Shopify CEO谈AI如何成为大规模裁员的替罪羊，以及加拿大的“特朗普妄想症”
- Shopify CEO Tobi Lütke认为，今天的裁员源于疫情期间过度招聘，而不是AI。 “AI会被拿来归咎于一切……它是完美的吉拉尔式替罪羊，因为它无法反击。”他自己的规划则把真正的判断说得很清楚：Shopify目前约有7,500–8,000名员工，5年后“我真正希望的仍是7,500到8,000人——但生产力达到100倍水平”。
- 代理式AI的拐点被他定在12月和Opus。 Shopify“超过50%的代码”由AI生成，而且“还在转向更高比例……我们很多最优秀的工程师今年都没写过代码。12月改变了一切，Opus改变了一切。”名为River的AI生活在Slack里，“负责Shopify工程工作中荒谬之多的一部分”，并在公开频道中接受指挥。
- Lütke公开推翻了自己此前的判断。 他原以为没有行业先验、原生使用AI的毕业生会占上风，但实际胜出的是资深工程师，因为“所有工程师都严重低估了引导的重要性”——引导本身就是高层次编程。值得关注的新角色是context engineering（上下文工程），这是他帮助普及的说法；他认为，这条上升路径最终会吞并专业化的产品、设计和工程岗位。
- Shopify自身市场的看多逻辑，是一个“创业黄金时代”。 他认为创业“绝对是最抗AI的工作，也是最受益于AI的工作——想想这两点的组合”。当“充当你联合创始人的得力AI助手会直接告诉你”时，原有先验将不再重要。
- 他在财富与慈善问题上刻意采取反主流立场。 “一个实体拥有的财富越多，就越应该接受审视”——捐赠者也不例外。“捐钱如果没有带来正确的结果，就谈不上美德……我们的慈善资金太多了。”非营利组织选择退出“地球上发明过的最佳适应度函数机制”。Stebbings强烈反驳，认为这种审视会扼杀捐赠；Lütke没有让步。
- 政府“极其不擅长自己做的事”。 任何被政府接管的事情“成本都会变成10倍”。欧洲的解决方案应从“摆脱气候邪教”开始；国家的职责是设计具有良好正外部性的游戏，然后退出——基础设施和将暴力外包给国家是例外。
- 加拿大患有“令人震惊的特朗普妄想症”。 超过60%的人认为美国比俄罗斯或中国更具风险，他直言这是错的。加拿大历史上一个行之有效的策略是“通过帮助美国获胜来获胜”；与此同时，应“他妈的把管道建起来”，发展矿山和炼化，因为加拿大“本应是地球上最富有的国家”。
- 中国因素不容忽视。 给AI设置儿童年龄门槛不会阻止使用——“他们会下载中国模型，你永远、永远不会再看到一篇关于天安门广场的高中作文。”中国模型携带集体主义世界观，而“所有政治都只是集体主义与个人主义之争”。
1. 创业者“本质上都是疯子”——长期视角才是全部优势
- Stebbings先提出自己的二分法：害怕失败，还是渴望胜利。Lütke的回答“连我自己都不满意”——他还在摸索真正答案，但核心观点是：害怕失败的人和害怕胜利的人，“显然都会以更短期的视角看待事情”。更长的视角会改变你面对几乎每个挑战时所采取的基本框架，因为培养人才、把最难的任务交给他们所产生的更多复利，最终都会沉淀到你正在建设的事业里。
- 他对领导力进行了祛魅：“世界真的需要明白，创业公司的人本质上都是疯子……人们以为优秀领导力应有的所有美学，都来自电影。”读一读IBM的Watson传记：“你觉得现在商业界有暴君？没有。你根本不知道过去这些事情是怎么运转的。”
- “我做这份工作，是为了让其他人拥有我希望自己能做的工作。”他并不想当CEO，但后来发现自己必须成为CEO，因为产品需求和公司需求之间“极其分歧”，直到你采取3年视角。经营一家产品公司就是一场“每日棉花糖测试”，长期主义意味着“愿意接受一段时间的糟糕数字”。
2. 8型人格会被“合谋排斥”——后乔布斯时代的创始人实验
- 他的九型人格分类法——他开玩笑说要用2019年前的测试，因为后来“变得政治正确了”——显示，20年的高管大多是成就型，另有少数喜欢深挖细节的5号，通常在CFO岗位。Lütke是8号：会说“这就是垃圾，怎么包装都改变不了事实”。8号“对其他人的职业生涯都很危险”，他们得不到晋升，选择离开，然后创办公司。
- 从“苹果请乔布斯回归”以来，这场实验一直在进行：董事会给创始人更长的任期，相当于把“一个没人能摆脱、只会把胡扯称为胡扯的人”装进公司——事实证明，这“对公司真的非常、非常好”。Shopify的8号人格比例“高得惊人”，因为“我确实很主动地寻找他们”。
- 他的领导力物理学：如今他不用“混乱”，而改用“温度”来描述——“伟大的领导者必须是放热体，必须成为公司的热源……在室温下很难锻造出任何新东西。”他只把时间花在出问题的地方，并在Obsidian里保存受他尊敬之人的善意话语，把它当成应对数周持续坏消息的“激励管道”。
- 谈到犯错，他异常坦率：“我犯错的次数，超过绝大多数人一生会犯错的次数……我会做出可证伪的陈述，因为我希望别人证伪它们。”他最公开的一次失误，是在AI变强之前全面押注物流和仓储。他曾说最初的决定是正确的；如今改口：“我可能确实判断错了，也许就该承认这一点。那次很糟，因为它影响了人们的生活。”
3. IPO定价机制：银行家的客户是簿记里的投资者，不是你
- 他的信任排序是：值得信任的上市公司 > 值得信任的私营公司 > 不值得信任的私营公司 > 不值得信任的上市公司。创始人往往会逃避上市带来的信任折价——“这很有意思，因为最终要到达最好的位置，其实就必须经历这一步”。Shopify在2015年上市时营收约2亿美元，正是为了“让那些押注我们的公开市场投资者成就自己的职业生涯……现在他们信任我们”。Collisons的1,500亿美元身价“并不是你该效仿的好参照，因为大多数人都不是他们”。
- 在80层楼的定价会议上，银行家的客户“其实是簿记里的投资者……他们的工作，是给簿记中的所有人拿到尽可能便宜的交易——只是这一点被巧妙地遮蔽了”。他事先研究过银行家的薪酬机制，于是告诉他们：“我的受托责任从T-minus 10开始。”他顶着全场把价格定在区间之上；首笔交易仍上涨了约10美元——“市场是出色的分布式大脑，它会告诉你真实价格是什么”。公司上市时市值约16.7亿美元，此后某些时点“上涨了100倍”；当年带他上市的银行家，如今是他的CFO。
- 他已经“23天没看过”股票代码，可能实际更久。“股票代码只是其他人用来交易的游戏……我研究的是公司的公平市场价值。”他认为最错误的一次，是疫情初期股票代码上涨了“很多——大概超过100%”时；“我们不可能在那一刻突然聪明了100%”。他还讲了一个金矿CEO的故事：当产品没有差异化时，“技术本身就是投资者关系”——魅力可以把估值倍数往上推，让你买下其他矿山。
4. 裁员源于疫情期间过度招聘——“AI是完美的吉拉尔式替罪羊”
- Shopify目前约有7,500–8,000人。5年后，“我真正希望的仍是7,500到8,000人——但生产力达到100倍水平”。人数持平不等于收缩；真正的失败是“受机会限制”，而Shopify让创业更普遍的使命，仍有大量空间可以拓展。
- “你现在看到的不是AI裁员。那只是一些真正动作迟缓的公司，如今才开始纠正过度招聘。AI会被拿来归咎于一切——它是完美的吉拉尔式替罪羊。”与此同时，他说，自己的行业“多年来一直在对所有人进行AI恐惧的煤气灯操纵，而科幻作品在此之前已经做了60年”。
- Stebbings反驳这种乌托邦式视角：Claude已经能很好地处理执行导向的任务。Lütke回应：“所有那些工作只是替别人完成任务的人，都没有好工作。成为一个自动化任务队列，并不是一份好工作。”购买力“会疯狂上升”，产品会变得极其便宜，人们也会获得更多选择。
- 他愿意押注的结果是：“创业黄金时代——绝对是最抗AI的工作，也是最受益于AI的工作。想想这两点的组合。”原有先验将不再重要：当“充当你联合创始人的得力AI助手会直接告诉你”时，你不再需要家族企业背景。
5. 人类极其擅长发明工作：F1、终端与绅士程序员
- Stebbings给出的数据被Lütke称为一个精彩的起点：今天收入最高的10种工作中，有8种在20年前还不存在——“10年后会这样，再过10年也会这样”。
- 他最喜欢的例子是一级方程式：Mercedes有800人负责打造一台发动机，每家制造商都一样；他们“生活在工业机械工程永恒的黄金时代”，每个周末都接受成绩检验。“这是一项以规则手册命名的运动……这是我们凭空创造出来的，而它的存在让所有人的生活更丰富。”Stebbings补充了叙事的一面：《Drive to Survive》拯救了一项正在衰败的运动——规则手册加上 spectacle，创造出没人预先估算到的价值。
