Tobi Lütke

摘要

Tobias "Tobi" Lütke（1980 年生于西德 Rhineland-Palatinate 的 Koblenz）是 Shopify 的联合创始人兼 CEO。他 6 岁从父母那里收到一台 Schneider CPC 电脑，11–12 岁开始改写游戏代码、动手改装硬件；十年级（约 16 岁）后辍学，进入 Carl-Benz-Schule Koblenz 走德国双元制（duales Ausbildungssystem），在 Siemens 完成约三年的程序员学徒训练。2003 年因伴侣 Fiona McKean 来到加拿大，定居 Ottawa；2004 年与 Scott Lake、Daniel Weinand 共同创办在线滑雪板店 Snowdevil；2006 年把支撑 Snowdevil 的 Ruby on Rails 后台代码独立出来，作为 SaaS 产品上线，先后名为 Jaded Pixel，最终命名 Shopify。

2008 年 Scott Lake 离开后他从 CTO 接任 CEO；2010 年从 Bessemer Venture Partners 拿到 $7M Series A；2014 年被 The Globe and Mail 评为加拿大年度 CEO；2015 年 5 月 21 日 Shopify 在 NYSE 与 TSX 同步上市，募资 $131M，估值约 $1.27B；2020 年 5 月在 COVID 早期宣布 "Digital by Default"；2021 年市值突破 $2000 亿，登上 Bloomberg Businessweek 12 月封面；2022 年起股价暴跌 75% 以上，他公开承认自己赌错了 pandemic acceleration；2022 年 7 月裁员约 10%（约 1,000 人），2023 年 5 月再裁约 20%（约 2,000 人）并把 Logistics 业务（Deliverr + 6 River Systems）卖给 Flexport，换得约 13% Flexport 股权；2023 年 1 月 "Chaos Monkey" 一次性删除 12,000 个例会；2025 年 3 月 20 日发布内部 AI 备忘录"Reflexive AI usage is now a baseline expectation"，4 月在 X 公开。截至 2025 年，Shopify 年营收 $11.56B（同比 +30%），GMV 约 $378B，拥有 7,600 名员工，约占美国电商零售 8.6%，处理量约相当于加拿大 GDP 的三分之一。Tobi 个人持有约 7% 股权，通过 dual-class + Founder Share 结构控制约 40% 投票权；2025 年 8 月 Forbes 估算其个人净资产 $123 亿。2026 年 5 月，他披露 Shopify 内部超过 50% 的代码已由 AI 生成，工程任务大量交给住在 Slack 里的内部 AI agent "River"（20VC, 2026-05-04）。2026 年 6 月 17 日 Editions Summer '26 把 Universal Commerce Protocol 在全部店铺默认开启；8 月 5 日 Q2 财报营收 $3,583M（+34%）、GMV $115.6B（+32%），AI 驱动的流量与订单各自同比翻三倍——但 AI 占 GMV 的绝对分母始终未披露。同年 7 月 25–26 日，他在 X 上对"按所缴所得税加权投票、退休者不再享有选举权"的提案回复 "Good system."，并称参加市政会议的市民是 "useful idiots"，引发加拿大主流媒体广泛批评（见批判视角 §7）。

他的核心思想脉络可以串成几条骨干线索：编程是他的"happy place"，是其他一切技能的"trunk"；他坚持每年至少几次清空日历、回去和工程师一起 coding 到深夜。软件不应让人觉得自己愚蠢——这是他反复表达的产品伦理。Trust Battery：把人际信任建模为 0–100% 的电量计。Subtraction：删除比添加更难，他多次主动"subtraction event"——删例会、卖业务、辞退执行层。Calm progress：在追求长期复利的同时不允许 burnout。Infinite game（James Carse）：Shopify 是他的无限游戏，目标不是"赢"，而是让创业本身变得更容易。Founder-product fit > product-market fit：他认为最好的公司由对问题有切肤之痛的创始人创造。Reorgs work because reorging is valuable：组织结构的优势立刻显现而劣势缓慢累积，所以应当周期性地调整。

他不是一个寻常意义上的"硅谷 CEO"。他在 Ottawa 而非 Bay Area；他骑自行车而不是开他的 Tesla；他公开说自己学到的商业智慧"more from playing StarCraft than from business books"；他赞助一支电竞战队 Shopify Rebellion，与 Ruby on Rails 创造者 David Heinemeier Hansson 一起在 IMSA LMP2 赛事中作为业余车手参赛；他的孩子没有零花钱——只能"pitch"得到付费任务；他每周日晚上看新创业者使用 Shopify 的录像；他与 150+ 位 Shopify 商家保持 WhatsApp 对话直接收集反馈。

一、生平时间线

第一章：Koblenz 童年与早期编程 (1980–1996)

Tobias Lütke 1980 年生于西德 Rhineland-Palatinate 联邦州的 Koblenz——一座位于莱茵河与摩泽尔河交汇处的小城。他和家人在那里度过整个童年与少年。Wikipedia 与 John Coogan 的 Shopify 故事都给出同一组核心日期：

• 1980：出生
• 约 1986（6 岁）：父母送给他第一台 Schneider CPC（CPC 系列在欧洲是 Amstrad 在德国的本地化版本）
• 约 1991–1992（11–12 岁）：开始改游戏代码、动手改装硬件
• 约 1996（16 岁）：在 Carl-Benz-Schule Koblenz（Koblenz 市内的职业学校）完成十年级后离开普通学校体系，进入 Berufsausbildung（双元制学徒）
• 约 1996–1999：在 Siemens 完成程序员学徒训练

童年的"computer was a portal"

Tobi 在 Tyler Cowen 那期访谈里把 Koblenz 描述为一个对计算几乎不感兴趣的环境。他说自己上学的体验是糟糕的，是个 "terrible student"——不是因为不爱学，而是因为传统学习方式与他的认知风格不吻合。他在《The Apprentice Programmer》（自己的博客 Too-biased，2013 年 3 月 3 日）里直接说自己是 kinesthetic learner——需要动手，需要立即的反馈。

"I had difficulties with traditional academic learning. I'm a kinesthetic learner. Programming clicked because it provided immediate feedback."

"我对传统学校式的学习有困难。我是动觉型学习者。编程之所以一下子击中我，是因为它能给出即时反馈。"

Tobi Lütke, The Apprentice Programmer, 2013-03-03

2022 年 11 月在 The Knowledge Project 上 Tobi 第一次公开把"传统学校学习有困难"具体落到诊断与课程设计的不匹配上：

"I got diagnosed with all sorts of learning disabilities. It was a pretty clear-cut way of ADHD. I grew up dyslexic, so just reading was hard. But basically, I was bored. The curriculum in Germany was taught — here's all these solutions you might find useful later in life with very, very little time spent talking about the particular problems those solutions would help you with. If I don't understand the problem I'm trying to solve, it's very hard for me to learn a solution."

"我被诊断出各种学习障碍。基本就是典型的 ADHD。我有读写障碍，所以光是阅读就很难。但根本上，我只是觉得无聊。德国的课程是这样讲的——'这是一堆你将来可能用得上的解决方案'，几乎不花时间讲清楚这些方案是为解决什么问题。如果我不理解一个问题，我就很难去学它的解。"

The Knowledge Project, 2022-11-15

这是他后来反复回到"先理解问题，再寻找方案"的认知根源——也是他多年后用来批评 Shopify 内部"迷恋解决方案而非问题"的同一条线。

他多次提到：六岁那台 Schneider CPC 是他个人世界的入口。他不是用它玩——他拆解、修改它：

"By ages 11–12, I began modifying game code and computer hardware."

"11、12 岁时我开始改游戏代码、改装电脑硬件。"

Wikipedia / 综合早期访谈摘要

"I only succeeded in this because I was a computer programmer."

"我能做成这一切，唯一的原因是我是程序员。"

Tobi Lütke, John Coogan profile

2023 年 1 月在 Microsoft CTO Kevin Scott 的 Behind The Tech 上，Tobi 把这台 Amstrad 的意义给出了最完整的现象学版本——并把它直接联到他后来作为 CEO 反复使用的"agency"一词：

"I got really, really, really lucky. My parents got a computer for me for the house. I was six years old. It was like a basic computer. If you think C64, then you sort of get the right idea. It's a German product called an Amstrad. ... When you're a 10-year-old boy growing up in a somewhat rigid schooling system like Germany's, like we had hours of Latin a day — you might have some free time, but what you don't have is agency. And there's nothing in the world that you can absolutely control in a way that it does this thing and it will do it forever."

"我真的真的真的非常幸运。我父母给家里买了一台电脑给我用。当时我六岁。那是一台德国产品，叫 Amstrad。……你想一下，一个十岁男孩长在德国那种偏僵硬的教育体系里——我们每天有几个小时 Latin 课——你可能有一些空闲时间，但你没有'能动性'。世界上没有任何东西是你能完全控制的、能让它做某件事并且永远做这件事的。"

Behind The Tech with Kevin Scott, 2023-01-03

"agency"在 Tobi 后来的 Shopify 语境里反复出现——他对孩子的反黑盒原则、对员工的高 agency 文化、对加密货币和远程办公的 framing，都可以追溯到 Latin 课上那个没有 agency 的十岁的他。

第一个内部工具：自学拉丁文的 shell 脚本

Tobi 写过的"第一个内部工具"是一个为对抗自己 Latin 课成绩下滑而做的间隔重复 shell 脚本——这是 Shopify 后来"内部工具改变行为"哲学的童年起源：

"I just hated doing the vocabulary tests. I was bad at it. So at some point I wrote software for teaching myself, basically similar to, I suppose, what's called spaced repetition now. It's a little basic shell script. ... I think I got a D on a test, which was starting to be problematic. I was just like, A's for the rest of my life. ... Maybe my first internal tool. It changes you."

"我就是讨厌做词汇测试。我也做不好。所以某一刻我写了一个软件来教自己——基本就是现在所说的'间隔重复'。一个很基础的 shell 脚本。……我有一次测试得了 D，开始有点麻烦了。我就想，从此之后我要全部拿 A。……可能是我的第一个'内部工具'。它会改变你。"

Behind The Tech, 2023-01-03

"为什么要做程序员才能开店？" 的早期种子

他对优化的本能从童年就显露。Tim Ferriss 那期访谈里他自己讲过这段：

"I counted steps as a child to find the shortest route to school."

"小时候我会数自己走到学校的步数，找最短的那条路。"

The Tim Ferriss Show #359, 2019-02-07

"If I have to do something once, that's fine. If I have to do it twice, I'm annoyed. If three times, I automate it."

"一次做完一件事，没问题。要做两次，我会烦。要做三次，我就会自动化它。"

The Tim Ferriss Show #359

这句"做三次就要自动化"后来变成 Shopify 工程文化里反复被引用的一句话。

Linux mailing list 与德国程序员的入门方式

在 Koblenz 时，1990 年代中后期没有 Stack Overflow、没有 GitHub、互联网仍然是拨号——他自学的方式是订阅 Linux kernel mailing list：

"Grew up in Koblenz where computing wasn't valued. Learned by reading Linux kernel mailing lists."

"我在 Koblenz 长大，那里没人在乎计算机这件事。我是靠读 Linux kernel 邮件列表自学的。"

Tobi Lütke, Conversations with Tyler, 2024-09-18

他把这种"从源头读最严肃技术文档"的训练称为他后来一切技术品味的来源。Linux kernel mailing list 是世界上最高强度的开源工程讨论场——一个 13–16 岁少年几乎用十年时间被这种成人级别的工程话语训练了。

"everything is interesting" 的家庭口号

虽然 Wikipedia 没有详细记载父母的职业，但 Shawn Ryan 那期访谈里 Tobi 自述了家庭文化：

"The real motto of our family is really that it's a super simple sentence. It's just everything is interesting."

"我们家真正的家训就是一句非常简单的话——'万事皆有趣'。"

The Shawn Ryan Show #261, 2025-12-11

这句话后来成了他自己育儿（三个 10–15 岁的儿子）和 Shopify 招聘哲学的延续——找那些对世界本身感到好奇的人，而不是只对特定技能感兴趣的人。在 Behind The Tech 上他把这句家训和工程师的"反黑盒"美学连成了一条育儿原则：

"I hate black boxes. I really, really dislike black boxes. It's actually like a parenting principle in our house. We do not like black boxes in the house. Everything is understandable and everything is way more interesting than it seems."

"我讨厌黑盒。我真的真的讨厌黑盒。这其实是我家里的一条育儿原则——我们家不喜欢黑盒。一切都是可以理解的，一切都比表面看起来有趣得多。"

Behind The Tech, 2023-01-03

"Age 17: dropped out of high school to apprentice at Siemens."

"17 岁，他从普通高中辍学，去 Siemens 当学徒。"

John Coogan, The Story of Tobi Lutke

第二章：Siemens 学徒制 (1996–1999)

德国的 duales Ausbildungssystem（双元制教育系统）是 Tobi 反复称赞的少数"我希望北美也有"的德国制度之一。它的结构是：学徒同时在 Berufsschule（职业学校）和工作单位（这里是 Siemens）学习，每周分配在两边的时间，通常持续 2.5–3.5 年，毕业后取得 IHK（工商会）认证的职业资格。Tobi 选择的方向是 Fachinformatiker 路径下的程序员（Wikipedia 写为 "computer programmer track"），学徒地点是 Siemens AG 在 Koblenz 的工厂。

学徒生活的具体形态

Tobi 在 Tim Ferriss #359 那期访谈中详细描述了学徒生活的轮转方式——这一点比起任何关于 Shopify 的故事更早塑造了他对"组织"的理解：

• 在不同部门轮岗：cafeteria（食堂）、accounting（会计）、inventory（库存）、reception（前台）
• 每个部门待几周到几个月，被要求理解工作流程
• 学校与工厂分时进行
• 通过 immersion 而不是 lecture 学习

"Rotated through different departments: cafeteria, accounting, inventory, reception. Learned by immersion rather than lecture. Advocates apprenticeship over traditional college paths."

"在不同部门间轮转：食堂、会计、库存、前台。靠浸入式学习，而不是听课。我至今仍主张学徒制比传统大学更值得学习。"

Tim Ferriss #359 / 综合 The Apprentice Programmer

这套机制的内核是：你不是在某个抽象的"学习者"角色里——你是 Siemens 的一员，做的事情虽然有训练成分但是真实有效的工作。这种设计让 17 岁的 Tobi 在心智上比同龄学校里的同学早成熟好几年。

导师 Jürgen Saar

Siemens 时期最关键的人物是他的程序员学徒导师 Jürgen Saar。Tobi 在多个访谈里把 Jürgen 描述为"塑造我整个领导哲学"的人：

"Jürgen Saar, my programming apprentice mentor at Siemens, sent 17-year-old me alone to customer meetings."

"我在 Siemens 的程序员学徒导师 Jürgen Saar，把 17 岁的我一个人派去见客户。"

The Tim Ferriss Show #359

"He created scenarios where Tobi had to exceed his perceived capabilities. Modeled growth facilitation by creating scenarios where he had to exceed his perceived capabilities."

"他给我设置了一些超出我自我认知能力的场景。他示范了'怎样通过制造略超能力的处境来让一个人成长'。"

The Apprentice Programmer, 2013-03-03 (引用自 Tim Ferriss summary)

在 The Knowledge Project 上 Tobi 给出了他遇见 Jürgen 的具体场景——以及为什么他作为一个 ADHD/dyslexic 孩子会被这个"叛逆者"吸引。这条第一人称叙述揭示了他后来招聘标准 "too good for anyone to complain" 的源头：

"I found this one team that just seemed really fun. It was run by Jürgen Starr. He was a fantastic engineer programmer. He would always come on his BMW motorcycle to work and have long hair and was not wearing a suit like he was supposed to. So, like a total rebel, but so good that no one could complain. I immediately gravitated to him and his little group of rebels."

"我找到一个团队，看起来特别有意思。是 Jürgen Starr 带的。他是个超棒的工程师/程序员。他每天骑 BMW 摩托车上班，留着长发，没按规定穿西装。完全是个叛逆者，但他做事好得没人挑得出毛病。我立刻被他和他那群小'叛徒'吸引过去了。"

The Knowledge Project, 2022-11-15

Jürgen 的方法后来直接进入 Shopify 招聘语言：

"I have a history of being right about people's potential more than they are themselves."

"在判断别人的潜力上，我的命中率比他们自己更高——我有这样的过往记录。"

Lenny's Podcast, 2025-02-02

他认为人在被相信的时候才会真正成长——而 Jürgen 是最早让他经历这件事的人。

"the apprentice programmer" 这一身份的持久影响

Tobi 在 2013 年 3 月 3 日的博客 The Apprentice Programmer 里把 Siemens 的经历放在他整个人生哲学的中心位置。他讲了三件 Siemens 学徒制留下的烙印：

1. 学习对他来说必须有立刻的工作场景——脱离场景的概念学习他做不下去
2. 被信任和被推到能力边缘是成长的两个必要条件——Jürgen 把他派去客户那里，是相信他能够顶住
3. 职业身份比学历更重要——Siemens 不发"学位"，发"行业资格"；这件事让他后来对硅谷"Stanford / MIT 学位即是入场券"的招聘文化保持距离

"The apprenticeship shaped his entire leadership philosophy: push people beyond comfort, provide real-world challenges, trust in potential over credentials."

"学徒制塑造了我整个领导哲学：把人推过自己的舒适区，给他们真实世界的挑战，相信潜力大于看简历。"

Tobi Lütke, The Apprentice Programmer / Tim Ferriss #359

这条线后来直接体现在 Shopify 的两个长期实践：（1）招聘时让候选人讲生命故事和决策时刻，而不是问技术题；（2）Tobi 自己愿意把刚加入的工程师推到他们觉得"还做不到"的项目上。

"Germans are direct about quality standards — no shit-sandwich configuration needed for criticism. Germans prefer lengthy planning before shipping; North Americans favor rapid iteration. Germany lacks tech giants due to pessimism about the future and low status of startups."

"德国人对质量标准很直接——批评的时候不用'三明治式包装'。德国人喜欢在交付之前长期规划；北美人喜欢快速迭代。德国之所以没有科技巨头，是因为对未来的悲观态度和创业者的低社会地位。"

Conversations with Tyler, 2024-09-18

第三章：搬迁加拿大与 Snowdevil 起源 (1999–2004)

2003 年是 Tobi 人生的分水岭：他离开德国，搬迁到 Ontario 的 Ottawa——加拿大首都，人口约一百万，远离传统科技中心。Wikipedia 写得简洁："Immigrated to Canada in 2003."

但他对德国创业生态的判断不是抽象观察——他自己在德国试过一次。在 2024 年 Conversations with Tyler 上 Tobi 罕见地承认了一件别处从未公开的事：

"I started a small company in Germany, didn't do anything. So it's not like people didn't do this. I came to Canada again, this time it worked."

"我在德国开过一家小公司，没成什么气候。所以并不是没人在德国创过业。我后来到加拿大，这次就成了。"

Conversations with Tyler, 2024-09-18

这一次失败的德国创业是后来他多次给出的"为什么加拿大移民这件事对 Shopify 不是巧合"的具体根据。他对德国的诊断也在同一期变得更精确：

"Tech is something that Americans do from the perspective of Germans. There is very much a pessimism about the future that I think means you cannot build tech companies because you kind of have to be optimistic about the future."

"从德国人的视角看，科技是美国人才会做的事情。德国有一种对未来非常深的悲观主义——我认为这导致你不可能建科技公司，因为搞科技多少必须对未来抱有乐观。"

Conversations with Tyler, 2024-09-18

因为 Fiona McKean

搬家的直接动因是他的伴侣 Fiona McKean。她是加拿大人，曾任加拿大外交官；他们大约在 Tobi 二十出头时相识，2003 年时已交往，加拿大对她而言是回家。Tobi 在 Tyler Cowen 那期里直白地说：

"Moved to Canada because his wife studied there."

"我搬到加拿大是因为我太太当时在那里读书。"

Conversations with Tyler, 2024-09-18

20VC 那期访谈里他多次把 Fiona 视为 Shopify 故事的隐藏主角：

"Wife Fiona encouraged him to start a company; Tobi considers Shopify a 'team production' where she doesn't get enough credit."

"Fiona 鼓励我创业；我把 Shopify 看作是一次'团队生产'，她得到的功劳远远不够。"

20VC #6727, 2023-04-03

"Started dating wife when both were above 20 years old, more than 20 years ago."

"我和我太太开始交往时我们都已经超过 20 岁，那是 20 多年前的事了。"

20VC #6727

Fiona 的父母 Bruce 和 Dale McKean 也是 Shopify 早期的天使投资人。他们后来一同购回并重启了 Ontario Chaffey's Locks 附近的历史度假酒店 The Opinicon（2015 年）。

Ottawa 这座城市本身

Tobi 选择 Ottawa（而不是多伦多或温哥华）有偶然性，但这个偶然后来变成了他主动的战略立场。Ottawa 是加拿大政府所在地，主要经济结构是公务员和军方相关产业，技术圈很小。Tobi 在 Lenny 的访谈里讲了 Ottawa 的尺度限制：

"Ottawa, Canada — a city of 1 million people, insufficient talent pool for a 10,000-person company."

"Ottawa 是一座只有 100 万人口的加拿大城市，无法支撑一家一万人量级的公司所需的人才池。"

Lenny's Podcast, 2025-02-02

这个事实后来直接影响了 Shopify 在 2020 年彻底转向 "Digital by Default" 的决定。

Snowboarding 与 Snowdevil 的起源（2004）

Tobi 在德国就已经是滑雪板爱好者；到 Ottawa 后冬天的雪场更近，他和当地朋友 Scott Lake 决定开一家在线滑雪板店——Snowdevil.ca。同时加入的还有 Tobi 的德国老朋友 Daniel Weinand（来自德国，受邀来 Ottawa 帮忙）。

"2004: Co-founded Snowdevil online snowboard shop with Daniel Weinand and Scott Lake."

"2004 年，Tobi 和 Daniel Weinand、Scott Lake 一起创办了在线滑雪板店 Snowdevil。"

Wikipedia

但他们立刻撞上了一个让 Tobi 持续愤怒的事实——当时所有的电商平台软件都很糟糕。他们试用了 Yahoo Stores、Miva、osCommerce 等当时的主流方案：

"When the snowboarding company started, existing e-commerce solutions weren't good enough — they couldn't make the storefront look the way we wanted, payment integration was broken, the admin was unusable."

"做滑雪板店时，现有的电商方案都不够好——我们想要的店面外观做不出来，支付集成是坏的，后台简直没法用。"

Tobi Lütke, Stripe Sessions, 2025-10-06（综述）

这个挫折并非偶然——它是 Tobi 后来反复讲的"为什么电商平台必须重做"的具体场景：

"I built Shopify to avoid people having to use shitty software, not to add to the pile."

"我做 Shopify 是为了让人不必再用糟糕的软件，而不是给糟糕软件这一堆里再加一份。"

Stripe Sessions, 2025-10-06

自学 Ruby on Rails

Snowdevil 起步时（2004），Ruby on Rails 刚由 David Heinemeier Hansson 在丹麦发布（2004 年 7 月开源），全球用户极少，官方文档大量是日文（Ruby 语言本身由日本程序员松本行弘 Yukihiro Matsumoto 创立）。Tobi 不会日语，但他无视这一点，硬啃了下来：

"Discovered Ruby programming language through self-study (couldn't read Japanese documentation, figured it out anyway)."

"我是自学 Ruby 的（我读不了日文文档，但我硬是把它弄明白了）。"

Globe and Mail, 2014

他对 Ruby 的痴迷，后来变成了对编程语言哲学的反复表达：

"Fell in love with programming languages that didn't try to control him."

"我爱上的，是那些不会试图控制使用者的编程语言。"

The Apprentice Programmer

Ruby 与他在 Linux mailing list 学到的精神一脉相承：给程序员最大自由度。这成了他后来对所有产品（包括 Shopify 自己）的设计判断标准。

"Built Shopify as an immigrant without Canada's 'supposed to make small plans' cultural narrative."

"我作为移民来建 Shopify，所以我没有沾上加拿大那种'我们就该规划小一点'的文化叙事。"

Tobi Lütke, The Cheeky Pint, 2025（与 John Collison）

第四章：Shopify pivot (2006)

2006 年 6 月 Snowdevil 进入夏季淡季——卖滑雪板的生意在北半球 4–9 月几乎是停滞状态。Tobi 此时已经为 Snowdevil 写了一个完整的 Ruby on Rails 后端（包含商品管理、模板系统、订单流），他和 Scott Lake 看着这个软件，意识到一件他在 Shawn Ryan 那期访谈里精确还原的事：

"Holy shit, I just went from being a builder to being an entrepreneur today, this moment."

"妈的，我刚刚就在这一刻——从一个 builder 变成了一个 entrepreneur。"

The Shawn Ryan Show #261, 2025-12-11

"Why do I need to be a programmer to then become a retailer? That doesn't make sense."

"我为什么必须先是个程序员，才能成为零售商？这没道理。"

The Shawn Ryan Show #261

"Shopify itself was the solution to the very problem that they stated. It was only 40,000 online stores because no one made it easy to start them."

"Shopify 自己就是 VC 当时所提那个问题的解决方案——网上之所以只有 40,000 家店，是因为没人让开店变简单。"

The Shawn Ryan Show #261

从 Snowdevil 到 Jaded Pixel 到 Shopify

正式拆分流程是：

• 2006 年中：Snowdevil 暂停滑雪板业务（季节性原因）
• 2006 年中：将公司更名为 Jaded Pixel（最初的注册名）
• 2006 年 6 月公开发布 Shopify（这是产品名；公司名仍为 Jaded Pixel 一段时间，后来才公司改名）
• 2006 年 10 月：第一个完整月营收 $8,000

"Original company name: 'Jaded Pixel' (rebranded to Shopify)."

"公司原名是 Jaded Pixel，后来更名为 Shopify。"

John Coogan, The Story of Tobi Lutke

"October 2006 revenue: $8,000."

"2006 年 10 月营收：8,000 美元。"

Globe and Mail, 2014

Tobi 和 Scott Lake 对早期定价很纠结。最初的方案是收商家销售额的 3%：

"Initial pricing model: Revenue percentage (3% of sales) — high fees discouraged growth-minded merchants. 2007: Switched to flat-rate pricing; growth accelerated."

"最初定价模型是抽销售额的 3%——这种高费率劝退了想做大的商家。2007 年改为定额订阅模型，增长立刻加速。"

John Coogan, The Story of Tobi Lutke

三人创始团队

Shopify 真正的"创始人三角"是：

• Tobi Lütke（CTO 兼工程一切）
• Scott Lake（CEO，加拿大本地人）
• Daniel Weinand（设计 + UI；曾是 Tobi 在德国的朋友，被叫到 Ottawa 加入）

Daniel Weinand 后来对 Tobi 的描述成为最常被引用的句子之一：

"Sometimes it feels like he is going a few years into the future and coming back with positive information."

"有时候我感觉他像是先去了几年后的未来，然后带着正面信息回来。"

Daniel Weinand, Globe and Mail, 2014

这句话是 Acquired ACQ2 那期标题 "How to Live in Everyone Else's Future" 的智识源头——Tobi 自己后来把这种"住在别人的未来"明确成一种学习方法（详见第八章 Build Phase 2 与第十章 AI 时代）。

第一笔种子投资

2007 年 Tobi 通过 Ottawa 当地的天使 John Phillips 和 Fiona 父母 Bruce/Dale McKean 等约 10 位朋友家人募到 $200,000：

"Launched with $200,000 from 10 friends and family members."

"我们靠 10 位朋友和家人凑的 20 万美元起步。"

Globe and Mail, 2014

"First major funding: $250K from angel investor John Phillips."

"第一笔重要资金是天使投资人 John Phillips 给的 25 万美元。"

John Coogan, The Story of Tobi Lutke

John Phillips 后来反复出现在 Shopify 公司故事里，他在 Globe and Mail 给 Tobi 的评价是：

"It is staggering how good he is with people. He is absolutely heroic."

"他和人打交道的能力让人惊愕。绝对是英雄级的。"

John Phillips, Globe and Mail, 2014

"We thought existing e-commerce platforms would work for us, but they didn't. So we built our own. After a while we realized: the platform itself was the actual product."

"我们以为现成的电商平台够用，结果不够。我们就自己建了一个。过了一段时间我们意识到——这个平台本身才是真正的产品。"

Tobi Lütke (paraphrase), Shopify IPO Letter, 2015 / Founderoo

第一人称起源叙事（How I Built This, 2019）

2026-07-03 refresh 折入 How I Built This with Guy Raz（2019-08-05）——这是白名单 Tier-1 节目里 Tobi 亲口讲的规范版起源故事，此前档案里最早的生平阶段主要靠 Wikipedia / 博客 / 第三方转述，缺一手长访谈。（注：该期后半段还有 Barb Heilman / Rebecca Davison 的 "Unbuckle Me" 故事，属 HIBT 惯例的第二段，与 Tobi 无关，不折入。）

Tobi 在这里第一次给出了 Shopify 使命的最形象比喻——把创业的"垂直墙"变成"斜坡"，并把"第一笔销售"定义为身份转变的临界点：

"Our North Star is the moment a merchant gets their first sale. It is a profound, identity-changing experience. My goal is to take the 'vertical walls' of entrepreneurship — the technical complexities of payments, shipping, and inventory — and turn them into manageable inclines. Every time we remove a friction point, we create more entrepreneurs."

"我们的北极星，是一个商家拿到第一笔销售的那一刻。那是一种深刻的、改变身份的体验。我的目标是把创业的那些'垂直的墙'——支付、物流、库存的技术复杂度——变成可以爬的斜坡。我们每消除一个摩擦点，就创造出更多创业者。"

How I Built This, 2019-08-05

"Until you sell something, you are just someone who builds something. It is that moment when the order comes through that makes you an entrepreneur."

"在你卖出东西之前，你只是一个 build 东西的人。就是订单进来的那一刻，让你成为了创业者。"

How I Built This, 2019-08-05 [00:22:05]

这条与 §一 第四章 Shawn Ryan 那句 "I just went from being a builder to being an entrepreneur today"（2025）互为印证——但 HIBT 版本把它落在商家（而非他自己）身上：他是从服务对象的第一笔销售来定义 Shopify 的成功。

一个此前档案完全缺失的产品决定——故意不在商家店面上打 Shopify 的品牌：

"We don't even put the 'Shopify' brand on our merchants' stores because the goal is to make our merchants look good, not ourselves."

