Patrick Collison

摘要

Patrick Collison（生于 1988 年 9 月 9 日，爱尔兰利默里克）是 Stripe 的联合创始人兼 CEO。Stripe 在 2026 年 2 月估值 1590 亿美元，2024 年处理了 1.4 万亿美元的支付总额（约相当于全球 GDP 的 1.3%）。他是世界上最年轻的自手起家亿万富翁之一（2016 年，28 岁），也是 Arc Institute（生物医学研究机构，2021）、Fast Grants（COVID-19 紧急科研资助，2020）、Stripe Press（2018）、Frontier（9.25 亿美元碳清除 AMC，2022）的共同创始人，与经济学家 Tyler Cowen 共同提出 "Progress Studies"（进步研究）研究纲领。

他不是一个寻常意义上的"科技 CEO"。他在乡间长大，12 岁时跟当地修道院的修士每周一次学习古希腊语；16 岁用 Lisp 写了一门叫 Croma 的编程语言并赢得爱尔兰青年科学家大奖；19 岁第一次卖掉公司；20 岁辍学；22 岁创立 Stripe。他的兴趣从 API 设计、数据模型、Conway's Law，延伸到经济增长、住房政策、生物基础模型、Enlightenment 的脆弱性。他偏爱旧书甚于新书；他坚持认为公司应该"plug into existing rails"（嵌入既有体系）而非取代之；他在团队大会上反复强调"multi-decadal abstractions"（跨越数十年的抽象设计）；他公开拒绝 IPO，称之为"a solution in search of a problem"。

他的核心思想轴线，可以概括成一句话："progress is amazing, influenceable, and understudied."（进步是惊人的、可以影响的、但研究得太少。）他几乎所有的活动——Stripe、Atlas、Press、Climate、Arc、Fast Grants——都是这一信念的具体投影。

一、生平时间线

第一章：乡间童年与启蒙（1988–2003）

Patrick Collison 1988 年 9 月 9 日生于爱尔兰利默里克（Limerick），在邻郡 Tipperary 的小村 Dromineer 长大——按他自己的描述，是"in very rural Ireland, the middle of the countryside surrounded by farms and fields"（爱尔兰的乡下，被农田和田野包围）。他是家中长子，下面有两个弟弟：John（小他两岁，后来成为 Stripe 的联合创始人）和 Tommy。

家庭背景极为关键：父亲 Denis Collison 是电子工程师，同时经营一家 24 间客房的湖畔小旅馆；母亲 Lily 是微生物学家，自己创立了一家做企业质量管理培训的公司。这意味着两位父母都是创业者——这是后来 Patrick 反复强调的事实：

"starting a company was really not some strange, foreign thing. It was pretty normal."

"创业并不是什么奇怪、陌生的事。它挺平常的。"

Tim Ferriss Show #353, 2018

母亲一边带孩子一边做企业培训，父亲一边做工程一边经营旅馆。两兄弟看到的世界是：自己开张做事是日常生活的一部分，不是某种奇异、遥远的可能性。

童年的另一个核心特质是 "free-range parenting"——给予孩子极大的自由和自主权。Patrick 描述他和 John "had to figure out ourselves what was going to be entertaining and interesting and fun." 父母 "really give us agency and autonomy and treated us as adults"（真正给我们自主权，把我们当成年人对待）。最极端的一次：弟弟在美国接受手术时，父母把 Patrick 和 John 单独留在家里"很长一段时间"。

家庭每年的欧洲漫游：

"would take the ferry to Europe... bring some tents or a little caravan. And we'd just go driving around the continent staying at all these different campgrounds."

"会坐渡轮去欧洲……带上几顶帐篷或一辆小房车，就这样开车在大陆上四处转，住在各种不同的露营地。"

"the most vivid memories I have mostly of these trips... it's like particular books that I read."

"我对这些旅行最鲜明的记忆……主要是那些我读过的特定的书。"

Tim Ferriss Show #353

他 8 岁时第一次在利默里克大学上了电脑课，10 岁开始正式编程。他在爱尔兰语小学（Gaelscoil Aonach Urmhumhan，位于 Nenagh）读完小学，然后入读 Castletroy College。

最具传奇色彩的一段插曲，是他 12-13 岁时学习古希腊语。当时他对荷马表达了兴趣，母亲就 "went and found somebody at literally a local monastery who was willing to teach ancient Greek." 之后两年，每周放学后他都要去修道院上一次课，跟修士学古希腊语。

这段经历——加上欧洲露营、自由阅读、自我摸索——构成了他后来反复强调的一个底层信念：

"nobody is going to teach you to think for yourself."