- Ghostty可能会有类似的故事：一位创始人卖掉公司后，“亲手打造了一款他认为必须存在、且工艺完美的软件”——这款终端应用被Lütke同时运行15个实例，用来“告诉我的clankers该在Shopify构建什么”，而他在电话会议上处理法律必需事项。他更大的判断是：“软件过去太昂贵，这让整个世界变得非常贫瘠。”
- 他把底层经济学讲清楚，是因为“至少一半人完全搞不懂这一切是怎么运作的”：增长不会创造财富；“增长是把资源中的财富带入经济体”，公司则把无限的假设价值收拢成真实产品，让人们用金钱投票。
6. 用财富造成的结果来审视财富——别再把愤怒归错对象
- 对美国仇富情绪，他认为这股情绪“可能正处于高水位，但仍然低到足以自我修复”。他的真实立场是：“一个实体，甚至一个个人，拥有的财富和资源越多，就越应该接受审视。”真正失灵的是：判断谁在增加社会价值、谁没有增加价值的过程“完全坏掉了；它被那些彻底不怀好意的人掌控”，尤其是国家控制、扭曲事实的媒体。
- 关于Elon的讨论就是一个例证：那天早上他没有碰方向盘，Model Y“相较汽车行业的常规产品，价格极其亲民”，车后备厢里还有Starlink——Elon是“一个人的发动机”，实现了巨大的价值，只拿走其中“非常小的一部分”。“当你查看地球上最富有的人名单时，他们全部都创造过某种东西，这难道不酷吗？”
- 对“他们一定是偷来的”这种叙事，他的回答是：“靠偷窃不可能赚到10亿美元”——但随即补充，“也许可以，也许确实存在一些例外”。通过创办公司积累财富的人创造了产品，并得到消费者的投票认可；这其实是最民主的存在，远比任何选举民主。Lütke拥有Shopify 6%的股份，94%属于其他人，约1,000万人日常依靠Shopify工作。继承或离婚所得的财富则是另一回事：“你是财富的托管人，需要以有益于社会的方式使用它。”
7. “我们的慈善资金太多了”——本期最激烈的冲突
- Stebbings回忆，Jeff Bezos的妻子曾因低效地捐出一大笔钱而遭到批评——“我的天”。Lütke面无表情地回应：“这才是好事。这正是我希望更多出现的情况。”他的原则是：捐钱如果没有带来正确的结果，即使听起来很好，也不算美德；必须停止让听起来正确的事情免于审视。
- 其机制是：每一美元的支出都是一张选票——“真正的民主，发生在分布式资本配置之中”。非营利组织宣布退出“地球上发明过的最佳适应度函数机制——也是唯一一种曾大规模让人摆脱贫困的机制”，转而用“个人的号召力”替代“组织的能力”，吸引来的是“最会耍滑头的人”，而不是建设者。
- Stebbings的反驳值得保留：他没有时间对每个组织做尽调，所以这种审视会“让所有捐赠都停下来，原本能拿到钱的好组织也拿不到”。Lütke坚持原立场：“你给我找出好组织——它们拿到了很多钱。相信我，钱正在被这些体系里的坏人攫取。”他甚至预先回应常见反驳：Carnegie的图书馆毕竟是“一百多年前的事……但确实非常成功”。
- 这条思想脉络很可能来自Thomas Sowell，他引用后者的遗憾：社会在过去50年里一直在“用听起来好的东西替代真正有效的东西”。他还用裙带关系做压力测试——试图找出其可取之处，最后得出“没有”——但裙带关系“可能比某些疯狂的东西更好”，例如按交叉性定制的学术招聘启事，因为作为黄金标准的双盲择优，已经极其罕见。
8. 政府的成本是10倍——定义游戏，然后退出
- 相比Adam Smith和自由放任主义，他更支持Friedrich List和普鲁士学派：政府的职责是“定义以社会繁荣为正外部性的游戏”，然后彻底退出，让竞争完成其余工作。因为“政府极其不擅长自己做的事……一旦政府接管任何事情，成本就会变成10倍”。如果宣布开设国营杂货店，“就会有几百个像我这样的人不得不去创办公司，弥补这一刻凭空蒸发的财富”。
- 他承认两种例外：把暴力外包给国家——“人类做过的最鼓舞人心的事情之一”，因为它让私人财产成为可能，而私人财产是他所谈论的一切的基础；以及基础设施，它在社会价值上“可能是存在过的最赚钱的东西”（老机场“正在创造数万亿美元的利润”），但即使是长期经营的企业也无法建设，因为时间尺度不匹配。
- 如果他担任“欧洲总统”：“你首先得摆脱气候邪教”——绿色政党建立在反核神话之上，工厂会因为周边某条小溪里曾经有青蛙繁殖而被阻止。接着定义优秀的内部市场游戏——增长“没有速度上限”——之后再讨论如何用财富塑造经济。另一个机制是：进入本地社区的1美元会通过工资支出“循环7次”，这正是本地企业重要的原因。
9. 加拿大的“特朗普妄想症”——以及天安门作文的警告
- 加拿大“拥有地球上未来20年里我们做任何事情所需的全部资源……加拿大本应成为地球上最富有的国家，这纯粹是一个选择。它只需要愿意建设一些矿山——但不知为何这件事存在争议。”加拿大人“过度适应于友善”，结果反而“导致不友善的谎言”——通过隐瞒事实，让人以为一切都还好。
- 谈到Carney，他说：“这是我听过最伟大的演讲之一……但我不认为它完整、可信地反映了地面现实。”如今超过60%的加拿大人认为美国比俄罗斯或中国更具风险——“当然是错的”。加拿大历史上“唯一的胜利策略是：通过帮助美国获胜来获胜”。他赞成所有多元化动作——“欧洲，完美；亚洲，走起”——但“我真的不认为我们需要做选择”：“他妈的把管道建起来”，在国内完成炼化。因为从海狸皮运到伦敦、再以帽子形式运回来开始，加拿大的故事一直是“我们拿到资源，别人赚走钱”（除了Canada Goose、Lululemon、Shopify之外，举一个加拿大终端产品）。
- 对Stebbings提出的中国共产党资助的开源模型正悄悄支撑美国大量初创公司，Lütke认为中国威胁“既被低估，也被高估”。更大的危险是政府对AI设置年龄门槛：孩子不会停止使用，“他们会下载中国模型，你永远、永远不会再看到一篇关于天安门广场的高中作文”。中国模型，尤其切换成中文后，携带“非常集体主义的世界观”——“所有政治都只是集体主义与个人主义之争，其他一切都是人们用来制造烟幕的东西”。
- 关于社交媒体和孩子，他给出的回答是真诚的非回答：“非常巧，我的孩子完全不感兴趣”——他们没有手机，从没提出过要求；家里甚至流传着一个玩笑：拿到手机就会失去PC。但他也特别指出自己有3个男孩——这是一个他主动标注的数据点，因为他认为社交媒体对其他人的影响“非常不同”。
10. “12月改变了一切。Opus改变了一切。”
- 数据是：“超过50%的合理比例”Shopify代码由AI生成，“还在转向更高比例……从12月开始，我们很多最优秀的工程师今年都没写过代码。12月改变了一切。Opus改变了一切。”
- River是一种“生活在Slack里的”AI，“负责Shopify工程工作中荒谬之多的一部分”，在公开频道接受指挥，让“每个人都从每个人那里学习”。这个单体代码库叫World——“河流塑造世界……我们先把她造出来，再问她想叫什么。River是她自己命名的。”这有点疯狂。
- 他明确表示自己改变了看法：原本以为没有行业先验、原生使用AI的毕业生会靠重新发明一切胜出。事实错了——“所有工程师都严重低估了引导的重要性……这就是高层次的编程”，资深工程师凭借经验“能在极短时间内完成不可思议的工作”。即便如此，一名13岁的Waterloo实习生也“非常厉害——简直难以置信”，他的母亲正在听Lütke授课。即将出现的角色是上下文工程，或一条协调智能行动者的“产品构建者”上升路径；有工程管理经验的人尤其擅长，因为“早在Claude Code出现之前，他们就已经提示智能代理很久了”。
- 这门手艺会像机械表一样存续——机械表“卖得比以往更多；增长不是替代，而是叠加”。他的个人优势来自德国工程学派：先设计数据如何在磁盘上持久化，然后“其他一切都可以在上面进行vibe code……也许未来几个月这一点会改变”。他最愿意号召大家遵循的建议是：“你完全可以直接去做……行动会带来信息”，但硬约束是不能让任何人受害；在零和环境里不要做实验，而要“想办法获得正和结果”。
Tobi, I always enjoy our discussions. I've been looking forward to this. I have many, many questions printed out for us today. Thank you so much for joining me.