"我们甚至不在商家的店面上放'Shopify'这个品牌，因为目标是让我们的商家显得好，而不是让我们自己显得好。"

How I Built This, 2019-08-05

[推断] 这条"隐去自己品牌"的决定，是 §6.4 "arm the rebels / anti-Amazon"叙事在产品层的最早、最具体的落点——Shopify 把商家的品牌置于自己品牌之上，与 Amazon marketplace 把所有商家收编进 Amazon 品牌之下正好相反。

第五章：Ruby on Rails 与开源工程师身份 (2005–2010)

这一章在生平时间线中并行于第三、四、六章——Tobi 的 Ruby on Rails 开源工程师身份是和 Shopify 公司的早期增长同时形成的。但因其重要性独立成章。

Rails core 团队成员（2004 起）

Wikipedia 给出的事实：

"Part of Ruby on Rails framework core team (since 2004)."

"自 2004 年起，他是 Ruby on Rails 框架核心团队成员。"

Wikipedia

David Heinemeier Hansson（DHH，Rails 创造者）在 2004 年 7 月开源 Rails；Tobi 在自学 Rails 给 Snowdevil 做后台时迅速参与了开源项目。这段时间他和 DHH 建立了至今 20+ 年的关系——20 年后他们一起在 IMSA LMP2 赛事中作为 co-driver 出场。

他贡献的关键开源项目

Wikipedia 直接列出了他参与/创造的几个广为流传的 Ruby/Rails 库：

"Created Typo weblog engine, Liquid template language, Active Merchant, DelayedJob."

"他创造了 Typo 博客引擎、Liquid 模板语言、Active Merchant、DelayedJob。"

Wikipedia

每一个都有具体故事：

• Typo（约 2005）：用于学习 Rails 的博客引擎；Tobi 自述就是为了"测试 Rails 能做什么"而写的玩具，最终被 10,000+ 人安装

"Typo blogging platform: Created to test Ruby on Rails capabilities; achieved over 10,000 installs."

"Typo 博客平台：为测试 Ruby on Rails 的能力而创造，最终安装量超过 10,000。"

John Coogan, The Story of Tobi Lutke

• Liquid（约 2006）：Shopify 模板语言；后被 Mailchimp、Jekyll、众多平台采用，成为业界事实标准的"安全模板语言"。Liquid 解决的核心问题是"让非技术用户在 Shopify 模板中写表达式而不能执行任意 Ruby 代码"——这个安全沙箱设计后来变成 Liquid 被广泛采纳的原因
• ActiveMerchant（约 2007）：Ruby 库，封装多家支付网关（Authorize.Net、PayPal、Stripe 等）。Shopify 自身需要支持几十个支付网关，他抽象成开源库——后来被全球数千个 Rails 应用使用
• DelayedJob（约 2008）：异步任务队列；Shopify 内部需要后台处理订单/邮件，Tobi 写了这个并开源；后来 Resque、Sidekiq 等都受其启发

工程哲学：Ruby 的优雅审美

Tobi 在 AI + a16z（这里指广义引用——具体在 Lenny's Podcast）和 Founders 那期里反复表达对编程语言风格的态度：

"Fell in love with programming languages that didn't try to control him."

"我爱上的是那些不会试图控制使用者的编程语言。"

The Apprentice Programmer

2019 年 How I Built This 上 Tobi 给出了他为什么选 Ruby 的最形象版本——把编程语言比作画笔，把选语言比作选择用什么工具在脑中的画面转移到画布上：

"I realized I was likely going to be the only programmer working on it, so I didn't need to choose a 'popular' language like Java. I chose Ruby because it was the best tool for the job and it was fun. Coding is like painting; the language is the mechanism for translating the picture in your brain onto the canvas. Ruby matched the way my brain constructed the relationships in software, whereas other languages felt like trying to paint a masterpiece with crayons."

"我意识到自己很可能会是唯一一个写它的程序员，所以我不需要选 Java 那种'流行'语言。我选 Ruby，因为它是最趁手的工具，而且它有趣。写代码就像画画——语言是把你脑中那幅画搬到画布上的机制。Ruby 契合我大脑构建软件里各种关系的方式，而别的语言感觉就像想用蜡笔画出一幅杰作。"

How I Built This, 2019-08-05 [00:约 Q6]

这条与他 2024 年 Invest Like the Best EP.393 对 Ruby "poetry mode / 为程序员幸福感优化"的赞美（§11.6）是同一审美的两个时点——2019 年他用"蜡笔画杰作"，2024 年他用"Matsumoto 31 年前的美学决定"，指向的是同一件事：语言不是中性的，它编码了对使用者的态度。

"Coding is his happy place; will clear 3 days of calendar to build with teams until midnight."

"写代码是他的'happy place'；他会清空连续 3 天的日历，和团队一起编程到午夜。"

Lenny's Podcast, 2025-02-02

"Still actively writes code as CEO."

"即使作为 CEO，他至今依然写代码。"

a16z Podcast, 2024-12-16

CEO 仍然写代码这件事在硅谷不算极端罕见，但在公开市场公司中很罕见。Tobi 给出了他自己的理由——那是为了避免靠 briefing 做决策：

"Hands-on management style including writing code — prevents guessing and allows faster problem-solving."

"亲自写代码的管理方式——可以防止靠猜测做决策，也能更快解决问题。"

a16z Podcast

"Deep understanding of technical details prevents need for extensive briefings."

"对技术细节的深度理解，让我不需要先听一堆 briefing 才能做决策。"

a16z Podcast

Shopify 早期作为 Rails 标志案例

到 2010 年前后，Shopify 是全球最大的 Rails 应用之一。Stripe Sessions 那期里 John Collison 与 Tobi 直接讨论了 Shopify 当下的代码规模：

"Shopify codebase: ~20M lines of Ruby (core) + ~8-10M lines of TypeScript (admin); largest Ruby app and one of largest TypeScript apps in existence."

"Shopify 的代码库：核心约 2,000 万行 Ruby + 后台约 800–1,000 万行 TypeScript；这是世界上最大的 Ruby 应用，也是最大的 TypeScript 应用之一。"

Stripe Sessions, 2025-10-06

也就是说 Shopify 至今依然以 Tobi 在 2004–2006 年自学 Ruby/Rails 做出的技术决定作为基础。20 年后，他和 DHH 一起在赛车场上做 co-driver——这种连续性在科技行业极其罕见。

"Shopify as a prominent Rails success story. Tobi's contributions to Rails core team gave him credibility in the developer community. Connection to DHH (David Heinemeier Hansson) dates back to this era — they would later race together in IMSA."

"Shopify 是 Rails 的标志性成功案例。Tobi 对 Rails 核心团队的贡献，给了他在开发者社区的信誉。他与 DHH 的关系从这个年代开始，二十年后他们在 IMSA 同场赛车。"

Signal v. Noise, 约 2009

第六章：Shopify 早期艰难期 (2006–2010)

2006–2010 年是 Shopify 几乎倒闭的年代。这段时间几乎所有 Shopify 早期员工后来回忆都有一句类似的话："我们当时都不知道下个月还在不在。"

现金流崩溃与 Tobi 的隐瞒

Globe and Mail 2014 年的报道里 Tobi 自己讲过：

"I was sitting in meetings talking about year-long projects when I knew we didn't have four weeks of money left."

"我坐在会议里讨论一年期的项目，但我心里知道我们的钱撑不过 4 周。"

Globe and Mail, 2014

"Hid the company's precarious financial situation from employees."

"我把公司岌岌可危的财务情况对员工瞒下来了。"

Globe and Mail, 2014

婚礼前一天的定价模型转向（约 2007）

最戏剧化的一段是他在结婚前一天将公司业务模型从"抽销售额"改成"月订阅"：

"The day before his wedding, switched Shopify's entire business model from percentage-of-sales to monthly subscriptions. Fielded angry customer calls into the night."

"在结婚的前一天，他把 Shopify 整个商业模式从抽佣改成了按月订阅，一直到深夜还在接愤怒客户的电话。"

Globe and Mail, 2014

这次定价转向是 Shopify 增长曲线最重要的转折之一：抽佣定价让用 Shopify 增长越快的商家越亏，这是反向激励；改成月订阅后商家越成功 Shopify 也越赢，但单笔订阅的费用不会随销售额放大。

"Initial pricing model: Revenue percentage (3% of sales) — high fees discouraged growth-minded merchants. 2007: Switched to flat-rate pricing; growth accelerated."

"最初定价是抽销售额 3%——劝退了那些想做大的商家。2007 年改为定额订阅，增长立刻加速。"

John Coogan

Scott Lake 离开 + Tobi 接任 CEO（2008）

Shopify 创始 CEO Scott Lake 在 2008 年离开。当时 Shopify 月收入已经超过 $60K——但增长压力让 Scott 与 Tobi 的方向产生分歧。

"2008: Monthly revenue exceeded $60,000 when co-founder Lake departed."

"2008 年，联合创始人 Lake 离开时，月营收已经超过 6 万美元。"

Globe and Mail, 2014

Tobi 当时只想写代码，根本不想做 CEO：

"Initially resisted the business side, wanted only to code."

"我一开始拒绝处理商业那一面——我只想写代码。"

Tim Ferriss #359

但 John Phillips（前面提到的天使投资人）的一句话改变了他：

"You will never find anyone who will care about Shopify as much as you do."

"你永远找不到一个像你一样在乎 Shopify 的人。"

John Phillips（投资人对 Tobi 说的话）, Tim Ferriss #359

这句话后来被 Tobi 反复引用——它是他 2008 年下定决心做 CEO 的关键。

"几周内拿到 MBA" 的现学现卖

Tobi 形容当 CEO 那段时间像被扔进深水：

"When I took over as CEO, I had to essentially get an MBA in a couple of weeks."

"我刚接任 CEO 时，得在几周里硬学完一个 MBA。"

Globe and Mail, 2014

"Visited Silicon Valley VCs, wrote down unfamiliar terms, researched them on Wikipedia in a hostel."

"我去硅谷见 VC，把听不懂的词记下来，回到青旅在 Wikipedia 上查。"

Tim Ferriss #359

瓶颈期（2008–2009）：自己当 CEO 拖慢公司

Tobi 在 Tim Ferriss #359 中坦白他自己是公司增长的瓶颈：

"Held back company growth by making decisions too slowly. Ran on cash, avoiding marketing spend until comfortable."

"我自己当 CEO 时拖慢了公司——决策太慢。我们靠手头现金运转，因为我不敢花营销钱，直到觉得舒服为止。"

Tim Ferriss #359

这种"决策瓶颈"自我察觉，后来变成了他对 CEO 角色的核心理解——CEO 的稀缺资源是判断速度和判断质量的乘积。

在 2019 年 How I Built This 上他把这段"我把公司当程序来管"的自我诊断讲得更完整——这是他把"人不是计算机"这一课学到手的原点：

"I initially held the company back because I treated it like a computer program where I could just control every input and output. I eventually realized that humans are not like computers; they don't just do what you tell them until you stop."

"一开始我拖住了公司，因为我把它当成一个计算机程序来对待——以为我可以控制每一个输入和输出。我最终意识到，人不是计算机；他们不会像你告诉计算机的那样，一直照做直到你叫停。"

How I Built This, 2019-08-05

"When you tell computers what to do, they will keep doing it forever until you stop. It turns out humans are not like that."

"你告诉计算机做什么，它会一直做下去直到你叫停。结果发现人类不是这样。"

How I Built This, 2019-08-05 [00:01:20]

2008 金融危机作为增长催化剂

How I Built This 补上了一个此前档案只从公司编年史角度记录、缺第一人称因果解释的关键节点——2008 危机为什么反而救了 Shopify：

"We expected the market to fall off a cliff, but the opposite happened. Many people who were newly laid off decided to reach for independence and start their own businesses using Shopify. The crisis acted as a catalyst for entrepreneurship... This surge in new users helped us reach cash-flow neutrality in 2009, which allowed us to move past the 'survival' phase."

"我们本以为市场会跌下悬崖，结果恰恰相反。很多刚被裁员的人决定去争取独立、用 Shopify 开自己的生意。这场危机成了创业的催化剂……这一波新用户帮我们在 2009 年实现了现金流打平，让我们得以走出'求生'阶段。"

How I Built This, 2019-08-05

[推断] 这段"衰退→失业→创业潮→Shopify 增长"的因果，正是他 2020 年 COVID 早期押注"pandemic 永久加速电商"的认知先例——2008 年那次赌对了（衰退利好开店），2020 年他把同一模式外推到疫情，却赌错了（§九）。同一个 mental model 一次成功一次失败，是理解他 2022 年那句"I got this wrong"的重要背景。

创业者的情绪双态："unbeatable" 与 "complete dread"

HIBT 那期里 Tobi 给出了他对早期创业心理状态的最坦率描述——尤其考虑到他当时是靠家人（Fiona 父母）的钱在运转：

"I existed in two states: one where I felt we were unbeatable, and one of complete dread where I felt we were going to die tonight. The bottoms were deep, and the implications of failure were significant because I was responsible for my team's livelihoods and my family's savings... Over time, the frequency of those mood swings shifted from months to days, but the weight of the responsibility never truly went away."

"我活在两种状态里：一种是觉得我们不可战胜，另一种是彻底的恐惧、觉得我们今晚就要死了。谷底很深，失败的后果很重，因为我要为团队的生计和家人的积蓄负责……随着时间推移，这种情绪摆动的频率从以月计变成以天计，但那份责任的重量从未真正消失。"

How I Built This, 2019-08-05 [00:44:44]

这条与 §10.9 "关不掉的 background process"（2025 Shawn Ryan）是同一根神经在不同人生阶段的两次读数——早期是"unbeatable / dread"的高振幅摆动，后期收敛成一个"低优先级但关不掉"的持续进程。

"90% 是运气"：对时机的诚实归因

"I would say it's 90% luck. I had an incredibly supportive family in Canada that supported us when we had nothing. The timing was also perfect; this is not a company that could have been started even two years later. It needed to be ready exactly when the financial crisis hit. I don't know why I ended up 'winning the lottery' multiple times in a row, but the most important thing is to be honest about that reality and enjoy the journey."

"我会说这里面 90% 是运气。我在加拿大有一个极其支持我的家庭，在我们一无所有的时候撑着我们。时机也堪称完美——这不是一家哪怕晚两年都还能被创办出来的公司。它必须恰好在金融危机来袭时准备就绪。我不知道自己为什么能连着中好几次'彩票'，但最重要的是对这个现实保持诚实，并享受这段旅程。"

How I Built This, 2019-08-05 [00:约 Q8]

[推断] 这种"高度归因于运气/时机"的姿态，与 §三 主题 D "公开纠错"是同一种 epistemic humility 的两面——对成功归因于运气，对失败归因于自己（"I got this wrong"），构成一个非对称但内部一致的归因框架。

"Build a Business" 比赛押上最后 $100K（2009）

2009 年公司面对潜在破产，Tobi 做了一个反直觉的决定：把最后 $100K 用作 Build a Business 创业大赛奖金。

"Facing potential bankruptcy, bet Shopify's last $100K on a competition incentivizing entrepreneurs."

"在面对潜在破产时，我们把 Shopify 最后的 10 万美元押在了一个鼓励创业者的大赛上。"

Tim Ferriss #359

"Tim Ferriss business competition ($100K prize). Result: $3M+ revenue generated by competitors."

"和 Tim Ferriss 联办的商业大赛（10 万美元奖金），结果参赛者总共创造了 300 万美元以上的销售额。"

John Coogan, The Story of Tobi Lutke

获胜者 DODOcase（用复古装订工艺做 iPad 套）成了 Shopify 之后最常被引用的成功客户故事——它示范了 Shopify 如何把一群不会编程的工匠变成在线零售商。

"Winners like DODOcase (iPad cases from revived bookbinders) created ripple effects beyond prize money. Winning entrepreneurs shared storytelling sophistication, not just product genius."

"DODOcase 这样的获胜者（用复活的书籍装订工艺做 iPad 套）创造了远超奖金的涟漪效应。获胜创业者的共性不是产品天才，而是会讲故事的能力。"

Tim Ferriss #359

"High Output Management (Andy Grove) — positioned business as engineering exercise, made CEO role less scary."

"Andy Grove 的 High Output Management 把'经营公司'重塑成'工程练习'——这让 CEO 这个角色变得没那么可怕了。"

Tim Ferriss #359

"First year results: Team grew to 10 people, $1M revenue, cash flow positive. Early customer: Tesla."

"他接任 CEO 第一年的结果：团队 10 人、年营收 100 万美元、现金流转正。早期客户之一是 Tesla。"

John Coogan, The Story of Tobi Lutke

第七章：增长与上市 (2010–2015)

2010 是 Shopify 从"勉强活下来"到"开始 hyper-scale"的转折年。

Series A: $7M from Bessemer（2010 年 12 月）

"December 2010: Series A — $7M from Bessemer Venture Partners, FirstMark Capital, Felicis Ventures."

"2010 年 12 月，A 轮 700 万美元，由 Bessemer Venture Partners、FirstMark Capital、Felicis Ventures 投资。"

Wikipedia (Shopify)

"October 2011: Series B — $15M. December 2013: Series C — $100M."

"2011 年 10 月 B 轮 1500 万美元；2013 年 12 月 C 轮 1 亿美元。"

Wikipedia (Shopify)

Bessemer 的 Jeremy Levine 后来在 Globe and Mail 给 Tobi 的评价是：

"We certainly think Tobi is in that same class as founders of Skype, LinkedIn, Yelp."

"我们确信 Tobi 属于和 Skype、LinkedIn、Yelp 创始人同一档次的人。"

Jeremy Levine (Bessemer), Globe and Mail, 2014

故意慢下来：5 个营销实验都成功（2009–2010）

Founderoo 的 Shopify Playbook 详细记录了这一段：

"2009-2010: Critical turning point — intentional slowdown to test 5 marketing programs (all succeeded). Rather than chase growth blindly, Lütke systematized marketing experimentation. Set aside money to try five different marketing programs. Every single marketing idea worked — validating need for aggressive expansion."

"2009–2010 是关键转折期：他刻意放慢扩张，先做 5 个营销实验（全部成功）。他没有盲目追求增长，而是把营销实验系统化——专门留出预算尝试 5 种不同的营销路径。每一个都见效——这反而证明真正的瓶颈是资本，不是需求。"

Founderoo, 2024

"Once venture capital arrived, every test campaign worked — problem was capital constraint, not product-market fit."

"VC 钱进来之后，每一次营销实验都成功了——所以瓶颈是资本约束，不是 product-market fit 的问题。"

Tim Ferriss #359

这是 Tobi 决策模式的早期典型：先小规模做实验把不确定性变成确定性，再大规模复制——而不是相反。

App Store + Theme Store + 平台战略（2009–2010）

"June 2009: API platform and App Store."

"2009 年 6 月发布 API 平台和 App Store。"

Wikipedia (Shopify)

App Store 与 Theme Store 把 Shopify 从一个"软件产品"变成了一个"平台"。Tobi 在 Masters of Scale 那期里给出他对 platform 的核心信条：

"I believe the secret to massive scale is compressed in three words: Be a platform."

"我相信大规模扩张的秘诀，可以浓缩成三个词：be a platform。"

Masters of Scale, 2020-12-15

"Bill Gates: 'You are not a platform until the people who are building on you make more money than you do."

"Bill Gates 说：'除非在你之上构建的人挣到的钱比你多，否则你算不上一个平台。"

Masters of Scale

"Developers earn more collectively than Shopify charges in subscriptions."

"开发者一起赚到的钱已经超过 Shopify 收的订阅费。"

Masters of Scale

Shopify Payments + POS（2013）

"August 2013: Shopify Payments (Stripe partnership); point-of-sale system."

"2013 年 8 月发布 Shopify Payments（与 Stripe 合作），同期发布 POS 系统。"

Wikipedia (Shopify)

Shopify Payments 是与 Stripe 深度合作的产物——这是 Tobi 与 John & Patrick Collison 合作的开端。这条合作线后来在 13 年里没有断裂，至今仍是 Shopify 重要的支付通道：

"13-year partnership maintained through focus on respective core competencies."

"我们这段 13 年的合作之所以维持下来，是因为各自专注在各自的核心能力上。"

The Cheeky Pint, 2025（Tobi 与 John Collison）

Shopify Plus 启动（2014 年 2 月）

"February 2014: Shopify Plus for enterprise."

"2014 年 2 月发布面向企业级的 Shopify Plus。"

Wikipedia (Shopify)

Shopify Plus 是大客户线，这条线后来撑起了 Shopify 在 2020 年代主营收的大头（详见第八、第十章）。

Globe and Mail "CEO of the Year"（2014）

"2014: Named 'CEO of the Year' by The Globe and Mail."

"2014 年，被《Globe and Mail》评为加拿大年度 CEO。"

Wikipedia

那篇报道（本档案 source 07）至今是关于 Tobi 早期最详细的人物特写——它捕捉到了 IPO 前夜的状态：

"Short and boyishly slender, almost elfin in appearance, typically dressed in shirt sleeves or hoodies. Commutes by bicycle rather than driving his Tesla. A quiet, studious introvert who enjoys media attention as much as an owl likes the midday sun."

"身材矮小、男孩气、近乎精灵形象，通常穿着衬衫或卫衣。骑自行车上下班，宁愿不用他的 Tesla。是个安静、好学的内向者——他对媒体关注的兴趣，就像猫头鹰对正午太阳的兴趣一样大。"

Globe and Mail, 2014

IPO（2015 年 5 月 21 日）

"May 21, 2015: IPO raising over $131M; stock jumped 51-65%."

"2015 年 5 月 21 日 IPO，募资超过 1.31 亿美元，上市当日股价上涨 51–65%。"

Wikipedia (Shopify)

IPO 估值 $1.27B（每股 $17）。Tobi 写了一封 IPO 致股东信（本档案 source 22, 43）作为公司公开身份的奠基文件：

"Make commerce better for everyone."

"让商业对每一个人都变得更好。"

Shopify IPO Letter, 2015

"We do not chase revenue as the primary driver of our business."

"我们不把收入当作业务的主要驱动力。"

Shopify IPO Letter, 2015

"Simplicity isn't simple. It takes tremendous care, discipline, and craftsmanship."

"简单一点也不简单。它需要巨大的关怀、纪律和工艺。"

Shopify IPO Letter, 2015

"I want Shopify to be a company that sees the next century."

"我希望 Shopify 是一家能看到下个世纪的公司。"

Shopify IPO Letter, 2015 / We need to talk about carbon blog, 2019

Stratechery 的战略框架（2022 年 Ben Thompson 总结）

Ben Thompson 在 2022 年回顾 Shopify 的战略演化（本档案 source 09）时给出了三大支柱：

1. Shop Pay —— 第一次把 Shopify 推向消费者品牌触点（统一支付/送货信息），让一键支付变得可能
2. Fulfillment Network —— 从对接第三方物流，转向自建仓储；承诺到 2024 年每年投资 $1B
3. Advertising —— 从依赖 Facebook 广告，转向用聚合的转化数据成为大规模广告购买方

"As Shopify becomes increasingly reliant on Facebook and Amazon, must differentiate through vertical integration — payments, fulfillment, advertising."

"随着 Shopify 越来越依赖 Facebook 和 Amazon，它必须通过纵向整合（支付、履约、广告）实现差异化。"

Ben Thompson, Stratechery, 2022

第二条（Fulfillment）后来在 2022 年走向极端、并在 2023 年被 Tobi 亲手 unwind（详见第九章）。

"Our entire motto becoming a public company was: We are not going public. We are creating a public version of Shopify."

"我们成为上市公司时的整体口号是：我们不是去 go public，我们是在创造一个 public 版本的 Shopify。"

Founders (David Senra), 2026-01-18

2026 年 5 月的 20VC 访谈补上了 IPO 定价过程的一段重要细节：路演前他研究了投行的薪酬结构，得出的结论是银行家的真正客户是 order book 里的买方机构——

"I realized early on that the investment bankers were not my customers; I was their crop... I told them, 'I understand how you get your bonus, and I want you to get it, but I am becoming a public company CEO, so I am going to optimize for the company."

"我很早就意识到，投资银行家不是我的客户；我是他们的庄稼……我告诉他们：'我知道你们的 bonus 怎么来，我也希望你们拿到，但我即将成为上市公司 CEO，所以我要为公司优化。"

20VC, 2026-05-04

他把发行价定到 $14–16 区间之上（最终 $17），首笔交易比发行价高 $10。当年主导 Shopify IPO 的 Morgan Stanley 银行家 Jeff Hoffmeister 自 2022 年 10 月起出任 Shopify CFO——他在同一访谈里说 "the banker who took me public is my CFO today"。

第八章：Build Phase 2 — Shop Pay, Mobile, Plus (2015–2020)

2015–2020 是 Shopify 全速做平台扩张的时段。这个时期最重要的几条主线：Shop Pay、Plus 的企业化、Sustainability、Ottawa 的"加拿大科技救星"叙事。

Shop Pay 上线（2017 年 4 月）

"April 2017: Shop Pay (renamed from Shopify Pay in 2020)."

"2017 年 4 月，Shop Pay 上线（2020 年从 Shopify Pay 改名）。"

Wikipedia (Shopify)

Shop Pay 是 Shopify 第一次直接面对终端消费者的产品。Stratechery 给出的框架：

"Shopify's first major consumer brand touchpoint. Pre-populating payment and delivery information reduces friction to single-tap purchases. Establishes direct customer relationships beyond just merchant-facing software."

"Shopify 第一次的重要消费者品牌触点。预填支付和送货信息，把购买摩擦减到一键。它让 Shopify 与终端消费者建立直接关系，而不仅是面向商家的软件。"

Ben Thompson, Stratechery, 2022

后来 The Cheeky Pint 那期 Tobi 自述 Shop Pay 现在的转化率优势：

"Shop Pay alone delivers 16% conversion increase."

"仅 Shop Pay 一项就能带来 16% 的转化率提升。"

Tobi Lütke, The Cheeky Pint, 2025

Build a Bigger Business 等大型平台计划

Shopify 在这段时期持续推动 platform DNA：扩展 App Store（2009 起），Theme Store（2010 起），开发者收入分成。后来在 2021 年走出极端慷慨的一步：

"2021: Removed 20% revenue share for developers earning under $1M annually."

"2021 年起，对年收入低于 100 万美元的开发者，免收 20% 的分成。"

Wikipedia (Shopify)

"2024: Over 10,000 apps; by 2025: 16,000+ apps."

"到 2024 年 App Store 超过一万个 App；2025 年超过 16,000 个。"

Wikipedia (Shopify)

"Developer earnings: Over $230M collectively in 2020."

"2020 年，开发者集体从 Shopify 平台赚到的钱超过 2.3 亿美元。"

Wikipedia (Shopify)

2018 Meritorious Service Cross + 2019 Bloomberg "加拿大科技救星"

"2018: Received Meritorious Service Cross."

"2018 年获得加拿大 Meritorious Service Cross 勋章。"

Wikipedia

"2019: Donated over $1 million to Team Trees (largest donor)."

"2019 年向 Team Trees 捐款超过 100 万美元（单笔最大捐赠人）。"

Wikipedia

2019 年 12 月 Bloomberg 把 Tobi 称为加拿大 tech 圈的"savior"（本档案 source 35）。Shopify 在 2020 年初一度超过加拿大皇家银行（Royal Bank of Canada）成为加拿大市值最高的公司。

Sustainability Fund + 100 年视角的公开化（2019 年 12 月）

Tobi 2019 年 12 月 22 日在自己的博客 Too-biased 上发了一篇关于碳的长文：

"I want Shopify to be a company that sees the next century."

"我希望 Shopify 是一家能看到下个世纪的公司。"

Tobi Lütke, We need to talk about carbon, 2019-12-22

"Shopify Sustainability Fund: $5M+ annual commitment to carbon sequestration. Focus on emerging technologies for carbon removal."

"Shopify Sustainability Fund：每年承诺 500 万美元以上用于碳封存，重点支持新兴的碳清除技术。"

Tobi Lütke, We need to talk about carbon, 2019-12-22

这是科技行业最早一批主动承诺为新兴 carbon removal 技术付费的企业基金；它后来催生了 Stripe Climate / Frontier 等更大型的 advance market commitment。Tobi 与 Patrick Collison 在这一议题上互为同盟。

"Shopify went remote: Ottawa couldn't support 10,000-person company; hybrid was worse than all-remote (unequal experience); COVID accelerated a decision Tobi would have made within 1-2 years anyway."

"Shopify 全面远程化：Ottawa 撑不起一万人公司；混合模式比纯远程还糟（体验不平等）；COVID 把一个我本来 1–2 年内就要做的决定提前了。"

Lenny's Podcast, 2025-02-02

"We choose to jump in the driver's seat, instead of being passengers to the changes ahead. We cannot go back to the way things were. This isn't a choice; this is the future."

"我们选择跳到驾驶座，而不是当眼前这些变化的乘客。我们回不去了。这不是一个选择，这就是未来。"

Tobi Lütke (Twitter), 2020-05-21

四年后在 Invest Like the Best EP.393 上 Tobi 把这次反转的 mental model 浓缩成一句话——它揭示了他对远程办公的判断不是"对疫情的妥协"，而是一次主动的方向倒转：

"Don't port the office online. Let's port the internet into a company. And then we're like, let's go."

"不要把办公室搬到网上。我们要做的是把互联网搬到公司里。然后——就开始干。"

Invest Like the Best EP.393, 2024-10-22

行业里所有公司当时的默认 mental model 是"office is the center, internet is the bridge that brings remote workers in"。Tobi 这一句把整个隐喻反过来。

Acquisitions: 6 River Systems + Deliverr（埋下 logistics 雷）

"2019: 6 River Systems ($450M)."

"2019 年以 4.5 亿美元收购 6 River Systems。"

Wikipedia (Shopify)

6 River Systems 是仓库机器人公司，购入后成为 Shopify Fulfillment Network 的硬件后端。这是 Shopify 第一次大规模进入"重资产 logistics"——这条线后来在 2022 年因 Deliverr 收购（21 亿美元）扩大成赌注，2023 年又被全部卖给 Flexport（详见第九、十章）。

第九章：COVID 高峰与战略误判 (2020–2022)

2020–2022 是 Shopify 的"高峰—戛然而止"双重时刻。这段时间是 Tobi 公开承认自己策略失误的关键窗口。

2020 年 COVID 电商爆发：Shopify 月增 62%

"62% surge in new store registrations within six weeks of pandemic onset."

"疫情爆发后六周内，新店注册量激增 62%。"

Fortune, 2020-05-22

"Particular growth among unexpected players: grocers and farmers going digital. Significant order growth from customers within 15 miles of physical stores — hyperlocal e-commerce potential."

"意外的增长来自杂货商和农户开始数字化。来自实体店 15 英里范围内消费者的订单增长尤为强劲——表明 hyperlocal 电商的潜力。"

Fortune, 2020-05-22

Shopify 那段时间发布了一系列产品：

• Shopify Balance：商户银行卡 + 现金流工具
• Curbside Pickup：路边自提
• Shopify Fulfillment Network：履约网络快速 buildout

双倍员工 + 大幅 hire（2020–2021）

"Shopify had more than doubled its employee base since end of 2019."