"没有人会教你独立思考。"

Tim Ferriss Show #353

第二章：青年科学家时代（2003–2007）

2004 年，15 岁的 Patrick 在欧洲青年科学家大赛（EU Young Scientist competition）上认识了一位瑞士-美国选手 Silvana Konermann。当时他不会想到，17 年后他将与这位生物化学家共同创立 Arc Institute；18 年后他将与她结婚。

同年他参加了第 40 届 BT Young Scientist Exhibition，提交一个叫 "Isaac" 的 AI 项目，获得亚军。

2005 年 1 月，第 41 届 BT Young Scientist Exhibition，16 岁的他凭一个叫 "Croma" 的项目夺冠——一门基于 Lisp 的编程语言。爱尔兰总统 Mary McAleese 颁发奖金 7,500 欧元和 Waterford Crystal 奖杯。

"Collison attributes the success of his company to his win in the Young Scientist and Technology Exhibition."

"Collison 把他公司的成功归因于他在青年科学家与技术大赛上的获奖。"

Wikipedia

Croma 的 Lisp 渊源不是巧合。他后来在 EconTalk 上回忆：

"Had I not decided to learn the programming language Lisp, or had I not attended a specific conference at Stanford when I was 16, I likely would not have applied to college in the U.S. or dropped out to start a company."

"如果当年我没决定学习 Lisp 这门编程语言，或者 16 岁时没参加 Stanford 的那场特定会议，我大概不会申请美国的大学，也不会辍学去创业。"

EconTalk, 2019

那次斯坦福的 Lisp 大会几乎完全改写了他的人生轨迹——一次小概率的青少年学术兴趣，把他带到了硅谷，让他申请美国大学，最终走向辍学创业。这是他后来反复用来支撑"opportunity is unevenly distributed"（机会分布极不均匀）这一观点的个人例证。

这段时期的思想伏笔

• Croma 用 Lisp 写，证明他从最早就着迷于"可以修改自己结构的编程语言"——这条线后来直接通向他对 API 设计、Mathematica、Smalltalk、Conway's Law 的执着。
• Seymour Papert 的 Mindstorms（基于 Logo 的编程教育书）出现在他个人书架的"特别推荐"列表上——Logo 是 Lisp 的儿童版方言。Croma 的设计美学几乎可以追溯到这本书。
• 与 Silvana Konermann 的相识发生在一场科学竞赛，这条线 17 年后开花成 Arc Institute——他对"科学竞赛 → 终身科学投入"的这条路径有深刻的个人理解。

第三章：Auctomatic 与 MIT（2007–2009）

2007 年，Patrick 18 岁、John 16 岁。两兄弟在利默里克成立了一家名为 Shuppa（爱尔兰语 siopa "店铺"的玩词）的初创公司，做的是面向 eBay 重度卖家的工具——批量上传商品和库存管理。

他们去找爱尔兰国家创业基金 Enterprise Ireland 申请资金，被拒绝。这次拒绝直接导致他们申请 Y Combinator，被录取，于 2007 年夏迁往加州。这是 Stripe 故事最隐秘但最重要的伏笔之一——后来 Patrick 在 2012 年用一篇博客回望"Stripe 能否在爱尔兰创立"时，第一条就是"weaker investment climate with fewer experienced investors"（投资环境弱、有经验的投资人少）。

到加州后他们与牛津毕业生 Harjeet Taggar 和 Kulveer Taggar（也是 YC 创始人）的相似项目合并，更名 Auctomatic。

2008 年 3 月 21 日（Good Friday，复活节前的星期五），Auctomatic 以约 300 万欧元（约 500 万美元）卖给加拿大上市公司 Live Current Media。Patrick 19 岁，John 17 岁——一夜之间成为百万富翁。

"The brothers were unable to secure funding from Enterprise Ireland before relocating."

