatever." [63:38] It's easier to justify, you know, [63:39] expanding into other TAMs. [63:41] Like, I got rejected from plenty of VCs in 2008 [63:44] when I went around, and some of them said they would invest, [63:47] and then it was usually terrible terms [63:50] of, like, ratchets or multiples, [63:51] and I had to move the company to Silicon Valley, [63:53] which, again, I would have if they wouldn't have told me. [63:55] Yes. [63:56] Like, once you tell me to do something, [63:58] I don't want it anymore, [64:01] but, anyway, so, like, [64:02] and I caught up with many of the partners later. [64:05] In one case, we had a really interesting conversation. [64:08] Just like you asked me, what did we miss back then? [64:10] Yeah. [64:11] And just, like, well, [64:12] the reason why we didn't invest is, like you said, [64:16] there's 40,000 online stores in the world right now in 2008, [64:20] and, well, you can get half of them, [64:23] but that's not actually that big of a business, [64:26] and what was missed is, like, well, [64:28] Shopify itself was a solution to that problem. [64:31] Like, the reason why no one was building online stores [64:33] is because the friction was so high, [64:37] and so it's a little bit of that, right? [64:40] Like, you got to, like, yes, [64:42] it's not an interesting business plan. [64:45] The pitch isn't vague [64:45] or specifically coming from me. [64:46] This, but more. [64:47] Yeah, yeah, coming from me, it's probably especially bad, [64:50] but, like, it's also kind of, man, [64:53] the best products in the world come from people [64:55] just, like, really, really, really, really [64:58] being deep on something. [64:59] Like, in the programming world, we are using Cursor, [65:02] which is basically VS Code. [65:03] VS Code is Erich Gamma's fourth editor, right? [65:07] It's like he spent his entire life making these things. [65:09] That's craftsmanship. [65:10] Yeah, it's like he wrote the "Design Patterns" book [65:12] that many of us read at some point [65:15] literally about programming text editors, right? [65:17] So it's like, yeah, sometimes you just need to [65:20] dedicate a career to a field [65:24] and get deeper into it than anyone else [65:26] and celebrate craftsmanship along the way [65:29] and kind of try to build the best thing. [65:32] Yeah, but again, it's hard for people [65:35] to give credit to that idea and the model, [65:38] or, like, people are not capable of being [65:40] sufficiently optimistic about a business that really works. [65:43] It blows through all their projections. [65:45] Yeah, the question is also, like, [65:46] we've seen the story so many times [65:48] that, like, some step function changes, like at some point, [65:51] like maybe with Shop Pay [65:53] and, like, all these kind of things with Shopify, [65:55] somehow it became something very different [65:57] than what it otherwise sort of, [65:58] what you would've visualized, [66:00] and, I mean, this is in no area more true than [66:04] in NVIDIA's case, where, like, [66:06] I mean, these CUDA cores, the tensor cores were around. [66:09] They were on these cards which you put, which everyone put, [66:13] they're part of a bomb, [66:15] and very little used by any engine. [66:19] Yes. [66:19] CarMax's new engine probably used them [66:21] for something useful, but they were just there, [66:23] and CUDA was not even, like, [66:25] that was around for 10 years before AlexNet got trained. [66:28] Yeah. [66:29] Like, it's an incredible vision to just, like, [66:31] doggedly pursue this, because, you know, [66:33] in the future, there will be value if you do this, [66:36] and it makes you wonder. [66:38] I think that's probably true for almost every space. [66:40] I think every space, driven to that level, [66:44] could become so big. [66:45] Isn't there something here where [66:48] more founders should be willing to follow their nose [66:51] of what's just a good product? [66:56] Because, again, I think that [66:58] what can happen as companies scale is [67:00] founders get successful by building good product, [67:03] and then they hire a bunch of people and finance people [67:05] and run the planning process and everything like that, [67:07] and it's like, okay, well, [67:08] what's the revenue projection for this? [67:10] And, you know, things like that, [67:11] whereas actually, like, when we built [67:15] Stripe Atlas for incorporation, [67:16] there was no amazing projections behind that, [67:19] that, one day, this will be a giant business, [67:20] and it's not a giant business for us [67:22] by any kind of revenue metric, [67:23] but we think it's very valuable [67:25] because people start their business on Stripe [67:27] and we help them with something foundational, [67:29] like incorporating the company [67:30] and getting it off the ground, [67:31] and it all kind of ties together [67:32] in this very immeasurable, nebulous way, [67:35] and kind of like you're saying with CUDA, you know, [67:37] that was hard to justify at that moment in time, [67:40] but it seemed like the right technical decision, [67:42] and there's something maybe where [67:44] founders should be a bit more willing to follow their noses [67:48] on what is right for the product and right for the user, [67:51] and it'll all come out in the wash in years down the line. [67:54] Yeah, I'm a great believer that [67:55] if you build things that are of value, [67:59] then people will figure that out. [68:01] It's like, it's very, very hard [68:02] to keep people away from things that make their lives [68:06] significantly easier, better, or more joyful, [68:09] and I mean, I get the front-row seat to this [68:12] on my platform, right? [68:13] Like, it's crazy how fast some success stories. [68:16] One of my favorite things we do is [68:19] we started making these really [68:22] sort of nice awards for people [68:25] when they hit certain sales milestones, [68:28] and, you know, some of those, like when I go travel, [68:33] I check out who's just got them, [68:35] and just, like, sometimes deliver them, [68:37] and Harley does this much more than