"Shopify 自 2019 年底以来员工数翻了一倍以上。"

Globe and Mail + Bloomberg, 2022-07-26

Bloomberg Businessweek 12 月封面 + 200 亿美元的 "arming the rebels"（2021）

2021 年 12 月 Bloomberg Businessweek 推出 Shopify 封面故事，提出了那个标志性的句子：

"Amazon is trying to build an empire, and Shopify is trying to arm the rebels."

"Amazon 在试图建立一个帝国，Shopify 在试图武装反叛者。"

Bloomberg Businessweek, 2021-12-23

文章发表时 Shopify 估值约 $200B，Tobi 净资产数十亿美元，加拿大科技板块以 Shopify 为核心。但 Bloomberg 在后续报道中自述这是一次 "cover curse"：

"Shopify's CEO Was on the Cover of Businessweek. Then Its Stock Fell 75%."

"Shopify 的 CEO 上了《Businessweek》封面，然后股价跌了 75%。"

Bloomberg, 2022

Founder Share controversy（2022 年 6 月）

2022 年 6 月 Shopify 通过的 triple-class share 提案给 Tobi 个人 40% 投票权，即使他持股稀释到 1.1% 也不变（详见第十一章 / Critique）。

"Created triple-class share system giving Tobi Lütke 40% voting control indefinitely. Despite owning only 6.2% of company equity. Even if equity stake diluted to as low as 1.1%, voting power maintained as long as he remains with Shopify."

"三类股结构，让 Tobi 永久持有 40% 投票权。他持股仅 6.2%。即便股份稀释到 1.1%，只要他仍在 Shopify，投票权不变。"

Harvard Law / Glass Lewis 分析, 2022-06

"53.68% of non-Lütke shareholders voted in favor. 46.32% voted against — nearly half opposed."

"排除 Tobi 自己以外的股东中，53.68% 赞成、46.32% 反对——几乎一半反对。"

Harvard Law / Glass Lewis 分析, 2022-06

2022 年 7 月 26 日：第一次裁员 + 公开承认赌错（10%, 约 1,000 人）

"Cut 10% of workforce, about 1,000 people. Shares sank roughly 16% on Toronto Stock Exchange. Stock had fallen more than 75% year-to-date by that point."

"裁员 10%，约 1,000 人。Toronto Stock Exchange 当日股价下跌约 16%。当时年初至当日累计跌幅已超过 75%。"

Globe and Mail, 2022-07-26

裁员邮件中 Tobi 写下了他之后被反复引用的承认：

"Shopify had bet that increasing mix of online spending would 'permanently leap ahead by 5 or even 10 years.' It's now clear that bet didn't pay off — consumer online shopping habits are reverting to roughly pre-COVID trajectory."

"Shopify 之前赌：在线支出占比会'永久性地向前跳 5 年甚至 10 年'。现在明显这个赌不成立——消费者的网购习惯已经大致回到 COVID 前的轨道。"

Tobi Lütke (内部备忘), 2022-07-26

"Ultimately, placing this bet was my call to make and I got this wrong."

"归根到底，做这个赌注是我的决定，我决策错了。"

Tobi Lütke (内部备忘), 2022-07-26

关于股价的态度：fair market value ≠ stock price

20VC 那期里 Tobi 谈过 2021–2022 股价剧烈波动期间他对内部的沟通：

"The stock price is not the company. We all work on something called the fair market value of Shopify, which is like the sum total of products and skill sets of people."

"股价不是公司。我们所有人在做的事是 Shopify 的'fair market value'——它等于产品总和加员工技能总和。"

20VC #6727, 2023-04-03

"Used period to restructure executive team and return to hands-on project review. Implemented GSD (Get Stuff Done) system reviewing all projects against product principles."

"我利用这段时期重组高管团队，回到亲自做项目复盘。我推动了 GSD 系统，按产品原则复盘所有项目。"

Acquired ACQ2, 2025-09-18

反思的范本：Founders 那期里"Shopify went bankrupt and I bought it on a fire sale"

2026 年 1 月 Founders (David Senra) 那期里，Tobi 把 2022 年的精神状态形容为：

"Shopify went bankrupt and I bought it on a fire sale and I'm marching in on day one here — that's the energy I brought to the post-crisis reset."

"我当时的内心戏是：Shopify 破产了，我把它从清仓拍卖买了下来，今天是我作为新主进场的第一天——这就是我带给危机后 reset 的能量。"

Founders (David Senra), 2026-01-18

"Post-IPO, discovered 60% of internal projects had gone off-mission; canceled them and turned over every executive over the course of a year."

"上市之后我发现 60% 的内部项目其实已经偏离使命；我把它们全部停掉，并在一年里换掉了每一位高管。"

Founders (David Senra), 2026-01-18

"Personally reviewed every project in 16-hour days, canceled ~60% of projects, turned over all executives within a year."

"我亲自用每天 16 小时复盘每一个项目，砍掉了大约 60%，一年内换掉了所有高管。"

Founders (David Senra), 2026-01-18

第十章：Calm Progress 时代 (2023–2026)

2023 年起 Shopify 进入"主动收缩 + 重新聚焦"的阶段。Tobi 把这个时期的核心精神命名为 calm progress——和"快速增长"对立，强调可持续、复利、不被业绩驱动 panic。

2023 年 1 月 3 日：Chaos Monkey 删 12,000 个例会

"Eliminated all recurring meetings with more than 2 participants. No meetings scheduled on Wednesdays (company-wide). Large meetings (50+ people) limited to Thursdays, 11 a.m.–5 p.m. ET. Maximum one large meeting per week. 2-week moratorium on adding new meetings to calendars."

"取消所有 2 人以上的固定例会；周三全公司禁止安排会议；50 人以上大型会议仅限周四 11:00–17:00 ET，每周最多一次；新会议加入有 2 周禁令。"

Fortune, 2023-01-03

"12,000 calendar events deleted at a stroke (10,000 recurring + 2,000 ad hoc). Projected cuts: 322,000 hours and 474,000 discrete events for all of 2023. By March, tracked 25% more projects toward completion."

"一次性删掉 12,000 个日历事件（10,000 个固定例会 + 2,000 个临时会议）。预测 2023 年累计减少 32.2 万小时、47.4 万次离散会议。到 3 月，追踪到完成项目数增加 25%。"

Fortune, 2023-01-03

"The best thing founders can do is subtraction. It's much easier to add things than remove things. If you say yes to a thing, you actually say no to every other thing you could have done with that period of time."

"创始人能做的最好的事是 subtraction。增加东西比拿掉东西容易得多。你对一件事说 yes，其实就是对那段时间能做的所有其它事情说 no。"

Tobi Lütke, Fortune, 2023-01-03

"It seems very hard for people to subtract from things, while it's very easy to add them — which only leads one way."

"对人来说从既有的东西里减一项似乎非常难，而加一项很容易——这只会导向一个方向。"

Tobi Lütke, Bloomberg, 2023-01-03

COO Kaz Nejatian 同期补充：

"Meetings are a bug. We can either go slow and deliberate, or fast and chaotic. We are going fast and chaotic."

"会议是 bug。我们要么慢而谨慎，要么快而混乱。我们选快而混乱。"

Kaz Nejatian (VP Product & COO), Fortune, 2023-01-03

2023 年 5 月 4 日：Logistics 卖给 Flexport + 第二次大裁员（约 20%, 约 2,000 人）

这次反转之所以引人注目，部分原因在于 Tobi 一年前还在公开为 SFN 战略辩护。2022 年 5 月在 Invest Like the Best EP.278 上他这样描述当时的方向：

"Now we are building warehouses and robots and all these kinds of things. I didn't think that would be part of the journey, but it clearly is."

"现在我们正在建仓库、造机器人、做所有这些事情。我从没想过这些会成为这段旅程的一部分，但显然就是这样了。"

Invest Like the Best EP.278, 2022-05-24

这是 Tobi 对 SFN 战略反转前一年的精确定格——一年后他公开把 Logistics 卖给 Flexport。同一期他自己也给出了反沉没成本的纪律预设——这条纪律最终在 2023 年被用在了他自己一年前坚持的方向上：

"Shopify is not immune to a sunk cost fallacy, but I think it's better than most companies. We don't really have sacred cows in these ways. If something makes sense, it makes sense."

"Shopify 并非对沉没成本免疫，但我认为我们比大多数公司做得好。我们没有什么神圣不可碰的事物。如果一件事说得通，它就说得通。"

Invest Like the Best EP.278, 2022-05-24

Tobi 在内部备忘 Important Team and Business Changes 中提出了"main quest vs side quest"框架（本档案 source 10）：

"Main quest' = Shopify's core mission: democratizing and simplifying commerce. 'Side quests' = distracting initiatives that dilute focus. Larger companies accumulate distractions during stable periods, then struggle to adapt when conditions change. Logistics was 'clearly a worthwhile side quest' but had become one nonetheless."

"Main quest' 是 Shopify 核心使命：让商业平民化、变简单。'Side quest' 是稀释焦点的项目。大公司在平稳期积累干扰，环境变化时就难以适应。物流'显然是一个值得做的 side quest'——但它仍然只是一个 side quest。"

Tobi Lütke 内部备忘, 2023-05-04

"Workforce reduced by approximately 20% (~2,000 people)."

"裁员约 20%，约 2,000 人。"

Shopify 公开公告, 2023-05-04

"Shopify Logistics sold to Flexport. Shopify received 13% equity interest in Flexport plus existing equity."

"Shopify Logistics 卖给 Flexport。Shopify 获得 Flexport 13% 股权，加之前已有的股权。"

Shopify 公开公告, 2023-05-04

具体被卖出的资产：

"Deliverr e-commerce fulfillment technology. 6 River Systems warehousing robotics unit (sold separately to Ocado). People, technology, and services related to logistics operations."

"Deliverr 履约技术，6 River Systems 仓储机器人（独立卖给 Ocado），以及物流业务相关的人员、技术和服务。"

TechCrunch, 2023-05-04

时间线对比：

"2019: Acquired 6 River Systems for $450M. 2022: Acquired Deliverr for $2.1B. 2023: Sold both units — massive strategic reversal within 1-4 years of acquisition."

"2019 年 4.5 亿美元收购 6 River Systems；2022 年 21 亿美元收购 Deliverr；2023 年两块都卖出——在 1–4 年内做了重大战略反转。"

TechCrunch, 2023-05-04

NPC 风波：游戏隐喻引发批评

TechCrunch 同日的批评文章（本档案 source 18）指出：

"CEO used video game terminology ('main quests' vs 'side quests') to explain cuts. Described affected employees' work as 'figuring out which weren't contributing to the main game story'. Author argued this inadvertently reduced laid-off workers to non-player characters (NPCs)."

"CEO 用电子游戏的术语（'主线任务' vs '支线任务'）解释裁员。把被裁员工的工作描述成'判断哪些不在主线故事里'。作者批评这种说法把他们贬为 NPC（非玩家角色）。"

TechCrunch, 2023-05-04

"Failed to express regret or acknowledge his own past strategic choices. Unlike other CEOs handling similar reductions, didn't take sufficient personal responsibility. Second major layoff in less than a year."

"他没有表达足够的遗憾，也没有承认自己过去的战略选择是关键原因。与处理类似裁员的其他 CEO 相比，他承担的个人责任不足。这是他不到一年内的第二次大规模裁员。"

TechCrunch, 2023-05-04

2025 年 3 月 20 日：AI 备忘录（公开 4 月 7 日）

这是 2025 年最被广泛传播的 CEO 备忘录之一。2025 年 5 月 HR Heretics 那期对 memo 做了行业层面的诊断分析，并把 memo 的核心四条结构化提炼出来：

"Reflexive AI usage is now a baseline expectation, not a suggestion, not encouraged, but expected. AI must dominate the prototype phase of all get-stuff-done projects. Teams must prove AI cannot do the work before requesting headcount, and everyone means everyone, including the bosses, including Lutke and his executive team."

"反射性使用 AI 现在是基础期望，不是建议、不是鼓励——是要求。AI 必须主导所有 GSD 项目的原型阶段。任何团队在申请增员之前，必须先证明 AI 无法完成这件工作。所有人都意味着所有人——包括 boss、包括 Lütke 本人和他的执行团队。"

HR Heretics, 2025-05-01（对 memo 的结构化提炼）

"Using AI effectively is now a fundamental expectation of everyone at Shopify."

"在 Shopify 有效使用 AI，已经是对每个人的基本期望。"

Tobi Lütke 内部备忘, 2025-03-20

"Frankly, I don't think it's feasible to opt out of learning the skill of applying AI in your craft."

"坦白说，我不认为'选择不学'怎么把 AI 应用到自己的职业'是个可行选项。"

Tobi Lütke 内部备忘, 2025-03-20

"Teams must demonstrate why they cannot get what they want done using AI before requesting additional headcount."

"团队必须先证明 AI 无法完成他们想做的事，然后才能申请额外的招聘 headcount。"

Tobi Lütke 内部备忘, 2025-03-20

"In a company growing 20-40% year over year, you must improve by at least that every year."

"在一个年增长 20–40% 的公司里，你每年至少必须提升那么多。"

Tobi Lütke 内部备忘, 2025-03-20

"Reflexive AI usage is now a baseline expectation at Shopify."

"对 AI 的反射式使用，已经是 Shopify 的基线期望。"

Tobi Lütke 内部备忘, 2025-03-20

2025 年财务恢复

"2024: $8.88B revenue (25.78% YoY). 2025: $11.556B revenue (30.14% YoY). GMV: $378.44B in 2025."

"2024 年营收 88.8 亿美元（同比 +25.78%）；2025 年营收 115.56 亿美元（同比 +30.14%）；2025 年 GMV 3784.4 亿美元。"

Shopify Q4 2025 Financial Results

"Stock surged 38% in 2024. Company launched $2B buyback program. Shopify's 8.6% share of U.S. retail e-commerce sales. Market cap approaching $200B."

"2024 年股价上涨 38%；公司启动 20 亿美元回购计划；Shopify 占美国零售电商销售约 8.6%；市值接近 2,000 亿美元。"

Shopify Q4 2025 Financial Results

"Transferred U.S. listing from NYSE to Nasdaq (March 2025). Became Nasdaq-100 component (May 2025)."

"2025 年 3 月美股上市从 NYSE 转到 Nasdaq；2025 年 5 月成为 Nasdaq-100 成分股。"

Wikipedia (Shopify)

2026 年上半年：AI 内化进入深水区（River agent、50%+ AI 代码）

2026 年 5 月先后两场深度访谈（20VC 2026-05-04、Uncapped with Jack Altman 2026-05-20）给出了 AI memo 一年后的落地刻度：Shopify 内部超过 50% 的代码已由 AI 生成且占比仍在上升；按 Tobi 的说法，最好的一批工程师从 2025 年 12 月起没有再手写过代码。工程组织里出现了一个新"居民"：

"We have an AI agent named 'River' that lives in Slack and handles a massive amount of our engineering. A monorepo is a 'world,' and rivers shape the world."

"我们有一个叫 'River' 的 AI agent，住在 Slack 里，承担了我们工程工作中相当大的量。Monorepo 是一个'世界'，而河流塑造世界。"

20VC, 2026-05-04

他自己写代码的方式也转向 "steering"：聚焦数据结构与数据如何持久化到磁盘，其余交给 AI 构建。组织机制上的配套：绩效评估转向 "net impact"（用 AI 放大产出的人客观上影响力更大）、unlimited token 政策鼓励全员 tinker、全公司六周一轮的 review cycle 作为 pace floor：

"Pace must be induced, otherwise, it is received... if you don't do that, you end up being run by quarterly PowerPoints, which is a sign that you are in trouble."

"节奏必须被主动注入，否则就只能被动接受……如果你不这么做，最终就会被季度 PowerPoint 牵着走——那是公司出问题的信号。"

Uncapped with Jack Altman, 2026-05-20

回看 2025 年那封 AI memo 的推行机制，他在 Uncapped 里给出了一个此前材料中缺失的自我解释——为什么他可以不经多年文化铺垫直接强推：

"I used my 'founder credibility tokens'—the social credit built up over 20 years—to push through a major shift. When AI became clearly transformative, I knew I had to act. I didn't want to wait for years of small cultural changes; I used a memo to set a new pace."

"我动用了我的'创始人信用代币'——20 年积累的社会信用——来推动一次重大转变。当 AI 的变革性变得清晰时，我知道我必须行动。我不想等待多年的渐进文化变化；我用一封 memo 直接设定了新节奏。"

Uncapped with Jack Altman, 2026-05-20

"People overestimate founders, but they massively underestimate what you can do when the founder is still present and in charge."

"人们高估了创始人，但他们严重低估了当创始人仍然在场、仍然掌舵时你能做成的事。"

Uncapped with Jack Altman, 2026-05-20 [00:00:00]

2026 年 5 月 27 日：Toronto Tech Week——"the one-person company is bullshit"

Toronto Tech Week 2026 Homecoming 上，Tobi 与 Lulu Cheng Meservey（Rostra 创始人、Shopify 董事会成员）在 700+ 听众面前做 fireside（播客版 2026-06-04 上线，开场是 "The Era of the Entrepreneur" 主题短片）。最被引用的表态是对"一人独角兽"流行叙事的直接反对——他承认 AI 让单人 billion-dollar 公司在技术上成为可能，但质疑其合理性：

"I think the one-person company is bullshit. Why the fuck would you not spend some of that money to have someone else around?"

"我认为一人公司是扯淡。你他妈为什么不花掉其中一点钱，让身边多一个人呢？"

Tobi Lütke, Toronto Tech Week Homecoming, 2026-05-27（BetaKit 报道）

"The shape of companies is going to change. Companies will be smaller, but there will be vastly more of them."

"公司的形态将会改变。公司会更小，但数量会多得多。"

Tobi Lütke, Toronto Tech Week Homecoming, 2026-05-27（BetaKit 报道）

他举 Wispr（AI 语音听写）为例：从 7 名员工扩到约 60 人、朝 ~$2B 估值走。对加拿大的话术延续 "go-for-bronze culture" 批评，但落点从批判转向动员：

"There's a meta game of global companies [that] exists, and you plug in from wherever you are. There are some places which make it harder. Canada is not one of them."

"存在一个全球公司的元游戏，你可以从任何地方接入。有些地方会让这件事更难。加拿大不在其列。"

Tobi Lütke, Toronto Tech Week Homecoming, 2026-05-27（BetaKit 报道）

[推断] 这与他 2024 年 Conversations with Tyler、2026 年 5 月 20VC 里的 beaver-pelts 批判并不矛盾：对加拿大政策与文化照旧尖锐（"it is a choice not to build"），但对创业者个体的 message 是"位置不是借口"。

这场 fireside 的完整音频（podwise id 8206351）已于 2026-06-26 处理完成，补回若干此前只有 BetaKit 二手转述的 Tobi 逐字——它们把这次活动从"一人公司金句"扩成一篇完整的创业哲学：

"Most people compete in the middle of the bell curve... If you aim for 'best,' there is actually surprisingly little competition because most people don't allow themselves to go for it. You can negotiate the world... but you can never compromise on being the best at that specific thing."

"大多数人在钟形曲线的中间竞争……如果你瞄准'最好'，竞争反而出奇地少，因为多数人不允许自己去争。你可以谈判'世界'（先在某个 niche 做到最好），但你永远不能在'做到最好'这件事本身上妥协。"

Tobi Lütke, Toronto Tech Week fireside, 2025-06 [Q1]

"I am unable to learn the solutions to any problem that I have never had. You can trick yourself into having all problems, at which point you can learn everything."

"我无法学会任何我从未遇到过的问题的解。你可以'骗'自己去拥有所有问题——到那一步，你就能学会一切。"

Tobi Lütke, Toronto Tech Week fireside, 2025-06 [00:04:14]

他在这里第一次把 Shopify 的起家归因于 craftsmanship（手艺），并把它对立于"北美学术-功利"传统——这条与 §二「关于产品」的设计哲学直接咬合：

"The killer feature of Shopify in 2006 wasn't that it had the deepest features; it was that it was beautiful and legible. We tamed the complexity of business through design."

"Shopify 在 2006 年的杀手级特性不是功能最深，而是它美且可读。我们用设计驯服了商业的复杂度。"

Tobi Lütke, Toronto Tech Week fireside, 2025-06 [Q4]

"Craftsmanship is about making things for real humans. It is about putting in the reps... Ornamentation isn't frivolous; it's deeply human and invites people to explore."

"手艺是为真实的人做东西，是不断地 putting in the reps（做一千个让第一千零一个更好）……装饰并不轻浮，它深刻地属于人性，邀请人去探索。"

Tobi Lütke, Toronto Tech Week fireside, 2025-06 [Q4]

以及他反"永久底层阶级"悲观论的乐观主义，落在一个可证伪的口径上：

"Every period of history that people describe as a golden age is simply a time where GDP grew 4% or faster."

"历史上每一个被称为'黄金时代'的时期，本质就是 GDP 增长达到或超过 4% 的时期。"

Tobi Lütke, Toronto Tech Week fireside, 2025-06 [00:44:56]

（"一人公司是扯淡"在完整音频里的版本更克制——"It is possible to build a billion-dollar one-person company today, but why would you not spend some of that money to have people around you? Life is too short."，与 §9.13 "小团队、不是无团队"互为印证。）

⚠️ 披露性语境（2026-08-05 补记）：这场 fireside 的提问方 Lulu Cheng Meservey 自 2024 年 6 月起即为 Shopify 在任董事（2026-06-16 AGM 连任确认）。也就是说这不是独立记者提问，而是一位董事在公开场合与 CEO 对谈。本档案仍采用其中的 Tobi 逐字，但读者应把提问的选择与追问的力度放在这一语境下理解。

2026 年 6 月 16 日：AGM——97.72% 连任，但 say-on-pay 只有 81.24%

2026 年股东大会十位董事全部当选，Tobi 以 97.72% 支持率连任并保留 Chair 职务。两个值得记的数字：

• Say-on-pay 仅 81.24% 赞成（18.76% 反对）——对一家业绩强劲的公司而言偏软。
• 一项要求 Shopify 制定 AI 政策的股东提案以 13.86% 赞成 / 86.14% 反对被否决。

[推断] 后一个数字值得与 §六.6（AI memo 作为招聘过滤器的批判性解读）以及 The Walrus 记录的员工侧摩擦对读：在 AI 强制令这件事上，压力来自内部而非股东——股东以近 6:1 的比例拒绝给管理层加上 AI 治理约束。这也间接印证了 §2.13 的 founder legitimacy 机制在实际投票中的效力。

Lead Independent Director Joseph Natale 的董事支持率最低（90.45%）。

2026 年 6 月 17 日：Editions Summer '26——UCP 从"可选"变成"默认开启"

150+ 项更新中的两项结构性变化：Universal Commerce Protocol (UCP) 在每一个 Shopify 店铺默认开启（无需任何配置）；Catalog 让合资格商品自动分发进 ChatGPT、Microsoft Copilot、Google AI Mode、Gemini 与 Shop app，Tobi 称其已覆盖 "billions of products"。UCP 的支持方除共同开发者 Google 外还包括 Amazon、Meta、Microsoft、Salesforce、Stripe、Etsy、Target、Wayfair。

[推断] 这是 2026-01 UCP/Google 合作与 §9.6（把每个店面变成 agent-ready）的执行里程碑，不是新论点——真正的变化是从 opt-in 变成 default-on：他把"商家要不要接入 agent 时代"这个决定从商家手里收了回来，与 §1.5 "pit of success"（让正确的事成为默认）是同一个设计动作。

2026 年 7 月 25–26 日：加权投票言论风波

⚠️ 这是本档案窗口内 Tobi 最具争议的一次公开表达，且与 Shopify 业务无关。

7 月 25 日晚至 26 日，Tobi 在 X 上回复了退休 TD Bank 高管 Eric Thor 的一条提案。Thor 主张按个人所缴所得税加权投票：不缴所得税者 0 票，缴 $500,000 以上者 5 票。Tobi 的回复是两个词：

"Good system."

"好制度。"

@tobi on X, 2026-07-26

同一线程里他把这条延伸到退休人群：

"Now you are a dependent and that means no voting, just like dependents under age. Enjoy the deal, let people with a stake in the future decide."

"现在你是一个受抚养者，那就意味着没有投票权，就像未成年的受抚养者一样。享受这个交易吧，让对未来有利害关系的人来决定。"

@tobi on X, 2026-07-25/26

另一条帖子里他评论多伦多的市政会议：

"Exactly what just happened in Toronto. Retired, Unemployed, Unemployable, and useful idiots are the only people who go to these town hall sessions."

"多伦多刚刚发生的正是如此。退休的、失业的、没有就业能力的，以及有用的白痴——只有这些人会去参加这些市政厅会议。"

@tobi on X, 2026-07-25

事实核查状态：以上三条经 CBC、BNN Bloomberg（Canadian Press）、Fortune、CTV News、Radio-Canada、blogTO、MobileSyrup、Daily Hive、HRD Canada 等 8 家以上可信媒体交叉证实。x.com 本身对抓取返回 402/403，故全部为媒体转引而非原帖直读。Shopify 未回应媒体询问；截至 2026-08-05 未见 Tobi 撤回或澄清。加拿大有请愿要求撤销其 2018 年获颁的 Meritorious Service Cross。

⚠️ 未采信：单一 Substack 来源称他被质疑后辩解自己想要的上限是 5 票而非上百票、目的是 "reward productive people with leverage"。未获第二家佐证，本档案不采信。

详见批判视角 §7。

2026 年 8 月 5 日：Q2 财报——AI 指标首次量化，但分母未公开

营收 $3,583M（同比 +34%）、GMV $115,567M（+32%）、经营利润 $488M、自由现金流 $654M（18% margin）、净利 $1,502M；股价当日涨约 18%，Q3 指引营收增速 low-thirties。

仅在电话会上（未进新闻稿与 10-Q）披露的 AI 口径：AI 驱动的流量与订单各自同比翻三倍；Sidekick 本季处理约 3,400 万次商家对话，日活商家使用量同比 3.6x；AI 归因购买中 75% 落在 top 100 品类之外；约 50% 的 AI 引荐 session 直接落到商品页（传统搜索约 20%）。

⚠️ 两条纪律性注记：（1）Shopify 未披露 AI 驱动 GMV 的绝对占比——上述全部是增长率，分母未公开，因此无法判断 AI 对生意的实际权重；（2）Tobi 本人未在这场电话会上发言（发言人为 IR、President Harley Finkelstein、CFO Jeff Hoffmeister）。网络上流传的一句归给 Tobi 的 "AI has become our native language..." 经核对原始文字稿确认不存在，系合成产物，本档案不采纳。

可归属的相关表述来自 Finkelstein 而非 Tobi：

"Tobi, who I think has been thinking about this longer than most and certainly put Shopify on the path of being AI reflexive before almost anyone else."

"Tobi，我认为他思考这件事比大多数人都久，而且几乎在所有人之前就把 Shopify 放上了 AI reflexive 这条路。"

Harley Finkelstein, Shopify Q2 2026 财报电话会, 2026-08-05

[推断] 75% 的 AI 归因购买落在 top 100 品类之外，是 §9.14 那条赌注目前唯一的量化证据——如果 AI 购物真的把需求推向长尾而非集中到头部商品，那么"AI 让个体商家更值钱"这一支就还站得住；反之则是 §9.14 里他自己描述的那条"Shopify 不再被需要"的路径。这个比率值得在后续 refresh 中持续跟踪。

2025 年 7 月 2 日：Toronto Tech Week 与 Chamath Palihapitiya 的另一场——能源、内部工具、软件产业复合体

2026-07-03 refresh 折入的这一场是 Toronto Tech Week 另一个独立 session（Tobi 与 Chamath Palihapitiya 对谈，2025-07-02），不同于上面与 Lulu Cheng Meservey 的 fireside。⚠️ 多嘉宾场，须严格归属：本节只折入明确记作 Tobi 的答案（Q2/Q3/Q4/Q5/Q6 中 Tobi 的部分）；"software industrial complex"（软件产业复合体）与 "app crappers / vibe coding 是垃圾"这两条最抢眼的判断在转录里明确归属 Chamath 而非 Tobi，故不作为 Tobi 观点折入本档案。

Tobi 在这里给出了一条此前档案缺失的、非常"他"的软件品味判断——他要软件像一棵没有可见年轮的树，不出卖创造它的组织的组织架构图：

"I try to ensure the 'tree rings' of the company don't show. I want Shopify to feel like it was created yesterday, utilizing all the modern technologies available, rather than having some parts feel modern and others feel like legacy code. I hate using products that betray the org chart of the company that created them or reveal which parts of the product were built by high-quality teams versus cost centers. Software should feel cohesive, not like a patchwork of different eras."

"我努力让公司的'年轮'不显露出来。我想让 Shopify 感觉像是昨天才被创造出来的、用尽了所有现有的现代技术，而不是某些部分感觉现代、另一些感觉像遗留代码。我讨厌用那些出卖了'创造它的公司的组织架构图'的产品——那些暴露出哪部分是高质量团队做的、哪部分是成本中心做的产品。软件应该感觉是浑然一体的，而不是不同时代的补丁拼凑。"

Tobi Lütke, Toronto Tech Week (with Chamath Palihapitiya), 2025-07-02 [Q2]

这条把 §1.1 "软件不应让人感到愚蠢"往前推了一层：不仅软件对用户要友好，软件对自身的内部一致性也是一种伦理——"betray the org chart" 直接呼应他 §主题 A "从代码层而非 MBA 抽象层看 Shopify" 的连续性。

关于 AI 时代的择路，他重申了那条一以贯之的信条，但这次把它锚在"不要跳进 buzzword"上：

"You have to be very careful not to just jump into a buzzword, as that often leads to a dead end. You cannot fall in love with a solution; you must fall in love with a problem and remain flexible on the solutions. If you are deeply committed to solving a problem, you can remain flexible on the tools and technologies you use to solve it."

"你必须非常小心，不要一头扎进一个 buzzword——那往往通向死胡同。你不能爱上一个解决方案；你必须爱上一个问题，并对解决方案保持灵活。如果你深深地投入于解决一个问题，你就可以对使用的工具和技术保持灵活。"

Tobi Lütke, Toronto Tech Week (with Chamath Palihapitiya), 2025-07-02 [00:06:27 / Q1]

内部工具作为"诚实反馈闭环"——这是 §1.9 "internal tools shape behavior"（2025 The Cheeky Pint）的更上游论证，Tobi 把它讲成一条关于 agency 的论证：

"Building internal tools allows you to become your own first customer, which creates a cycle of honest, immediate feedback that you cannot get otherwise. If you adopt someone else's software solution, you lose agency over your culture because you are forced to adapt your operations to their specific implementation."