Yeah, I was looking forward to catching up again. I enjoyed our conversation last time. You know me, so this is incredibly ambitious of you to have lots of notes. Let's try to get as much as we can, but chances are the first question is going to send us into some kind of rabbit hole that hopefully winds up more interesting than expected.
1. Fear of Losing vs Hunger to Win
The first question is not, “Tell me how you started Shopify.” I mean, Jesus, I love these where they start with that, and it's like, “Have you heard my thousand other shows that I've done?” I do want to start there. I found with entrepreneurs that you can basically put them into 2 buckets: one is motivated by the fear of losing, and another is inspired by that incredible hunger to win. For me in particular, I hate losing more than anything, and that's what drives me. What would it be for you?
2. Why "Not-for-Profit" Should Make You Suspicious
The fear of losing. My own answer to your question here is unsatisfactory to myself, partly because I'm in the middle of trying to figure this out, and any thoughts would be helpful there. But I do feel like there's a marked difference with this, because people who fear losing and people who fear winning are clearly going to look at things much more short term. It's the next iteration that matters the most, and I find that is a limiting thing in a lot of cases, especially with partnerships, especially in business, and especially with working together and careers.
It just makes everything change when you're taking a longer perspective. It starts with sitting down with people and helping them become the best version of themselves, puzzling out something for them, or giving them the hardest possible task that they could possibly manage in order for them to grow. All these kinds of things will have significantly more compound advantage if you take a longer perspective, because more of the benefits of these actions will accrue to whatever it is you're building.
I think it really changes the base framing you take to almost every challenge you encounter. It's funny, because I think this is what my colleagues and friends here tell me: once you understand that I'm the source of the energy that I'm tapping into in building, the rest of Shopify starts clicking a little bit more into place. It's a clarifying thing, because most people who have worked there for a while can sort of ask, “How would Tobi think about this?” They say that doing this actually makes things click into place and makes the next meeting with me a lot more predictable.
I would say there are more orientations. It's just that many of those don't lend themselves to becoming executives in companies. I think this is partly why there's a winnowing effect that happens as people are going through their careers, which discards people who come from different motivational backgrounds.
Is that winnowing effect changing in a world of AI, with different skills and different mindsets perhaps required to progress than were required before?
I think it's actually been changing in a very recognizable way. There's this thing called the Enneagram, which is reasonably commonly used and quite good. Unfortunately, I think it's sort of gone woke and changed its terms, so you would have to find a test from 2019 to get the real thing. But if you do it, it sort of separates people into different propensities and whatnot.
With this particular test, if I look through the executives I've worked with over the last 20 years, you end up with mostly Achievers—basically, people who have never seen a ladder not worth climbing. There are a few Fives, who are the people who go really, really, really deep. That's often a CFO. In most companies, especially professionally managed companies, there's nothing else in executive leadership.
I'm an Eight, which I almost already said by talking about my motivational source being this dissatisfaction, or the sort of identification of everything which I want to fix, at least in my little nook of the world. I don't think Eights are welcomed in companies because they're pains in the ass. They just say a thing. They say, “This is bullshit. If our bullshit does not smell better than other people's bullshit, then any amount of fancy dressing doesn't change what it is.”
They're dangerous to everyone else's careers around them. They usually don't get the promotions; they usually leave. What can happen is they start companies because they often see things. The effect of this experiment, I think, that's been run ever since Apple brought Steve back is that it created the boundary conditions for more financiers, capital, and boards to actually give founders a longer runway.
3. Why Builders Are "Eights" and Why Companies Conspire Against Them
The results were pretty good, because now you have diversity among executives. You have someone who no one can get rid of who just calls bullshit bullshit, and that turns out to be really, really good for companies. Shopify is remarkably high on Eights because I really seek them. I fundamentally like this kind of discussion. I propose nothing, and so I think this experiment is being run and can be made to be an active thing that people can act on. I think this can be quite actionable.
One of the most memorable statements I've heard on the show is from one of my best friends, Daniel Dines from UiPath. He said to me, “A lot of people think they want to be me, but it's very lonely alone in my head.” I was like, “I've never felt so seen in my life, and that's why I'm in therapy.” When you hear that, how do you feel?
I think there's no problem with that. I have lots of nice things. I have a really fun job. I think it's perfectly natural that a lot of people want to be me. They presumably do not want to spend the last 22 years the way I did, so I think it's a very natural thing.
4. People Who Build Companies Are Fundamentally Crazy
I'm a glorious mess. The world really needs to understand that the people who build companies are fundamentally crazy people. It's an unreasonable thing to do. There's an old trope that all change and all progress depend on the one unreasonable man, because a reasonable man does not just do things exactly the same way that everyone else is doing. That is a reasonable thing to do in any kind of situation.
You have to expect that. It's a package. It's high-variance people. All the aesthetics around what people believe to be good leadership come from movies, or from the very carefully publicly curated tip of the iceberg of how it actually worked. If you look at any biography of anyone, you realize that people are crazy. Read about Watson, the founder of IBM. You think there are tyrants in business right now? No, you don't know anything about how these things went back in the day. Absolute fireworks every meeting.
5. Why Tobi Didn't Want to Be CEO
I have this job so other people can have the jobs I wish I could have. I'm running interference. I didn't want to be CEO. I learned I had to become CEO because you can't run a product-driven company without also being in control of the company. The needs of product and company are extremely divergent unless you're taking a perspective 3 years out, at which point they kind of come to the same point and actually augment each other.
Basically, every single day is like a marshmallow test: are you going to eat the marshmallow now, or do you get 2 in the future? Frankly, what we know from the actual marshmallow tests is that people don't usually wait for the 2 marshmallows.
6. The Luxury of Long-Term Thinking as a Public Company
It's just not in human nature. Everyone's discontent, but there's some sense to this. If you want to create a great product, you have to take this sort of long-term position. Taking a long-term position doesn't mean just bringing up numbers in 3 years. It means being willing to have bad numbers for a while if you think this is going to lead to the right place, and that's a hard choice to make.
Do you think you have the luxury to do that today, being a public company, given the expectations that Wall Street puts on you?
7. Why Tobi Never Looks at the Stock Price
Oh yeah, of course. There's nothing better in the world than being a trusted public company. I mean, the second-best thing is being a trusted private company. It's much worse to be an untrusted public company than an untrusted private company.
The stack ranking is, unfortunately, people become a trusted private company with their investors and have to actually take a step down to being an untrusted public company by going public. This is why people don't go public. They just can't deal with that, which is funny because it's actually required to then get to the best spot in the end.
That's literally the reason why we did it 10 years ago, when we were tiny, because it just felt like, cool, let's go make the careers of public market investors who bet on us. Then they make lots of money, they become on our side, and now they trust us.
Do you think that chasm is harder from trusted to untrusted if you are much larger? We both know the Collison brothers, I'm sure. That chasm for them seemingly will be much larger, given their $150 billion valuation now.
Yeah. Again, I think any approach works if you have the Collison brothers, so I don't think it matters for them. They can just decide. I think they're a really bad proxy for what you should be doing, because most people aren't them.
I think it's a much saner strategy to do the flip to public as a minor financing. This is the thing: 10 years ago—more, actually, 11—in 2015, Shopify's revenue, I want to say—I'm sure people are going to laugh because I'm getting these numbers wrong—was $200 million or something like that.
I remember, man, the valuation for the first trade. You set the pricing to then figure out who comes into the book right after the roadshow. It's a high-stress meeting later that day, after spending a week just doing a presentation 100 times. You don't even know which city you're in.
Everyone's like, “Here's Fidelity. They did place this order here. You have a book, right?” You have to figure out what the price is. You set that price, which is literally arbitrary. As a CEO, you have to set that price, and you get all sorts of advice. The advice comes from the investment bankers, and you're not their customer. Their customers are actually the people in the book. Their job is to get everyone in the book the cheapest possible deal.
That is, of course, beautifully obscured in the process: who is the customer of whom? Luckily, I studied the way the compensation packages work for the investment bankers before I went on the roadshow, so this was all pretty transparent.
I'm sure they were thrilled about that.
Yeah, I don't think they liked it. I was also very clear about this. I said, “Hey, I understand how you're going to get your bonus. I want you to get your bonus, but I'm becoming a public company CEO, so I'm going to start my fiduciary duty T-minus 10. I'm going to optimize for the company, not for you.”