me, [68:39] and, like, it's such a cool thing to— [68:41] That's awesome. [68:42] To talk to them, and, man, like, [68:45] some of these businesses that hit [68:47] like a hundred thousand or a million sales [68:49] are like 18 months old. [68:51] Yeah. It's just like [68:52] they just cracked the code or something, [68:55] and with the most, you know, talk about 1,000 true fans. [68:58] It's like millions. [69:00] It's like a tsunami created [69:01] because people are just so starved for this, like, [69:05] authentic, like unabashed, [69:09] like something needed to be done. [69:11] Something was wrong in the world. [69:12] Yeah. [69:13] It involved a product not existing, [69:15] and that product not existing was such a bother to someone [69:19] that they made it their mission to create it, [69:21] and they went through trials and tribulations. [69:24] Usually, they're often documenting the journey. [69:27] Yes. [69:29] Like, allowing other people to ride shotgun, [69:31] to sort of, over time, try on their vision of the world, [69:35] which, again, is not to change the world, [69:37] but it's to add the thing that was missing, [69:39] which is like, it makes things 1% better, [69:42] not like sort of everything's new, [69:45] and they just agree with that [69:46] and they love the authenticity of it. [69:48] They love that someone [69:50] that they now have a little bit of a relationship with, [69:53] like, kind of pursued this thing [69:54] that they think it was worth doing, [69:57] and then the product comes out and they tell everyone [69:59] and then more people buy it, and it's just like, [70:03] there's so much good that comes out of this, [70:05] and that was all value [70:06] which was there for everyone to take, right? [70:08] Like, every product that sells well was a discovery. [70:10] It's like it was sort of invented, [70:12] but also kind of discovered, and so, [70:16] but it's that person that did it, [70:17] and they deserve the economic value that comes to them. [70:21] Everything about this, it's like, [70:22] it's the best parts of capitalism all rolled up [70:25] for everyone to inspect and review. [70:27] Yeah. [70:28] And we get to have a front-row seat from data, [70:30] and front-row seat sometimes [70:31] with giving these sort of celebration artifacts, [70:35] and, anyway, so this is, [70:38] obviously, the greatest thing is [70:40] just like the people who using the platforms [70:42] are actually really inspiring people. [70:44] Well, you know, I feel like [70:45] it must be particularly satisfying for you, [70:47] because there's all these edge case [70:48] failure modes of capitalism [70:49] of, you know, rent extraction or oligopoly [70:51] or regulatory capture, everything like that, [70:53] and you just have a front row seat [70:54] to the pure competitive differentiation, [70:59] people succeeding on the merits, [71:01] and it's kind of a very pure form. [71:05] I feel like you have that too. [71:07] Yeah. [71:08] And it's like, we should write postcards from, [71:11] you know, to the rest of the world. [71:13] It's like there's something really to this. [71:15] Like, you know, this real world thing, right? [71:19] It's like, the one everyone says, [71:21] your product wouldn't work in, like, [71:24] which sounds so dystopian. [71:25] It's like, it does really sound quite dystopian, [71:29] so I don't want to have anything to do with it. [71:31] Yeah. I'd just much rather, like, [71:33] hey, just build this whatever word, whatever that is, [71:37] and invite everyone over, and if it's good, it works, [71:40] and then people will join. [71:42] Like, it's actually that simple, [71:44] And I think it's better to live in a world with, like, [71:48] people going direct, people communicating, right? [71:50] People being real people. [71:51] Yeah. [71:53] There's very, very rarely, it's very PR department, [71:56] or, like, even there's probably a marketing strategy, [71:59] but the marketing strategy works, like, based on, you know, [72:01] just like new ideas and organically [72:03] and with very few people being highly enabled, [72:07] building products that they care about. [72:08] We had Mackenzie from Ambrook on here, [72:11] which is SaaS for financial infrastructure for farms, [72:15] but she was saying one of the big trends [72:17] in agriculture right now is direct-to-consumer stuff [72:19] where, you know, you're actually [72:20] transacting directly with your farmer. [72:22] I'm sure a lot of them are running on Shopify as well. [72:28] What have you learned on the Coinbase board? [72:30] I never had another job, really, in my apprenticeship, [72:33] and mostly learned what not to do as a company, [72:37] and then sort of had the opportunity to start Shopify [72:40] just trying literally the opposite, [72:42] and that's basically the operations I've been running on. [72:43] It's been amazingly successful, so thank you, Siemens, [72:46] but, like, it's so hard, especially building it in Canada, [72:50] like in Ottawa and Toronto. [72:52] It's funny. [72:53] It always had this effect of [72:54] I came to the Valley a lot when I, well, [72:56] when I had time and money. [72:58] You know, I had lots of meetings and then talked with people [73:01] and they described to me how real companies do these things, [73:04] and then I went home and tried to implement it, [73:05] and only years later did I realize [73:08] very few of the companies describing what they're doing [73:12] actually did the things that they told me. [73:15] Everyone gave me sort of an idealized highlight reel, [73:18] and I compared that to my average, [73:19] and then I implemented [73:21] how I think you could do a better job [73:23] on the stuff that they described, [73:24] and sometimes ended up with good things, [73:26] and I was very, very happy with this, [73:28] and then now, obviously, [73:30] I'm describing when this goes really right, [73:32] an just as often, this went really wrong, right? [73:35] So just, like, Shopify ended up being [73:37] this sort of lopsided thing that is, like, [73:40] probably ludicrously over-engineered [73:43] but advanced at certain