"构建内部工具让你成为自己的第一个客户，从而创造出一个诚实、即时的反馈闭环——这是你用别的方式得不到的。如果你采用别人的软件方案，你就失去了对自己文化的 agency，因为你被迫让自己的运营去适配他们那套特定的实现。"

Tobi Lütke, Toronto Tech Week (with Chamath Palihapitiya), 2025-07-02 [Q3]

这条把他反复出现的 "agency" 词（§一 童年 Amstrad 词源）落到组织层：买别人的软件 = 交出对自己文化的能动性。它也与他 §1.9 自建 HR/薪酬系统而不用 Workday 是同一条实践线。

能源—繁荣的判断——这是 Tobi 此前档案里完全没有的宏观立场，[推断] 与他 §7.5–7.6 的 Pinker 式乐观、GDP-4%-黄金时代论同源：

"The common narrative that using less power is a moral virtue is factually incorrect. There has never been a low-power-using prosperous society. Human prosperity is essentially a function of power multiplied by productivity. Guilt-tripping society about power usage is a moral hazard that holds back economic growth."

"用更少的电是一种道德美德'这个流行叙事，在事实上是错的。从来没有一个低耗电的繁荣社会。人类的繁荣本质上是'功率 × 生产率'的函数。为用电这件事对社会进行道德绑架，是一种阻碍经济增长的道德风险。"

Tobi Lütke, Toronto Tech Week (with Chamath Palihapitiya), 2025-07-02 [Q4]

注意这与他 2019 年主动设立 Shopify Sustainability Fund、为 carbon removal 付费（§八）之间存在一个值得记录的张力：他既主动为碳清除买单，又反对"少用电=美德"的叙事。[推断] 他的调和方式大概是"用更多干净的电 + 主动清除碳"，而非"用更少的电"——即把繁荣与去碳都押在"生产率/技术"这一侧，而不是"节制"这一侧。

关于加拿大——延续 §8.3 的批评但落点更具体到资本速度（capital velocity）：

"We need to shift away from relying on government-sponsored balance sheets and foster private actors motivated purely by profit... Canada has all the resources required to be a leader, but it requires a commitment to building companies from first principles and a willingness to demand more of each other."

"我们需要从依赖政府背书的资产负债表，转向培育纯粹由利润驱动的私人行动者……加拿大拥有成为领导者所需的一切资源，但这需要一种从第一性原理出发建公司的承诺，以及一种彼此要求更高的意愿。"

Tobi Lütke, Toronto Tech Week (with Chamath Palihapitiya), 2025-07-02 [Q6]

被主持人问"如果重回 22 岁会怎么用时间"，他给出了与 §7.1 "programming 是树干"完全一致的答案，但加了一条自我批评（他后悔太早离开技术一线）：

"I would prioritize learning to code and staying technical. The biggest mistake I made was moving away from being technical too early; it left me playing catch-up to the people I was starting companies with... Think of your own skillset as a product that you offer to the market."

"我会优先学会写代码并保持技术性。我犯过的最大错误就是太早离开技术一线；这让我一直在追赶那些和我一起创业的人……把你自己的技能组合当成一个你向市场提供的产品。"

Tobi Lütke, Toronto Tech Week (with Chamath Palihapitiya), 2025-07-02 [Q5]

[推断] 这条"太早离开技术是最大错误"与 §一 第六章他 2008 年不情愿接任 CEO、以及他今天仍写代码（§10.13）、2025 AI memo 要求全员写代码（§9）连成一条自洽的悔悟—矫正弧：他后半段职业生涯反复用"回到 coding"来对冲当年"太早离开技术"的代价。

（session summary 另提到 "mouse in the tub"（把溺水的老鼠捞起给它片刻喘息、极大提高其后续存活率）作为"小而及时的干预能指数级提升创始人韧性"的隐喻；但该比喻在可核验的 highlights/QA 转录里未逐字出现、且归属不完全清晰，故仅在此备注，不作为 Tobi 逐字引用折入。）

"We're making every Shopify store agent-ready by default."

"我们让每一家 Shopify 店面默认 agent-ready（对 AI agent 友好）。"

Tobi Lütke, Shopify Editions Dev 2025 keynote

"Companies falling in love with solutions rather than problems are depriving themselves of agency."

"迷恋解决方案而非问题的公司，正在剥夺自己的 agency。"

Tobi Lütke, BetaKit / Toronto Tech Week Homecoming, 2025-06

二、深度洞察集

1. 关于产品与软件工艺

1.1 软件不应让人感到愚蠢

这是 Tobi 反复表达的产品伦理底线，也是他多年来最一致的产品声明：

"When software goes bad, it makes people feel dumb and machines do not get to negatively influence people in my mind."

"软件出问题时，会让用户觉得自己很蠢——而在我看来，机器没有资格让人感觉糟糕。"

Lenny's Podcast, 2025-02-02

"Computers don't get to make people feel dumb, that is not their role, they are tools."

"电脑没资格让人觉得自己蠢——那不是它的角色，它是工具。"

The Shawn Ryan Show #261

"I built Shopify to avoid people having to use shitty software, not to add to the pile."

"我建 Shopify 是为了让人不必再用糟糕的软件——而不是给那堆糟糕里再加一份。"

Stripe Sessions, 2025-10-06

"Computers should never constrain people. The computer's role is making humans better."

"计算机永远不应该限制人。它的角色是让人变得更好。"

Acquired ACQ2, 2025-09-18

1.2 产品质量 = 创造它的人有多在乎

"Every product in the world, the quality at the end of the day is simply a reflection of how much the people who created it gave a shit about product."

"世界上每一个产品，它最终的质量只是反映了——做这个产品的人有多在乎产品本身。"

Lenny's Podcast, 2025-02-02

"It is not possible to make great products if the people who work on it do not give a shit."

"如果做产品的人不在乎，做不出伟大的产品。"

Lenny's Podcast, 2025-02-02

"Do not engage in product work on products that you don't care about."

"不要在你不在乎的产品上做产品工作。"

Lenny's Podcast, 2025-02-02

1.3 Live in everyone else's future

这是 Acquired ACQ2 那期的标题，也是 Tobi 一以贯之的学习方法。它的具体操作：使用最早的 beta 软件、为新模型做自己的 eval、把生活构造成提前进入"未来日常"。

"Living in 'everyone else's relative future' by adopting beta software early. Building personal eval frameworks (test harnesses) to assess AI model capabilities. Context engineering as the fundamental skill for using AI effectively."

"通过早期使用 beta 软件来'住在别人的相对未来'里。自己搭建个人 eval 框架（test harness）评估 AI 模型能力。Context engineering 是使用 AI 的根本技能。"

Acquired ACQ2, 2025-09-18

"You just go experience, develop taste for it, and figure out how to make your own decisions."

"去亲自体验，培养自己的品味，然后学会自己做决定。"

Acquired ACQ2, 2025-09-18

"Sometimes it feels like he is going a few years into the future and coming back with positive information."

"有时候我感觉他像是先去了几年后的未来，然后带着正面信息回来。"

Daniel Weinand 评 Tobi, Globe and Mail, 2014

1.4 Build infrastructure as composable primitives, not solve stated problems

"At some point you understand it well enough to not solve the problems that people state but build infrastructure in a form of primitives that can be composed to solve all the problems."

"到某个时刻，你对这个领域足够理解，不再去解决人们提出的问题，而是用一组可以被组合的 primitives 来构建基础设施，让所有问题都可解。"

Stripe Sessions, 2025-10-06

1.5 Design system 作为 "pit of success"

"This design system helps everyone fall into some big pit of success because a lot of the knowledge of how to build a great UX is encoded in the design system."

"design system 让每个人都掉进一个'成功大坑'——因为它把大量优秀 UX 的知识编码在系统里。"

20VC #6727, 2023-04-03

1.6 Simplicity 是工艺而非省略

"Simplicity isn't simple. It takes tremendous care, discipline, and craftsmanship."

"简单一点也不简单。它需要巨大的关怀、纪律与工艺。"

Shopify IPO Letter, 2015

1.7 "Goal-oriented software" 是软件演化的终点

"Goal-oriented software is actually what we always wanted."

"面向目标的软件，其实正是我们一直想要的。"

Conversations with Tyler, 2024-09-18

1.8 Counterfactual 思维：从今天的角度重建

"If this would have been possible from the beginning, we would have built this entirely different thing."

"如果一开始就能做到这个，我们会做出一个完全不同的东西。"

a16z Podcast, 2024-12-16

"My biggest fear is that Shopify winds up being a fantastic solution to a problem that people no longer have."

"我最大的恐惧是：Shopify 最后变成了一个对人们已经不再有的问题的绝妙解决方案。"

a16z Podcast, 2024-12-16

1.9 软件 vs 政策：软件部署快、政策部署慢

"Software changes can deploy instantly; policy rollouts require consultative processes. Shopify built custom HR and compensation systems rather than using Workday."

"软件改动可以即刻部署；政策改动需要咨询流程。Shopify 自建了 HR 和薪酬系统，不用 Workday。"

The Cheeky Pint, 2025

"Internal tools shape behavior more effectively than policies."

"内部工具改变行为，比政策有效得多。"

The Cheeky Pint, 2025

1.10 Innovation 是蓝领的、增量的

"Innovation is actually much more blue collar, it's much more vocational, it's the frequent incremental improvement of the things that we care about."

"创新其实更像蓝领工作，是工艺级的，是对我们在乎的东西做频繁、小幅度的改进。"

Shopify Blog (Tobi 的话, 2024-12)

"Making better shoes or more beautiful tables."

"做出更好的鞋子，或者更漂亮的桌子。"

Shopify Blog (Tobi 的话, 2024-12)

在 Behind The Tech 上 Tobi 给出了关于编程本质的一句态度声明——把"computer science"这个名字的使用本身判为一种身份焦虑：

"We've always had an identity crisis. We always wanted to be someone else. We have significant science envy. We even attempt to call it computer science. But clearly it's a vocational art."

"我们一直有身份危机——我们一直想成为别人。我们有非常严重的'科学羡慕症'。我们甚至试图把它叫做 computer science（计算机科学）。但很显然，它是一门志业之艺。"

Behind The Tech, 2023-01-03

1.11 Friction 是商业里最被低估的力量

2020 年 5 月在 Invest Like the Best EP.173 上，Tobi 给出了他对 friction 的一句总论——也是他后来反复在 Shopify 产品决策里运用的根本框架：

"Friction is probably the most powerful under-acknowledged force that exists in the world of business and products in general. A lot of what people think the world is just like is actually products of accidental introduced friction or sometimes willingly introduced friction."

"摩擦力可能是商业和产品世界里最强大却最被低估的力量。很多人以为'世界本来就是这个样子'的现象，其实是意外引入的、有时是故意引入的摩擦力造成的。"

Invest Like the Best EP.173, 2020-05-12

他给出的具体例子是机场餐盘归还点的故事——同一段访谈里他把这个故事讲得最完整：

"Instead of trying to get people to change their mind, let's just make it simpler for them to do the thing that we want them to do. Our floor plan design just involves putting the tray stations to all exits."

"与其试图改变人的想法，不如让他们想做的事变得更简单。我们的楼层设计就是把餐盘归还点放在每一个出口。"

Invest Like the Best EP.173, 2020-05-12

1.12 流程的两条充分条件

同一期 Tobi 给出了他对官僚化的操作化抵抗规则——一条二分法的准入门槛：

"We hold ourselves to the standards that process has to make something new possible or something existing 10x easier before introducing it."

"我们对自己定的标准是：流程要么让一件原本不可能的事变成可能，要么让一件已有的事变得简单 10 倍——否则我们不引入它。"

Invest Like the Best EP.173, 2020-05-12

这条规则与 §2.11 "三类 process" 是同一条线，但作为引入新流程时的二元闸门更为操作化。

2. 关于组织与文化

2.1 Trust Battery（信任电量）

这是 Tobi 最广为传播的管理 mental model，首次系统讲述于 Knowledge Project #41（2018）：

"Mental model for interpersonal relationships as a gradient rather than binary. Two new hires start at approximately 50% trust. Interactions either charge or drain the battery."

"把人际信任建模为一个 0–100% 的电量计，不是 binary。两个新员工互相起步约 50%；每次互动要么充电、要么放电。"

Knowledge Project #41, 2018-09-18

"Enables team communication about trust without becoming personal. Good decisions increase toward 80% autonomy; failure 'drains the battery' quickly."

"它让团队可以用一个不带个人攻击意味的语言来谈信任。好的决策把电量充到 80% 自治区；失败会迅速放电。"

Knowledge Project #41

2026-07-03 refresh 补回该期（The Knowledge Project #41，2018-09-18）的一手逐字——此前本节只有二手 summary。Tobi 亲口给出的 Trust Battery 定义强调它的去人格化沟通用途，并配了一个具体到"团队会议"的例子：

"Trust Battery is just a mental model for how to think about the relationship between people... People sort of think about trust as almost an on-off kind of thing. But it's actually clearly a gradient. It's useful to have this metaphor between people because it allows you to sort of talk about the trust that exists between two people without actually becoming personal. It's so much easier to say, 'hey, I love working with you and the kind of work you do, but you don't show up to the team meetings. I just want you to know those two things offset each other."

"Trust Battery 只是一个思考人与人关系的心智模型……人们往往把信任当成一种近乎 on-off 的东西。但它显然是一个渐变。这个隐喻在人与人之间很有用，因为它让你能谈论两个人之间的信任，而不必让谈话变得针对个人。说这句话就容易多了：'嘿，我很喜欢和你共事、喜欢你做的东西，但你不来参加团队会议。我只是想让你知道，这两件事会互相抵消。"

The Knowledge Project #41, 2018-09-18

2.2 Subtraction 是创始人的超能力

"The best thing founders can do is subtraction. It's much easier to add things than remove things."

"创始人能做的最好的事是 subtraction。增加东西比减去东西容易得多。"

Fortune, 2023-01-03

"Subtraction (removing initiatives) is the founder's superpower."

"做减法（去掉项目）是创始人的超能力。"

Knowledge Project #152

"Periodically removes things from the organization, which he finds 'really fun' though 'some people didn't'."

"他周期性地从组织里移除东西——他自己说'非常有意思'，但'有些人不这么觉得'。"

20VC #6727, 2023-04-03

2.3 Micromanagement is good

20VC 那期 Tobi 提出了一个反直觉的观点：

"There's probably no singular idea that has destroyed more business value on Earth than the idea that micromanagement is a bad thing."

"在这个地球上，可能没有哪一个观点比'micromanagement 是坏事'毁掉过更多商业价值。"

20VC #6727, 2023-04-03

"That Dilbert-esque view of a leader, where pointy head boss comes and moves the cheese around — that is not micromanagement. That is just being extremely bad as being a leader."

"Dilbert 漫画里那个尖头老板换来换去 cheese 的画面——那不是 micromanagement，那只是个极差的领导。"

20VC #6727

他的核心区分：micromanagement = caring deeply about details；bad leadership = uninformed interference。

2.4 Constraints breed creativity

"Artists always talk about a tyranny of a blank canvas. Creativity always comes from constraints."

"艺术家总在抱怨'空白画布的暴政'。创造力永远来自约束。"

20VC #6727

"The best teams are greedy for constraints. They want to understand how can we restrict our space so that we get started."

"最好的团队对约束是贪婪的——他们想搞清楚要怎么限制空间，才能动起来。"

20VC #6727

2.5 反 KPI、反 OKR

"Goodhart's Law just says any metric that becomes a goal ceases to be a good metric."

"Goodhart 法则的意思是：任何变成目标的指标，就不再是一个好的指标。"

Lenny's Podcast, 2025-02-02

"The overlap of most valuable things you can do and the things that happen to be fully quantifiable is maybe 20%."

"你能做的最有价值的事，和恰好能完全量化的事，重叠区可能只有 20%。"

Lenny's Podcast, 2025-02-02

"Shopify deliberately bans traditional KPIs. Metrics should support decisions, not drive them."

"Shopify 故意禁止传统 KPI。指标应当辅助决策，不是驱动决策。"

Lenny's Newsletter, 2024

2.6 Tobi Tornado：把变革压缩到极短时间窗口

"Tobi Tornado: compresses change management into short timeframes; stops projects immediately when they're wrong rather than letting people waste time."

"Tobi Tornado：把变革管理压到极短的窗口里；项目错了就立刻停，不让人浪费时间。"

Lenny's Podcast, 2025-02-02

"We have a fairly limited time in our careers."

"我们在职业生涯里能用的时间是相当有限的。"

Lenny's Podcast

2.7 Reorgs 总是有效（因为优势先到，劣势后到）

"All reorgs work because reorging is valuable — advantages of any structure appear immediately while disadvantages emerge over time."

"所有的 reorg 都有效——因为重组本身有价值。任何结构的优点立刻显现，而缺点要过一段时间才出现。"

Stripe Sessions, 2025-10-06

2.8 公司是一种被研究不足的"社会技术"

"A company fundamentally allows you to run the counterfactual to the world you see around you."

"公司本质上让你能跑'反事实'——和你眼前的世界对照。"

Founders (David Senra), 2026-01-18

"Companies are understudied technologies that create social acceptance for people to pursue a mission together."

"公司是被研究不足的技术——它制造了一种社会接受度，让一群人能共同追求一个使命。"

Stripe Sessions, 2025-10-06

2.9 直接反馈与 Crocker's Rules

"When he looks at the fruits of someone's labour, he says what he thinks. Even if what he thinks is: 'This is shit."

"他看到别人工作的成果时，会说出他真实的想法。即便他想的是：'这就是 shit。"

Globe and Mail, 2014

"Aims for direct feedback without 'shit sandwich' politeness. References Lee Daniel Crocker's Wikipedia principle: take 100% responsibility for your mental state."

"他追求不加'三明治式包装'的直接反馈，引用 Lee Daniel Crocker 在 Wikipedia 上的原则：你要为自己的心理状态负 100% 的责任。"

Tim Ferriss #359

"Learned to frame 'This is shit' as 'Is this really the best way we can do this?"

"他学会了把'这就是 shit'重写成'这真的是我们能做到的最好版本吗？"

Tim Ferriss #359

2.10 Bureaucracy 来自激励错位

"Bureaucracy emerges when local incentives misalign with company missions. It's not process itself but processes serving unclear purposes."

"官僚化是当局部激励和公司使命错位时出现的。问题不是流程本身，而是流程在服务一个不清晰的目的。"

Knowledge Project #152

"Solution: design systems where individual effort compounds toward organizational benefit."

"解决之道：设计能让个体努力复利到组织收益的系统。"

Knowledge Project #152

2.11 三类 process

"Processes that make previously impossible things possible (valuable). Processes that simplify possible tasks (valuable). Everything else — 99.9% of corporate processes that contradict common sense (wasteful)."

"（1）让本来不可能的事变可能的流程——有价值；（2）让可行的事更简单的流程——有价值；（3）其余的——大约占企业流程的 99.9%——是反常识的浪费。"

Knowledge Project #41

2.12 失败 = 成功地发现了一种行不通的做法

"Frames failures as 'discovery of things that did not work,' not wasted mistakes. Eliminated 'failure' vocabulary in favor of 'successful discovery of something that didn't work'."

"他把失败视为'发现一种行不通的做法'，而不是浪费的错误。他在词汇上把'failure'换成了'某种 doesn't-work 的成功发现'。"

Knowledge Project #41

"The failure case is not failure. The failure case is not learning anything."

"真正的失败案例不是失败本身。真正的失败是什么都没学到。"

The Shawn Ryan Show #261

在 The Knowledge Project 上 Tobi 给出了这条"重命名失败"的完整版本，加上一条反风险厌恶的逆向论证——在 Shopify，"不冒险"才是真正的风险：

"At Shopify, it's very clear. It's very important that we explain to people it's not failure. It's the successful discovery of something that did not work if we underperform. We know actually more data. We'll try again."

"在 Shopify 这一点非常明确，很重要——我们告诉大家：如果表现不及预期，那不是失败，那是'成功地发现了某种行不通的方法'。我们因此多了数据。我们再来一次。"

The Knowledge Project, 2022-11-15

"It's actually risky to not try higher. A lot more people have performance matched out [euphemism for being managed out] if they don't take risks. For your career, Shopify is actually more risky to not aim high."

"其实'不冲得更高'才是有风险的。被'performance-matched-out'淘汰的人，远多于那些冒险失败的人。在 Shopify，对你的职业生涯而言，不瞄准高目标才是更冒险的。"

The Knowledge Project, 2022-11-15

2.13 Founder 拥有"legitimacy"作为资源

"Founders possess 'legitimacy' as a resource — the founding story deposits social credit that enables bold decisions. Non-founder CEOs must rely on planning and stakeholder alignment instead."

"创始人有一种叫 legitimacy 的资源——创始故事沉积了社交信用，让大胆决策成为可能。非创始人 CEO 必须依靠规划和 stakeholder alignment 来代替它。"

Knowledge Project #152

2025 年 6 月在 Rick Rubin 的 Tetragrammaton 上，他第一次把这条给出了完整的机制版本——关键不在"创始人更有权威"，而在权威到底附着在什么东西上：

"Let's call them founder-run companies and managerial-run companies. Managerial-run companies, the job of CEO is totally different. The job is to create a plan. The CEO doesn't actually have the legitimacy to decide exactly what the company does. People think so, but it's not actually like that. The legitimacy is a thing that's built. And what the CEO does, it creates the stakeholders a plan and then creates legitimacy in the plan. And the plan is then followed by everyone. It's the plan that has the authority. The CEO just has the ability to influence the plan heavily."

"我们把它们叫做创始人主导的公司和职业经理人主导的公司。在经理人主导的公司里，CEO 的工作完全是另一回事：工作是做出一个计划。CEO 其实并没有决定公司到底做什么的 legitimacy。人们以为有，但实际上不是这样。Legitimacy 是被建构出来的东西。CEO 做的事，是给 stakeholder 一个计划，然后为这个计划建立 legitimacy。然后所有人follow 这个计划。拥有权威的是计划本身。CEO 只是拥有大幅影响这个计划的能力。"

Tetragrammaton with Rick Rubin, 2025-06-25（00:56:30）

这直接推出了他对"为什么大公司接不住 AI"的解释——不是认知问题，是修改成本问题：

"That also makes these companies very, very hard to adjust, right? Like, so if you want to pivot or something, you need to make a new plan and explain why you didn't screw up with the last plan. So this is why large companies very often have a hard time incorporating AI quickly or something like this, or during COVID."

"这也让这些公司极其难以调整。你要 pivot 的话，你得做一个新计划，并且解释你上一个计划为什么不算搞砸了。所以这就是为什么大公司常常很难快速把 AI 纳进来——或者在 COVID 的时候也一样。"

Tetragrammaton, 2025-06-25（00:57:06）

"Founder-run companies are different. The fact that the slot is filled, that the person is there who has all, in my case, the 20 or 21 years of being there for every decision... when COVID happens, the day that happened, or when AI comes out, I'm actually not beholden to the roadmap — because the roadmap exists because we're figuring out what the exact best things are to work on right now. But if reality has changed, that is clearly now something different."

"创始人主导的公司不一样。那个位置被填满了，那个人在场，在我这里是 20 或 21 年里每一个决策都在场……所以 COVID 发生的那一天，或者 AI 出来的时候，我其实并不受 roadmap 约束——因为 roadmap 之所以存在，是因为我们在搞清楚此刻最该做的到底是什么。但如果现实变了，那答案显然就变成别的了。"

Tetragrammaton, 2025-06-25（00:57:22）

值得注意的是他在同一段里明确拒绝把这条讲成创始人优越论：

"I think they have underappreciated how different they are, founder-run companies and managerial-run companies, and neither is better than the other. In fact, in most business sectors, I think founder-run companies are a mess, a hot mess of chaos and hard to predict and they're harder to work for. Things change on a dime. For the right people, this is the best."

"我觉得大家低估了这两者差别有多大——创始人主导的公司和经理人主导的公司——而两者没有谁更好。事实上在大多数行业里，我认为创始人主导的公司是一团乱麻，一团热腾腾的混乱，难以预测，也更难为其工作。事情说变就变。对合适的人来说，这是最好的。"

Tetragrammaton, 2025-06-25（00:58:01）

[推断] 这段是本档案里少见的、他主动给自己的结构性优势划边界的表述——他没有说创始人控制是对的，只说它换来的是"不受计划约束"这一种特定能力，代价是混乱与不可预测。把它与 §四.3（Founder Share 与 40% 永久投票权的治理批评）对读，可以看出他为这套制度辩护的论据是适应速度，而不是所有权正当性。

2.14 World of Warcraft 公会是更好的协调研究对象

"I think you can learn more by studying World of Warcraft raiding guilds than you can learn from studying companies about coordination."

"我觉得研究 World of Warcraft 副本公会，比研究公司，能学到更多关于协调的东西。"

20VC #6727

2.15 "没有 Shopify culture"：内部多元文化而非统一文化

2018 年 The Knowledge Project #41 上 Tobi 给出了一条与硅谷"strong culture"主流叙事相反的立场——他反对把 culture 当成可被 manager"拥有"和管理的东西，也反对存在一个可复制的、单一的"Shopify culture"：

"I have mixed opinions on culture. I think we're getting dangerously close to the point where a lot of people think that culture is this sort of manageable thing that managers have to kind of own. I think culture is just the sum total of all the people who work with each other. Shopify is super happy with internal multiculturalism. Every one of our offices has a different culture, and that's okay. There's no Shopify culture. I'm super happy about everyone being different, everyone to show up as their authentic self at work, not some sanitized conclusion of what people should be like."

"我对 culture 有复杂的看法。我觉得我们正危险地接近这样一个点——很多人以为 culture 是一种可被管理的东西、是 manager 必须去'拥有'的东西。我认为 culture 只是所有一起工作的人的总和。Shopify 很乐于接受内部的多元文化——我们每一个办公室都有不同的文化，这没关系。没有一个'Shopify culture'。我很乐意每个人都不一样、每个人都以真实的自己出现在工作里，而不是某种被消毒过的、'人应该是什么样'的结论。"

The Knowledge Project #41, 2018-09-18

这条早于 2020 年 Digital by Default 两年——[推断] 它预示了他后来对远程 + 分布式办公的接受度：如果本就"没有单一 culture、每个 office 各不相同"，那么放弃总部这件事对他而言并不构成文化损失。

2.16 环境即质量标准：连热水器的 UX 都不能烂

同一期（#41）Tobi 给出了一条把"产品质量标准"直接编码进物理办公环境的原则——他把它作为一种 nudge，而非制度：

"I ask everyone to do world-class software, build something significantly better than anything ever made in this particular space. But if you arrive in the office and the first thing you do is get hot water for your tea and you're faced with some absolutely insane user experience, I can't really ask everyone to do better. A much better way is just make sure that everything around us has the same controls of quality that I want to see in our own products. People are so much more affected by their environment than we like to believe."

"我要求每个人做世界级的软件，做出比这个领域里做过的任何东西都好得多的东西。但如果你到了办公室，第一件事是给自己泡茶接热水，却撞上一个糟糕透顶的用户体验，我就没法真的要求所有人做得更好。更好的做法是：确保我们周围的一切，都带着我希望在自己产品里看到的那种质量把控。人受环境的影响，远比我们愿意承认的要大。"

The Knowledge Project #41, 2018-09-18

"I'm a huge fan of Richard Thaler. Danny Kahneman and Amos Tversky are my spirit animals... You change the environment to make it so that people, their common sense compels them to do the right thing."

"我是 Richard Thaler 的铁粉。Danny Kahneman 和 Amos Tversky 是我的 spirit animals……你改变环境，让人们的常识本身就驱使他们去做对的事。"

The Knowledge Project #41, 2018-09-18

这条与 §1.11 friction（"让人想做的事变简单"）和 §2.10 bureaucracy 是同一条线，但落在"用环境而非政策改变行为"的最早一手表述上——比 §1.9 "internal tools shape behavior more effectively than policies"（2025 The Cheeky Pint）早了七年。

2.17 组织韧性 = 反应速度，不是防止坏事发生

"If you want an organization that lasts, you need to be okay with bad things happening. The quality of an organization is not how good it is at preventing bad things from happening — although that's a useful skill — but how quickly it reacts to bad things happening. We decided to make plan B plan A: we closed our office, deprogrammed everyone's fob one day, sent an email the evening before that said 'hey, we're going to be homeless for a month."

"如果你想要一个能持久的组织，你必须接受坏事会发生。一个组织的质量，不在于它多擅长防止坏事发生——尽管那是有用的技能——而在于它对坏事发生的反应有多快。我们决定把 plan B 变成 plan A：我们关掉办公室，某一天直接注销了所有人的门禁卡，前一晚发了封邮件说'嘿，我们接下来一个月要无家可归了。"

The Knowledge Project #41, 2018-09-18

[推断] 这条 2018 年的"主动制造 plan-B-成为-plan-A"演练，与他 2020 年"Digital by Default"、2023 年 Chaos Monkey 一次性删例会是同一种手法——用一次刻意的、不可回头的冲击来测试并重建组织，而不是渐进调整。

2.18 "团队，不是家庭"：公司作为 common carrier

2020–2021 年是 Shopify 内部"极度分心与撕裂"的时期——员工普遍希望公司对每一个社会议题公开表态。Tobi 写下那份著名的内部备忘 "Shopify is a team, not a family"，把公司重新定义为一个中立的 "common carrier"（公共承运人）：它托管从政界人物到形形色色创业者的所有人，而"预先筛选谁有资格使用我们的工具"根本不是它的角色——只要生意合法，Shopify 就支持。2023 年 All-In Summit 上被 David Sacks 追问这份 memo 时，他把这条原则还原成一个工具隐喻：

"Your screwdriver is not going to tell you that you can't screw something into a wall because you're a Republican voter."