They totally respected that. In fact, the banker who took me public is my CFO right now, 10 years later, which is really funny.
Picture the 80th floor somewhere in Manhattan: a room with the biggest table you've ever seen, lots and lots of people, paper everywhere, prospectuses, calls, a high-tech iPad with the order book. Everyone looks at this thing, people are making changes, and you have to freeze at a certain point. You pick a price, and everyone puts in a limit order that caps out for the first trade.
I just thought it was crazy. Literally, absolutely everyone in the room was saying, “No, you've got to give allocations, because those are the people who stay in your stock and they're long-term. They don't immediately flip.”
I thought, “Okay, well, whatever.” I raised it past the range because we had a lot of pick-ups. I knew where my growth was going to come from. I was unafraid, and I picked a number that I thought was a really good one. The first trade was $10 high or something. It's like, fucking hell.
Clearly, I was pretty happy with my performance at this moment. The nice, beautiful thing about markets is that they are excellent distributed brains, and they tell you what it really was.
Anyway, the point was, we went out at, I believe, a $1.67 billion valuation. At various times—not right now, I think, but I don't know, I don't check—it's been 100 times higher. As you can imagine, that makes a lot of careers. The people on the road who then didn't fucking place an order for a couple of dollars past their range for a good company probably got their promotion at all the places. They're the senior partners now, so they love Shopify.
Tobi, you're one of the most thoughtful CEOs that I know, and I've interviewed 1,000. So the catalog is quite wide at this point. When we think about today, how many people do you have at Shopify? A round number.
Just about 7,500 to 8,000 people.
7,500 to 8,000. How many people will you have in 5 years' time, do you think?
My real hope is 7,500 to 8,000. At a 100x productivity level, right? I do think there are problems of scale. It would be sad to be in a company that is opportunity-limited, right?
Actually, maybe I shouldn't say this. Most companies probably are opportunity-limited. They're probably not ambitious enough, and I mean this in a product sense. Remember, Shopify has a fairly broad mission: We want to make entrepreneurship more common, right? Inside of this ambition, we can do a lot.
I just think we're going into a golden age of entrepreneurship. It's by far the most AI-safe job. It's by far the most AI-benefiting job. Think about that combination. Everyone who has a Plan B of some kind—some kind of ambition, like, “Hey, I'm going to make products”—is going to be like, “Yeah, but let's fucking go.”
All priors don't matter anymore, because you didn't need to come from a family of entrepreneurs and have been at Dad's office or your grandma's printing shop—in my case—to understand that business formation is something that real humans can do, and that it gives employment and these kinds of things. You can actually learn that anyway, and then, for all the various tasks, your handy sidekick AI, which acts as your co-founder, will just tell you.
If we're staying flat, in the wake of all the layoffs that we see today, is that efficiency gains from AI, or is that overhiring in the COVID era?
Yeah, it's overhiring. AI will change the mix of jobs, no doubt. What you see right now is not AI layoffs. Those are just companies that are really slow and overhired, just like everyone else, doing it now.
What you will see is AI being blamed for absolutely everything. It's the perfect scapegoat. It's the perfect Girardian scapegoat for everything, because it can't fight back.
You can find communities that I think are entirely wrong, built around various people who are talking very eloquently about the dangers of AI displacing jobs. My industry has been gaslighting everyone into AI fear, and science fiction has done it for the 60 years before that, too. I don't even know who's supposed to be on what side.
It's just incredible. It's objectively, mind-blowingly making my life better in every which way—everything.
Do you think you have a utopian lens? When I look at what Claude can do when it comes to social media design, I'm being blunt: a lot of tasks are not actually that difficult, but they are execution-oriented. I know I've done them, so people can't be upset with me for saying it. Claude does them very well now.
How do you keep that utopian mind with the recognition that most of these tasks are easily replaceable with agents?
Look, none of the people who are doing jobs that are just tasks for other people have good jobs. Those are not good jobs. Being a task in an automated task queue is not a great job. Great jobs are things where you have agency and can produce things and make things.
Do you think everyone can have a great job, though?
Look, lots of people who are rich right now choose not to work. This is a very, very common thing. Lots of my friends are telling me, “What the fuck are you doing, Tobi?” And I'm just like, “Hey, I'm born and bred for working.”
I’m like, don’t ask a raincloud to do the job of the sun, right? This is my task, and this is what I do. Maybe some people aren’t like that. Lots of people find that what they do on the weekend, or in the evening, or for some extra dollars—giving lessons, for instance—is vastly more productive for society than the cashier job at the supermarket, or these kinds of things.
I think people will get options. I think products will get enormously cheap, and our purchasing power is going to go up like crazy. People will be able to make choices. We will find new ways of working together. We will find much more liquid ways of moving money around for value generated.
I know multiple people who have built companies, sold the companies really, really well, and became essentially gentleman scientists or gentleman programmers. Most of us who use Claude Code use it through Ghostty on a Mac. Usually, Ghostty is a terminal application, which is perfect.
It’s such a weird thing to have spent your time doing this, but the author chose, after he sold his company, to make by hand a perfectly crafted piece of software that he thought needed to exist. Now everyone agrees with him that it needed to exist, because he built everything on it. I have about 15 Ghostty terminals on my desktop all around me right now, which I’m ignoring because we’re having a fun conversation here.
But if I weren’t, I would be on a phone call with some legal necessity and just telling, through Ghostty, my clankers what needs building at Shopify. That’s the world we live in. I think we’re all going to figure out remarkable ways of doing things. Think about just how good we are at coming up with new jobs. We are incredible at this, and we’ll never stop being incredible at this.
Of the top 10 best-paid jobs today, 8 did not exist 20 years ago.
It’s a brilliant start, and that is going to be true in 10 years, too, and 10 years after that.
My favorite example of this is, again, as a motorsport guy, Formula 1. Formula 1 is one of the biggest spectacles, providing some of the best jobs. There are 800 people building an engine at Mercedes and at every one of the engine manufacturers. Eight hundred people are living in the perpetual golden age of this type of industrial mechanical engineering. Everything matters. It’s all skill. They’re getting scored every weekend on how they’re doing compared to everyone else. They’re basically in an engineering world championship.
That’s the same for the aerodynamics team, the chassis team, and so on. It’s a sport named after a rule book. Think about this: Formula 1 is literally the name. It’s not the name of a sport; it’s the name of a rule book, which is called Formula 1 because the sport falls out of a rule book, and it’s this thick. You can get it. I have a printed version of it. It’s absolutely fascinating.
There are 800 people there, another 800 people working on the car, and then the drivers and all that. We just made this up, right? All of our lives are richer because it exists. I think—
You know, my favorite thing about Formula 1, actually, is the importance of storytelling. If you followed Formula 1 for 10 years, you’ll remember that it was Bernie Ecclestone and, bluntly, an aging sport with middle-aged to older men being the only people who liked it, with a really decaying audience. Drive to Survive saved the sport in many respects.
100%. It brought out the best part, which was accessible to the people who really, really, really worked hard on pulling out the story.
If you watch Rush—you know, I grew up in Germany, so Niki Lauda was a hero in the Germanic world. He was everywhere. What a story that is, right? If you haven’t listened to this and haven’t watched the movie, definitely watch Rush. It’s one of the greatest stories. The sport has had its heroes, its nemeses, and its rivalries in such a way, but it couldn’t get them out until it was combined with media—essentially Hollywood—and Liberty Media.
Think about how much growth there is, and therefore how much entertainment and delight, just by putting pieces together in a new way. An aging sport about a rule book, plus spectacle, ends up being something that’s of way more value than you would ever imagine. I just feel like we will find a lot of that.
The world is very impoverished by how expensive it was to make software at any given time. There is a limiting factor in the world for why things can’t grow faster, right? Of course, growth creates wealth. Well, growth doesn’t create wealth. Growth brings wealth from resources, essentially, into the economies. That’s what it does.
If you build a company, you are causing wealth to be created—to be brought from the hypothetical things that could exist into the actual things. Once things are actual, people can use money to move wealth around, and this is where our economies grow.
I feel like this is worth saying because an enormous amount—at least half of the population—is utterly confused about how this all works and makes very, very bad choices because of this.
Do you worry that the U.S. hates wealth more than ever before? You said that about your friends who stopped working. I have many of those friends. But there is a hatred toward wealth that has never been so visible in my eyes as an outsider looking in at the U.S.
I think we are probably at a high-water mark for the United States, but it’s still sufficiently low that it’s fine and can easily get into self-healing.
Look, I think the more wealth and resources an entity, or even an individual, has, the more deserving of scrutiny they are. The thing that isn’t working right now is that the process of figuring out who’s additive and who isn’t is completely broken. It’s owned by people acting in utterly bad faith.