things, [73:45] and then, like, incredibly primitive [73:47] at things that no one fails usually, [73:49] because in a more liquid labor market [73:53] where more people build companies, [73:55] the information is just more readily available. [73:57] Yeah. [73:58] What was awesome was, like, [73:59] joining Coinbase as a board, because like I said, [74:02] here's another founder-run company [74:04] by an incredible entrepreneur that's, like, [74:06] put together, like, again, not by committee, [74:10] but by, you know, like, a clear desire [74:13] to build, like, an important company [74:15] and important institution and just holding it up [74:18] and seeing, like, what they were incredibly strong at [74:22] compared to, like, it gave me, [74:23] I learned so much about my own company, [74:25] and I could, like, you know— [74:26] Have a reference point? [74:27] Have a reference point. Right. [74:28] Like, so it's quite a gift. [74:31] I don't know when is the right time [74:33] for an successful entrepreneur to join another company. [74:36] I wouldn't do it too early, but, like, [74:38] it eventually becomes quite valuable, [74:41] and so Coinbase is such a remarkable company. [74:43] You had Brian on. He's such an incredible entrepreneur. [74:47] Very first principles. [74:48] And that was a company under brutal attack by regulators [74:52] in a way that, like, [74:54] very few companies of all time have ever faced, [74:57] that's like, been through, like, [74:59] the government wanted them dead [75:01] and pretty much said so, right? [75:03] So I learned so much about [75:05] the posture that Brian took to it. [75:07] He will never get enough credit for the courage, [75:11] and, like, the intestinal fortitude he's shown there, [75:15] because many of the details are internal [75:17] to the company and the board, [75:19] but, like, God damn it, he's good. [75:22] Yes, yes. [75:23] I do not think I could have [75:24] navigated my company through some of the barrage. [75:28] Yeah, Brian strikes me as [75:29] someone who is very hard to manage up to. [75:31] You know, that's always the risk for CEOs [75:33] is that people are trying to, you know, [75:34] construct some Potemkin village, [75:36] and, you know, he was describing [75:38] the policy work that they were doing, [75:41] and he was saying, you know, well, you know, [75:44] the company's meant to be the good cop [75:45] and the trade group will be the bad cop, [75:47] and then he'd be in the trade group and it's like, [75:48] wait, these are also the good cops, [75:49] and so he was saying, "Okay, here's our new bad cop plan." [75:52] But, again, he was just kind of refusing to go along with [75:54] the things that everyone was serving up to him as the plan. [75:57] I think great entrepreneurs are ungovernable, [76:00] and the problems often arise when people [76:04] somehow manage to figure out how to, yeah. [76:08] I think studying failed companies [76:09] is actually often more useful than— [76:11] Totally. I agree. [76:12] You agree? Yeah. [76:13] Like, what you find invariably is that [76:16] the company knew exactly the stuff that was the problem. [76:19] It's just, like, somehow, [76:21] the CEOs got disconnected from that. [76:23] Yeah. It's not so much that I want to study failures. [76:26] It's that I really want to avoid making the same mistakes [76:31] that other people have well-intentionally made [76:34] and, like, found, [76:35] "Oh, yeah, no, that's a blind canyon up there." [76:37] We should at least be able to make new, exciting mistakes. [76:40] Rob, the guy here who runs corp dev, you know, [76:42] he is very experienced in it, [76:43] and he's done hundreds of acquisitions previously, [76:47] but it makes him way better as a corp dev leader [76:50] because it turns out that [76:52] there's lots of ways for deals to go sideways, [76:54] and if you just have seen a lot of them before, [76:56] you can keep tabs on them [76:58] and be finding ways to steer away from them, [77:00] things like that, [77:01] and I think just avoiding the well-trodden mistakes [77:04] and if you're going to make mistakes, [77:05] at least make new mistakes that haven't been discovered yet [77:08] is kind of underrated as a strategy. [77:09] It's totally underrated, but completely vital, [77:12] because you die to a thousand paper cuts as a company. [77:15] So, like, you kind of have to first avoid, like, [77:18] it becoming a thousand, [77:20] and then, second, you will make mistakes. [77:22] Make them original so there's alpha in it [77:24] so that you accrue some additional knowledge [77:26] which, now, it's unique to you. [77:28] How do you like to be ungovernable? [77:29] How do you like to avoid being managed up to? [77:31] You probably can't share all your secrets [77:33] because part of them are part of the specialty. [77:35] I just think going direct a lot inside of a company too. [77:38] It's just like, I mean, this is— [77:40] The tyranny of the org chart [77:41] and information flowing up, being polished at each step. [77:43] Yeah, I'm just like, I don't think, like, [77:46] up and down is sort of not the right mental model, [77:48] and now it is good. [77:49] It's like, certain decisions are best made [77:52] at the outskirts of a company closest to the customers, [77:56] but I'm a big believer in [77:58] many decisions have to be brought inwards as well. [78:00] Like, this is why, you know, [78:01] product reviews are, like, very centralized. [78:03] It's like a way of efficiently moving information [78:06] completely in the bend of org charts. [78:08] I'm a fan of functionally organized companies. [78:12] The trade-offs, it is significantly harder for executives. [78:18] It's very unpopular for executives because, obviously, [78:21] they just don't own, like, the entire shebang [78:23] that they can, you know, make all the decisions in, [78:25] and that requires a lot more collaboration, [78:28] which feels wrong and feels slow, [78:29] but, like, you make vastly better products [78:33] in functional organizations, at least for Shopify, [78:38] and, you know, that helps. [78:41] I just, like, know