"你的螺丝刀不会因为你是共和党选民，就告诉你不能把螺丝拧进墙里。"

All-In Summit（fireside w/ David Sacks & Jason Calacanis）, 2023-09-13 [00:08:57]

他把这份 memo 的功能描述为"简化内部文化的复杂度"——提醒团队"我们在这里是造工具的，不是当政治仲裁者（political arbiter）"；公司的使命只有一件：帮人 reach for independence。

[推断] 这条"对内中立"与 §8.10 / 批判章 §4 的"对外中立"（2017–2018 Breitbart：产品即言论）是同一根神经的两个层面——一个管"谁能上架"，一个管"公司要不要表态"。两者都从他反复表达的 mission 收敛而来：Shopify 只做一件事，其余一律视为 side quest（§9.13 / §10 main-quest）。他用"team not family"把这种收敛落到雇佣关系本身——家庭无条件包容成员的一切，而团队只对共同目标负责。

⚠️ 需要与批判章 §7（2026-07 加权投票言论）并读：他在 2020–2021 年要求员工把政治留在门外，而 2026 年他本人在 X 上就选举权发表了高度党派化的主张。这两件事之间的张力，他未公开回应。

2.19 Chaos engineering：让失败变频繁，因为稀有的失败最危险

这是本档案里少数能追到具体动作的组织原则——而且它的起点是 Tobi 本人手动去服务器农场里关掉服务。2025 年在 Tetragrammaton 上他把这条从头讲了一遍。

起点是一个他不接受的说法——"硬件坏了是天意"：

"There's downtime because some piece of hardware failed. And everyone is going, well, look, a piece of hardware failed. It's an act of God, so to speak. It just happens. I'm like, well, if it happens, then that means we have to engineer around it. So we have to build the redundancies to make that not affect the system. And what I hear back is that it's just hard to test because it happens rarely. I'm like, cool, let's make it —"

"会宕机是因为某个硬件坏了。所有人都说：看，硬件坏了，这算是天意吧，它就是会发生。我说：好，如果它会发生，那就意味着我们必须围绕它做工程设计。我们得建冗余，让它不影响系统。而我得到的回答是：这个很难测试，因为它很少发生。我说：好极了，那我们就让它——"

Tetragrammaton with Rick Rubin, 2025-06-25（01:16:33）

"So I log into our server farm and turn off my service. And I just started doing that. And it was better than doing it when something fails — it usually is at night and then pages sometimes are slept through and then suddenly you have a real problem. So let's make this actually happen more frequently, because if failure is more common, it's actually easier to reason about. It's actually bad that failure is rare. If it's a real part of a system, you want it to be frequent."

"于是我登录我们的服务器农场，把我的服务关掉。我就这么开始做了。这比等到真出事再做要好——真出事通常在半夜，呼叫有时候会被睡过去，然后你就真有麻烦了。所以我们让它更频繁地发生吧，因为失败越常见，反而越容易推理。失败很稀有这件事其实是坏事。如果失败是系统真实的一部分，你会希望它频繁。"

Tetragrammaton, 2025-06-25（01:17:11）

他到很久以后才知道这个直觉有名字：

"Years later, [Taleb] wrote a fantastic book on this exact thing called Anti-Fragile. He coined a name for this kind of thing. He said, hey, after you break a bone, after it heals, that part of the bone is the strongest part of the bone. And no language on planet Earth actually has a word for that kind of thing. No culture has discovered that certain things get better by breaking them or disrupting them, even though the concept is clearly all around us."

"多年以后，[Taleb] 就这件事写了一本极好的书，叫《Anti-Fragile》。他给这类东西起了个名字。他说：你看，骨头断了之后，愈合之后，那一段反而是整根骨头最强的地方。而地球上没有任何一种语言真的有一个词来指这种东西。没有任何文化发现过'某些东西会因为被打破、被扰动而变得更好'——尽管这个概念明明到处都是。"

Tetragrammaton, 2025-06-25（01:17:44）

最后一个细节暴露了这套做法在组织里的真实处境：

"Very quickly, the operations people figured out that they can just set alerts on me logging into servers."

"很快，运维的人就想明白了：他们可以直接在'Tobi 登录服务器'这件事上设告警。"

Tetragrammaton, 2025-06-25（01:18:13）

他给这条做法的上位理由是他反复出现的那句判断——静止即死亡：

"That's a solve for things being too derivative, too static. I think stasis is death. When things are static, that means you're not advancing. And if you're not advancing, that means you're sliding — backwards, usually."

"这是针对'太因循、太静止'的解法。我认为静止就是死亡。事情静止，意味着你没在前进；而你没在前进，通常意味着你在往后滑。"

Tetragrammaton, 2025-06-25（01:16:16）

[推断] 这条给 §4.7（"增长 = 不变化的稳定 = 死亡"）和 2023 年 1 月删 12,000 个例会的"Chaos Monkey"命名提供了来源解释：那次组织手术借用的不是 Netflix 的行业术语，而是他自己十几年前在 Shopify 基础设施上养成的、亲手制造故障的工程习惯——把"人为注入扰动"从服务器搬到了日历上。同一逻辑也解释了 §2.6 Tobi Tornado（把变革压进极短窗口）：与其等扰动稀疏地、在最坏时刻到来，不如自己定时引爆。

3. 关于决策与判断

3.1 Calm progress 与 infinite game

"Calm progress" 一词出自 Knowledge Project #152（约 2023）的标题。它的内核：在 infinite game（James Carse, Finite and Infinite Games）的视角里，目标不是赢，而是让游戏持续。

"Drawing from James Carse: finite games (with winning conditions) vs infinite games (no endpoint). Shopify is his infinite game — advancing entrepreneurship endlessly."

"出自 James Carse：有限游戏（有获胜条件）与无限游戏（没有终点）。Shopify 是他的 infinite game——无止境地推进创业。"

Knowledge Project #152

"100-year vision: at that timescale, the software product is irrelevant but the mission survives — entrepreneurship is precious."

"百年视角下，软件产品本身不重要，但使命会存活——创业是珍贵的。"

Lenny's Podcast, 2025-02-02

"Short-term: 'Don't die.' After that, focus on the positional game on an infinite, fog-covered chessboard."

"短期：'别死。'之后是一盘无限的、被雾笼罩的棋盘上的位置博弈。"

Lenny's Podcast

3.2 决策容易，找输入难

"Decision-making is easy; finding inputs is hard."

"做决策是简单的；找到正确的输入很难。"

Knowledge Project #152

"Most decision errors stem from blind spots — missing information or unconsidered consequences — not flawed reasoning."

"决策错误大多来自盲点——遗漏的信息或没考虑到的后果——而不是推理瑕疵。"

Knowledge Project #152

"Capture assumptions underlying major strategic choices; periodically rerun the decision function as world variables change."

"把重大战略决策背后的假设捕捉下来；当世界的变量变化时，定期重新跑这个决策函数。"

Knowledge Project #152

在 2022 年 11 月那期 Knowledge Project 上 Tobi 把这条诊断说得更尖锐——他直接说决策理论行业研究错了对象：

"I think decision-making is misstudied. I think decision-making is the easiest thing in the world. It is blindingly obvious how to make a decision. What isn't is finding the inputs. What is the context? What are all the things that affect the decision? Developing that is hard."

"我认为决策这件事被研究错了。决策本身是世界上最容易的事——怎么做决策显而易见。难的是找输入。什么是上下文？哪些东西影响这个决策？把这些搞清楚——那才难。"

The Knowledge Project, 2022-11-15

含义：行业里的"决策框架"大都关注最后一步（选项 A vs B），但真正的工作量在前面所有步骤（什么算输入、信息从哪来）。这是为什么他强调 RTS scouting（详见 §2.7.18）、与 150+ 商家的 WhatsApp 群——所有这些都是在解"输入难"问题。

3.3 决策可逆性是首要分类维度

"I think the most important thing that people have to understand is how un-doable is a decision? Reversible decisions should be made quickly. Irreversible ones require deep understanding."

"我觉得人们要理解的最重要的事是：这个决定有多不可撤销？可撤销的决定应该快做；不可撤销的决定需要深入理解。"

Knowledge Project #41

3.4 没有"正确决策"，只有"事后被证明正确的语境"

"There's no right decision, there's only context in which decisions turn out to be correct."

"没有'正确决策'，只有'事后被证明决策正确的语境'。"

Founders (David Senra), 2026-01-18

"The cheat code to always being right is just change your opinion every time you get better information."

"永远正确的作弊码是：每次拿到更好的信息就改主意。"

Founders (David Senra), 2026-01-18

3.5 公开承认错误的范本：2022 年那封裁员信

"Ultimately, placing this bet was my call to make and I got this wrong."

"归根到底，下这个赌注是我的决定，我决策错了。"

Tobi Lütke 内部备忘, 2022-07-26

3.6 反"best practices"

"Best practices simply mean don't take risk."

"Best practice 的意思就是别冒险。"

Knowledge Project #152

"True innovation requires deliberately departing from orthodoxy. Companies must psychologically safeguard intelligent risk-taking."

"真正的创新要求刻意脱离正统。公司必须为聪明的冒险提供心理保护。"

Knowledge Project #152

3.7 默认 contrarian

"If most people are doing it a certain way, I, by default, don't want to do it that way."

"如果大多数人在按某种方式做事，我默认就不想那样做。"

Lenny's Podcast, 2025-02-02

"You try to make your company as different from every other company as possible, which is actually the opposite of what most companies, most CEOs optimize for."

"你应该尽量让自己的公司和其他每一家都不同——这恰恰和大多数公司、大多数 CEO 在优化的方向相反。"

Stripe Sessions, 2025-10-06

在 The Knowledge Project 上 Tobi 罕见地承认这条默认逆向——而不是"权衡后逆共识"——是他的默认起点：

"If most people are doing it a certain way, I by default don't want to do it that way. If you want to do something world-class, you can't do it like everyone else."

"如果大多数人都用某种方式做某件事，我默认就不想用那种方式去做。如果你想做出世界级的东西，你不可能用所有人都用的方式去做。"

The Knowledge Project, 2022-11-15

3.8 First-principles algorithm

"First principles algorithm: (1) analyze path dependencies of existing solutions, (2) identify what they overfit for that may no longer be necessary."

"First-principles 算法：(1) 分析既有方案的 path dependency；(2) 识别它过拟合于哪些不再必要的东西。"

Lenny's Podcast, 2025-02-02

"If you do the same thing, you get the same results. Differentiation requires new ideas."

"做同样的事会得到同样的结果。差异化需要新的想法。"

Founders (David Senra), 2026-01-18

3.9 Decide as the company 10 years hence would wish you had decided

"Decide as the company 10 years hence would wish you had decided."

"做决策时，要像未来 10 年后的公司希望你今天做的那样去决策。"

Knowledge Project #152 框架引用

3.10 故意说"high-conviction, incorrect things"激发 truth-seeking

"Deliberately says 'high-conviction, incorrect things' in meetings to prompt corrections — applies Twitter's truth-seeking mechanism to in-person conversations."

"他故意在会上说'有高确信度但其实错的话'，以激发同事纠正——把 Twitter 那种 truth-seeking 机制搬到线下对话。"

Conversations with Tyler, 2024-09-18

3.11 Ambiguity 比文字化更重要的时刻

"Actively avoiding to write something down is sometimes the most important thing you can possibly do."

"刻意不把某件事写下来，有时候是你能做的最重要的事。"

Founders (David Senra), 2026-01-18

3.12 In/out 而非 up/down 的决策层级

"In/out vs up/down: decisions aren't hierarchical but about proximity — some best made at outskirts (near customers), others centralized."

"in/out 比 up/down 更准确：决策不是层级的，而是关于距离——有些决策最好在边缘（靠近客户）做，有些应当集中。"

Stripe Sessions, 2025-10-06

"Pushes choices toward edges closest to customers. Brings critical decisions inward."

"把选择推向最靠近客户的边缘。把关键决策拉到中心。"

The Cheeky Pint, 2025

4. 关于增长与 scaling

4.1 增长的瓶颈通常是资本，不是 product-market fit

"Once venture capital arrived, every test campaign worked — problem was capital constraint, not product-market fit."

"VC 钱进来之后，每一次营销实验都成功了——所以瓶颈是资本约束，不是 product-market fit。"

Tim Ferriss #359

"2009-2010: Critical turning point — intentional slowdown to test 5 marketing programs (all succeeded)."

"2009–2010 是关键转折期——刻意放慢扩张去测试 5 种营销项目（都成功了）。"

Founderoo

4.2 客户数 vs 客户价值：做新创业者，不是抢现有零售商

"Chose to focus on new entrepreneurs rather than competing for existing retailer market share — VCs initially missed that Shopify would grow the market itself."

"我选择服务新的创业者，而不是抢现有零售商的市场份额——VC 一开始没看到 Shopify 自己会让市场长大。"

The Shawn Ryan Show #261

"VCs initially rejected Shopify, believing addressable market (40K-50K stores) was too small. Lütke recognized they were wrong about market size, not the problem. By making store creation accessible, Shopify unlocked vastly larger entrepreneur base — 'reshaping the meta'."

"VC 一开始拒绝 Shopify，认为 4–5 万家店的市场太小。Tobi 知道他们错的是市场规模，不是问题本身。把开店变得可及，Shopify 释放出大得多的创业者基数——'重塑了 meta'。"

NonGaap, 2023

4.3 增长 = 让客户成功

"Best thing we can do to grow Shopify is make our customers more successful."

"Shopify 增长的最好方式，就是让我们的客户更成功。"

Conversations with Tyler, 2024-09-18

"On a long enough timeline, playing positive sum games with your customers is the ultimate growth hack."

"在足够长的时间尺度上，与客户玩正和游戏是最终极的增长 hack。"

Lenny's Podcast, 2025-02-02

4.4 Lowering complexity 直接生产更多企业

"Lowering complexity directly causes more entrepreneurship: every UX simplification means more businesses exist."

"降低复杂度直接产生更多创业——每一次 UX 简化都意味着更多企业存在。"

Lenny's Podcast, 2025-02-02

"Mission framed as making entrepreneurship simpler — every UX simplification directly causes more entrepreneurship."

"使命被定义为让创业更简单——每一次 UX 简化都直接产生更多创业。"

Stripe Sessions, 2025-10-06

4.5 Be a platform：让生态赚得比自己多

"I believe the secret to massive scale is compressed in three words: Be a platform."

"我相信大规模扩张的秘诀可以浓缩成三个词：be a platform。"

Masters of Scale, 2020-12-15

"You are not a platform until the people who are building on you make more money than you do."

"除非在你之上构建的人赚到的钱比你多，否则你算不上一个 platform。"

Bill Gates 引用 / Tobi 转述, Masters of Scale

"We're all doing this together. Shopify could not accomplish anything that it does without this incredible partner ecosystem."

"我们都是一起做这件事。如果没有这个 incredible partner ecosystem，Shopify 什么都干不成。"

Tobi Lütke, Masters of Scale

4.6 Late-game dominance（StarCraft Zerg 隐喻）

"Lütke plays as Zerg — the 'swarm' race that grows exponentially through ecosystem expansion. Shopify mirrors this: independent merchants collectively overwhelm larger competitors."

"Tobi 在 StarCraft 里玩 Zerg——通过生态扩张呈指数增长的'蜂群'种族。Shopify 是对应：大量独立商家加在一起压倒大型对手。"

NonGaap, 2023

"Late-Game Dominance: Zerg struggles early but becomes nearly unbeatable in extended games through superior economy."

"后期统治：Zerg 早期挣扎，但在长期游戏里靠经济优势近乎不可战胜。"

NonGaap, 2023

4.7 增长是不变化的稳定 = 死亡

"In a company growing 20-40% year over year, you must improve by at least that every year."

"在年增长 20–40% 的公司里，你每年至少必须以同样的速度提升。"

Tobi Lütke 内部 AI 备忘, 2025-03-20

"Red queen race"——必须不停跑才能站在原地。

4.8 增长不该被指标驱动

"Anti-KPI Stance: Shopify deliberately bans traditional KPIs. Metrics should support decisions, not drive them."

"反 KPI 立场：Shopify 故意禁止传统 KPI。指标应当辅助决策，不是驱动决策。"

Lenny's Newsletter

"If metrics drop but everyone says they're having more fun, metrics will follow."

"如果指标下降但每个人都说更开心，指标会跟上来。"

Lenny's Podcast, 2025-02-02

5. 关于 hiring、talent 与德国学徒制

5.1 学徒制是被低估的人才路径

"The German dual education system: alternating between school and workplace. Rotated through different departments: cafeteria, accounting, inventory, reception. Learned by immersion rather than lecture. Advocates apprenticeship over traditional college paths."

"德国双元制教育系统：在学校和工作单位间轮转。我在不同部门间轮岗：食堂、会计、库存、前台——靠浸入式学习，而不是课堂。我至今主张学徒制比传统大学更值得。"

Tim Ferriss #359 / The Apprentice Programmer

5.2 Jürgen Saar 的方法：把人推过自己感知的能力边界

"Jürgen Saar, my programming apprentice mentor at Siemens, sent 17-year-old me alone to customer meetings."

"我在 Siemens 的程序员学徒导师 Jürgen Saar，把 17 岁的我一个人派去见客户。"

Tim Ferriss #359

"He created scenarios where Tobi had to exceed his perceived capabilities."

"他制造一些场景，让我必须超过自己感知中的能力上限。"

The Apprentice Programmer, 2013-03-03

5.3 招聘看人生故事，不看简历

"Hiring Approach: Requests candidates' life stories and decision-making moments rather than traditional interview questions."

"招聘方法：让候选人讲生命故事与决策时刻，而不是问传统面试题。"

The Profile Dossier, ~2023

5.4 Hiring for potential（在 secondary markets 尤其有效）

"In secondary markets (outside Bay Area), hiring for future potential yields greater returns than current skill. Employee retention is significantly higher outside tech hubs."

"在二级市场（湾区之外），按未来潜力招人的回报比按当前技能招更高。在科技中心之外，员工留存率显著更高。"

Knowledge Project #41

5.5 "Better fit elsewhere" 框架

"Mediocre performers should be freed: 'Someone who is a mediocre performer at your podcast would likely be absolutely badass somewhere else."

"中等表现者应该被释放：'在你这里中等的人，在别的地方可能完全是一个 badass。"

The Shawn Ryan Show #261

5.6 关于潜能：每个人都远离"上限"

"Everyone is 'far from reaching their maximum potential' — reminding people of this constantly is valuable."

"每个人都'远未到达自己潜能的上限'——不断提醒身边人这一点是有价值的。"

Lenny's Podcast, 2025-02-02

"There is not a single person on this planet who is even close to being at their maximum potential."

"这个星球上没有任何一个人接近自己的潜能上限。"

Lenny's Podcast

5.7 Courage > IQ

"The rarest thing in the world is not even creativity or genius. It's courage. So let's lower the net amount of courage needed."

"世界上最稀缺的不是创造力，也不是天才——是勇气。所以让我们降低系统中需要的总勇气量。"

Lenny's Podcast, 2025-02-02

"Disagreeing with the CEO is immediately rewarded with more trust."

"在我面前不同意 CEO，会立刻得到更多信任。"

Lenny's Podcast

5.8 不做最高薪资 bidder

"Avoids being highest compensation bid because 'people who optimize for that above all else actually don't do well at the company'."

"刻意不做最高薪 bidder——因为'把这件事放在其他一切之上来优化的人，反而在我们公司过得不好'。"

Founders (David Senra), 2026-01-18

"Replaced complex stock options with radical simplicity. Transparent annual salary + quarterly sliders allowing employees to choose cash, equity, or charitable allocations."

"把复杂的 stock options 用激进的简单替换：透明的年薪 + 季度滑杆，员工自己选现金/股票/慈善。"

Knowledge Project #152

"Eliminates arbitrary windfalls based on grant timing disconnected from actual performance."

"消除了那种依赖授予时间运气、与实际业绩脱钩的随机财富。"

Knowledge Project #152

这套 compensation flexibility 上线的当年（2022 年 9 月），他在 This Week in Startups 上讲了推行它的动机和他自己对它的不确定——这段值得单独记，因为它是他少见的"公开承认自己在拿组织做未验证实验"的表述：

"I have a particular propensity for wanting to rethink things, especially if everyone does the same thing. I get really suspicious of it. There's usually a way to do it better. By the way, I don't know if this is better, but I will report back. There are clear downsides that people can level at it. For instance, it's clearly not as retentive as a free option grant."

"我有一种特别的倾向，就是想重新想一遍事情，尤其是当所有人都在做同一件事的时候。我会变得非常怀疑。通常总有更好的做法。顺便说一句，我不知道这样是不是更好，但我会回来汇报。它有明显的缺点，别人可以拿来指摘。比如说，它的留人效果显然不如白送的期权授予。"

This Week in Startups E1568（Jason Calacanis）, 2022-09-24（00:31:57）

而他之所以愿意付"留人效果更差"这个代价，是因为他不认为"留住"本身是好的——他把离职拆成了三种：

"I wrote one on leaving well, which is: there's basically three things you can do in a company. You can leave too early, which sucks for you. You can leave too late, which actually sucks for you and for the company. And you can leave right on time, which works for everyone. And the sort of retentive component of stock options — like the golden handcuffs — they often just simply cause the bad sides of these things. People who actually really want to be somewhere else stay around because they have to. And honestly, that just doesn't work for the company or you. You become more bitter over it. I think everyone should be working where they can be a superstar."

"我写过一篇讲'好好地离开'的。在一家公司里你基本上只有三种可能：你可以走得太早，这对你不好；你可以走得太晚，这对你和公司都不好；你也可以正好在对的时候走，这对所有人都好。而期权里那种留人的成分——就是所谓的金手铐——它们往往就只是把这些事的坏的一面放大。真心想去别处的人因为不得不留而留下来。老实说，这对公司和对你都行不通。你会越来越苦涩。我认为每个人都应该在自己能当 superstar 的地方工作。"

This Week in Startups E1568, 2022-09-24（00:32:54）

访谈者 Jason Calacanis 当场把这条总结成"给员工 agency 就把亲子关系换成了同侪关系"，Tobi 回答 "That's exactly what I'm going for. You're spot on."——这与 §2.18 的 "team, not a family" 是同一条线的两个时点（2022 年谈薪酬结构、2023 年谈政治中立），都在拒绝把公司类比成家庭。

5.9 Compensation incentive failure = CEO skills issue

"If incentive system tells you to behave in a way that is not congruent with the company's goals, that is my skills issue in designing the conversation, not your fault."

"如果激励系统让你做出与公司目标不一致的行为，那是我设计对话的能力问题，不是你的错。"

Founders (David Senra), 2026-01-18

5.10 Be surrounded by people you admire

"We want to be surrounded by people we admire."

"我们希望被自己敬佩的人包围。"

Founders (David Senra), 2026-01-18

"My job is to make a company worthy for the best and brightest to work for."

"我的工作是把公司打造成值得最好、最聪明的人来效力的地方。"

Founders (David Senra), 2026-01-18

5.11 Manager / crafter 双轨

"Distinguished between 'crafter tracks' (individual contributors) and 'manager tracks' (team builders). Restructuring aims to restore focus on outcomes, reduce 'heavy layers of process, approvals, meetings'."

"区分 crafter（独立贡献者）和 manager（团队组织者）双轨。重组目标是回到结果导向，减少'层层流程、审批和会议'。"

Shopify 2023-05-04 内部备忘

5.12 Recruit founders to fix post-IPO crisis

"Recruited founders and individual contributors from engineering to replace executives, citing that 'every one of those things worked'."

"我从工程组里招创业者和 individual contributor 来替换高管——每一次都奏效。"

Founders (David Senra), 2026-01-18

5.13 流体智力 vs 晶体智力：他真正在面试里找的东西

2025 年在 Tetragrammaton 上，Rick Rubin 问他"招资深的人 vs 招聪明的新手，差别在哪"。Tobi 的回答给出了本档案里唯一一条可复现的面试技术，以及一条关于职业衰减的冷峻观察。

先是定义：

"There are two kinds of intelligences, I think. There's fluid intelligence and then there's crystallized intelligence. Fluid intelligence is essentially a sort of curiosity-driven, like, I want to figure this out. It's just: how quickly can you pick something up? Do you like playing with ideas?"

"我认为有两种智力：流体智力和晶体智力。流体智力本质上是好奇心驱动的那种——我想把这个搞明白。它就是：你能多快学会一样东西？你喜不喜欢玩弄想法？"

Tetragrammaton with Rick Rubin, 2025-06-25（01:13:11）

"Then there's crystallized [intelligence], which really is wisdom. If you've been programming for a long time, you've picked up a lot of patterns, you know what not to do. In so many situations, you can say the words, 'oh, this is another one of those,' and already have an idea for how to solve it because you've encountered the problem before."

"然后是晶体智力，它其实就是智慧。如果你编程很久了，你积累了大量模式，你知道什么不该做。在非常多的情形下你可以直接说'哦，这又是那一类'，并且已经有解法的雏形了，因为你以前遇到过这个问题。"

Tetragrammaton, 2025-06-25（01:14:12）

具体的面试手法——他不问经验，他人为地破坏对方的经验：

"In my interviews with people, I try to find a thing that they are very passionate about and then I make changes to what they're passionate about in such a way that they have to reimagine something. If someone is really into soccer, for instance, I'm like: okay, we all wake up, this crazy thing happened, Earth's core stopped rotating quite as quickly, and gravity is now 0.7 of a G, so 30% less. Which positions in the game of soccer end up being more important? I have no idea what the right answer is, but I want to play with ideas. That's fluid intelligence in this moment."

"在面试里，我会找到一件他们非常有热情的事，然后对这件事做出改动，改到他们必须重新想象某样东西。比如某人特别喜欢足球，我就说：好，我们某天醒来，发生了一件疯狂的事，地核转速慢下来了，重力现在是 0.7 个 G，少了 30%。足球比赛里哪些位置会变得更重要？我根本不知道正确答案是什么，但我想玩弄想法。这个瞬间发生的就是流体智力。"

Tetragrammaton, 2025-06-25（01:13:33）

然后是最尖锐的一段——AI 让这两种智力的相对价值突然重估：

"In the world of AI, AI comes and suddenly everything is new. Suddenly the patterns end up not actually mattering so much. And I had this incredibly interesting experience where some of our best engineers, we put on a project that really, really, really needed to succeed — AI projects — and they kind of didn't contribute. And we're like, well, hold on, why is that happening? And then we put an intern on it, and everything just goes incredibly well."

"在 AI 的世界里，AI 来了，突然之间一切都是新的。突然之间那些模式其实没那么重要了。我有过一次极其有意思的经历：我们把一些最好的工程师放到一个非常非常需要成功的项目上——AI 项目——结果他们几乎没有产出。我们说：等一下，这是怎么回事？然后我们放了一个实习生上去，一切进行得好得不可思议。"

Tetragrammaton, 2025-06-25（01:14:48）

"What I've observed in these 20 years of building Shopify is that some people are okay with the overall amount of intelligence they bring to bear to be the same. And slowly it's all converting from fluid to crystallized. They are day-to-day just as helpful, but they've stopped cultivating their skills. They just stopped reading the books, tinkering with new technologies. And you just don't know. And then everything changes and suddenly you find out very quickly. But some people keep it and then they compound."

"我在建 Shopify 这 20 年里观察到：有些人可以接受自己带来的智力总量保持不变。而它在缓慢地全部从流体转成晶体。他们日常一样有用，但他们停止了培育自己的技能。他们只是不再读书、不再折腾新技术了。而你是看不出来的。然后一切改变了，你会非常快地发现。但有些人守住了流体的那部分，于是他们复利。"

Tetragrammaton, 2025-06-25（01:15:23）

[推断] 这条把 §5.4（hiring for potential）、§5.6（每个人都远离潜能上限）和 §9.3（AI memo 要求 reflexive usage）串成一条因果链：AI memo 的强制性不是为了提高工具熟练度，而是因为他相信流体智力是会静默流失的资产——"你是看不出来的，然后一切改变了，你会非常快地发现"。从这个角度读，那份 memo 更像是一次全公司范围的流体智力压力测试，而非技能培训。这也给 §六.6（memo 作为招聘过滤器的批判性解读）提供了 Tobi 自己一侧的对应机制说明。

6. 关于商业模式

6.1 Subscription + Transaction 双层

Shopify 的商业模式自 2007 年改成订阅以来一直是两层：

• Subscription Solutions：固定月费 SaaS（基础 $29 起，Plus $2,000+，Enterprise 1,000+）
• Merchant Solutions：交易相关收入（Shopify Payments 抽成、Shipping、Capital、Apps）

"2025: Subscription solutions: $2.75B (up from $2.35B in 2024); Merchant solutions: $8.80B (up from $6.53B). Gross profit: $5.55B (up from $4.47B). Operating income: $1.46B (up from $1.07B)."

"2025 年订阅服务 27.5 亿美元（2024 为 23.5 亿）；商家服务 88 亿美元（2024 为 65.3 亿）；毛利 55.5 亿美元；营业利润 14.6 亿美元。"

Shopify Q4 2025 Financial Results

6.2 平台分账：保留极少

"Left all economics on the table for third-party app developers."

"我们把所有经济利益都留在桌面上让给第三方 app 开发者。"

Masters of Scale, 2020-12-15

"2021: Removed 20% revenue share for developers earning under $1M annually."

"2021 年：取消年收入低于 100 万美元的开发者的 20% 分账。"

Wikipedia (Shopify)

6.3 不为收入而做产品

"We do not chase revenue as the primary driver of our business."

"我们不把收入当作业务的主要驱动力。"

Shopify IPO Letter, 2015

"Long-term value over short-term revenue opportunities."

"长期价值优先于短期收入机会。"

Shopify IPO Letter, 2015

6.4 反 marketplace、做 anti-Amazon

"Amazon is trying to build an empire, and Shopify is trying to arm the rebels."

"Amazon 在试图建一个帝国，Shopify 在试图武装反叛者。"

Bloomberg Businessweek, 2021-12-23

"Shopify positioned as the primary alternative to Amazon's marketplace model. Merchants retain their own customer relationships, brand identity, and data. Shopify provides infrastructure; Amazon provides traffic."

"Shopify 是 Amazon marketplace 模型的主要替代——商家保留自己的客户关系、品牌、数据。Shopify 提供基础设施；Amazon 提供流量。"

Bloomberg Businessweek, 2021-12-23

"Amazon pursues customers and brings merchants onto its platform on its own terms; Shopify gives merchants opportunity to differentiate."

"Amazon 追求消费者并按自己的条款把商家拉上来；Shopify 给商家差异化的机会。"

Ben Thompson, Stratechery, 2022

6.5 Stripe 13 年合作

"13-year partnership maintained through focus on respective core competencies. Draws parallels to NVIDIA's focused pursuit of GPUs for decades before AI applications emerged. Founder-driven companies outperform those optimizing for diversification."

"13 年的合作之所以维持，是因为各自专注核心能力。类比 NVIDIA 几十年专注 GPU 直到 AI 应用爆发。创始人驱动的公司比追求多元化的公司表现更好。"

The Cheeky Pint, 2025

2023 年 All-In Summit 上 David Sacks 问他，为什么平台上会共存 86 个不同的支付网关。Tobi 的解释是：碎片化来自 interchange（交换费）与监管固化的"拜占庭式"世界，许多网关是建在过时技术上的遗留系统；而 Stripe、Adyen 这类现代供应商真正的差别，在于处理幕后复杂度的能力。他由此给出一条一手工程判断——支付真正的差异化变量不是费率，而是性能：

"Performance is the silent differentiator; latency in payment processing is a negative externality that directly impacts a merchant's sales. When a merchant switches gateways, everything about their business performance changes."

"性能是那个沉默的差异化因素；支付处理里的延迟是一种负外部性，会直接打到商家的销售额上。当一个商家换掉支付网关，他生意的所有表现都会随之改变。"

All-In Summit, 2023-09-13

[推断] 这条"acceptance rate 与 latency 最终都以商家可见的营收结算"的判断，正是他对 [[patrick-collison]] 的 Stripe 保持 13 年合作的工程理由——在这一层，毫秒和百分点都会换算成商家的钱。

6.6 Shop Pay 是消费者品牌入口

"Shop Pay alone delivers 16% conversion increase."