What do you mean by that, Tobi?
I mean there’s simply no chance that we shouldn’t be deeply, deeply, deeply—indeed, daily—thankful for what Elon is doing on the planet. He’s so singular. My car—I did not touch the steering wheel of my car this morning. It just drove me. It also didn’t use any gasoline. I don’t particularly care about this, but that’s just, like, “Holy shit, that’s cool,” to have that option.
It’s also an amazing car. I have a Model Y. I love it. By the way, the Model Y that I love is a car that is incredibly affordable compared to the norms of cars. It’s better than any of the whatever Kias I could potentially buy. I have no interest in those because they’re not as good a car.
Then I’m going on a longer trip with that car. I also put a Starlink unit into the back, and if someone else drives, I can work from there. That’s also Elon.
Usually, I open my phone and talk to planet Earth increasingly. I don’t know if you’re on X, but what you’ll have seen is that you’re getting replies from people in Japan. It turns out, like in The Hitchhiker’s Guide to the Galaxy, the Babel fish is central to the story. It’s just like, “Cool. You know what? The biggest shared good-or-bad system on planet Earth is X right now.”
It’s the biggest one-to-many system, at least in terms of the biggest follower accounts. There are no barriers anymore. Someone says something really, really smart in French or Arabic or any other language, and I see it in my feed. We can have a conversation in English, and they see it in their language, and so on and so on.
I could spend the rest of this podcast talking about the incredible gifts that he’s made. Of course, he’s captured some very small amount—a percentage point—of all the hypothetical value that exists all around us. He actualized a tremendous amount and captured some of this for himself, which he then reuses to do more of it. He’s a one-man engine.
I’m marveling at this in a way that’s like, “Holy shit, how cool is it that we live on a planet where, when you go through the list of the richest people on planet Earth, all of them have built something?” That’s incredible.
What I really mean is not about Elon. What I really mean is the misattribution of whom we should be angry at. It’s okay to have a lot of questions about rich people who inherited their wealth and then used it poorly, or came to it by divorce, for instance. If you have not created the wealth, you very much are a custodian of the wealth, and you need to use it gainfully for society.
I think what really sucks in the world is that we are prioritizing things that sound good over things that work, in a way that real people don’t. Everyone knows exactly what’s good when you talk to them. It’s just that media—especially state-run media and sort of distorting media, like the stuff that’s been around for a while—is on a beat that is very intentionally trying to obscure what to be upset with.
Well, so if we just combine those two together, media is kind of distorting what to be angry about. I remember when Jeff Bezos’s wife gave some extremely large amount of money—I can’t remember what it was—and they then chastised her for giving it away inefficiently, or to the wrong organizations, or to the wrong people. I was like, “My God.”
But that’s the good part. This is what I want more of. Maybe we’re on different sides on this, but, again, giving somebody money, ideally through a transaction, like, say someone starts a Shopify store and sells a new thing—ethically sourced leather jackets, whatever. When you purchase this, presumably it’s going to be slightly more expensive than leather.
But what you do is make a vote for it. You make a vote that you think this is a product that should exist, and you vote with your money, right? What will happen afterward is that it creates resources for the company, which means it’s a viable product. They can employ people.
It’s a vote for everything that made this. It’s also a vote for the supply chain that created it. Ethically sourced leather now has an additional customer that’s viable. Therefore, the numbers go up. The factories can go—the production can ramp—and eventually it’s going to come down in price.
Therefore, it might even—in fact, with further scale, there’s the potential to make it cheaper than the alternative that some people have a lot of issues with. The point is, every dollar everyone spends is a vote. Real democracy actually happens through capital allocation that’s distributed.
This is also why we need markets and why capitalism works, because capitalism causes markets. Capitalism doesn’t work intrinsically. It just happens to be the thing that causes the most intelligence to be imbued into the markets because it’s the least distorted. When money is used, you should figure out what the second- and third-order effects of this are.
If you’re very Internet, this is sort of hot off the presses, but it’s pissing me off to such an unbelievable degree that I just can’t stop thinking about it. Maybe a really disarming example of this is that I went to a Ruby conference in 2004 or 2005, maybe the first Ruby conference—I think it was the first programming conference I ever went to.
I stayed for the weekend because, at the same hotel, there was another tech conference, which was a conference called—it doesn’t really matter—but it was an object-oriented programming conference. The first talk was, “Well, guys, I think you won, so I’m not quite sure what you’re supposed to talk about, because everyone’s writing object-oriented code right now.”
It was hilarious. I loved the honesty. This is the kind of honesty you can only find among nerds. I don’t know if it ended up being held again, but I think everyone was just doing a victory lap over their cause.
In the real world, we don’t do that. When we run out of a cause, we cause that cause to continue.
Do you not think that by placing such scrutiny on giving, you just prevent people from doing it in the first place? Giving money is not virtuous unless it causes the right things, even if it sounds good.
We need to get away from making good-sounding things beyond scrutiny. You need to look at the data. You really need to figure out what effects you’re causing in the world, right?
What I worry about is that there are people like—respectfully, you, and I don’t mean to make this personal—but you don’t have time to do the due diligence on the organizations that you’re giving money to. You’re building an incredible business, changing a generation—
So then you just don’t bother.
But if I said to you, “Here’s an amazing organization. Just trust me on it,” the scrutiny would say, “Don’t freaking do it.” You’d be like, “Harry, but you don’t know,” so it stops all giving, and then good organizations that would have got money don’t get it.
Whatever good organizations there are, they’re getting a lot of money. Trust me, money is being captured by really bad actors in these systems. You will very likely tell me about Andrew Carnegie’s libraries very soon, because that is the go-to example. That was a hundred-something years ago.
That’s the thing that really worked—built hospitals. That wasn’t Carnegie. Carnegie built libraries, a complete banger, right?
No. When I thought of good, I just thought of Jensen Huang and Bloom Energy. How are we going to fund compute?
This is the thing: the world has too many charity dollars. The problem with charity dollars is that they can’t be given to anything that has a self-healing fitness function, which is a very fancy way of saying the words “for-profit”—markets voting on whether a thing should exist.
And “not-for-profit,” the weird term everyone should be deeply suspicious about, basically says, “The best fitness-function mechanism that planet Earth has ever invented, the thing that’s the only thing that’s ever raised anyone out of poverty, really, at scale, is the thing you’re not participating in.”
So you’re documenting that you’re opting out of markets, but you will not tell us what your actual fitness function is. A fitness function is just: How do you know if a game you’re in is worth playing and continuing? Therefore, the fitness function is going to be determined by anyone. It replaces the merit of an organization with the pull of individuals.
Who is attracted to these places? The people who want to do something for which you have to use pull, and the smoothsters. It’s not the builders who are doing it. By the way, this is not a bug; this is the system that’s set up. Fine. Sometimes it works, and, yes, there are great charities.
8. Tobi's Take on Government Intervention & the Prussian School of Economics
And do you believe in government intervention? I’m in Europe, where we have a lot of government intervention.
No, not really. Well, okay, so let me rephrase this: yes. I’m a fan of Friedrich List, who was a Prussian. Long before James Hunt and Niki Lauda’s rivalry, there were Adam Smith, laissez-faire, and Friedrich List and the Prussian school of economics.
What the Prussian school says differently from laissez-faire is that the role of government is very simple: it’s to define games that have, as externalities, societal thriving, and then completely get out of the game. The players in this game can then cause, through means of competition, extremely positive externalities to accrue to society, which, again, happens automatically in profit-seeking markets.
What a company does is take the potential value that exists—which is infinite, absolutely infinite—and corral it into products that other people then vote for. Again, the voting mechanism causes a larger supply chain and more of a supply of these products to come into society. That creates wealth, and the wealth is part of society.
Any dollar that’s created in a place that brings new dollars into the system cycles 7 times through society, in a local way. So a dollar flowing into a community cycles 7 times through the payroll of people there. This is why it’s so important to have local businesses. Then it goes into the company, through payroll, and out to the barista’s salary at the local Starbucks, and so on.
Money itself is amplifying. This is through mechanisms of trade and also through debt creation by banks. So am I for government intervention? Governments are extremely bad at what they do. Everyone needs to know that: the moment a government takes over anything, it will cost 10 times as much.
That’s really, really bad, because you can literally say, the second someone says, “Let’s run grocery stores as a government,” there are going to be a couple hundred people like me who have to go and build companies to make up the wealth that simply evaporates in that moment, right?
One of the most inspiring things that humanity has ever done is outsource violence to governments, because that creates the environment in which personal property can exist. Personal property is the boundary condition and foundation of everything else I’m talking about.