a lot about the company. [78:43] Like, being so many details. [78:45] Like, I try to be in AI stuff first [78:48] so that I can, you know, [78:51] I always see it as a gift when something new happens. [78:53] Yeah. [78:54] Because especially, like, [78:56] I'm very good at quickly figuring out stuff, [78:58] and then I sit there with knowledge, [79:01] and I can now, like, look at, okay, [79:04] who else gets there fast? [79:05] And those are my future leaders. [79:07] Like, that's a very implicit testing. [79:09] Oh, that's interesting. [79:10] You're kind of testing some adaptability [79:11] or something like that. Yeah. [79:12] Like, the biggest difference [79:13] between people inside of any company [79:15] or just like in general in industry [79:16] is, like, some people fall in love with solutions. [79:19] Some people fall in love with problems, [79:21] and I just, like, fundamentally, [79:23] I'm a person who appreciates [79:24] the people who fall in love with problems. [79:25] They're very hard to distinguish [79:27] during fairly stable times, right? [79:28] Like, it's very hard to know who is who, [79:30] because, well, if you don't change anything, [79:32] they look exactly the same. [79:33] I like inducing change in general to suss this out. [79:36] Like, I don't know. [79:38] How do you induce change? That sounds euphemistic. [79:41] I don't know. Just, like, literally like that. [79:42] Like, do stuff. [79:44] I think reorgs are sort of a standard way. [79:48] All reorgs work because reorging is valuable. [79:50] It doesn't matter [79:51] what your organization structure is afterwards, [79:53] but, well, I mean, it does, [79:54] but, like, there's some benefits [79:56] to any organization structure. [79:57] It shakes things out of the stasis. [79:59] Well, and the advantages of any organization structure [80:02] accrue immediately. [80:03] Their disadvantages only appear over time. [80:05] So, if you change a lot, you do multiple things, [80:08] you're also, like, I think heightening your misalignment [80:11] with the people you don't want in your company, [80:12] is an enormously important thing. [80:14] So, like, if you reorg— Like Brian Armstrong. [80:16] Yeah, exactly, so if you reorg a lot, you end up, [80:20] the people self-select out who don't like that, [80:22] which is perfect. [80:24] Like, it's probably the sane position to take, even. [80:26] Yeah. [80:27] If a company thrives under a lot of change [80:28] and is exposing itself to a lot of change, [80:31] then it's important that you're clear about that. [80:33] I think you try to make your company [80:35] as different from every other company as possible, [80:37] which is actually, I think, [80:38] the opposite of what most companies, most CEOs optimize for. [80:41] And probably most professional management. [80:43] Right. [80:44] I think a lot people want to make their company [80:47] as similar as possible to all other companies, [80:50] but I think that's, I mean, to me, [80:51] that's a lack of diversity, right? [80:52] Like, there should be diversity in companies [80:54] because there is diversity in people [80:56] and people need to find a product. [80:59] In what ways is Shopify most different to other companies? [81:02] I mean, the sort of internal tool culture [81:05] is maybe an example of this. [81:09] It's quite clearly R&D-run [81:12] in a way that quite a few companies are, [81:13] but they don't acknowledge or obscure it. [81:15] We actually say it and celebrate it [81:17] and say we actually are of the opinion at this point [81:20] that everyone should be able to code, [81:21] at least sort of since vibe coding's around. [81:23] This has gone from, you know, [81:25] have to spend all of your teenage years in a dark room [81:29] cultivating a rare skill [81:31] to spending a couple of weekends with YouTube. [81:34] Like, let's go. Like, this is not a problem anymore. [81:36] Speaking of teenage years, I'm always very interested in [81:39] the on-ramp to entrepreneurship [81:42] and the on-ramps to programming, and I'm curious. [81:46] Some people decry that [81:47] programming has gotten harder to on-ramp into, [81:50] where, you know, the first stuff I built [81:52] was, like, janky PHP webpages, [81:54] but, obviously, PHP is actually very nice, [81:56] because, you know, you're just writing code [81:57] directly in the webpage [81:58] and you refresh it and that's the execution environment, [82:00] and now— [82:01] It's the high-water mark of our industry. [82:02] Well, so is it? [82:03] And so have we made programming too hard [82:05] to get into? Yes, totally. [82:07] Okay. [82:08] To tell my wonderful board member [82:09] David Heinemeier Hansson, [82:10] like, the high-water mark of the world of programming, [82:14] at least web programming, [82:15] was clearly when the FTP'd PHP file up [82:19] and it was instantaneously deployed. [82:21] How long does it take code to deploy these days? [82:23] I know. [82:24] This is all choices now, right? [82:25] Yeah. Like, I mean, [82:26] we've made some choices which are worth making. [82:27] Maybe we'll do it again. [82:29] Yes. [82:30] But, like, there was an immediacy [82:32] and a visceralness to programming, [82:34] and that clearly has been lost. [82:36] Yes. So, lots of things [82:38] have gotten incredibly amazing, [82:40] and I always said, like, we used to say that, [82:43] after two years, [82:44] it's like someone pours cement into a code base [82:45] and no one changes anything. [82:46] We figured out how to avoid those horrors. [82:49] I think we have a very mature discipline. [82:55] We've always, as programmers. I'm an engineer too. [82:57] We've always had, like, an identity crisis. [83:00] We always wanted to be someone else. [83:03] We have significant science envy. [83:06] We even attempt to call it computer science. [83:10] There are some people who do computer science [83:12] in the machine learning world, so it's a real thing, [83:14] but, clearly, it's a vocational art. [83:17] It's not most computer science degrees. [83:18] Yeah, and this is also how we eliminated [83:20] things like aesthetics