"仅 Shop Pay 一项就能带来 16% 的转化率提升。"

The Cheeky Pint, 2025

"Over 150M Shop Pay users (as of 2024)."

"Shop Pay 用户超过 1.5 亿（2024 年）。"

Wikipedia (Shopify)

6.7 Merchants for new entrepreneurs（不是已有零售商）

"Tobi built for new entrepreneurs who could use lunch breaks to build businesses."

"Tobi 是为'用午休时间就能开店的新创业者'而做的。"

Lenny's Podcast, 2025-02-02

"Existing e-commerce served existing retailers with complex logic."

"已有电商方案是服务有复杂业务逻辑的成熟零售商。"

Lenny's Podcast

6.8 Ads 不是威胁，是 channel

"Not viewing Facebook as a threat but a channel: 'Facebook and Instagram are the communications channels of choice for many entrepreneurs'."

"我不把 Facebook 看成威胁，而是 channel：'Facebook 和 Instagram 是很多创业者首选的沟通渠道'。"

Fortune, 2020-05-22

"Personalized ads are win-win: monetize free platforms while showing relevant products to users."

"个性化广告是双赢：让免费平台赚钱，同时给用户看到相关产品。"

Stripe Sessions, 2025-10-06

"Meta/Shopify advertising relationship 'has created more businesses than any government policy in history'."

"Meta–Shopify 的广告合作'创造的企业比历史上任何政府政策都多'。"

Stripe Sessions, 2025-10-06

6.9 Stablecoins 作为下一层支付

"Working to enable stablecoin payments across all merchants. Merchants settle directly into USD through Bridge. Predicts stablecoins will eventually feel as natural as QR codes."

"我们在让所有商家都能接受 stablecoin 支付——通过 Bridge 直接结算成美元。我预测 stablecoin 最终会像 QR 码一样自然。"

The Cheeky Pint, 2025

6.10 客户粘性的本质

"Quality products solve consumerism: 'The thing that solves consumerism is quality products which you want to keep using'."

"高质量产品本身就解决消费主义：'解决消费主义的东西，是那些你愿意继续使用的高质量产品。"

Stripe Sessions, 2025-10-06

"Consumerism comes from people throwing away things that they probably shouldn't. But they throw away things because they hate the things that they have."

"消费主义来自人们扔掉本不该扔的东西。他们扔，是因为讨厌手里那些东西。"

Stripe Sessions, 2025-10-06

6.11 读经济：消费者 trade-down 一档，而非退出

被 David Sacks 问到 Shopify 这个巨大 vantage 看到的消费与衰退信号时，Tobi 先给了一条诚实的限定——Shopify 不是经济的完美观测者，因为它的商家卖的是"人们想要"的东西，而不是"人们需要"的东西：

"The products sold on Shopify are not the products people need; they are the products people want."

"在 Shopify 上卖的产品，不是人们需要的东西，而是人们想要的东西。"

All-In Summit, 2023-09-13 [00:02:36]

但也正因如此，Shopify 看到的是需求侧最细腻的一层。他给出的衰退读法不是"消费者离场"，而是整体下移一档质量档位：从中端品牌换到 fast fashion，从平台上的 challenger brand 换到 Walmart 这类大众零售商。宏观数字本身往往看不出"崩塌"，但数字底下的动态一直在变——消费者在不断调整自己花钱的方式，而这比传统宏观图表更能揭示经济的真实健康度。

[推断] 这条"看数字底下的动态、而非数字本身"的读法，与 §7.17（注意力是最稀缺资源）、§7.18（侦察式获取信息）同源——都是他工程师式"不信聚合指标、要看底层信号"的认知习惯在宏观层面的一次应用。

6.12 互联网上你只能拥有两样东西：你的域名和你的邮件列表

这是他给"为什么不做 marketplace"最凝练的一次论证——不是从竞争格局出发，而是从所有权出发：

"Your store is your own thing. It's sovereignty. There's only two things you can own on the internet: your own domain and your mailing list. Everything else is leased. And so Shopify is the thing you put on your own domain. It's your design. You make the full margin."

"你的店是你自己的东西。这是主权。在互联网上你只能拥有两样东西：你自己的域名，和你的邮件列表。其余一切都是租来的。所以 Shopify 是你放在你自己域名上的那个东西。是你的设计。你拿到全部的 margin。"

Tetragrammaton with Rick Rubin, 2025-06-25（00:23:39）

他在同一段里对比了 marketplace 的经济学："在 marketplace 上卖，费用像税一样，通常 15% 左右，薄如剃刀——很多品类根本在这个数字下运转不了，有的还更高。"

他也主动承认了这个二分法在现实中是模糊的——这一点很少被引用：

"I'm going to complicate things a little bit, because often when they buy something from Amazon, that order flows into a Shopify store and then Shopify sells it out. If you buy something off Instagram, there's a lot of crossover. Software is complicated. It's never quite like this one thing it does."

"我要把事情稍微搞复杂一点：很多时候人们从 Amazon 买东西，那个订单流进的是一个 Shopify 店，再由 Shopify 发出去。你从 Instagram 买东西，也有大量的交叉。软件是复杂的。它从来不是只做那一件事的样子。"

Tetragrammaton, 2025-06-25（00:24:01）

两个月后（2025-12）他在 Intrepid Growth Partners 上用几乎同样的措辞讲了一遍，可见这是稳定表述而非临场发挥：

"We kind of nailed it with online stores, because they have these little outposts of sovereignty for customers. It's one of the very few things you can own on the internet: it's your website, your design, your brand."

"我们在在线店铺这件事上算是做对了，因为它们是给客户的一个个主权小前哨。这是互联网上你能拥有的极少数东西之一：你的网站、你的设计、你的品牌。"

Intrepid Growth Partners（Ajay Agrawal 主持）, 2025-12-19（00:06:11）

[推断] 这条与 §6.4（反 marketplace / 做 anti-Amazon）是同一立场的两种论证路径：§6.4 是竞争性的（不与自己的商家竞争），这条是产权性的（租户 vs 业主）。后者更根本，也更能解释他为什么在 §9.14 里能坦然设想"Shopify 不再被需要"——他捍卫的是商家的所有权结构，不是 Shopify 的中介地位。

6.13 消费主义不是因为人们太爱东西，而是因为人们讨厌拿到的东西

一条对"消费主义"的非常规重构。前提是他对旧零售渠道的诊断：

"If you made a product, you had meetings with the merchandising executives at Walmart and Target, and they had limited shelf space... The channel by which products could be sold was one which optimized away everything that was interesting for some, but made it appealing to the most."

"如果你做了一个产品，你得去和 Walmart、Target 的采购主管开会，而他们的货架空间有限……产品得以被销售的那个渠道，会把'对一部分人有意思'的东西全部优化掉，只留下'对最多数人还算有吸引力'的东西。"

Tetragrammaton with Rick Rubin, 2025-06-25（00:32:46）

结论：

"A lot of the consumerism that people sort of vilify — it's not because people love stuff too much. It's because people hate the stuff they get. The quality is not there."

"人们妖魔化的那种消费主义，很多并不是因为人们太爱东西。是因为人们讨厌他们拿到的东西。质量不在那里。"

Tetragrammaton, 2025-06-25（00:33:15）

他在 2025-12 的 Intrepid 场上原样重复了这句（01:15:47），并在那里明确加了一句"我认为这（对消费主义的批评）是成立的批评"——即他不否认现象，只重新归因。

[推断] 这条把 §1.2（产品质量 = 创造它的人有多在乎）从产品伦理升级成了一条宏观命题：如果人们买得太多是因为买到的东西不够好，那么"降低创业门槛"就不只是经济政策，而是环保与消费主义的解法。这也解释了他为什么把长尾（"如果世界上有一千个人感兴趣，那就能是一门好生意"）当作道德论据而非市场论据。

6.14 Shopify Capital：平台数据比银行更早、更准地看见小企业

2023 年在 The Network State Podcast 上，他给出了 Shopify 做贷款的理由——不是新收入线，而是被迫复刻缺失的机构：

"So we have to kind of replicate these institutions inside of our own mini economy. Now, we can do it because you're so early part of the journey of our merchants... we see the traffic build up, we have inventory visibility. So we can underwrite people massively better than what banks could do. But this is only a path-dependent consequence of us being more motivated to do it, because we are doing it as a complement to the mission rather than [as a] bank — I mean, because banks have to for their charter."

"所以我们不得不在我们自己的迷你经济体内部复刻这些机构。我们做得到，是因为我们处在商家旅程非常早的位置……我们看得到流量怎么起来的，我们有库存的可见度。所以我们对人的授信可以做得比银行好太多。但这只是一个路径依赖的结果——我们更有动力去做，因为我们是把它当作使命的补充在做，而不是当作银行——我是说，银行是因为牌照义务才做。"

The Network State Podcast #2（Balaji Srinivasan 主持，Kaz Nejatian 同场）, 2023-02-15（00:25:34）

同场的 Kaz Nejatian（时任 Shopify COO）补了一条个人化的注脚：他在做成第四家公司之前有三家 POS 创业失败——"如果我当时是一个硅谷不理解的行业的创始人，第一次失败就会让我破产，我不会去做第二、第三、第四次。"

[推断] 这条补上了 §6.1（subscription + transaction 双层）没覆盖的第三层：资本。它也和 §7.22 的 leaky bucket 命题咬合——如果经济活力取决于小企业补充速度，那么"资本获取"就是漏桶最大的那个洞，而平台的数据尾气恰好是唯一能低成本堵上它的东西。⚠️ 需要注意这是 Tobi 的自述，未经第三方对 Shopify Capital 违约率的独立核验。

7. 关于哲学与系统思维

7.1 Programming 是树干

"Tree metaphor for skills: programming is the trunk, business skills are branches, observable outputs are leaves."

"技能的树木隐喻：编程是树干，商业能力是枝，可观察到的输出是叶。"

The Shawn Ryan Show #261

"I only succeeded in this because I was a computer programmer."

"我能成功，唯一的原因是我是程序员。"

Tobi Lütke, John Coogan profile

7.2 Companies as social technologies

"Views companies as technology themselves — social structures enabling groups to pursue shared missions."

"公司本身就是一种技术——一种让群体追求共同使命的社会结构。"

The Cheeky Pint, 2025

"Companies are path-dependent solutions to social and legal problems — the modern company is not that old and will evolve."

"公司是路径依赖的产物，针对社会和法律问题的解决方案——现代公司其实并不古老，它会演化。"

Founders (David Senra), 2026-01-18

7.3 反 cargo culting

"Cargo culting: 'We copy behaviors to try to create the same outcomes' without understanding underlying principles."

"Cargo culting：'我们复制行为以求得到同样的结果'——但没理解底层原理。"

The Shawn Ryan Show #261

7.4 Verschlimmbesserung（让事情变更糟的"改进"）

"Verschlimmbesserung — improving something so badly you make it worse."

"Verschlimmbesserung——把一件事'改得'更糟。"

Conversations with Tyler, 2024-09-18

"Many corporate initiatives are Verschlimmbesserungen — well-intentioned changes that create new problems."

"许多企业举措是 Verschlimmbesserung——好意但制造了新问题。"

Inc, ~2024

Tobi 在 Tyler Cowen 那期把这个词与 Chesterton's Fence 直接连了起来——给出了他对"为什么改革常常变更糟"的完整因果链条：

"Verschlimmbesserung is such a good word. What that means is by trying to improve something, you made it worse. It's just so wonderful because you see it often. Chesterton's fence — people don't often know what parts of large systems are important parts that have a lot of cultural or technical understanding coded in them."

"Verschlimmbesserung 真是个好词。意思是：你想改进一个东西，结果反而把它弄糟了。这个词太美妙了，因为你经常会看到这种事——Chesterton 的栅栏——人们经常不知道一个庞大系统里哪些部分其实是承载了大量文化或技术理解的重要部件。"

Conversations with Tyler, 2024-09-18

同一期他给出了一个把"语言学差异"操作化为"管理工具"的具体做法——用德语里有或没有某些词来给 Shopify 团队教学：

"Entitlement is not a word that Germans have, which I find really, really interesting. Sometimes in Shopify, I have to explain to people when I sit down, it's like, hey, gratitude and entitlement are two sides of the spectrum."

"Entitlement'（理所应得）这个词德语里没有，这一点我觉得非常非常有意思。有时候在 Shopify 我得给人解释——感恩和理所应得是同一个光谱的两端。"

Conversations with Tyler, 2024-09-18

7.5 Today is the dystopia of the future

"Today is the dystopia of the future — this mindset motivates building toward progress."

"今天就是未来的反乌托邦——这种心态驱动你为进步而建造。"

Lenny's Podcast, 2025-02-02

7.6 Steven Pinker 式的乐观主义

"Steven Pinker's book just drops every single stat you have seen quoted as going worse and shows you it actually gets better."

"Steven Pinker 的书把你听过的每一个'变差'的数据，逐条拿出来，告诉你它其实在变好。"

20VC #6727, 2023-04-03

"The hedonistic treadmill is the engine of humanity."

"享乐适应（hedonistic treadmill）是人类的发动机。"

20VC #6727

7.7 反"efficiency"绝对化

"Efficiency obsession is detrimental: forces actions based on quantifiable metrics, overlooking taste and quality."

"对'效率'的偏执是有害的——它逼你只按可量化指标行动，忽略了品味和质量。"

Stripe Sessions, 2025-10-06

7.8 Karl Popper / 最佳问题是无法解决的问题

"The best gift in life is finding a beautiful problem that you can never solve."

"人生最好的礼物，是找到一个你永远解决不完的美丽问题。"

Stripe Sessions, 2025-10-06

7.9 Decentralization 是真正的"人民的力量"

"It's actually the true power of the people is decentralization. And the false one is communism."

"真正属于人民的力量是去中心化；假的那种是共产主义。"

a16z Podcast, 2024-12-16

7.10 阅读是终生 multiplier

"If you don't read books, you live a lifetime. If you read books, you live a thousand."

"如果你不读书，你只活一辈子。如果你读书，你活一千辈子。"

Founders (David Senra), 2026-01-18

"Books are 'cheat codes for real life'."

"书是'现实生活的作弊码'。"

Founders (David Senra), 2026-01-18

值得记的是：这句"cheat codes for real life"其实早在 2018 年 The Knowledge Project #41 就已成型——同一期 Tobi 也罕见地披露了他读书起步极晚的事实（因 dyslexia，头 20 年几乎不读书），以及他破冰的第一本书：

"It came pretty late. I'm actually dyslexic, so it's hard for me to read, and I read very slowly. So I didn't read any books for the first 20 years of my life. Then at some point I challenged myself to read a book fully, and that was a wonderful book called Cryptonomicon by Neal Stephenson. It's still one of my favorite books... I found books are the closest you'll ever come to finding cheat codes for real life."

"我读书起步相当晚。我其实有阅读障碍，所以读起来很难、非常慢。人生头 20 年我几乎没读过书。后来某一刻我给自己下了个挑战——完整读完一本书——那是一本很棒的书，Neal Stephenson 的 Cryptonomicon。它至今仍是我最喜欢的书之一……我发现，书是你能找到的、最接近'现实生活作弊码'的东西。"

The Knowledge Project #41, 2018-09-18

[推断] 这与 §一 童年"kinesthetic learner / dyslexic"的线索闭合：他后来对书的热爱不是天赋顺延，而是一次刻意的、逆着障碍的自我训练——和他"把人推过自己感知能力边界"的 Jürgen Saar 方法论是同构的。

在 Tyler Cowen 那期 Tobi 公开了一条他罕见的、对自己阅读纪律的具体规则——刻意保留母语的认知通道：

"Occasionally, I must read one book in German a year as a sort of self-policy, which I have violated last year, which I was not super happy about. I'm actually a fan of rewriting books, like every 25, 50 years for the next two generations down the line, because they just get hard to access."

"我有一条自律：每年至少要读一本德语原书。去年我违反了这条规则，我对此并不开心。其实我支持每 25 年、50 年重新翻译一次经典——给下两代人——因为它们会变得难以读懂。"

Conversations with Tyler, 2024-09-18

7.11 Tobi 推崇的几本书

• High Output Management (Andy Grove) — 把经营公司重塑为工程练习
• Influence (Robert Cialdini) — 故事与 framing 的力量
• The Box (Marc Levinson) — Malcolm McLean 的集装箱：解决"跨运输模式搬运货物"的问题，而不是各自模式的优化
• Mindset (Carol Dweck) — 内化 growth mindset
• A Conflict of Visions (Thomas Sowell)
• Seeing Like a State (James C. Scott)
• The Machiavellians (James Burnham)
• Brandon Sanderson 系列 — 把奇幻世界当组织的镜像

"High Output Management (Andy Grove) — positioned business as engineering exercise, made CEO role less scary."

"High Output Management 把生意重塑成工程练习，让 CEO 这个角色没那么可怕。"

Tim Ferriss #359

"The Box (Marc Levinson) — Malcolm McLean's shipping container: solved the actual problem (moving goods across transport modes), not optimizing individual modes."

"The Box 讲 Malcolm McLean 的集装箱——他解决的是'跨交通模式搬货物'的真正问题，而不是优化某一种交通方式本身。"

Tim Ferriss #359

7.12 SAT solver 用于公司组织

"Uses SAT solvers and GitHub to model organizational structure computationally (Shopify OS project)."

"使用 SAT solver 和 GitHub 来计算式地建模组织结构（Shopify OS 项目）。"

Founders (David Senra), 2026-01-18

7.13 Studies Walt Disney as understudied tech leader

"Studied Walt Disney as someone understudied in technology."

"他把 Walt Disney 当作'在科技史里被研究不足'的人物来学习。"

20VC #6727

7.14 Affirmations 重写 self-identity

"Uses affirmations: believes they can 'rewrite one's self-identity in the frontal cortex'."

"他用 affirmation——他相信可以'在前额叶里重写自我身份'。"

Founders (David Senra), 2026-01-18

7.15 Spaced repetition 给自己

"Uses scheduled messages to himself for spaced repetition of key ideas."

"他给自己安排定时短信，对关键想法做间隔重复。"

Founders (David Senra), 2026-01-18

7.16 局部最优陷阱：好选择反而把你卡进死胡同

在 Invest Like the Best EP.393 上 Tobi 给出了他对"为什么需要不断重新推导决策"的最上游解释——这条比"反 sunk cost"更深一层：

"Most of the time, you end up in a bad part of a tree, in a local maxima of a path-dependent environment by only making good choices. People think that making a good choice inoculates you from making mistakes, or that the presence of a kind of downside of an idea ends up disqualifying the idea. Both of those things are incorrect."

"大多数时候，你最后陷在决策树的某个糟糕分支里、陷在一个路径依赖环境的局部最优里——靠的是不停做好的选择。人们以为'做了一个好选择'就能让自己免于犯错，或者'一个想法有缺点'就足以否决它。这两种想法都错了。"

Invest Like the Best EP.393, 2024-10-22

含义：好决策不能保证你处于正确位置。这与他在 §3.2 "决策容易，找输入难" 是同一条线，但落在"为什么必须周期性 reorg、为什么必须 subtraction"的更深层逻辑上。

7.17 注意力作为商业里最稀缺的资源

2020 年 5 月在 Invest Like the Best EP.173 上，Tobi 给出了他对商业本质的一句 RTS 类比：

"There is another resource that is way more limited and is way more important, which is your attention. It ends up being a real-time strategy game, especially played at a very high level one-on-one. It's really that you're tested of how good are you at paying attention to what's going on?"

"还有一种资源比金钱稀缺得多、也重要得多——那就是你的注意力。商业最终变成一个即时战略游戏，特别是在高水平的一对一对抗里。真正考验的是你有多擅长留意正在发生的事情。"

Invest Like the Best EP.173, 2020-05-12

更尖锐的延伸：商业竞争的高阶状态，是主动消耗对手的注意力：

"At a later level it gets to the point where you actively do things to your opponent which are designed to just drain their well of attention. You're making stuff they have to pay attention to so that they can actually do as good a job as they would otherwise do on their normal game."

"到更高的阶段，你会主动对对手做一些事情——专门设计来掏空对方的注意力井。你制造他们不得不关注的东西，让他们没法在正常的'本职动作'上保持原本的水准。"

Invest Like the Best EP.173, 2020-05-12

7.18 侦察作为信息获取的方法

同一期 Tobi 把他作为 CEO 的工作方式机制化成了一个 RTS 隐喻——刻意去找原本不会主动到达自己的信息：

"I basically do the equivalent of what I described in the scouting analogy with real-time strategy games. I try to get as many information I'm not entitled to, which wouldn't have made it to me in any other way so that I can make better decisions about things."

"我基本就是在做我刚刚用 RTS 游戏'侦察'类比里描述的那件事。我尽量收集那些'我本来无权得到'的信息——这些信息在常规渠道下不会传到我这里——这样我才能做更好的决策。"

Invest Like the Best EP.173, 2020-05-12

这条与他与 150+ Shopify 商家保持 WhatsApp 群、每周日晚看新创业者使用录像，是同一条工作方式——信息密集型 CEO 的具体机制。

7.19 资本分配 = 揭示出来的真正策略

2022 年 5 月在 Invest Like the Best EP.278 上 Tobi 给出了他对策略与执行关系的一句尖锐总结：

"Capital allocation is the name of the entire game. Funding a team that's working on something is a huge expense. At some point, show me your roadmap and I'll show you your actual strategy."

"资本分配是整场游戏的核心。给一个团队拨经费做某件事是巨大支出。某种程度上——给我看你的 roadmap，我就能告诉你你真正的策略是什么。"

Invest Like the Best EP.278, 2022-05-24

"strategy doc 写的版本" 与 "roadmap 反映的版本" 之间的差异，就是公司里"宣布的策略"与"实际策略"的差异。

7.20 加密货币：克制的逆共识判断

2022 年 5 月（Luna 崩盘前后）在 Invest Like the Best EP.278 上 Tobi 给出了一个 sharp 的中间立场判断——既不全盘看空也不狂热：

"I think a lot of crypto is like big red button that doesn't do a lot. That is not a statement about the future potential of the entire space, far from it. It will unlock completely new potential in somewhat a similar way how the 90s internet did."

"我觉得加密世界里很多东西像是'一个不起多大作用的大红按钮'。这不是在评判整个领域的未来潜力——远不是。它会像 90 年代的互联网一样解锁全新的可能性。"

Invest Like the Best EP.278, 2022-05-24

这条态度与 §6.9 关于 stablecoin 的实际部署是连贯的——Tobi 对 crypto 的判断从一开始就是长期看好底层 + 短期对大多数项目克制。

7.21 Charity 怀疑论：self-healing fitness function

"The problem with charity dollars is that they cannot be given to anything that has a self-healing fitness function—which is a fancy way of saying 'for-profit.' ... When you opt out of markets, you are opting out of the best fitness mechanism humanity has ever invented."

"charity 美元的问题在于：它们没法投给任何拥有自愈式适应度函数的东西——这是'营利性'的花哨说法……当你退出市场，你就退出了人类发明过的最好的适应度机制。"

20VC, 2026-05-04

"Giving money is not virtuous unless it causes the right things."

"捐钱本身并无美德，除非它促成了正确的事情。"

20VC, 2026-05-04

配套论点：non-profit 常以个人的魅力与人脉替代组织的 merit，"attracts the smoothest talkers rather than the builders"（吸引的是最会说话的人，而非 builders）。[推断] 这或可解释他个人慈善行为的形态——倾向 Team Trees 这类可量化、结果导向的捐赠，而非设立基金会。

7.22 漏桶理论：经济活力 = 小企业被补充的速度

2023 年在 The Network State Podcast 上，他给出了本档案里最完整的一次"为什么创业本身是宏观变量"的论证。起点是一个反直觉的事实——小企业死亡通常与经营能力无关：

"Small businesses tend to go out of business often for no fault of their own — like circumstances with the people who run the businesses, the family. Changes might be a tragic death, which prevents any kind of continuity, because small businesses are built around their founders in most ways. It's a leaky bucket. And I think the vibrancy of almost every economy depends on how quickly this leaky bucket is being refilled. And standard of living tends to go up if the bucket is filled faster than it leaks out."

"小企业倒闭往往不是自己的错——是经营者本人的处境、家庭的变故。可能是一场不幸的死亡，让任何形式的延续都不可能，因为小企业在大多数意义上都是围着创始人建起来的。这是一只漏桶。而我认为几乎每个经济体的活力，都取决于这只漏桶被重新灌满的速度。如果灌进去比漏出来快，生活水平就倾向于上升。"

The Network State Podcast #2, 2023-02-15（00:04:50）

由此推出他对政策的一个明确排序——摩擦 > 政策：

"We believe that the entrepreneurship in economies actually relates a lot more to the friction of starting businesses than the policies, because the demand for doing it and wanting to do it is constant. The opportunity is variable... The thing that's preventing business from succeeding is often just the sort of Byzantine, almost Kafkaesque complexity of starting these businesses."

"我们相信，一个经济体里的创业活动，和创业的摩擦的关系，其实远大于和政策的关系——因为想做这件事的需求是恒定的。变动的是机会……阻止生意成功的，往往只是开办这些生意时那种拜占庭式的、近乎卡夫卡式的复杂性。"

The Network State Podcast #2, 2023-02-15（00:06:28）

配套的一条人类学式断言（他自己也标为猜测）：

"This is a sort of meta-emergent phenomenon that just exists partly because some people just cannot work for other people. There's probably some kind of gene that gives people the authority problems that necessitate them reaching for independence themselves. This has been true for all of human history. Sometimes it becomes illegal in places — and those places don't last. The fate of entrepreneurs is a bit of a Where's Waldo story throughout human history: it tends to be not recognized or not talked about a lot, but it always plays a very important role."

"这是一种元涌现现象，它之所以存在，部分是因为有些人就是没法给别人打工。大概有某种基因，给了人们那种权威问题，逼得他们自己去够独立。这在整个人类史上都成立。有些地方它会被定为非法——那些地方都长不了。创业者的命运在人类史上有点像一个'找沃利'的故事：往往不被承认、不被大量谈论，但它始终扮演着非常重要的角色。"

The Network State Podcast #2, 2023-02-15（00:05:38）

[推断] 漏桶模型是他"arming the rebels"使命最接近可证伪形式的一次表述：它把 Shopify 的成功指标从"我们有多大"换成了"补充速率有多快"，因而与 §4.2（做新创业者而非抢现有零售商）、§4.4（降低复杂度直接生产更多企业）在同一根轴上。也可与 §8.3 的加拿大生产率批评对读——他对加拿大的不满，本质是认为那里的桶漏得快而灌得慢。

7.23 "Lifestyle business" 是硅谷的建构——而 YC 路径才是"含着银汤匙"

这是他自陈的 "pet topic"，也是本档案里他对创投世界语言最直接的一次反击：

"Maybe one thing to just underline, partly because it's sort of one of my pet topics, is just the concept of lifestyle business. The term 'lifestyle business' is like a construct of Silicon Valley. It's actually pretty mean. I don't know how we renamed businesses to lifestyle businesses just so we could call ventures businesses — and how this succeeded."

"也许有一件事值得强调，部分因为这算是我的一个执念话题，就是 'lifestyle business' 这个概念。'Lifestyle business' 这个词是硅谷的一个建构。它其实相当刻薄。我不知道我们是怎么把'生意'重命名成了'生活方式生意'，就为了能把'风险投资项目'叫做'生意'——而且这居然还成功了。"

The Network State Podcast #2, 2023-02-15（00:11:20）

被 Balaji 追问词源（是不是出自 Basecamp 而非硅谷）后，他没有坚持归因，而是把论点推到了更锋利的地方：

"Maybe without laying the blame at anyone's doorstep — it is a pretty funny thing, because businesses were around for a long time and we didn't call them lifestyle businesses. If anything, you're looking at sort of a techie path of businesses, which is: you start a business by an application to Y Combinator, you get all this teaching, and then afterwards people give you money. That sounds a lot more like silver-spoon businesses to me than what entrepreneurs and SMBs do, because that's a lot more hardcore."

"也许不必把责任推到谁头上——但这挺好笑的，因为生意存在了很久很久，我们并没有管它们叫生活方式生意。真要说的话，你看到的是一条技术圈的创业路径：你通过向 Y Combinator 递申请来开始一门生意，你得到全套教学，然后之后人们给你钱。在我听来那才更像是含着银汤匙的生意——比创业者和中小企业做的事更像，因为后者要硬核得多。"

The Network State Podcast #2, 2023-02-15（00:12:09）

他对自己商家的描述回到了 §5.10（被你敬佩的人包围）的同一句公式：

"Our customers are entrepreneurs... They are heterodox thinkers who just don't take no for an answer and make things happen. I think everyone ends up becoming sort of the average of the five people they spend the most time with — [so] load that number with as many founders as you can."

"我们的客户就是创业者……他们是异端思考者，不接受'不行'这个答案，并且真的把事情做成。我认为每个人最终都会变成他花时间最多的那五个人的平均值——所以尽你所能，把这个数字塞满创业者。"

The Network State Podcast #2, 2023-02-15（00:13:02 / 00:13:26）

7.24 Builders vs critics：电波掌握在批评者手里

同一场里他给出了一条罕见的、关于"叙事分配"的政治经济学观察——而且他明确不是在反对批评：

"One of the biggest issues in our world is that the builders are special in a way that we need to write back into the story. Founders are courageous people who are going on journeys, discovering things, and then bringing back their lessons. We are currently in a world that treats the people who build things and the people who critique things equally — in fact, the airwaves are completely owned by the people who critique things, and because they get to write the story, they're basically more powerful."

"我们世界最大的问题之一，是 builders 的特殊性需要被重新写回故事里。创始人是勇敢的人，他们踏上旅程、发现事物、然后把教训带回来。我们现在所处的世界，把造东西的人和批评东西的人同等对待——事实上，电波完全被批评东西的人所拥有；而因为他们掌握了书写故事的权力，他们基本上更有权力。"

The Network State Podcast #2, 2023-02-15（01:49:05）

"I'm not against criticism. In fact, I think high-effort criticism is one of the most important things to inspire builders to do their thing. But critics are capped in the amount of value they can create for society by holding builders accountable or inspiring builders to do better. So they're really a supportive tissue for builders."