We mentioned fears or concerns about where our society is worrying. One thing that worries me is, bluntly, Europe falling into this kind of chasm of irrelevance. This is why I started Project Europe, which you kindly supported.
You’ve got the US and Canada, and you also have Asia. What would your advice be? Or, if I placed you in charge of Europe for the future, what would you do to encourage growth and innovation in a relatively stagnating economy?
I have lots of thoughts on this. Of course, Canada has all the resources that planet Earth needs for literally everything that we are going to do for the next 20 years. It’s purely a choice that Canada ought to be the richest country on planet Earth.
So it can become that every moment it wants. It’s actually quite easy. It just has to be willing to build some mines, right? Which somehow is controversial.
9. Canada's "Trump Derangement Syndrome"
Mark Carney, who I’m proud of for pushing back on Trump. I’ve had him on the show multiple times. I would like to see him as a friend. Are you proud of him as a Canadian for bluntly standing up in an elegant and independent way?
I can’t say I’m proud of him for that. It’s one of the greatest speeches I’ve ever heard, but I don’t think it’s a fully credible witness to the reality on the ground.
I also love a speech that is an ambition for a country. I personally disagree with the stance toward the United States. I think Canadians are massively overfit to niceness. It’s kind of the best thing and the worst thing in a country to prioritize niceness, because it’s certainly pleasant, but the problem is that it leads to unkind lies. It means you’re lying by omission. You’re lying by saying things are fine when they aren’t, right? Speaking the truth is a much more important test.
Unfortunately, the way that intersects with the Trump administration is really brutal, because it creates this sort of positive feedback loop inside Canada. He is so unbelievably not nice in any way you look at it. There’s a Trump Derangement Syndrome in Canada that is just stunning. More than 60% of people, I think—the numbers have gone up since—think that the United States is the largest risk to Canada, much larger than Russia or China.
And you think that’s wrong?
Of course it’s wrong. Canada is a small economy attached to a hegemon. There’s been one winning strategy in Canada’s history, which is: win by helping America win.
Carney is saying that it’s over. Everything he’s doing, I’m agreeing with: “Let’s diversify the economy.” Bang, great idea. “Let’s get closer to Europe.” Perfect, let’s go. “Asia, let’s go.” I’m for all those things. I just really don’t think we need to choose.
The obvious way for prosperity here is to build the shit out of pipelines, build the shit out of our industry, and get the resources that everyone needs. Start actually refining the stuff ourselves, because we have the most educated workforce on planet Earth here and we can literally do everything.
Canada started from trapping beaver pelts, sending them to London, getting them turned into hats, and then buying them back, right? That’s the Canadian sales story: we get the resources, other people make the money. We’ve never changed from that. We don’t make the end products ourselves, and so it’s very, very rare that Canada does anything. It’s Canada Goose, Lululemon, Shopify, maybe Molson Lite. What other products exist from Canada?
You mentioned that we overexaggerate the concern around the US and the threat that they pose. The thing that I think we were all not talking about is how every great American company is using frontier American models and then going, “Well, that’s too expensive, so let’s use open-source Chinese models to get as close as we can to the benchmark set by US frontier models.”
10. The Real Chinese Threat
These CCP-funded models are basically powering a lot of the US startup ecosystem in many ways. Do you think the Chinese threat is underestimated?
I think it’s both underestimated and overestimated. The much bigger threat is that in a lot of countries—I think this happened in Spain already, and I think even in the United States—social media aside, the moment the government has the ability to decide at which age children can use technologies, they will.
AI has so many naysayers. We’re going to tell kids they can’t use AI. Kids will not stop using AI. What they will do is download Chinese models, and you will never, ever, ever get another high school essay about Tiananmen Square as a consequence of that.
Is that truly what we want to create here, where children are incentivized to get a monoculture because models are trained and RLHF’d into a perspective? Chinese models, especially if you switch them to Chinese and then ask them to translate something else, have a very collectivist worldview, which makes sense because it’s a collectivist country.
This is the predominant fight. All politics is just collectivism versus individualism. Everything else is people smoke-screening this stuff.
Do you let your kids have social media? I was going to ask this of one of the biggest technology leaders. I can tell you after the show, that’s one question we won’t let you ask.
This is going to be super unsatisfying to people, because I feel people hope that I would have a very strong opinion on this that’s actionable, and I don’t. For the craziest coincidence, my kids are totally uninterested in social media. They just don’t completely give a shit.
They follow links to social media sometimes, but they have no interest in having accounts, and they don’t have phones. They’ve never asked for them. It’s bizarre.
I saw them when they were very small. We have PCs, because we do video games as a family. I once probably just ran my mouth. When phones came up, I think Tristan said that a bunch of kids in his class were starting to get phones. They’re super distracted, and it’s cutting into their book-reading time, which tells you a lot about Tristan.
Then I said, “At some point, you will ask me for a phone, and you’re going to have to cash in your PC, because you can only have one of those two things.” Somehow that became a meme in the house, and they’re so into their PCs that it just never comes up.
Honestly, I can’t add much to this conversation. Also, three boys, which I think is a very important data point there, because I think the situation with social media, for what it is, is very different there, with gender having a different effect on them.
11. What Tobi Would Do to Fix Europe
Going back to our question, I’m making you the new president of Europe, which is a brilliantly American title and doesn’t exist, obviously, as I’m aware, for anyone who’s about to chastise me in the comments. What would you do to ensure Europe remains on the global stage?
Well, you have to get rid of the climate cult, first of all, because it’s one of the most evil things that’s been wrought on populations.
The climate cult—you mean what?
Random green parties having, as part of their founding myth, that somehow nuclear power plants are bad and shit like this. But it’s everywhere, right? We can’t build incredibly important infrastructure and factories because some fucking frog breeds once in some fucking creek on the perimeter. What the fuck is this?
If you were in a sane world and someone wrote a science-fiction book about the future and described the shit that you actually have to do to do something on this planet, the book would go nowhere because everyone would say, “This is just completely made up.”
But we all live in a world where people do not understand that there are very few people who have the capability to truly build things. They need to be enabled by the village to build those things, and then they also have to be held accountable by the village.
Again, I think Europe, especially now that the UK has left the European Union, needs to double down on Prussian economic theory, on Friedrich List and his co-conspirators, and just define excellent games that create internal markets.
A country is defined by its internal government services: security, safety, property rights, a body of law, policing, and these kinds of things. Governments have no money. They can’t make money. They can only take money by compulsion from people who create things or work.
That’s fine, because that is a funding mechanism by which they can provide these services that we deem valuable, build infrastructure, and so on. Infrastructure is a really interesting case because it’s probably the most profitable thing that exists and ever has existed.
Just building highways changes everything, and building an airport—everyone talks about how much an airport costs. Every airport ever built that’s actually connected into the sort of hub-and-spoke system on planet Earth is probably, especially the older ones, running trillions of dollars of profit in societal value creation.
Infrastructure is one of those things that businesses are really bad at making. It sucks that we’re committing businesses to having to build infrastructure, because it doesn’t actually fit into the timescales.
Even long-term-focused businesses can't really build infrastructure very well. So you have the odd SpaceX that can start a new city in a swamp. But that's just not the way to do it. Infrastructure building is very good, and in the end you define games, and you are therefore getting an economy that's growing. By the way, there's no speed limit for that, and afterwards you can have a discussion about how you take the riches from this economy—the additional profits—and then sculpt the economy that you want. This has been, ever since the Enlightenment, the path by which all progress happened.
12. Shopify's Biggest Mistake
What were you wrong on that cost you the most?
Most of the things you're wrong about are the paths not traveled, where many years later you understand that the counterfactual would have been a better position. There are lots and lots of deals. I think my most public mistake was going into full-on logistics and physical warehouse building just before AI got really good, and we didn't have the ability to do both of those things at the same time. So we actually had to abandon this thing.
I still think the initial decision was right, which is actually how I value my performance. But I think this is starting to be harder to make that case because I know now about information that I could have had and that would have been available to me back then, which would have discredited this path. So I probably just got it wrong and maybe should admit this. That one sucked because it affected people's lives with employment and so on.
But I'm wrong all the time. I've been wrong more times than most people will ever be wrong because I do a lot of reps on it. I make falsifiable statements because I want people to falsify them. I'm unafraid of a conversation. I also just don't care that much about what people think about me, honestly.
13. Great Leaders Must Be a Heat Source
It's just like, wow, that, with the weight of your decisions and the weight of your voice. When you speak, 7,000 people follow you, and with the frequency of your wrongness, that could be very costly. I think you've said before your job as a CEO is to inject chaos into an organization.
I prefer to call it temperature these days, just because chaos is probably too negatively connotated. What is heat? Heat is atoms that jiggle, right? With enough heat, these atoms—I mean, they usually come in the form of molecules—and with enough heat, the molecules fuse into something maybe worth having more. That's what all of chemistry does.