and beauty and stuff [83:23] from our code bases, from talking about it. [83:25] There are beautiful code bases with beautiful code. [83:29] It's code that, like, tells a story [83:32] and clearly communicates what it's there for [83:34] to everyone who reads it, ideally, [83:36] and is an artwork of sorts. [83:39] And so, yes, I think we have progressed, [83:41] I think we have also advanced, but I think we don't have [83:44] a clear vision to drive simplicity, [83:50] and I think I personally really deeply appreciate [83:54] and have always prioritized hiring [83:56] electrical engineers into programming roles [83:59] just because they just have a better idea [84:02] for the costs of inefficiency. [84:04] Waste, really. [84:05] You know, like, you put too many chips on a thing, [84:07] it's too expensive to manufacture. [84:09] You will just never sell it. [84:11] It's not a very good product, [84:13] where the zero marginal cost of copying software [84:16] just means all sins are forgiven, [84:18] and in fact, of course, [84:19] it still runs and does the thing, right? [84:20] You really have to fit the whole system in your head [84:22] as an electrical engineer. [84:23] That's right. Yeah. [84:24] And so, anyway, [84:25] so I think the immature, we don't know what great is. [84:30] We leave a lot of this up to people, [84:33] and therefore the outcomes are so different [84:36] depending on who's on the project. [84:38] I mean, again, we are all getting [84:39] more conscientious about it, [84:40] and I think we are in a counter-swing [84:42] to the sort of horrors of architecture astronauts [84:47] that have caused a lot of damage to, I think, our industry [84:51] over the last, especially like in the 2010s. [84:54] So we are getting in a better spot, [84:56] and then I don't know what AI does, [84:58] but the interesting thing is [85:00] I don't even think it's going to end up mattering [85:01] because AI is going to solve all technical depth, [85:03] so it's, like, maybe— [85:07] So, yeah. [85:07] Same question about on-ramps to entrepreneurship. [85:09] You know, as a kid, [85:10] you were selling things to your classmates, [85:12] and what I observe is [85:14] a lot of entrepreneurs have the same origin stories, [85:18] where they were entrepreneurial early, [85:20] either something like that [85:21] where they had some business in school. [85:24] Super commonly known is Patrick did web development, [85:26] and, you know, when you're a kid [85:27] getting paid thousands of euros to build a website, [85:30] it is like infinite money. [85:31] I didn't know such large amounts of money exist, [85:34] and obviously, to a small business, they're like, [85:36] "Wow, I'm totally taking advantage of this kid [85:38] for a mere, you know, few thousand euros." [85:42] And then another common one seems to be [85:43] drop-shipping arbitrage. [85:45] You know, this whole thing of, like, buying on AliExpress [85:47] and selling on eBay or something like that, [85:48] but you maybe have a window into this [85:50] with Shopify or something. [85:51] Do you have a sense of what the right, like, [85:53] if a 13, 14-year-old is listening to this [85:56] and interested in making money, what are the areas [85:59] where you think are kind of good to pursue today? [86:01] Yeah, I think it's just, [86:02] these reps are the important thing. [86:05] The mechanisms change, [86:06] and there's, like, entrepreneurship clubs [86:08] are really, like, using Shopify a lot, [86:10] and, like, I love that, and this is one of those areas [86:13] I would love to go much deeper on, [86:15] just because it's, like, such a, [86:17] it would be so fun to really, really, you know, [86:21] build directly for them. [86:22] The super important thing is just go through life, like, [86:24] understanding that, like, everything around you is built. [86:27] You wrote one of my favorite tweets of all, [86:29] probably like actually maybe a high-water mark on X, [86:32] like, on this topic. [86:33] Again, I— [86:34] The world is a museum of passion projects? [86:35] That's right, and I love that [86:37] because it's, like, if you can teach that to students, [86:40] it's worth more than [86:42] most of the entire curriculums of subjects, [86:45] and so I think the commonality behind entrepreneurs is that, [86:49] very early in life, [86:52] somehow, something happened that told them [86:55] that the world is dynamic and it changes. [86:59] Some of us, I had the enormous advantage of living [87:03] exactly sort of at the time [87:05] when we went from computers to no computers, like in homes, [87:09] and, like, I had one early and lots of my neighbors didn't. [87:12] So I had the counterfactual there of, [87:14] and I remember being just absolutely fascinated with them, [87:16] and then networks came in my teens, and internet. [87:21] So I had a front-row seat to a world. [87:23] Like, the world around me [87:25] in a small, sleepy town in Germany [87:29] was irreconcilable with what was becoming possible. [87:31] Like, it was so clear that so much would change, [87:34] and so I think that dissuaded me [87:36] of sleepwalking through the world. [87:38] It just made me, like, okay, like, that's fun. [87:41] Like, I am sitting on a piece of information [87:44] that everyone around me doesn't believe yet, [87:47] but it's clearly true and I'm going to bet on, [87:49] and so I'm going to take reps, looking at things around me [87:53] and figuring out how it could be done better, [87:54] and, like, that process is [87:57] I think the most valuable process to cultivate [88:00] as a young adult or teenager, and— [88:04] So the meta process is more important [88:05] than the actual domain you choose. [88:07] I agree. Yeah, absolutely. [88:09] I think, if you spot opportunities, [88:10] see what you can do with it. [88:11] It's a really important thing. [88:13] I think we overemphasize those stories [88:17] and de-emphasize the process that causes this. [88:20] It's building the habit of looking at things [88:23] and thinking about, A, [88:24] do I know how to make a better vision of this? [88:27] Which you should