"两兄弟在搬迁之前未能从 Enterprise Ireland 拿到资金。"

RTÉ News, 2008-03-27

Patrick 卖完公司后于 2008 年 5 月赴温哥华任 Live Current Media 的工程总监（Director of Engineering）。后来他申请并入读 MIT。John 则去 Harvard。

两人都在 2009 年前后辍学。Patrick 离开 MIT 时心里很清楚自己在追什么。

这段时期的关键决策

决策：拒绝 Enterprise Ireland 路径，转向 YC（2007）

• 当时背景：18 岁的爱尔兰兄弟，只有一个 eBay 工具的想法，国内基金会拒绝，YC 是当时还很新的"创业孵化器"实验。
• 关键假设：YC 提供的不只是资金，更是 "role models and cultural capital"（榜样和文化资本）——他多年后在 EconTalk 上反复强调，早期公司是 "fragile, delicate, and implausible"（脆弱、精致、不像样的），需要有机制托住它们。
• 结果：YC 给他们提供了硅谷网络、技术合作伙伴（Taggar 兄弟）、退出路径。这次成功让他们后来再做一次（Stripe 2009）时知道这条路通往哪里。

"It is vividly clear how fragile, delicate, and implausible successful companies are at the larval stage. We see the potency of mechanisms to shepherd them, provide them with role models, and offer them cultural capital."

"你能非常清楚地看到，那些后来成功的公司在'幼虫'阶段是多么脆弱、精致、不像样。我们也看到那些托住它们、为它们提供 role models 和文化资本的机制有多重要。"

EconTalk, 2019

第四章：Stripe 创立（2009–2011）

4.1 "Dark Energy" 检验

2009 年底，Patrick 和 John 都已离开学校、积累了一些资金、在硅谷的网络里游走。他们想做下一件事，反复回到同一个困惑——为什么没有一个简单的方式让开发者在网上收钱？ 他们尝试给自己的小项目接入支付，每次都要走银行、几周文件、复杂申请。Patrick 的话最直白：

"How could there not be an easy way to move money online? Like, it's not an obscure need."

"网上转账怎么可能没有简单的方式？这又不是什么冷僻的需求。"

"Whenever you wanted to move money, it was kind of jumping back in time a century. You'd go to a bank and fill out paperwork, and the forms were in Latin—or so it felt."

"每次你想转账，感觉像穿越回了一个世纪前。你得跑银行、填表，那些表格简直像是用拉丁文写的——至少给人这种感觉。"

"We thought something like Stripe must exist, and we were Googling for it... we eventually concluded there wasn't."

"我们以为类似 Stripe 的服务一定已经存在，于是不停地搜……最后才得出结论：根本没有。"

Berkeley Haas talk

这就是 Stripe 故事中最核心的一段心智过程——他后来在多个访谈中复述并精炼为一个方法论术语："dark energy" 检验：

"We looked for it for months, and its absence was so strange that we initially worried there was some 'dark energy' or regulatory constraint preventing it."

"我们找了好几个月。它的缺席太奇怪了，以至于我们一开始担心是不是有某种'dark energy'或监管约束在阻止它出现。"

"It seemed incredibly strange that a service like Stripe didn't already exist... we spent months investigating if there was some 'dark energy'—regulatory barriers or insurmountable constraints—preventing it."

"像 Stripe 这样的服务居然不存在，这件事极其反常……我们花了几个月去调查是否存在某种 'dark energy'——监管壁垒或者无法逾越的约束——在阻止它。"

Knowledge Project #32, 2018 / Knowledge Project 2022

也就是：当一件"显然应该存在"的事不存在时，先不要假设"机会就在那里等你"，先去排查是否有某种隐藏的结构性力量在阻止它。如果有（比如银行牌照稀缺造成的消费银行业难以创新），就别浪费时间；如果没有，就动手做。

兄弟俩花了几个月调研，结论是：开发者支付这件事不存在没有任何"dark energy"——只是没人做而已。

4.2 Buenos Aires 的第一原型

2009 年底，他们决定去阿根廷布宜诺斯艾利斯一边度假一边写代码：

"We didn't really know where to start and I definitely wouldn't pretend we did... basically we faked it."

"我们其实不知道从哪里开始，我也绝不会假装我们知道……基本上是边做边假装。"

"We decided to go on holiday. Just go hack somewhere... Buenos Aires... We spent about $10 a day and the weather was gorgeous."

"我们决定去度假，就找个地方写写代码……Buenos Aires……一天大概花 10 美元，天气好得不得了。"

"At the end of that we had the first prototype of Stripe... after a month."

"一个月后，我们就有了 Stripe 的第一个原型。"

Startup Grind, 2012

每天 10 美元的生活费，一个月就建起了第一版原型。这就是 Stripe 的诞生——不是在车库或 hackerspace，而是在南半球度假途中。

4.3 命名与早期客户

公司最初注册名是 SlashDevSlashPayments, Inc.（即 /dev/payments，Unix 设备路径的玩词）。产品域名 devpayments.com。这个名字反映了他们最早的自我定位：给开发者用的支付通道，开发者文化味十足。

"Stripe was originally incorporated as SlashDevSlashPayments, Inc."