I think it's very hard to forge anything new at room temperature. I think great leaders must be exothermic and must be a heat source for the company. It's a very different opinion from the custodial idea of leadership, which is also a valid position to take.
I do not need to inject heat into any area that has heat, right? That's not needed. There are some areas that are so red-hot because everyone's just into chasing innovation or incredible vision or whatever that my services are not needed there. I get to play a completely different role in these areas.
But those are usually not the ones I spend my time on because I spend my time on everything that's wrong, which is again why people should not overestimate how good these jobs are. I have a folder. One of my concessions to what you could potentially call vanity—although I don't know if it's quite vanity; maybe call it motivational plumbing—is that I have a part of my Obsidian folders that are just people I respect who said things about things I build that are meaningful to me: little jolts of positivity and mini-accomplishments.
I can handle a couple of days of this, but once it gets to weeks where I've only heard things that are going extraordinarily poorly, or people are making life decisions with incredible consequences—retiring, maybe, or spending more years on the beach because I somehow figured out that's a good idea—then there are deep consequences: months of work to figure out how to create the continuity of an area that otherwise was going great, or whatever, or much worse. You get a lot of people's health issues; it all affects companies. Sometimes you get into these really, really dark places, and then it's just a nice pick-me-up.
I'm not completely detached emotionally from this. In fact, I'm enormously emotionally attached to Shopify. I think that's actually a common thing with founders, too.
Which is why I'm surprised when you said earlier you don't look at the ticker. I pretty much only interview public-company CEOs now, and I've never met one like you.
Again, it depends on where you are. It depends on what kind of company you run. In some companies, it also has more relevance, right? If you run a gold mine, you have no differentiation in the product. You might have differentiation in the amount of mining rights.
I talked with a gold-mine guy. This is why this is my example here; I just thought it was funny. For them, technology is actually investor relations. If you're the CEO, your job is to be everywhere and play golf with everyone, and be extremely charismatic, because if you can somehow finagle your multiple to be slightly higher than the other gold mines, you just buy the other gold mines, right? This goes very quickly, so you consolidate.
Actually, your product is not the gold. Your product is actually your investor relations. That sounds like an environment in which I would not thrive.
Now I know why you're not a venture capitalist, Tobi.
I think I would be a very bad venture capitalist. I imagine—I don't know. I think I tend to think I can get on board with anything. It wouldn't come naturally to me, and therefore it probably would not be what I would do. I'm just too interested in making things. I would blog every day of my life.
14. Senior Engineers Don't Write Code Anymore
Tobi, I'm going to do a quick-fire round with you. I'll say a short statement, and you give me your immediate thoughts. What have you changed your mind on most in the last 12 months?
One thing I'm wrestling with is that I had to change my mind. I thought straight out of school, engineers would have a really big advantage with no prior and immediately being AI-native. Many of them are exceptional and are speed-running their careers in ways that are incredible. There are even younger people at universities. We had an intern who was 13 from Waterloo, which still blows my mind, and he was not even in his first semester. His mom has to accompany him for all the classes, which I love. By the way, he's very good; it's just unbelievable.
We're seeing incredible career-trajectory increases that I think are coming here. But I had to change my mind because initially I thought just having no prior would be better because we had to reinvent everything. It turns out the way agentic programming actually works is that you give an initial problem, you give a task—it might come from a ticket—and afterwards an engineer constantly talks with the AI to steer it in the right direction.
All engineers are massively underestimating how important the steering is. It's incredibly important. It's actually just the same as programming; it's just very high-level. It turns out that just having done a lot of reps and having seen a lot of those means senior engineers steer these things in such a way that they can accomplish incredible feats in very, very little time. I don't know what that means exactly, but it's something I had to change my mind on.
What role does not exist today that will be incredibly common in 5 years' time?
15. The New Role That Will Dominate in 5 Years
There's something about context engineering, a term that I had a hand in popularizing. I think that will become more recognizably a lead role in companies that will potentially subsume a bunch of other, more specialized roles right now.
Honestly, it turns out I think people who have done engineering management and management in general are actually excellent AI programmers because they have been prompting intelligent agents for much longer than Claude Code existed, and they are just good at communicating. But now I think increasingly their best work is to prompt other people and AIs directly as part of the same team. I think there’s going to be some ascension-path job that you can get into from product, design, or engineering, something like a product-builder role that involves the coordination of intelligent actors to build products. It won’t sound or look or be described as what I say, but that’s sort of the directionality here.
Honestly, people who are good communicators are usually good thinkers because good communication is usually the distillation of complex thoughts into unambiguous words. Even when I say this, it just sounds like context engineering, doesn’t it? So I think that is becoming a very, very important role.
If you’re building a large-scale AI system, we have one called River, which lives in Slack and does some ludicrous amount of Shopify’s engineering. People just talk to it over Slack, and it’s called River because Shopify has a very big monorepo. Our monorepo is called World, and rivers shape worlds, so I think that’s very poetic.
16. "River" Shopify's AI Engineer That Named Herself
River named herself. We first built her, then asked her what name she wanted, and she came up with that, which is kind of crazy. So River sits on Slack and does a lot of Shopify’s engineering, with people steering her through public Slack channels, which is awesome because then everyone learns from everyone else.
What percentage of code today at Shopify is AI-generated?
It’s a fair bit over 50%. It’s converting to much higher numbers. I don’t think we’ve updated the public numbers there, but many of our best engineers have not written code this year, ever since December. December changed everything. Opus changed everything.
Does that excite you as someone who loves the craft?
Oh, I love it. It’s so wonderful because I just love the craft, and the craft is not lost. I also love mechanical watches, which make no sense, right? But they’re selling more than ever, right? People always overestimate—again, this is what growth is. Growth is not subtraction of things or replacement. It is just adding; there’s more in the world.
We now have AI as well, on top of all the other wonderful products we can consume. Anyway, I can sit down and write code when I want to do this. Sometimes I do want to do this. Often it’s part of steering, right? I write the part of the code that concerns itself with the data structures and how the data is persisted on disk, which, unfortunately, is sort of a German school of engineering that’s different from the computer-science world of North America. I think North America is wrong.
I think you should think about how data is persisted on disk above everything else in your databases or so. Then you can plan the system with this. I still follow this, and I work on it either by describing it or by literally just making the data-access layer. Everything else can be vibe-coded on top because I know the result will be excellent as long as the data structures are used right. Maybe that changes in the next couple of months.
That’s an area where I can definitely steer the AIs to get better results than they can do themselves for a little while longer, and then afterwards I’ll do something else. Maybe I’m going to do it just for kicking it old-school, but what my engineering ability allows me to do is reason about these systems, right? I can be a partner and a coworker with AI because I can ask it very specific questions to go chase down.
I have those same questions all my life, in every meeting. I’m really wondering what the data structure is here, how we’re persisting this, what we save, what we discard, or what we rederive in the data warehouse, instead of just writing it down in the first place. Now I have my laptop open and River open, and I can ask a question and get a result. Because again, River can use the entire monorepo and just chase down whatever I’m asking her.
Sorry, random one. Would you encourage your children to go to university?
17. Would He Encourage His Kids to Go to University?
They seem uninterested right now. I think my oldest is probably going to make a play for it. I definitely would not encourage him to go to university just for being part of university. I would definitely encourage him to go get a degree, especially one that’s hard to get into, because then you’re surrounded by people who also got into it.
I think figuring out how to get in the room with people who really, really want to be in the room on the topic of something you care about is actually the cheat code for any career.
Would joining Stripe today not be better?
My bias would be that if you can be of value to a small company, that’s the thing, right? You have to have the skills to actually be of value, and the company agrees with you and sees that value. You should do it. But that’s actually a wonderful test for both sides.
An interview is a test—both sides interview each other, right? That’s what you should try, and I think it’s better than the alternative, because again, it’s much harder to get into a company worth working for than it is to get into university.
Do you believe in nepotism, Tobi? Respectfully, if one of your kids applied for a startup job, the surname is relatively obvious.
Remember, my kids have the genes of my wife and me. They probably don’t need it. I think nepotism comes in if someone needs a little bit of help, then I think so, but they’re fine. I don’t believe in nepotism.
I think, actually, nepotism has a bad rep, frankly. It’s annoying, but I’m starting to understand it also. I mean, nepotism is probably just actually bad. I’m a very big fan—this is probably discernible from what I’m saying here—of Thomas Saul. Basically everything he’s ever said is incredibly smart.
Among the many things worth quoting, I think in a book he wrote when he was already in his 80s, one of his regrets about society is that we have spent the last 50 years replacing things that work with things that sound good. Ever since reading this, I don’t have a knee-jerk reaction anymore against basically anything I’m hearing about that sounds terrible. I just try to figure out what the redeeming value of nepotism is. I think the answer is none.