do anyway, [88:30] and then we definitely underestimate [88:32] and under-celebrate and under-lionize— [88:35] The Beatles' Hamburg days. [88:36] Where we're going to, [88:38] and acting on this impulse would mean [88:40] I'm going to spend the next five years of my life doing this [88:44] and I'm willing to pay what that costs. [88:47] And for you, [88:48] was that just doing a lot of programming as a teenager? [88:50] Programming is how I just, like, made things different, [88:52] and, you know, I went to this classical language school. [88:59] I was bad at Latin. [89:00] We started with Latin. [89:01] I only got English in like eighth grade, hence my accent. [89:05] But your Latin accent is excellent. [89:06] Yeah, I imagine. [89:07] Well, no, because no one knows how that actually sounds, [89:09] and it's definitely not German. [89:13] So, everyone just, like, because of, yeah. [89:15] Anyway, I just hated doing the vocabulary tests. [89:18] I was bad at it, so at some point, [89:19] I wrote software for teaching myself, [89:21] basically similar to I suppose what's called [89:23] spaced repetition now, but like, it was just like— [89:25] Wow. [89:26] And it's like, a little basic shell script, right? [89:29] And it's stupid, but, like, [89:31] something about making this and then using it [89:34] made it interesting enough that I did it, [89:36] and so, like, it's a very clear moment where, like, [89:38] I went from, I think I got a D on a test, [89:40] which was, like, starting to be problematic. [89:42] I had just, like, A's for the rest of my life, right? [89:43] It's just like, that is a crazy experience, right? [89:46] Like, this is a thing that didn't exist [89:48] that, like, fits me personally, [89:49] maybe my first internal tool. [89:51] So, yeah, it changes you. [89:54] Exactly. [89:54] It made you passionate about building your own software. [89:56] Yes. You mentioned [89:57] spaced repetition. [89:59] The other interesting, you know, educational topic du jour [90:02] is mastery learning and the Bloom 2 sigma effect. [90:05] How do you put that into practice? [90:07] Do you with your own kids? [90:08] Yeah, so my kids, I have three boys, 15, 13, 11, [90:13] and they go to a school, a boys' school, [90:16] which we are quite fond of, and, I mean, they're my kids. [90:21] They're quite nerdy, [90:23] so, like, it's just like— [90:24] I've met them. [90:25] Yeah, yeah, yeah. [90:25] So I love them to bits. I can't wait to go home. [90:28] It's such a fun time. [90:30] Everyone should have kids, [90:33] and the particular mechanisms by, I don't know. [90:37] It's like the most important principle of, [90:41] like parenting principle I have is everything's interesting. [90:43] This is a sentence you will hear a lot in our household. [90:46] The kids say it. [90:48] No one is allowed to say they're bad at something. [90:49] They're only ever allowed to say [90:52] they are not good at something yet. [90:54] So that's basically the entire set of principles. [90:57] Yes, yes. We're not good at gross margins yet. [91:00] Exactly. Yes. [91:01] We choose not to. Yes. [91:03] You know, and making it a joke is funny too, right? [91:06] Like, it's just like, [91:08] I want them to have those things in their priorities [91:10] because it's really easy to say something is annoying, [91:13] but, like, it's much, much harder to figure out [91:16] what you would have to do [91:17] to make it interesting for yourself, [91:18] and I think there just isn't anything [91:21] that is not actually interesting. [91:23] You spent a long time in banking. [91:24] It doesn't sound terribly interesting to me initially, [91:26] but, like, I've learned enough about it. [91:28] I've read books about double-entry accounting [91:30] and the history thereof, [91:32] and it's just, like, absolutely fascinating, so. [91:33] Did we recommend the Calomiris book to you? [91:37] "Fragile by Design," [91:38] which is a history of various countries' banking systems. [91:40] I'm literally just reading it. [91:41] Oh, you are? Okay. [91:42] That's so— [91:43] And the Canada chapter is particularly good. [91:44] It's like a history of the Canadian banking system. [91:45] Like, very fond of Canada right now. [91:47] It just, like, needs a new version, I think. [91:51] Yeah, yeah, well, I found it interesting [91:53] how it kind of explains [91:54] the path dependence of how each country gets to where it is. [91:57] Speaking of Canada, what's your advice for Mark Carney? [91:59] Yeah, I mean, it's funny. [92:01] Our partnership is very, very storied and layered, [92:03] and somehow your board members know my prime minister. [92:07] So, that's fun. Look, I love Canada. [92:09] I lived more in Canada than Germany for longer. [92:13] I think it's a beautiful project. [92:15] It's a beautiful implementation [92:16] of different implementation from the United States [92:18] or from Europe of classical liberalism. [92:21] It's very much veered into socialism under Trudeau. [92:26] He pushed that very, very hard. [92:29] I am very much hoping that [92:32] Mark is helping all of Canada to figure out that [92:35] that's a successful discovery of something that doesn't work [92:38] because, obviously, that kind of push to socialism [92:41] always leads to poverty. [92:42] It's like— [92:43] The problem with socialism is, eventually, [92:44] you run out of other people's money. [92:45] Exactly, and so it's just like, [92:47] and it doesn't need it because it's such an incredibly, [92:50] like, I mean, I built Canada's largest company now, [92:53] at this point from my market cap, [92:55] as an immigrant. [92:56] I think that advantage I had as an immigrant is [92:58] I wasn't sort of aware that we are supposed to, [93:01] you know, make small plans. [93:03] So I made big plans, you know, [93:05] and held everyone accountable to that, and it turns out [93:08] you can build world-class companies if you do that, [93:10] and there doesn't need to be any kind of inferiority there, [93:14] and also, just like from first principles, it's like, [93:18] it's just like, I mean, first of all, [93:19] it is attached