"我不反对批评。事实上，我认为高投入的批评是激励 builders 做事最重要的东西之一。但批评者能为社会创造的价值，上限就在于让 builders 负责、或激励 builders 做得更好。所以他们其实是 builders 的支持性组织。"

The Network State Podcast #2, 2023-02-15（01:49:56）

⚠️ 这条需要与批判章 §7（2026-07 加权投票言论）对读：把"高投入批评"抬为必要、同时把批评者的价值定义为"支持性组织"的框架，在三年后延伸到他对参与市政 town hall 的市民的评价（"useful idiots"）时，失去了这里的克制。是否属于同一立场的激进化，本档案不做判定，但把两处并列存档。

8. 关于身份：移民、外来者、Ottawa

8.1 不在硅谷的 founder

Tobi 一直在 Ottawa 而非 Bay Area。这是他一切判断的物理底盘——而他多次解释这是优势，不是缺陷。

"Built Shopify as an immigrant without Canada's 'supposed to make small plans' cultural narrative."

"我作为移民来建 Shopify——没有沾上加拿大那种'我们应该规划小一点'的文化叙事。"

The Cheeky Pint, 2025

8.2 德国 vs 北美的工作文化对比

"Germans are direct about quality standards — no shit-sandwich configuration needed for criticism. Germans prefer lengthy planning before shipping; North Americans favor rapid iteration."

"德国人对质量标准很直接——批评时不需要'三明治式包装'。德国人喜欢在交付前长期规划；北美人喜欢快速迭代。"

Conversations with Tyler, 2024-09-18

"Germany lacks tech giants due to pessimism about the future and low status of startups: 'You have to be optimistic about the future. Otherwise, why would you want to contribute to progress?"

"德国缺乏科技巨头，是因为对未来悲观，加上创业者社会地位低：'你必须对未来乐观。否则你为什么会想为进步做贡献？"

Conversations with Tyler, 2024-09-18

Tyler Cowen 那期 Tobi 把德国对科技的态度精炼成一句外部视角观察——这是他德国创业失败 + 加拿大成功的根因诊断：

"Tech is something that Americans do from the perspective of Germans. There is very much a pessimism about the future that I think means you cannot build tech companies because you kind of have to be optimistic about the future."

"从德国人的视角看，科技是美国人才会做的事情。德国有一种对未来非常深的悲观主义——我认为这导致你不可能建科技公司，因为搞科技多少必须对未来抱有乐观。"

Conversations with Tyler, 2024-09-18

8.3 加拿大的"go-for-bronze mentality"

"Canada's 'go-for-bronze mentality' reflects being a smaller country attached to a larger one."

"加拿大那种'目标拿铜牌就行'的心态，反映了它作为附属于更大国家的小国身份。"

Conversations with Tyler, 2024-09-18

"Canada exports raw materials (oil, talent) rather than finished products."

"加拿大出口的是原材料（石油、人才），而不是成品。"

Conversations with Tyler, 2024-09-18

"Canada seems okay just exporting the raw materials for everything."

"加拿大似乎只把'原材料'出口给世界就满足了。"

Conversations with Tyler, 2024-09-18

Tyler Cowen 那期 Tobi 给出了对自己所在国家罕见的、有数据支撑的批评——他把加拿大生产率长期低迷归因于裁判/建造者比例失衡：

"Canada has an economy that really can hang with America. The productivity numbers are just really low, and I think the employment in the public sector, having grown the degree it has — if the ratio of referees to builders or critics to builders is going out of whack, that things grind to a halt."

"加拿大的经济本来完全能和美国对标。但生产率数字真的很低，我认为公共部门就业增长得这么厉害——如果裁判和建造者的比例、或批评者和建造者的比例失衡了，事情就会陷入停滞。"

Conversations with Tyler, 2024-09-18

他同时回到加拿大曾经的政策优势——基于积分的技能签证——并对加拿大主动放弃这些先验制度感到不解：

"Canada also implemented the thing that everyone's talking about, a skills-based visa program with point systems, which is well designed and has been doing a lot of work for Canada in the past. It's not quite clear to me why we walked away from these priors."

"加拿大早就实施了所有人现在都在谈的东西——基于技能的、积分制的签证制度——设计得很好，过去为加拿大做了不少功。我不太理解我们为什么把这些先前的判断丢掉了。"

Conversations with Tyler, 2024-09-18

8.4 Canadians have a lot to be proud of

2018 年 5 月 22 日 Tobi 在自己博客上发了一篇 The case for swagger: Canadians have a lot to be proud of，列举加拿大的科技发明：灯泡、电话、velcro、互联网光纤。这是他对加拿大谦逊文化的反弹。

"Canada is a 'beautiful project' and 'beautiful implementation of classical liberalism'."

"加拿大是一个'美丽的项目'，是'古典自由主义的一种美丽实现'。"

Stripe Sessions, 2025-10-06

8.5 Ottawa: 优势 + 局限

"Built Shopify in Ottawa — commuted by skating on the Rideau Canal."

"我在 Ottawa 建的 Shopify——冬天滑冰沿 Rideau Canal 通勤。"

Conversations with Tyler, 2024-09-18

"Ottawa, Canada — a city of 1 million people, insufficient talent pool for a 10,000-person company."

"Ottawa 是一座 100 万人的城市——人才池不够支撑一家一万人的公司。"

Lenny's Podcast, 2025-02-02

8.6 European 价值观对待债务

"European sensibility about money: 'I'm German, so I don't take on much debt."

"欧洲人对钱的敏感度：'我是德国人，所以我不背太多债。"

The Shawn Ryan Show #261

8.7 Open source 童年塑造 positive-sum 直觉

"German-born, describes himself as naturally predisposed to positive-sum thinking from growing up with open source software in the 1990s."

"德国出生，他自述自己天然倾向 positive-sum 思维——这来自 1990 年代成长于开源软件文化。"

Founders (David Senra), 2026-01-18

"Open source positive-sum mental model from childhood shaped business worldview."

"开源 positive-sum 的 mental model 在童年就形成，塑造了我整个商业世界观。"

The Shawn Ryan Show #261

8.8 加拿大 PM Mark Carney 的建议

"Advice to PM Mark Carney: depoliticize infrastructure decisions, leverage resource abundance."

"给加拿大总理 Mark Carney 的建议：让基础设施决策去政治化、利用资源丰富的优势。"

The Cheeky Pint, 2025

"Advocates for Shannon Economic Zone model."

"提倡爱尔兰 Shannon 经济区模式。"

Stripe Sessions, 2025-10-06

8.9 仍是 immigrant 的视角

Tobi 在多个访谈里反复强调，他作为德国移民、不是 Bay Area 的 native——这种"外来者"的距离感是他能反复挑战 industry conventional wisdom 的根因。他对硅谷的"best practices"有天然怀疑。

8.10 关于 Breitbart 与言论自由的德国移民视角

"Position reflects Tobi's German immigrant perspective on free expression."

"他对言论自由的立场，反映了他作为德国移民的视角。"

Tobi Lütke blog: In Support of Free Speech, 2018-08-13

9. 关于 AI

9.1 AI 是 underhyped

"I just think AI is actually underhyped right now and that seems very hard to believe."

"我觉得 AI 现在反而是被 underhype 的——这听起来很难相信。"

The Shawn Ryan Show #261

"AI is like changing everything. There's not a company that's safe in the world and I love it."

"AI 在改变一切。世界上没有一家公司是安全的——我喜欢这种状态。"

The Shawn Ryan Show #261

"The most exciting technology shift in my career."

"这是我职业生涯里最激动人心的技术转折。"

Shopify Blog (Tobi 的话, ~2024)

9.2 AI 不应该让人变笨，应让人变野心更大

"Want my software to inspire people and make them more ambitious than they otherwise would be, not less."

"我希望我的软件激励人，让他们比原本更有野心，而不是更弱。"

The Shawn Ryan Show #261

"AI as a tool to help people achieve their full potential rather than constraining them."

"AI 是帮助人达到完整潜能的工具，不是限制他们的工具。"

Acquired ACQ2, 2025-09-18

9.3 2025 年 3 月 AI 备忘录核心：reflexive AI usage

"Reflexive AI usage is now a baseline expectation at Shopify."

"对 AI 的反射式使用，已经是 Shopify 的基线期望。"

Tobi Lütke 内部 AI 备忘, 2025-03-20

"Frankly, I don't think it's feasible to opt out of learning the skill of applying AI in your craft."

"坦白说，我不认为'选择不学'怎么把 AI 应用到自己职业'是个可行选项。"

Tobi Lütke 内部 AI 备忘, 2025-03-20

"Teams must demonstrate why they cannot get what they want done using AI before requesting additional headcount."

"团队必须先证明 AI 无法完成他们想做的事，然后才能申请额外 headcount。"

Tobi Lütke 内部 AI 备忘, 2025-03-20

"AI usage questions added to performance and peer review questionnaires."

"AI 使用情况已经加入绩效和同行评审问卷。"

Tobi Lütke 内部 AI 备忘, 2025-03-20

Tobi 后来在 Acquired 上对自己 memo 的用意给出了一种equity 框架的解释——他主张强制 AI 使用是出于"对所有员工公平"，而非纯生产力目标：

"I sent a memo to my company and we set the expectation that we require that people reflexively reach for AI now. We require it because it's unfair not to because the people who do are otherwise going to be people who sequester all the best careers to themselves. I think it's fair to tell everyone that that's what it takes now."

"我给我的公司发了一份 memo，我们设定了一种期望：现在我们要求大家反射性地伸手去用 AI。我们之所以要求，是因为'不要求'反而不公平——那些用 AI 的人否则会把所有最好的职业机会都圈在自己手里。我认为告诉所有人'现在做这件事所需的就是这个'是公平的。"

Tobi 在 Acquired 上对 memo 的复述（HR Heretics 转引，2025-05-01）

注意他用的词是 "unfair not to require"——一种用平等主义为强制使用 AI 辩护的奇怪修辞。

9.4 100x 而非 10x 的人均生产力目标

Tobi 给出了一个具体的运营目标——他要的不是用 AI 替代人，而是用 AI 把每个 Shopify 员工放大 100 倍：

"My real hope is 7,500 to 8,000 at 100x productivity level."

"我真正的期望是 7500 到 8000 人，每人 100 倍的生产力。"

Tobi 在不同访谈里复述（HR Heretics 转引，2025-05-01）

含义：他要的不是缩小人，而是放大人——这与他在 §1 第八章 "Don't port the office online. Let's port the internet into a company" 的远程办公倒转 framing 在结构上完全一致。

9.4 监管 AI 像监管物理

"Advocates for permissive AI regulation: regulating AI at this stage is a bad idea, likening it to regulating physics."

"他主张对 AI 采取宽松监管——这一阶段监管 AI 是糟糕的主意，他把它类比为监管物理。"

a16z Podcast, 2024-12-16

9.5 Live in everyone else's future（再次）

"Building personal eval frameworks (test harnesses) to assess AI model capabilities."

"搭建个人 eval 框架（test harness）来评估 AI 模型的能力。"

Acquired ACQ2, 2025-09-18

"Context engineering as the fundamental skill for using AI effectively."

"Context engineering 是有效使用 AI 的根本技能。"

Acquired ACQ2, 2025-09-18

"Requirement that all employees reflexively use AI as first pass on projects."

"要求所有员工把 AI 当作项目的'第一次过滤'反射式地使用。"

Acquired ACQ2, 2025-09-18

9.6 Agentic commerce: 把每个 Shopify 店面变成 agent-ready

"We're making every Shopify store agent-ready by default."

"我们让每一家 Shopify 店面默认 agent-ready。"

Tobi Lütke, Shopify Editions Dev 2025

"Agentic commerce will be extremely valuable, potentially majority of internet commerce; Shopify building infrastructure for AI agents (OpenAI, Claude) to reason about products."

"Agentic commerce 会极其有价值，可能成为互联网商业的大部分；Shopify 正在为 AI agent（OpenAI、Claude）能够对产品进行推理而搭建基础设施。"

Stripe Sessions, 2025-10-06

"Universal Commerce Protocol (January 2026): Collaboration with Google. Standardized protocol for AI agents to interact with commerce."

"Universal Commerce Protocol（2026 年 1 月）：与 Google 合作。一个让 AI agent 与商业互动的标准化协议。"

Wikipedia (Shopify) / Shopify Editions

9.7 Sidekick

"Conversational AI assistant trained to understand Shopify's platform. E-commerce-focused ChatGPT alternative for merchants. Answers business questions, summarizes sales data, creates performance reports. Modifies shop designs via natural language."

"Sidekick 是一个对话式 AI 助手，被训练来理解 Shopify 平台。它是给商家的电商专属 ChatGPT。回答业务问题、总结销售数据、生成业绩报告，并支持自然语言修改店面设计。"

TechCrunch, 2023-07-26

"We believe that there isn't any corner of the internet that will benefit more from AI than the pursuit of people building and growing their own businesses."

"我们相信，互联网上没有哪个角落会比'人们建设和成长自己的生意'这件事，从 AI 中获益更多。"

Miqdad Jaffer (Shopify AI 产品负责人), TechCrunch

9.8 Agentic 工作 = Factorio

"Compares working with AI agents to playing Factorio — coordination and resource management."

"他把和 AI agent 协作比作玩 Factorio——协调与资源管理。"

Founders (David Senra), 2026-01-18

9.9 Anti-stagnation 而非纯效率

"Companies that don't embrace AI risk becoming irrelevant. AI adoption prevents the organizational entropy that kills large companies. Not just about efficiency — about maintaining the ability to adapt."

"不拥抱 AI 的公司会变得无关紧要。采纳 AI 是为了防止组织熵——那种杀死大公司的力量。这不是关于效率，是关于保持适应能力。"

BetaKit / Toronto Tech Week Homecoming, 2025-06

9.10 2018 年的 prescient 预演：Human Excellence

"Uses chess as primary case study. 'Humans plus machine are better together' — centaur chess concept. Argues fascination with human achievement remains irreplaceable."

"他以国际象棋为案例。'人 + 机器比单独的任何一方都更好'——半人马棋的概念。他论证人对人类成就的迷恋永远不可替代。"

Tobi Lütke blog, The Future Role of Human Excellence, 2018-07-17

9.11 "AI 是完美替罪羊"：对裁员潮的另一种诊断

2026 年 5 月，他对"AI 导致裁员"的流行叙事给出明确反对——当前裁员潮是 COVID 时期 overhiring 的修正，不是技术替代：

"It is definitely overhiring. AI will change the mix of jobs, but what we are seeing right now is not AI-driven layoffs... AI is just the perfect scapegoat because it cannot fight back."

"这肯定是过度招聘。AI 会改变工作的构成，但我们现在看到的不是 AI 驱动的裁员……AI 只是完美的替罪羊，因为它不会反击。"

20VC, 2026-05-04

"My industry has been gaslighting everyone into fearing AI for 60 years through science fiction, but objectively, AI is making my life better in every way."

"我所在的行业通过科幻作品，花了 60 年 gaslight 所有人去恐惧 AI；但客观上，AI 正在让我的生活在每个方面都变好。"

20VC, 2026-05-04

"Being an automated task queue is not a great job. Great jobs are things where you have agency and can produce and make things."

"做一个自动化的任务队列不是好工作。好工作是你拥有 agency、能够生产和创造东西的工作。"

20VC, 2026-05-04 [00:20:26]

[推断] 这个诊断与他对自己 2022/2023 两轮裁员的归因一致：他当年把裁员归因于自己对 pandemic acceleration 的误判（overhiring），而非技术替代——对行业的判断是同一框架的外推。

9.12 Context engineering 与 "product builder"：AI 时代的角色预测

"Context engineering' will become a recognized role that subsumes many specialized functions... Good communication is the distillation of complex thought, and that is exactly what context engineering requires."

"Context engineering'会成为一个被公认的角色，吞并许多专业职能……好的沟通就是对复杂思想的蒸馏，而这正是 context engineering 所要求的。"

20VC, 2026-05-04

配套判断：做过 engineering management 的人是天然的 AI programmer——他们多年来一直在"prompt"智能体（只不过对象是人）；未来还会出现 "product builder" 角色，协调人类与 AI 智能体共同造产品。在 Uncapped 里他给出了企业部署 AI 的方法论反转：

"If you invert the process—if you imagine AI had existed since Alan Turing—and you build your business from first principles with that assumption, you get much more interesting results."

"如果你反转这个过程——想象 AI 从 Alan Turing 时代起就已存在，并以这个假设从第一性原理出发构建你的业务——你会得到有趣得多的结果。"

Uncapped with Jack Altman, 2026-05-20

9.13 反"一人独角兽"叙事：小团队，不是无团队

见时间线第十章 2026-05-27 Toronto Tech Week 节。核心立场：AI 让一人 billion-dollar 公司在技术上可能，但 "the one-person company is bullshit"——公司会更小，数量会多得多。同月他在 Uncapped 里把 AI 时代团队设计的"基本分子"押在 3–5 人小团队上（AI 让每个人在每项技能上都是 7/10）。[推断] 两者是同一立场的两面：他押注的是"小团队"，不是"无团队"。

9.14 "宁可死于技术进步，也不要死于我做的一个坏决策"

2025 年 12 月，在 Intrepid Growth Partners 的一场由 Ajay Agrawal 主持、强化学习奠基人 Richard Sutton 与 Sendhil Mullainathan 等在场的对话里，Tobi 被推到了一个他很少被公开逼问的位置：如果 AI 最终把商业收敛成几个中心化的个人购物助手，Shopify 就不需要存在了。他的回答是本档案里他对自己公司命运最坦率的一次表述。

先是对去中心化处境的诚实评估：

"I cannot actually name a place other than commerce where the decentralization has actually held on, or even was thriving. And I can tell you it was a lot of work... it requires people making sufficient resources, making it a lifelong pursuit, and holding doggedly — because it's never going to be ideal. It would be vastly simpler for us to just put it all together and make a marketplace and then charge rent."

"除了商业之外，我实在举不出第二个去中心化真的守住了、甚至还繁荣着的领域。而我可以告诉你那需要大量的工作……它需要有人投入足够的资源，把它当作毕生的追求，并且死死地守住——因为它永远不会是理想的。对我们来说，把这一切合起来做成一个 marketplace 然后收租，会简单得多。"

Intrepid Growth Partners, 2025-12-19（01:12:43）

然后是那句关键的：

"But I am somewhat fatalistic about it. If AI is going to lead to that kind of thing, and we converge on 100% personal shoppers run through OpenAI, and that's that — then Shopify is no longer needed. I gave it my best, and that would not be the outcome I wanted. But it's vastly preferable to me to die to technological progress as a company rather than to a bad decision I made."

"但我对此有几分宿命论。如果 AI 会导向那种局面，我们收敛到 100% 的个人购物助手全部跑在 OpenAI 上，就是这样了——那 Shopify 就不再被需要了。我尽了全力，那不会是我想要的结局。但对我来说，作为一家公司死于技术进步，远远好过死于我做的一个坏决策。"

Intrepid Growth Partners, 2025-12-19（01:13:25）

而他押的反面（counterfactual）是"90 年代的网重新实现"：

"We are actually going to realize the web of the 90s. It was a lot of effort to make delightful web experiences. It's now trivial. In fact, the amount of creativity — especially if LLMs are going to be somewhat retrograde in creativity — that just puts an even higher premium on the creativity tokens created by real humans... everyone gets to be as creative as they can be with a data center of agents realizing all these ideas for them. And this could actually lead to an incredible Cambrian explosion of the kind of quirky and zany and fun web that we liked in the past."

"我们其实要把 90 年代的网真正实现出来。当年做出令人愉悦的网页体验需要大量工夫。现在它是微不足道的。事实上，创造力的量——尤其如果 LLM 在创造力上是有些倒退的——这只会让真人产出的'创造力 token'溢价更高……每个人都能尽其所能地有创造力，而有一整个数据中心的 agent 替他们把这些想法实现出来。这真的可能带来一次不可思议的寒武纪大爆发，爆发出我们过去喜欢的那种古怪、疯癫、好玩的网。"

Intrepid Growth Partners, 2025-12-19（01:14:21）

[推断] 这段是理解他 2025–2026 全部 AI 动作（AI memo、UCP、Catalog、agent-ready 默认开启）的关键框架：他并不认为 Shopify 在 AI 时代的存续是必然的，他把赌注押在"AI 让创造力稀缺化因而个体商家更值钱"这一条上，并明确接受另一条路径成立时公司出局。这与 §3.4（没有正确决策，只有事后被证明正确的语境）、§7.19（资本分配 = 揭示出来的真正策略）是同一套认识论；也与 §9.11（AI 是裁员替罪羊）互补——他拒绝把 AI 当作掩盖坏决策的借口，无论对别人还是对自己。

9.15 Assistive 而非 autonomous：把 AI 押在"人在环里"的具体机制

同一场对话给出了本档案里第一份具体到产品机制的 Shopify AI 清单，也交代了他为什么在整个行业冲向 autonomous agent 的时候押相反方向。

• SimGym——用基于历史数据合成的客户 persona 模拟 A/B 测试，让没有海量流量的小商家也能取得统计显著性。他的说法是这让"过去只有 Nike 这种每天数百万访客的公司才享有的迭代与优化能力"民主化了。
• Sidekick Pulse——不是再多一块 dashboard，而是主动给出 "next best action"：例如识别出 55 个只买过一次、之后没再回来的客户，并自动起草一封带 15% 折扣的召回邮件。
• Shopify Product Network——让小店用生态里其他商家的互补商品补上自己品类的缺口（卖滑板的可以在自己站上展示别家的头盔），以此抵消 marketplace 的规模经济优势，而不必自己去进货。

底层的研究立场，是他与在场 RL 研究者的分歧点：

"Most traditional AI research focuses on machine-only benchmarks — optimizing for correctness or autonomous performance. When you focus on assistive AI, you start asking different research questions, such as: how do we make the AI's errors more transparent to a human? Or: how do we optimize for a diversity of ideas rather than a single correct answer?"

"大多数传统 AI 研究关注的是纯机器基准——为正确性或自主性能做优化。当你转向 assistive AI，你开始问的是不同的研究问题，比如：我们怎么让 AI 的错误对人更透明？或者：我们怎么为想法的多样性而非单一正确答案做优化？"

Intrepid Growth Partners, 2025-12-19

以及训练目标的选择——用 lifetime value 而非交易额训练模型，理由是只有前者与商家生意的健康度对齐。他还指出 Shopify 在数据上有一个研究实验室没有的东西：从浏览到履约到退货的第一方全生命周期交易数据。

最后他把公司的野心边界讲得很明确：

"It's really not about building the most powerful AI company. I have plenty of friends who are doing this, so they can work that out between them. I'm trying to figure out how to make all the gifts that we're getting from the labs maximally valuable to people I really care about."

"这真的不是要建最强大的 AI 公司。我有一大堆朋友在做这件事，他们之间可以自己解决。我想搞清楚的是，怎么把我们从那些实验室拿到的所有礼物，对我真正在乎的那群人变得最大程度有用。"

Intrepid Growth Partners, 2025-12-19（00:00:05）

[推断] 这条给 §9.2（AI 不应让人变笨，应让人野心更大）和 §9.7（Sidekick）补上了工程层的对应物：assistive vs autonomous 不是产品口径的修辞，而是决定训练目标（LTV 而非 GMV）与研究问题（错误透明度、想法多样性）的分叉点。⚠️ 三个产品的效果均为 Tobi/Shopify 自述，无独立第三方评估；2026-08-05 Q2 财报也只披露 AI 相关指标的增长率而未披露分母（见第十章）。

10. 关于个人：家庭、阅读、赛车、滑雪板

10.1 家庭：Fiona、三个儿子（10–15 岁）

"Wife Fiona insists on family dinner at 6pm; three boys aged 10-15."

"我太太 Fiona 坚持每天 6 点全家共进晚餐；三个儿子，10 至 15 岁。"

The Shawn Ryan Show #261, 2025-12-11

"Conducts dinner appointments with his children, treating them as adults, phones down."

"我会和孩子单独约'晚餐时间'，把他们当成年人对待，手机收起来。"

Conversations with Tyler, 2024-09-18

"Wife Fiona encouraged him to start a company; Tobi considers Shopify a 'team production' where she doesn't get enough credit."

"Fiona 鼓励我创业；我把 Shopify 看成'团队制作'，她得到的功劳远不够。"

20VC #6727, 2023-04-03

"Spouse: Fiona McKean (former Canadian diplomat). Three children. In-laws (Bruce and Dale McKean) were early Shopify investors."

"妻子 Fiona McKean 是前加拿大外交官；他们有三个孩子；岳父岳母 Bruce 和 Dale McKean 是 Shopify 的早期投资人。"

Wikipedia

10.2 育儿：让孩子 pitch 而不给零花钱

"Kids don't get allowance but 'get to pitch' — they negotiate for paid tasks, learning entrepreneurship early."

"孩子没有零花钱，他们要'pitch'——通过谈判换取有偿任务，从小学创业。"

The Shawn Ryan Show #261

"Used video games as learning tools — to teach typing and problem-solving to his dyslexic children."

"他用电子游戏作为学习工具——教有阅读障碍的孩子打字和解决问题。"

Acquired ACQ2, 2025-09-18

"Better to break a bone as a kid when you heal quickly; the space of good, low-risk moves is too small to win."

"小时候摔断骨头反而好——愈合快。'低风险好走法'的空间太窄，靠它赢不了。"

The Shawn Ryan Show #261

10.3 滑雪板与 Snowdevil

Tobi 至今仍是滑雪板爱好者——这是 Snowdevil 的起源，也是 Shopify 的故事原点。

"The first Shopify store was our own. In 2004, took something they loved — snowboarding — and built a business around it."

"Shopify 的第一家店就是我们自己的店。2004 年，我们把热爱的滑雪板做成了生意。"

Shopify IPO Letter, 2015

10.4 IMSA 赛车与 DHH 同场

"FIA Bronze driver license. 2024: HSR Prototype Challenge (Amateur) with TWOth Autosport — 1st place, 3 wins from 6 races. 2025: IMSA SportsCar Championship LMP2 with Era Motorsport; co-drivers include David Heinemeier Hansson (Ruby on Rails creator). 2026: Racing for TDS Racing in IMSA."

"FIA 铜牌车手执照。2024 年与 TWOth Autosport 一起参加 HSR 原型车挑战赛业余组——夺冠，6 场 3 胜。2025 年与 Era Motorsport 一起参加 IMSA SportsCar Championship LMP2 组；co-driver 包括 Ruby on Rails 创造者 David Heinemeier Hansson。2026 年加入 TDS Racing 继续征战 IMSA。"

Wikipedia, Racer.com 2024-12-05

"Lütke has known and worked with co-driver David Heinemeier Hansson for 20 years through Ruby on Rails. Both are billionaire tech founders who race together — extremely unusual pairing."

"Tobi 通过 Ruby on Rails 认识并合作 DHH 已经 20 年。两个亿万富翁科技创始人一起同台赛车——极其罕见的搭档。"

Racer.com, 2024-12-05

"Motorsport enthusiast who races competitively — competes against himself, sees other cars as obstacles."

"他是赛车爱好者——他和自己竞争，把其他车视为障碍物。"

Stripe Sessions, 2025-10-06

10.5 The Opinicon

"2015: Purchased and relaunched Opinicon historic resort lodge near Chaffey's Locks, Ontario."

"2015 年，他和 Fiona 买下并重启了 Ontario Chaffey's Locks 附近历史悠久的 Opinicon 度假小屋。"

Wikipedia

10.6 StarCraft 与 Shopify Rebellion

"Shopify Rebellion: Esports organization (StarCraft II team, 2021)."

"Shopify Rebellion：电竞组织（2021 年从 StarCraft II 战队起步）。"

Wikipedia (Shopify)

"Let's change the game. Announcing Shopify Rebellion esports team. Starting with Starcraft2 because of course we do."

"我们来改变游戏。宣布 Shopify Rebellion 电竞战队成立——从 StarCraft 2 开始，因为我们当然会从这开始。"

Tobi Lütke (Twitter), 2021-02-19

"I firmly believe that I learned more about building businesses from playing Starcraft than I've learned from business books."

"我坚信我从玩 StarCraft 学到的关于做生意的东西，比从商业书里学到的多。"

Tobi Lütke, NonGaap / John Coogan profile

10.7 数字档案：15 年的输入习惯

"Maintains 15-year digital archive of keyboard logs, calendar, emails to analyze thinking evolution."

"维护 15 年的数字档案——键盘记录、日历、邮件——以分析自己思维的演化。"

Acquired ACQ2, 2025-09-18

10.8 一个小幸福：好键盘 + Uber Lux

"Enjoys luxury of Uber Lux rides — 'getting in a really nice car just makes me happier'."

"他承认一个小奢侈——坐 Uber Lux：'坐进一辆好车真的让我更开心。"

20VC #6727

10.9 关心人，但承认无法关掉 Shopify 的 background process

"Cannot fully turn off thinking about Shopify: 'a background process that can go on very low priority, but I don't think I can turn it off'."

"我无法完全停止想 Shopify：'这是一个可以调到非常低优先级的 background process，但我觉得我关不掉它。"

The Shawn Ryan Show #261

10.10 阅读 + 仙德森系奇幻

"Reads Brandon Sanderson; sees fantasy as organizational mirrors."

"他读 Brandon Sanderson 的奇幻小说；把奇幻世界当作组织的镜像来看。"

Conversations with Tyler, 2024-09-18

"Cites Lord of the Rings as a management book — Gandalf's perfectly-timed interventions."

"他把《指环王》当管理书——Gandalf 那种 perfectly-timed 的干预。"

Conversations with Tyler, 2024-09-18

10.11 Shawn Ryan 那期"happy place"

"Coding is his happy place; will clear 3 days of calendar to build with teams until midnight."

"写代码是他的 happy place；会清空 3 天日历，和团队一起编程到午夜。"

Lenny's Podcast, 2025-02-02

"Energy source: dissatisfaction with the status quo."

"能量来源：对现状的不满。"

Lenny's Podcast, 2025-02-02

"I channel this sort of dissatisfaction with the status quo as this unending well of energy that I can convert into building."

"我把这种对现状的不满当作一口永不干涸的能量井，可以转化为建造。"

The Shawn Ryan Show #261

10.12 与 Stewart Brand "Whole Earth" 思想脉的回声

虽然 Tobi 没有公开把 Stewart Brand 列为最大影响之一，但他对"长期主义、看见下一个世纪、把工具交给个人解放他们"的态度，与 Stewart Brand 在 Whole Earth Catalog 时代的精神高度一致。Brand 那句"We are as gods and might as well get good at it" 几乎可以放在 Tobi 任何一封 letter 的卷首。Tobi 自己用来形容 Shopify 角色的话是：

"Shopify as a 'fire flower from Mario' — empowering self-actualization rather than creating success."

"Shopify 像马里奥里的火焰花——让人自我实现，而不是替人制造成功。"

Knowledge Project #41

10.13 Hands-on：仍在写代码

"Continuous coding — Tobi remains deeply technical, coding alongside engineers."

"持续写代码——Tobi 至今仍是深度技术人，和工程师并肩 coding。"

Lenny's Newsletter, 2024

"Still actively writes code as CEO."