Merit, double-blind, is the gold standard. Unfortunately, merit, double-blind, is not the standard. This is very, very rare. So I think nepotism might be better than some of the crazy shit people have been doing to decide who should be appointed, especially in academia.
In Canada, we have a big thing with basically all the job postings for any kind of academic seats now being sort of a full-on intersectionality fever dream: find exactly a description that fits one person in terms of all the disadvantages they had in life. Which, I mean, always comes from a good place, and maybe it also ends up having some—maybe it’s an arbitrage sometimes—because if you know, actually, there are always countervailing forces that have prevented them from getting the jobs they ought to, and then on day 1 the person you do find is a complete banger in the job, hats off to you. You just played the system perfectly.
In any other case, you just made a choice of prioritizing things that are not merit, right? Which again goes against the Enlightenment values that literally underpin everything that anyone likes about planet Earth in Western society and the foundation of all the countries that essentially have net immigration. So I think we should probably do some of those.
Final one for you. What’s the best piece of advice you’ve ever been given? Mine is, “You’re never wrong to do the right thing, and the right thing is often the hard thing.”
I think the best advice I’ve been given—but I haven’t been giving it, but it’s sort of a rallying cry—I kind of want to coalesce on that sentence just because it’s memorable, and I think people can’t hear it enough: You can just do things.
At any point in time, the system exists to try to get to good outcomes. If you know what good outcomes are, you can step out of the system and give it a go. The cost is lower than ever, and action causes information. Again, one of the best things that can possibly happen is you figure out something that there are actually a lot of reasons why the system discourages it.
Now, there are limits to this, of course. This needs to be a completely victimless endeavor; you can’t have someone suffering because you’re trying a thing. That’s very important. People can’t have negative consequences from these kinds of things, because then you’re just being an asshole, dividing people and prioritizing your own gain over other people. Fundamentally, you’re engaging in a zero-sum thing.
In zero-sum environments, you should probably not run experiments. But if you are in a zero-sum environment, you should figure out how to get to a positive sum, and that’s much more often possible than people imagine.
Tobi, I'm not sure I've ever had quite such a broad-ranging discussion. I'm not sure if most of your interviews cover the future of climate, Trump's relationship with Canada, and the future of developers in organizations, but I've so enjoyed this. Thanks.
I wrestle with these things. If you get me at a different time, I'll talk about different things. It would be awesome if there were some people who really disagreed with everything I said but were still listening at this point, right? That's all I want. It's all anyone wants: let's get back to having conversations and allow people to be wrong. I'm wrong all the time; it doesn't matter to you.
So I think this is fine. We can lower the temperature. We can realize there are some priors in society, some things we've inherited, some things we've put in place, and some things which worked. I've made very, very few comments on what we ought to be doing. I mostly talked about creating the boundary conditions in which people can just do things right.
That's the genius of the world we inherited: our ancestors created, ever since the Enlightenment, a sandbox in which people can do things. They take responsibility for doing things. They are personally accountable for the outcomes of this. They go to jail if they do something bad, but they are allowed to have property if they are right.
Everyone who builds something is making an incredible Faustian bargain with society here, in a world that discriminates against builders in a way just on account of their products having worked. I own 6% of Shopify. That makes me incredibly wealthy, yes, but that also means 94% of it is owned by other people. All those people made a lot of money with the company.
That's not even talking about the customers and all the people employed by our customers. There are probably 10 million people working day to day on Shopify. Their job is to do the things that Shopify asks them to do, like send out products, deal with customer relations, build new websites, and so on. That was always there, and Shopify has brought it into existence in an actual way so it can be consumed.
Shopify isn't the only company that does this. People vote right now, and hopefully for a long time, every month by paying the subscription. Shopify's instantiation of this idea that needed to be pursued is the one that is valuable to them.
There is a belief in society that when there's a good, there must be a counterreaction. It's some kind of misapplication of Newtonian laws. Some things are just good. Some things are just super fucking good, and the downsides might even be describable, but they just pale in the face of all the good that comes.
It's like we said earlier, though, which is the demonization of wealth. A lot of people you see in America think that, to gain and have a billion dollars, they must have stolen from someone who doesn't have money.
Show me one. You can't get to a billion dollars by stealing money. Maybe you can—maybe there are some instances—but talk about those specifically. The people who get their wealth by building companies have not stolen anything. They have created a product that people voted for.
It's actually the most democratic thing that exists. It's vastly more democratic than any of the elections that we actually go and vote in. People shape the world around them by the way they allocate their money to products.
Every time you buy something from a local shop, you're voting for the welfare, the future, and the thriving of local shops in your area, not just for the product that you purchase. There's nothing bad about that shopkeeper. They are putting themselves out there and working extremely hard.
It's an irrational act to start a company. You're working extremely hard for yourself instead of working less hard for someone else, but you're actually creating something, and other people get to have employment there and have jobs that work according to their lifestyle. It's so weird that we are treating this with contempt.
Again, I'm very, very pro-constructive criticism, but constructive criticism is itself something that someone builds, right? There are people who create excellent constructive criticism, and I learn a lot from them. I'm fortunate that I get a lot of constructive criticism, even just as a public company CEO.
You have some people write incredible external memos that are really worth reading because they give you a perspective you didn't have. That's them building up a picture in the public mind about how to think about something.
But there's also lazy and bad-faith criticism, or just hot takes and this kind of stuff, that is actually incredibly corrosive and should be held in incredible contempt by people—and we don't, right? How many times have people who are saying something incredibly wrong looked through all the things people have published during COVID times, or even just climate-panic pieces? How many times have you actually posted revisions to this?
It's very rare, and so you realize it wasn't about truth, right? Otherwise, I would do it, because that's the correct thing to do if you were wrong. The ice caps did not melt. In fact, in the timeframe—in the years when they told us that the ice caps would be gone—they were not gone.
That doesn't mean there is no climate change. Climate change is happening all the time. I'm very pro-free speech. Free speech is a mechanism of self-healing for everything else related to property and classical liberalism, and so it's worth protecting.
But there's no absolute free speech, right? You can't scream “Fire!” in a cinema, in almost any place like that. You're going to the precinct afterwards, right? Somehow you can do it in the media, or in all sorts of other places, if the timeframe is long enough and it's fine.
Do you worry about that moving forwards, when the most common response on Twitter is, “Grok, is this true?”
No, I think this is why I'm so bullish about the future. There was an effort imbalance—a mismatch—in terms of bad-faith acting, because it's really easy to do. It's one assertion, right? Famously, there's a Stalinist kind of thing: for every man, there's a combination of words that will undo him.
It doesn't need to be true; it just exists. There's some accusation that's just plausible enough that the person disappears afterwards. I think this was specifically the case in the Soviet Union. It's always true when there's a lot of collectivism around, which is really just centralization of power with the state.
Fact-checking is very, very hard. It's asymmetric warfare. It's sort of information terrorism, to take a very absurdist position on it. Proving that it's not true is a lot more work than saying the thing is.
There aren't that many bad-faith actors. It's actually a small number, but they can be incredibly prolific and seed all kinds of important things. I think AI is just going to move the power back to the good guys, I believe—or that's actually the wrong framing. There's no good and bad. Everyone's good in their own local system, and some people have the wrong system.
This is why we need conversations to get them to the valuable ones in society. I mean this in the nicest way to you: you can't actually believe that there are malicious bad actors who will use AI to exploit.
I didn't say that. I'm sorry.
No, you're completely right. There's also a societal level and an individual level. On an individual level, which is the one we document, there's going to be so much shit happening—voice cloning, the hacking stuff. All software turns out to probably not be good enough in terms of security. That's going to be a whole lot of shit we will have to deal with, absolutely.
However, the same models can also write fixes for these kinds of things. In fact, being able to apply as much intelligence in the form of artificial intelligence to the side of preventing fraud—I think the balance will shift here. I think you're actually going to get this in a much, much better way.
Even claim testing—whether a statement is true—is something that can be done by a council of models, especially a Chinese-trained model, a German-trained model, and a French-trained model. Let's get everyone's opinion on something, give a different perspective, and synthesize it.
Building the thing I just described, I can do by going to one of those Ghostty terminals, taking the transcript of what I just said—the words—and 2 or 3 more prompts. I can put this on the internet, and it will work.
I would have to pay for it because the tokens are expensive right now. But suddenly, we have a new institution which is like, “Well, let’s see what everyone would think. What’s the council saying about this claim? Do some research.”
So we’re going into a world where I think fewer of the airwaves will be dominated by bored people who act in bad faith, because I think the other side has the ability to push back. The Claude and Grok thing is sort of a primordial version of this, is what I’m saying.