to, like, next door to [93:22] the best consumer market on Planet Earth. [93:24] Until 15 minutes ago, we were good friends, [93:26] and, you know, it has in the ground, basically, [93:32] like, it's number one, two, or three [93:34] on every single resource that is of value [93:40] in the next manufacturing build-up age. [93:42] So, like, we should probably get on [93:44] extracting some of those resources and do something. [93:46] I think it can be the richest country in the world [93:49] if it chose so, [93:50] and it just needs to make a decision that it wants that, [93:53] and I'm very optimistic about Mark, [93:59] you know, pushing into that direction, [94:00] and I hope he's going to push very, very far [94:03] because I think there's no advantage of waiting. [94:07] So, taking economic growth and GDP growth seriously, [94:09] which has obviously plateaued somewhat [94:11] in Canada in recent years, and then being willing to make [94:15] the potentially controversial policy decisions [94:17] and kind of changes that are required [94:19] to actually enable that. [94:19] Well, I mean the things that are [94:21] controversial policy decisions [94:22] are just, like, controversial for path dependence, not real. [94:26] It's like, Canada has, something crazy to tell people, [94:30] but it has this incredible heartburn over pipelines, [94:34] which are things that go underground next to highways, [94:37] and it's just like no one knows that there's a pipe. [94:41] Like, the only thing a pipeline does is [94:43] you don't need diesel trucks. [94:44] Yes, yes. That's the only thing, [94:45] and you can save a lot of diesel. [94:50] It also changes the cost basis, [94:51] and Canada has all these resources [94:53] and they need to somehow get to port, [94:55] and so it just feels like [94:57] these tiny little implementation details. [94:59] It's like as if computer manufacturers [95:01] would somehow get down on copper cables, right? [95:03] Like, it's a weird thing, and so de-politicizing [95:08] just like implementation details that professionals can do. [95:10] It's odd because Canada, for instance, [95:13] doesn't have this on nuclear power. [95:14] Like, where I live in Ontario, [95:17] it's like the grid is entirely clean. [95:19] It's like the Bruce nuclear power plant, [95:20] which we constantly built more reactors for, [95:23] which is like every time under budget and before schedule, [95:26] like six, seven gigawatts now and just humming, [95:29] and then the rest is, it's the cleanest grid in the world. [95:32] Yeah. [95:33] And so, like, the funny thing is [95:35] it doesn't even need it [95:36] because it has all the other stuff too [95:37] that it could just use, [95:39] but, like, we could have this tool. [95:41] So, like, I think that the blueprint is obvious. [95:43] What I actually, so my main advice to him [95:45] that is always, like, I want him to, in fact, [95:49] ideally talk to you about the Shannon Economic Zone, [95:51] because I feel like it's like a blueprint [95:53] for success in Canada, [95:54] because I think we can build data centers [95:56] or space ports and stuff like this much faster, [95:59] and my hope is that Canada can build up an economy [96:02] and a merit-ful market and so on, which makes it, like, [96:08] it can get back to doing what Canada does best, [96:11] which is succeed by helping America succeed, [96:14] and, you know, just work together and let's go. [96:17] Yeah, and as you say, energy could be a big part of that. [96:19] As you mentioned, nuclear. [96:20] I was thinking about hydro, [96:21] where, obviously, Canada has a ton of hydro already. [96:23] Like, you were saying [96:24] the seven-gigawatt nuclear power station, [96:26] but I think La Grande River is a seven-gigawatt project [96:30] and could be a lot more. [96:31] I think there's other phases that were not done. [96:33] Absolutely massive on every single angle you look. [96:37] Like, every single time I learn about a new resource [96:39] that we will probably need a lot of, [96:41] if I look, it's just there. [96:43] Yeah. Like, it's just unexplored. [96:44] Yeah. [96:46] Maybe last question. [96:48] We have your Daytona car here. [96:50] What is the draw of motor racing to you? [96:55] Yeah, I mean, I just love it, and it's surprising to me. [97:02] Honestly, this is one of the weirder— [97:05] Sorry. It is or it isn't surprising? [97:06] It is surprising to me. [97:07] So it's like, you know, so, [97:10] I love, like, kind of figuring out [97:12] my own limitations or propensities. [97:14] Like, I think it's, like, [97:16] it's a game everyone is playing, I think, to some degree. [97:18] Maybe it's a little less conscientiously. [97:20] I try on every hobby that anyone I admire gets into, [97:25] because it's like, maybe I'm interested in it too. [97:27] Yeah, yeah, yeah. [97:28] Who knows? [97:29] Like, again, everything is interesting. [97:30] Motorsports was, like, not really something [97:33] I had high likelihood or batting average on liking [97:35] because I didn't even have a driver license [97:36] after coming to Canada after my German one expired, [97:40] and so a couple of friends took me to, like, [97:43] I had went to a track and I went out [97:44] and I just immediately fell in love with it. [97:48] There's something just incredible, [97:50] and I love competing against myself, I guess. [97:51] Mhm. [97:53] I think that's the thing. [97:55] I race in races and I don't even try to win. [97:59] I'm still competing against myself [98:01] and I see all the other cars as obstacles, [98:04] and so there's just, like, this beautiful thing about, [98:09] it's technical and you connect it with this, like, [98:12] four basically playing card size patches of rubber [98:18] to the ground, [98:19] and there's only so much grip to get out of them, [98:23] and trying to get to like 99.9% of lateral acceleration [98:28] or turning, understanding the weight transfers of the cars [98:31] and all these kind of things, [98:32] and trying to get to the best lap time you can, [98:35] and, like, the difference between bad lap time [98:38] and the theoretical