"作为 CEO 至今仍在亲自写代码。"

a16z Podcast, 2024-12-16

11. 关于创始人模式与领导力

这一组洞察在 2024 年 10 月 Invest Like the Best EP.393 "Building Islands of Innovation" 上第一次被 Tobi 系统化地讲了出来——这是他在所有访谈里给出过的最完整、最自成体系的创始人理论。

11.1 创始人是放热型的人——一个热力学类比

"Founders are not room temperature people by definition. The main act of starting a company is you're trying to inject heat into planet earth. You're trying to do something like fighting against entropy. So I think founders are recruited out of a set of exothermic people to begin with."

"按定义，创始人不是常温的人。开公司这件事的核心动作，就是你试图往地球上注入热量——你试图做的事情，本质上像是在对抗熵增。所以我认为，创始人本来就是从一批'放热型'的人里挑出来的。"

Invest Like the Best EP.393, 2024-10-22

"I think companies run at room temperature because that's the entropic equilibrium of temperature. The best leaders are literally exothermic, they are just like this complete wellspring of heat into the system. Every atom jiggles faster around them."

"公司之所以会跑在常温，是因为常温就是温度的熵平衡。最优秀的领导者真的是放热的——他们就是热量源源不断地注入系统。他们身边的每一个原子都振动得更快。"

Invest Like the Best EP.393, 2024-10-22

这把"founder mode"从一个口号转成了一个物理类比：组织的默认状态是失温，"领导者"不是发号施令的人，而是热源本身。它与 §2.13 "founder legitimacy" 是同一个底层世界观，但这次给出了更紧的形象。

11.2 创始人的第二个职责：保护其他放热的人

"Here's the thing that founders end up doing, not just are they injecting heat from the top into the whole thing, they also protect the others. When things are going well, they make the politicking against them not work."

"创始人最终做的事情不仅是从顶部往整个系统注入热量——他们也保护其他人。当事情运转得好的时候，他们能让对这些人的政治攻击失效。"

Invest Like the Best EP.393, 2024-10-22

含义：组织里其他放热型的人天然会成为政治攻击的对象（"内卷者"对"做事者"），创始人的第二个职责是屏蔽他们。这条与他多年来"reorgs always work"、"subtraction"的实操是同一条逻辑链——创始人在做的事，是不停清扫熵增。

11.3 政治预算：最好的公司也有 20%，最差的有 60%

"The best companies in the world, people spend up to 20% of their time on politics. The worst companies is up to 60%. It's a very small margin in which you can operate."

"世界上最好的公司里，人们花在政治上的时间高达 20%。最差的公司高到 60%。可操作的区间其实非常窄。"

Invest Like the Best EP.393, 2024-10-22

注意：Tobi 没说政治应当是 0%。他认为公司里"政治"是不可消灭的，关键是把它约束在窄区间——不是理想主义地去除问题，而是把问题约束在工作区间内。

11.4 Code Red 作为升迁信号：跳过阶梯

"I pick people from individual crafter roles and put them in charge of large hundreds of people departments after they ran a Code Red. They skipped many, many rungs that other people think they should have really passed through on the ladder of a career. And I just don't agree with that."

"我会把那些刚刚跑完一次 Code Red 的个人贡献者直接放到管几百人部门的位置上。他们跳过了别人认为应该一步一步走完的阶梯。我就是不认同那种阶梯。"

Invest Like the Best EP.393, 2024-10-22

Tobi 用危机表现作为升迁信号，而不是任期或层级。这条与 §2.13、§5.4 "hiring for potential"是同一条线——但具体的升迁机制（Code Red → 跳级）是它最操作化的版本。

11.5 高 merit-to-ego ratio，不是低 ego

"Life's too short to surround yourself with mean energy vampires. High merit to ego ratio is important. Some people are incredibly good at what they do and they are allowed to be self-confident about them. People misconstrue this as arrogance, but it's arrogance if you're no good at a thing. If you're actually good at a thing, you're just communicating it."

"人生苦短，没必要把自己包围在恶意的能量吸血鬼里。重要的是 merit-to-ego 比。有些人在他们擅长的事上确实非常厉害，他们应当被允许对此自信。人们会误把这种自信当成傲慢——但'傲慢'只有在你做得不好时才叫傲慢。如果你真的擅长一件事，你只是在沟通这件事。"

Invest Like the Best EP.393, 2024-10-22

这条把"反 brilliant jerk"的常见叙事翻转：Tobi 反对的不是"高自信"，而是"低能力高自信"。配合 §5.10 "be surrounded by people you admire"——他要的不是低自我的执行者，而是能匹配自己实际能力的自我评估。

11.6 群岛而非公地：开源的反主流再诠释

同一期 Tobi 给出了他另一个原创概念——他把组织和开源运动都重新解读为"岛屿群"而非"协作公地"：

"How do you become more like an archipelago of islands that all come up with products of merit and then trade ideas and give gifts to each other?"

"你要怎样让公司更像一个由多个岛屿组成的群岛——每一个岛屿都生产出值得拥有的产品，然后彼此交换想法、互相赠送礼物？"

Invest Like the Best EP.393, 2024-10-22

"Open source is totally misunderstood. People always think about this as the commons produces something that's better, but literally it's the opposite. Open source is always island producing. It's actually the purest implementation of a market."

"开源被完全误解了。人们总是以为开源代表'公地'生产出更好的东西，但事实恰恰相反。开源永远是岛屿在生产。开源其实是最纯粹的市场实现。"

Invest Like the Best EP.393, 2024-10-22

具体到 Ruby——Shopify 整个公司的根基——Tobi 把它讲成一个单人小岛把礼物送给大陆的故事：

"Ruby optimizes for programmer happiness. It implements great pains, something called poetry mode, which allows you to just omit brackets here and semicolons. The system does stuff just to make engineers happy. That's an aesthetic. That's a decision that Mr. Matsumoto made 31 years ago."

"Ruby 为程序员的幸福感而优化。它煞费苦心实现了一个叫'诗歌模式'的东西——你可以省略括号和分号。整个系统就是为了让工程师开心而存在的。这是一种美学。这是 Matsumoto 先生 31 年前做的一个决定。"

Invest Like the Best EP.393, 2024-10-22

含义：Ruby/Rails 不是"社区"协作的产物——是 Matsumoto 一个人的小岛把礼物送给了大陆。Shopify 整个公司架在 Ruby 上，因此 Tobi 始终把这条岛屿—礼物经济学视为他的根基哲学，也是 Shopify 内部跨团队协作方式的隐喻。

11.7 4/5 创意小时的契约

2020 年 5 月在 Invest Like the Best EP.173 上 Tobi 给出了他对员工时间投入的一个量化承诺——他要的是 4/5 高质量小时，不是 80+ 小时的劣化输出：

"I think everyone has five creative hours in a day. And if you and I work together at Shopify, I would love four of those hours for Shopify. I think that's sort of a relationship that exists with a mission-based company."

"我认为每个人一天里有五小时是创意小时。如果你在 Shopify 跟我一起工作，我希望要走其中的四小时。我觉得这是 mission-driven 公司应当建立的那种关系。"

Invest Like the Best EP.173, 2020-05-12

这条与 §2.X 关于"反 hustle 但高 engagement"是同一条线，但用具体数字（4/5）操作化了。

11.8 创业的最不浪漫表达

2022 年 5 月在 Invest Like the Best EP.278 上 Tobi 给出了他对创业的最克制版本——不假装容易、不浪漫化代价：

"Entrepreneurship is fundamentally a little bit unreasonable. People point out that you end up spending 100 hours a week working for yourself so you don't have to work 40 hours for someone else."

"创业从根本上有点不合理。人们指出：你最后每周给自己干 100 小时，就为了不用给别人干 40 小时。"

Invest Like the Best EP.278, 2022-05-24

"Our observation with Shopify has always been that the journey is uphill. It's not easy. Shopify never claims it is. But if it becomes a technical climb, it filters out a lot of people."

"Shopify 一直以来的观察是：这段旅程是上坡。它并不轻松。Shopify 从不声称它轻松。但如果它变成一次技术攀登，就会筛掉很多人。"

Invest Like the Best EP.278, 2022-05-24

这两条加起来构成 Tobi 对 Shopify 使命的最克制版本：不假装创业容易、不浪漫化代价，只承诺把"技术攀登"转换为"上坡路"。

三、关键交叉主题

主题 A：工程师身份的连续性，从 C64 到 Rails 到 LLM

Tobi 从 6 岁拿到 Schneider CPC 起，技术身份就是他的"主语"。他在 Stripe Sessions 直接讲 Shopify 代码库现在 ~20M 行 Ruby + 8–10M 行 TypeScript——也就是说他 21 年前自学 Ruby on Rails 时的技术决定，今天仍然是 Shopify 全部基础设施的内核。这条线在任何科技 CEO 中都极其罕见：

• 1986：Schneider CPC，6 岁，第一台计算机
• 1996：Linux kernel mailing list 自学
• 2004：Ruby on Rails 自学（无视日文文档）
• 2004 起：Rails core team 成员
• 2005–2010：写 Typo / Liquid / ActiveMerchant / DelayedJob
• 2025：发 AI 备忘要求所有员工 reflexive AI usage、自己搭建 personal eval framework
• 2026：写代码仍然是 happy place，每年至少几次清空 3 天日历回去做 coder

这种连续性的影响是：他不是从 MBA 抽象层看 Shopify，他从代码层看 Shopify。他用 SAT solver 给公司 org 做建模，用 React 的"declare desired state"风格设计组织变革。他把"软件 vs 政策"作为对照——因为软件可以即时部署，政策不能。

主题 B：德国学徒制作为长期信念

德国双元制对他的塑造无处不在：

• 学习方法论：动觉、轮岗、立刻有客户—— Jürgen Saar 把 17 岁的他派去客户现场
• 招聘方法论：让候选人讲 life story，而不是问技术题
• 培养方法论：把人推过自己感知的能力上限——他自述"在判断别人潜力上，我比他们自己更准"
• 对 credentialism 的怀疑：他自己 17 岁辍学；Shopify 招人重 potential 大于 degree

主题 B 与主题 C（外来者视角）连在一起：他不在 Bay Area 的 talent gentrification 体系里，因此可以保持德国学徒制的招聘哲学。

主题 C：Outsider 视角作为竞争优势

"Built Shopify as an immigrant without Canada's 'supposed to make small plans' cultural narrative."

"我作为移民来建 Shopify，没有沾上加拿大那种'我们该规划小一点'的文化叙事。"

The Cheeky Pint, 2025

Tobi 是双重 outsider：（1）德国移民到加拿大，因此不浸泡于加拿大"目标拿铜牌"的文化；（2）Ottawa 而非 Bay Area，因此不浸泡于硅谷"best practices"。两者叠加让他对每一种 conventional wisdom 都保持怀疑。这条线显化为他的反 KPI、反 efficiency 绝对化、反 hybrid work、反 marketplace、反"not a programmer can't be a retailer"等几乎所有反共识的判断。

主题 D：直面错误的"Tobi 公开纠错"模式

Tobi 是少数公开纠正自己的科技 CEO：

• 2022 年 7 月："Ultimately, placing this bet was my call to make and I got this wrong."（pandemic 加速永久化的赌注）
• 2023 年 5 月：把 Logistics 卖给 Flexport，承认 Shopify 不该自己做物流
• Founders 2026 年 1 月："Shopify went bankrupt and I bought it on a fire sale and I'm marching in on day one"——他用这种心理姿态做 post-IPO reset

他的"the cheat code to always being right is just change your opinion every time you get better information" 是把这种 epistemic humility 升格为一条规则。这个模式背后是他对决策的方法论：决策本身不是关键，重做决策的能力才是关键——这与 Patrick Collison "决策速度优于精度，重做能力优于一次精度"的方法论同源。

主题 E：Calm progress 哲学的演化

Tobi 的领导哲学有清晰的三个阶段：

1. 2006–2015：survive, then scale——专注产品、定价、平台（App Store 2009）、IPO（2015）
2. 2016–2022：bigger bets, bigger errors——加速 hire、收购 6 River + Deliverr、digital by default、COVID 2x hire；最终在 2022–2023 撞墙
3. 2023–2026：calm progress + AI——12,000 例会删除、20% 裁员、Logistics 卖给 Flexport；3 月 2025 AI 备忘录；GSD 系统每 4 周复盘

这三个阶段对应不同的 mental model：阶段 1 主导是 Andy Grove 工程化经营；阶段 2 主导是 high-growth platform empire；阶段 3 主导是 infinite game + subtraction + AI multiplier。每一次过渡都伴随他自己的方法论升级（从 trust battery 1.0 到 GSD 到 Tobi Tornado 到 reflexive AI usage）。

主题 F：游戏作为商业语言

从 StarCraft（Zerg 蜂群、晚期统治、read-and-respond）到 World of Warcraft（公会协调）到 Factorio（与 AI agent 协作）到 main quest/side quest（分清核心 vs 干扰），Tobi 一以贯之地用游戏语言描述商业。这既是优势——给他独特的 mental models——也是风险（NPC 风波）。Shopify Rebellion 电竞战队是这种语言的物质化。这条线和他的"hands-on writes code"身份是一体两面：Tobi 的认知是工程师 + gamer，不是 MBA。

主题 G：Family + entrepreneurship 的边界设定

"Wife Fiona insists on family dinner at 6pm; three boys aged 10-15."

"Fiona 坚持每天 6 点全家共进晚餐；三个 10–15 岁的儿子。"

The Shawn Ryan Show #261

家庭对 Tobi 既是限制也是结构：每天 6 点 dinner、孩子 pitch 而非 allowance、把孩子当成年人对待。但他承认 Shopify "background process" 关不掉。这种"严格的硬边界 + 内心持续的工作思考"组合是他可持续 20 年的关键——calm progress 不是放慢，而是在硬性 limits 之内全力推进。

主题 H：Distance from Bezos 的双重含义

"Anti-Bezos" / "arming the rebels" 是 Shopify 的核心叙事，也是 Tobi 个人定位。但这个对立有两层：

1. 商业模式层：Amazon = marketplace（消费者首选），Shopify = platform（让商家保留 customer relationship）
2. 价值观层：Tobi 在多个访谈里把 Bezos 隐含放在他视为"not the model"的地方——SCMP 2022 的封面文章直接把他叫"the anti-Jeff Bezos"

但这种对立也有限度：Shopify 在 2021–2022 年的 logistics buildout 实际上在向 Amazon 模型靠拢；2023 年的 logistics 退出本质上是承认"Shopify 不应该把自己变成另一个 Amazon"。这是 Tobi 反复迭代自己定位的一个具体例子。

四、批判视角

1. 2022 年 logistics 战略误判与代价

Shopify 在 2019–2022 年累计投入超过 25.5 亿美元收购 6 River Systems（$450M, 2019）和 Deliverr（$2.1B, 2022），并按 Stratechery 的报道"承诺 $1B 年投入直到 2024 年"。2023 年 5 月把这块业务全部卖给 Flexport，仅换得约 13% 股权——相对账面投入而言是显著减值。Tobi 在 Founders 2026 那期承认上市后 "60% of internal projects had gone off-mission" 并把它们全部停掉。这次反转的代价不仅是 capital，更是品牌"arming the rebels" vs "build my own Amazon" 之间的张力暴露。

2. 双重裁员与员工信任

2022 年 7 月（10%, 1,000 人）+ 2023 年 5 月（20%, 2,000 人），不到 10 个月内的两次大规模裁员，对内部信任结构有深远影响。The Walrus 2025 年的长报道（本档案 source 38）记录了员工的失望感：

"Shopify was once 'a revered homegrown success' with ping-pong tables, free lunches, all-dogs-welcome policy. By July 2022, began walking back the utopian promise. Employee retreats, 'freedom to play around' — all gone with mass layoffs."

"Shopify 曾经是'被尊敬的本土成功'——有乒乓球台、免费午餐、欢迎所有狗的政策。从 2022 年 7 月起，乌托邦承诺开始回撤。员工 retreat、'自由玩耍'——都随着大规模裁员消失。"

The Walrus, 2025

"Exhausted and afraid' — constant layoff cycles created ongoing anxiety."

"疲惫又害怕'——不断的裁员周期制造持续焦虑。"

The Walrus, 2025

NPC 风波（详见第十章）暴露的也是同一个问题：Tobi 的游戏隐喻在内部传达出"我们在做战略 game"的距离感，让被裁员工感到自己只是"main game story"之外的人物。

3. 治理：Founder Share 与 40% 永久投票权

2022 年 6 月通过的 triple-class share 提案让 Tobi 个人持有 40% 投票权——即使他持股稀释到 1.1% 也不变。Glass Lewis 和 ISS 推荐反对：

"Limits shareholder rights and inadequately protects minority shareholder interests."

"限制股东权利，没有充分保护少数股东利益。"

Glass Lewis 投票推荐, 2022

"53.68% of non-Lütke shareholders voted in favor. 46.32% voted against — nearly half opposed."

"排除 Tobi 自己，53.68% 赞成，46.32% 反对——几乎一半反对。"

Harvard Law / Glass Lewis 分析

"Multi-class structures effectively silence ordinary shareholders. Passive fund investors cannot easily exit."

"多类股结构实质上让普通股东失声。被动型基金的投资人无法轻易退出。"

Harvard Law 分析, 2024-08-15

这是 Tobi 的"founder legitimacy" 哲学（详见 2.13）的代价：他持有这种制度上 entrenched 的 control，是因为他认为创始人的 legitimacy 是无法替代的；但代价是治理透明度受损。这是 Tobi 哲学最难调和的内部矛盾。

4. Breitbart / 言论自由与平台中立的张力

2017–2018 年 Tobi 在 Breitbart 商家事件中坚持 "platform neutrality"：

"Products are a form of speech, and free speech must be fiercely protected, even if we disagree."

"产品是一种言论，言论自由必须被强烈保护——即便我们不赞同。"

Tobi Lütke blog, 2018-08-13

但他在 2018 年 update 中已承认：

"Solely deferring to the law... is too idealistic and functionally unworkable. Legislative processes cannot match internet speed, leaving companies to create their own policies."

"仅依赖法律……太理想化、实际不可行。立法跟不上互联网速度，逼公司自己制定政策。"

Tobi Lütke blog, 2018-08-13

这条线显示 Tobi 思想里"libertarian + 实用主义"的张力：他从原则出发反对平台 mute/de-platform，但在面对实操时不得不让步到 acceptable use policy。批评者指出，Shopify 在这个问题上从未给出令两边满意的答案。

5. 小商家被低优先级对待的转向

The Walrus 2025 报道指出 Shopify 在 2022–2023 年起调整客户优先级：

"Deprioritized smaller merchants in favor of big businesses. Downgraded customer service for merchants with less than $10M in annual sales. Shifted from human support to AI-powered assistants for smaller merchants. Contradicts founding 'arm the rebels' philosophy."

"在策略上把更多精力放在大商家上，下调了年销售低于 1,000 万美元商家的人工客服等级，把小商家的支持转向 AI 助手。这与'武装反叛者'的初衷有所违背。"

The Walrus, 2025-05

这是对 Shopify 创立故事核心叙事的张力：Tobi 反复说要 lower the courage needed to start a business，但 2024–2025 年 Shopify Plus（企业级）已成为主营收驱动力；这意味着 Shopify 的使命表达和实际产品优先级出现了张力。批评者认为这是任何 SaaS 公司在大客户经济压力下的必然，但仍然是值得在档案里诚实记录的批判维度。

6. AI 备忘录的另一种解读：作为招聘过滤器

Tobi 2025 年 4 月公开的 AI 备忘录在 Shopify 内部和行业层面都被广泛讨论。本档案 §1 第十章 / §2.9 把它呈现为"工匠精神 + 教育 + equity"的延伸；但 2025 年 5 月 HR Heretics 那期对 memo 做了一个更尖锐的产业层诊断——HR 行业评论员 Nate B Jones 把 memo 重新解读为首要功能是人才筛选，其次才是生产力：

"By making AI usage a performance metric, Lutke was not trying to make Shopify more efficient per se. He was reshaping who would want to work there and who would thrive once they got to Shopify. The memo was effectively a filter on hiring as much as it was a talent message."

"通过把 AI 使用变成绩效指标，Lütke 并不是真的想让 Shopify 更高效。他是在重新塑造'谁会想来这儿工作'和'谁来了以后能够蓬勃发展'。这份 memo 本质上是一个招聘过滤器，几乎跟它作为人才信号同样重要。"

Nate B Jones（HR Heretics, 2025-05-01）

这条诊断对照 Tobi 一直以来坚持的"high agency life story"招聘标准，把 memo 放到了一条更长的连贯轨道里——他过去 20 年都在做"反向筛选"。同期评论员 Siadhal Magos 进一步把这条信号放到整个行业的大势里：

"It's a total inevitability that people who know how to orchestrate AI are going to be more effective and people who aren't sort of jumping on the train and learning these skills, they're not sowing the seeds that they need to sow for their future success."

"完全不可避免的是：那些会编排 AI 的人会更有效，那些没跳上车、没在学这些技能的人——他们没有为未来的成功播下应有的种子。"

Siadhal Magos（HR Heretics, 2025-05-01）

⚠️ 两种解读之间的张力——一种把 memo 呈现为工匠精神 + AI 反思的自然延伸（强调使能、教育、equity）；HR Heretics 的二次分析把它解读为首要是人才过滤器（强调淘汰、selection pressure、招聘信号）。这两种解读不一定矛盾——可以共存：对内是教育与 equity，对外是招聘信号。但这条张力本身值得在批判视角里记录——Tobi 给出的 "unfair not to require" 修辞与"招聘过滤器"诊断之间，是一个值得长期观察的语义裂隙。

7. 2026 年 7 月：加权投票言论与"利害关系式选举权"

事件本身见时间线第十章 2026-07-25/26 节（原话、媒体核查状态、未采信项均在那里）。这里记录它对本档案其余部分构成的张力。

（1）与他自己的组织原则冲突。 2020–2021 年他在内部 memo 里要求员工把政治留在门外，并用"你的螺丝刀不会因为你是共和党选民就不让你拧螺丝"作比（§2.18）。2026 年他本人在公开平台上就选举权本身发表了高度党派化的主张。他没有公开回应这一不对称。

（2）与"arming the rebels"使命的张力。 本档案 §7.22（漏桶）、§7.23（"lifestyle business"是刻薄的建构）、§4.2（做新创业者）的整条论证，建立在"创业者不该被既有制度按出身分层"之上——包括他自己对 YC 路径"含着银汤匙"的讥讽。而按缴税额分配选票，恰恰是一套以既有经济产出为准的分层制度。一个刚起步、尚未盈利的创业者，在 Thor 的方案下拿到的是 0 票。

（3）与治理批评的结构性回响。 批判章 §3 记录的 Founder Share 争议（40% 永久投票权 vs 约 7% 经济权益）与这条主张共享同一种逻辑形态——投票权应当与某种"利害关系/贡献"而非人头挂钩。本档案不认为二者可以互相推导（公司治理与国家选举权是不同的规范领域），但把这个形态上的呼应记录在案。

（4）与他 2023 年"builders vs critics"论述的对读。 §7.24 里他明确说"我不反对批评，高投入的批评是激励 builders 最重要的东西之一"，只是认为批评者的价值有上限。三年后，他把参加市政 town hall 的市民概括为"退休的、失业的、没有就业能力的，以及有用的白痴"。是同一立场的激进化，还是不同语境下的两件事，本档案不作判定，只并列存档。

（5）语境而非孤立失言。 这条表达不宜当作一次孤立的推文失手来读。他是 Build Canada 的支持者（一个由科技创业者组成的政策倡导组织，主张削减联邦开支、优先 AI、推行数字身份），其妻 Fiona McKean 在该组织董事会；The Logic 报道 Shopify 在 2026 年 Q1 单季于 Ottawa 的游说会面次数超过 2025 全年；BetaKit 另指出 Shopify 已终止其原住民创业项目并大幅削减 equitable commerce 项目。本档案 §八 / 主题 C 此前只把加拿大处理成经济批评的对象；2025–2026 的材料显示他已经是一个有组织的政治行为者。⚠️ Build Canada 的多数底层报道早于本轮时间窗（该组织 2025 年成立），此处为 backfill 线索，尚未做独立深挖。

五、数据来源

网络资料（54 篇 + 2 篇 _index/_summary，全部 cached 在 cache/web/）

播客访谈（24 篇深度访谈，全部 cached 在 cache/podcasts/）

已扫描但被过滤的内容（典型样本）

依据 CLAUDE.md 的播客质量过滤器，本档案构建过程中明确排除了：

• CNBC TV / Fox Business 短片访谈（< 15 分钟，PR 类）
• Bloomberg / Reuters 财报电话相关报道
• 公司业绩发布会片段
• 翻译转述节目
• 多人 panel/圆桌（如 Shopify 内部 podcast Context 中混合多人讨论的部分）
• "Top 10 Tobi quotes" 等剪辑合集
• Shopify 自有频道 Tobi Lütke x Maggie Fost: The Underrated Importance of Fun（id 7966313, 2025-12-16）——2026-07-03 triage 后 drop：题为 Tobi × Maggie Fost，但可核验的 highlights/QA 转录中所有实质答案均归属 Maggie Fost / 主持人，几乎无 Tobi 逐字可折入；其覆盖的主题（fun 作为人才磁石、craftsmanship、Editions、小团队）在本档案 §1/§2/§11.6 及 TTW fireside（8206351）已充分呈现。对一份 Tobi 档案而言，Tobi-可归属的边际增量过低，故不折入（own-channel + 归属稀薄双重原因）。

2026-08-05 refresh 新增排除 / 队列（多角度 sweep 捞出 20+ 条候选，去重与 triage 后）：

• 排除 — 重复 feed（同集不同 feed id，只留一个）：4536178（Tetragrammaton 同集未处理副本，正本 5616207）、8185761（How I Built This 2019 集 2026-06-08 重播，正本 128678）、8033548（Uncapped 正本 8083602）、7966002（20VC 正本 7909845）、6611551（Shawn Ryan 正本 6374206）、6885669（David Senra 正本 6886859）。
• 排除 — 已折入集的切片剪辑：1733404 / 1733407 / 1733409 / 1733411 / 1733417（20VC 2023 年 4 月的五段主题切片，均出自已折的 6727）。
• 排除 — 第三方转述、无 Tobi 一手内容：8406426（Capital Compounders Show「The Untold Story of Shopify」）、7912115（Transcripted.ai 的 20VC 浓缩版，转述节目）。
• 排除 — own-channel + 归属稀薄（沿用 7966313 的先例）：6586695（Shopify 内部节目 Context「Tobi Lütke x Vanessa Lee: From metafields to markets」，2025-11-20——实质答案主要归 Vanessa Lee）；7965295（ShopifyDevs Editions.dev 2024 双 CEO fireside w/ Klaviyo 的 Andrew Bialecki，2024-07-17——双主体归属混杂，且 infrastructure composability / platform 价值分配 / apprenticeship / IPO-as-clarity 四条均与 §1.4 / §4.5 / §5.1 / §七 IPO 重复）。
• 进 backfill_queue（确认存在且形态合格，但本轮 context 预算已满，下次 refresh 步骤 0 优先消化）：1057153（OMR Podcast「OMR Classic mit Shopify-Gründer Tobias Lütke」，2024-04-05，德语一手访谈——他母语状态下的表述、"办公室退化为通往数字世界的 Zubringer"、百万商家的集体议价力、influencer 去中介化）；7966333（Bloomberg Technology「Bloomberg Studio 1.0」完整访谈，2020-06-13——COVID 期 94% 线下销售被线上替代、25 公里本地商圈、"每 60 秒诞生一笔首单"口径）；1117361（Escape Velocity with Dan Martell #16，2019-12-05——neurodiversity 与公司决策质量，本档案目前 0 覆盖；"money as energy"）；2462075（Shopify Masters，2024-12-05，own-channel 但 Tobi 为唯一受访者——"创业门槛在二十年里发生倒置：从建基础设施的技术难题变成了获客的竞争难题"）。
• 提交 Podwise 处理（pending）：128618（How I Built This「How I Built Resilience: Live with Tobias Lütke and Jon Stein」，2020-05-16，Tier-1 白名单，COVID 早期 + Digital by Default 同期的现场录制）——2026-08-05 提交 process，下次 refresh 回收。

仅保留以独立、深度、长达 50+ 分钟的一对一访谈为主。

V2 待跟进

• ~~2026 年 5 月 27 日 Toronto Tech Week fireside 完整音频（podwise episode 8206351）~~ ✅ 已于 2026-06-26 处理完成并折入第十章时间线节点（go-for-best/craftsmanship/learn-by-collecting-problems/golden-age 等逐字）
• 2025 年 6 月 Toronto Tech Week Homecoming 完整视频文字稿（目前仅有 BetaKit 侧摘要）
• ~~2026 年 Universal Commerce Protocol 与 Google 合作的细节文档~~ 部分补上：2026-06-17 Editions Summer '26 确认 UCP 已 default-on + Catalog 分发范围（见第十章）；协议本身的技术规范文档仍缺
• backfill_queue（下次 refresh 优先，详见上方排除/队列节）：1057153 OMR 德语访谈 · 7966333 Bloomberg Studio 1.0 · 1117361 Escape Velocity（neurodiversity）· 2462075 Shopify Masters
• pending：128618 How I Built This "How I Built Resilience"（2020-05-16）——2026-08-05 已提交 process，下次 refresh 回收
• 2026-07 加权投票言论的后续：是否出现 Tobi 本人的澄清/撤回、Shopify 官方回应、Meritorious Service Cross 请愿的结果；单一来源的 "five votes max / reward productive people with leverage" 辩解是否被第二家媒体证实
• Build Canada 深挖：该组织的备忘录原文、Tobi 具体署名的政策主张、Fiona McKean 董事角色的一手材料——目前仅有 BetaKit / The Logic / The Tyee 的二手报道
• AI 归因指标的分母：Shopify 迄今只披露 AI 相关指标的增长率（流量/订单各 3x、Sidekick 3,400 万次对话），未披露 AI 驱动 GMV 占比；追踪后续季度是否公开绝对口径（关系到 §9.14 那条赌注的可证伪性）
• 任何与 Fiona McKean 的联合访谈（目前一手材料几乎都是 Tobi 单独访谈）
• 2015 年 IPO 时与 Scott Lake 的关系修复或正式叙事（Lake 离开后两人关系的公开材料稀缺）
• Tobi 在 IMSA 2026 赛季与 TDS Racing 的进展

文件路径

• people/tobi-lutke/profile.md — 本档案
• people/tobi-lutke/sources.json — 数据源元数据 + Notion page id
• people/tobi-lutke/changelog.md — 更新历史
• people/tobi-lutke/cache/web/01-...md … 54-...md + _index.md — 网络资料缓存
• people/tobi-lutke/cache/podcasts/{episode_id}.json — 播客 digest 缓存

六、迭代记录

见 changelog.md。