"Stripe 最初注册的公司名是 SlashDevSlashPayments, Inc."

Tim Ferriss Show #353

2011 年 1 月改名为 Stripe，并切换域名——但他们忘了通知已有客户：

"people woke up in the morning to go to devpayments.com and check their sales... and got redirected to stripe.com."

"人们早上起来想去 devpayments.com 查销售情况……结果被重定向到了 stripe.com。"

第一个公开使用 Stripe 的客户是 Ross Boucher（来自 280 North，YC 公司）：

"We sort of built this really nice API and interface... when you clicked credit account... we just went and called our friend... totally manual."

"我们做了一个挺漂亮的 API 和界面……当你点击 credit account 时……我们其实就是给朋友打个电话……完全是手动的。"

"Ross Bruchet was actually the very first user of Stripe... it would be really awesome if you actually transferred the money into my bank account."

"Ross Bruchet 其实是 Stripe 的第一个用户……如果你们能真的把钱转到我银行账户里就太棒了。"

Startup Grind, 2012

API 看起来很优雅，但后端是手动的——这是经典的 YC "do things that don't scale" 玩法。

4.4 定价决策：故意定高（5% + 30¢）

Stripe 早期价格是 5% + 30 美分——故意定高于行业标准（行业大约 2.9% + 30¢）。Patrick 的解释：

"We charged 5% plus 30 cents for every transaction... we don't want anyone to use us because of the costs."

"我们每笔交易收 5% 加 30 美分……我们不希望有人是因为价格便宜才来用我们。"

Startup Grind, 2012

逻辑是 Peter Thiel 式的：高价能筛选客户——只有真的因为开发体验差异化而非价格而选 Stripe 的客户才会来；同时高价倒逼你必须做出真正差异化的产品。后来他们把价格降到了行业标准 2.9% + 30¢，但这段定价 DNA 留在了 Stripe 的产品哲学中。

4.5 关键招聘：Greg Brockman

2010 年 Patrick 通过 MIT 同学网络招到了 Greg Brockman——刚从 MIT 辍学的化学奥赛银牌得主。Brockman 后来成为 Stripe 第一任 CTO（2013），五年内带 Stripe 从 5 人扩到 205 人。他于 2015 年 5 月离开 Stripe，半年后联合创立了 OpenAI。Patrick 失去他的时间点，正好是深度学习起飞的前夜——这是 Stripe 历史上一个 "if only" 的时刻。

4.6 技术栈选择：Ruby + MongoDB（故意主流化）

Patrick 在他之前的公司（Auctomatic 时期）用过 Smalltalk，并自己写过一个对象数据库。他对 SQL 不感冒，认为 SQL 在应用领域和存储原语之间有"翻译失配"（translational mismatch）。但开始 Stripe 时他做了一个明确的"reduce 异端度"的决定：

"I wrote a datastore for our prior company, an object-based datastore, and I didn't really like SQL. I thought there was too much of a translational kind of mismatch between the domain of the application and that which SQL natively makes expressible."

"我之前那家公司是我自己写的一个基于对象的数据存储。我不太喜欢 SQL，我觉得应用领域和 SQL 原生能表达的东西之间存在太多翻译性的失配。"

"With Stripe, we wanted to be more mainstream and a little bit less heterodox in our technology choices than our prior company. And so instead of using Smalltalk, okay, we weren't going to go to Java, but we went to Ruby... And similarly, rather than write our own object database, we went relatively more mainstream and used Mongo, which still gave a lot of flexibility, you know, by virtue of being a kind of a kind of object data store."

"做 Stripe 的时候，我们希望在技术选型上比上一家公司更主流一点、少一点异端。所以不再用 Smalltalk，当然我们也不会去用 Java，而是选了 Ruby……同样地，与其自己写一个对象数据库，我们走相对主流的路线，用了 Mongo——它本身也是一种对象数据存储，仍然给了我们很大的灵活性。"

AI + a16z, 2026

15 年后他公开评价这个决定：Ruby 和 Mongo 至今仍是 Stripe 的基础技术。他们后来花了大量精力让 Mongo 支持容错、分布式、持久化——结果是他们的核心 charge flow API 在去年达到 99.99986% 可用性（全年仅 44 秒不可用）。

4.7 公开发布与种子轮

2011 年 5 月，Stripe 完成 200 万美元种子轮，投资人是 Elon Musk、Peter Thiel、Sequoia Capital、Andreessen Horowitz、SV Angel——任何一个都堪称传奇。Patrick 后来在 EconTalk 强调：

"If we had relied on a committee to reach a consensus on our worthiness, Stripe would not exist."