possible lap time is just the, you know, [98:42] exact report card of all your inadequacies as a driver, [98:45] and, like, next lap, you try to do it slightly better [98:47] and better and better and just hone this craft. [98:49] It's just, like, the immediacy and the adrenalin. [98:53] It's just a brilliant thing that I discovered for myself. [98:57] So, I started taking it more seriously over a couple years, [99:01] and now, like, the Daytona this year, [99:04] and hopefully do the big 24-hour— [99:06] Which is a 24-hour relay race, right? [99:08] Like, you swap with other drivers. [99:09] Yeah, yeah. That's right. [99:10] So, one car, 24. [99:12] It's actually, if you watch the Formula 1 cinema movie, [99:16] that starts at that race, [99:18] which was really fun to see. Ah, okay. [99:19] It's like, I've done [99:20] the getting up in the middle of the night, [99:22] walk exactly down that first scene, [99:25] where Brad Pitt walks, like down to the pits, [99:26] and, oddly, our pit stop was even [99:29] exactly the one they went over in the movie, [99:32] and then get in the car, [99:34] and you're very, very tired in the middle of night, [99:35] so I have to say, Yeah. [99:37] I found out recently [99:38] they sold sponsorships for the fake F1 teams in that movie [99:43] and made a lot of money on that. [99:43] Brilliant. [99:44] It's an incredible bit of product placement. [99:46] I mean, like, if you think about it, [99:48] Formula 1 itself is such a made-up thing, right? [99:50] Totally. Yeah, yeah, yeah. [99:51] Like, it's so synthetic. [99:51] It's like we just decided to create a document, [99:55] which is a set of rules. [99:56] Yes. [99:56] And somehow, because that document existed, [99:58] 20 teams move cars around Planet Earth, [100:00] which is, like, also one of the world cases. [100:02] Like, I don't even believe in [100:04] AI unemployment to begin with. [100:07] It's not how things work. [100:08] But if you were worried about it, F1 is a good— [100:09] But if you were worried about it, [100:11] F1 is kind of proving that we can just make up [100:14] completely incredible pursuits, [100:15] especially with AI and robotics. [100:17] We can make it cheaper, and then, like, let's go. [100:19] Yeah, there's a great, [100:20] I mean, the history of F1 is super interesting. [100:22] There was a great Colossus podcast on the history of F1 [100:25] describing basically the Bernie Ecclestone era, [100:27] which came with, you know, the emergence of TV, obviously, [100:31] and there's even things like [100:32] the different race schedules were not standardized, [100:34] and there was, like, multiple, like, [100:36] not every team would show up to every race or whatever, [100:38] and so he really kind of harangued everyone [100:40] into kind of a standardized thing, [100:42] but then it feels like the modern era, [100:45] especially with Netflix, [100:46] I think they're very successfully internalized [100:47] that they're selling an entertainment product. [100:49] Yes. And some people [100:51] will watch every race every weekend, [100:54] but, like, I'm not going to watch every F1 race, [100:56] but I will watch "Drive to Survive." [100:57] Like, that is an appropriate short-form amount of F1 for me, [101:01] and so I think they've actually very savvily internalized [101:04] that F1 is an entertainment product [101:07] and it needs to be kind of consumable [101:09] in all the appropriate form factors. [101:12] And they just gave access to [101:14] the sort of intrigue and politics [101:15] and just characters behind it. [101:17] Like, Formula 1 drivers are crazy people, right? [101:20] It's a nutty thing to basically drive a billboard around [101:22] at 350 kilometers an hour. [101:24] So, like, the characters are larger than life, [101:26] and, like, absolutely driven. [101:28] They're the most zero-sum competitors you can imagine. [101:31] No one's there to just compete against themselves. [101:33] And part of how it works, [101:34] I remember hearing Zak Brown, the McLaren team principal, [101:38] talk about how Netflix is everywhere, [101:40] in, you know, the paddocks and around and things like that, [101:43] and he was saying, yes, you're aware that [101:44] you've signed a thing [101:45] that Netflix microphones will be everywhere, [101:47] but they're really savvy where they'll have, [101:49] like in a spy movie, [101:50] they'll have the directional microphones [101:52] trained on you from a super far-away distance, [101:54] and you just end up forgetting that, like, [101:57] you're on a hot mic. [101:58] Yes. Your entire day, [102:00] while trying to actually manage a race, [102:02] and so I think they get authentic content, [102:05] because, yeah, in theory, people know that they're on camera [102:07] and on a microphone all the time. [102:09] You actually just can't live your life that way. [102:11] You know, you just forget, [102:12] and so they actually get kind of interesting content. [102:15] I completely agree, and it's like, [102:16] the season's long and they do many races. [102:18] It's actually amazing what they do. [102:20] You know, like, they're fantastic athletes, [102:21] They lose like 5, 10 pounds, like— [102:23] Yeah, yeah. The Singapore Night Race. [102:26] Yeah, yeah, so it's pretty amazing, [102:27] and, like, it's a fantastic sport. [102:29] I actually love what Netflix did, [102:31] because, like, I mean, [102:32] Liberty Media is one of those companies [102:33] you kind of have to, like, John Malone. [102:36] Like, I mean, "Cable Cowboy" is already one of the best. [102:38] It was a great book. [102:39] It's such a great book, and then, like— [102:40] You know, he wrote a book. [102:42] It's coming out soon, imminently. [102:43] I did not know that. [102:44] Yeah, he has an autobiography, so. [102:46] That'll be fireworks. We'll do a book club. [102:47] Yeah, it'll be really good. I agree. [102:49] Well, look, thank you, Tobi. [102:51] This was fun. [102:51] This was super fun. [102:52] So it's really, really fortuitous [102:53] that you have an Irish pub just propped up in your office. [102:56] Exactly. Yeah, yeah.