"如果我们当年需要靠一个委员会就'我们值不值得'达成一致，Stripe 根本不会存在。"

EconTalk, 2019

"Most VCs were unenthusiastic about 'two teenagers entering the financial industry'; Stripe's existence depended on a few non-consensus bets (Graham, Thiel)."

"大多数 VC 对'两个青少年进入金融业'这件事并不热情；Stripe 的存在仰赖几笔反共识的下注（Graham、Thiel）。"

2011 年 9 月公开发布。

"It looked like a bad idea because it was a crowded market with significant regulatory barriers, we were young, and we had no obvious distribution mechanism. However, we were driven by the fact that something like Stripe should have existed."

"它看起来像是个糟糕的想法——市场拥挤、监管壁垒明显、我们年轻、没有明显的分发渠道。但驱动我们的事实是：像 Stripe 这样的东西*本来就应该*存在。"

Knowledge Project #32

第五章：扩张期（2012–2017）

5.1 主要产品线

• Stripe Connect (2012)：支持多方支付，让平台型业务能给数千个第三方收款
• Atlas (2016-02-24) at Mobile World Congress Barcelona
• Issuing, Radar (2018)
• Terminal (2019)

Patrick 自述 2016 年时 Stripe 已经为约 27% 美国人服务的商家做支付——巨大的覆盖。

5.2 关于 Atlas

Atlas 是把"在 Delaware 注册公司"这件复杂的事做成一个 $500 的产品：包含 C-corp 或 LLC 注册、EIN、创始人股权、83(b) 选举、$2,500 Stripe credits、$50K+ 合作伙伴折扣、Cooley LLP 法律文件。截至现在，Atlas 已经帮助来自 140+ 国家的 100,000+ 创业者创立了公司。

"The promise of the internet is that geography should be largely irrelevant. But that's not yet true: the majority of the world's population lives in a country where they don't have access to high-quality banking or payments infrastructure."

"互联网的承诺是地理位置应该基本无关紧要。但这件事还没成真：世界上大多数人生活的国家并没有高质量的银行或支付基础设施。"

"make simple incorporation as a service, so that any entrepreneur worldwide could have a corporate entity and a bank account spun up about as easily as they could get an EC2 server"

"把简单的公司注册做成一项服务，让世界上任何一位创业者都能像开一台 EC2 服务器那样轻松地拿到一个公司主体和一个银行账户。"

Patrick, MWC 2016 announcement

Atlas 内部数据：

"60% of founders who use Atlas say they wouldn't have started their company otherwise."

"60% 用过 Atlas 的创始人表示，如果没有 Atlas，他们就不会创业。"

Knowledge Project #32

Atlas 是 Stripe 第一次把"limit on Stripe's growth = number of successful startups"（Stripe 的天花板等于成功创业公司数量）这个内部洞察变成产品——后来这条逻辑被用在 Stripe Press、Climate、Frontier、Arc Institute。

5.3 早期战略思考：Stanford GSB 2016 访谈

2016 年他在 Stanford GSB "View From The Top" 节目里提出了几个标志性的早期 Stripe 思想：

"Silicon Valley is well known as a cradle of entrepreneurship, but it is just as successful in being a grave. Not only did these companies fail, but you haven't even heard of them."

"Silicon Valley 因为是创业的摇篮而闻名，但它做坟墓也同样成功。那些公司不仅失败了，而且你连听都没听说过。"

"It is possible for business to be positive sum; the existence of winners does not necessarily mean there will be losers."

"商业可以是正和的——有赢家并不必然意味着会有输家。"

"Stripe is a bet on human laziness, in the sense that people prefer to do things from their phone rather than going to physical places in person. Going to a car dealership is the platonic ideal of things we don't want to do."

"Stripe 是对人类懒惰的一次下注——人们更愿意在手机上做事，而不是亲自跑到某个实体地点。去汽车经销店简直是'我们不想做的事'的柏拉图式范本。"

"In order to endure, large companies have to be irrationally risk-averse. That is not a statement about coordination costs; it is a basic statement about survival."

"要长期存活，大公司必须非理性地厌恶风险。这不是在说协调成本，这是关于生存最基本的陈述。"

"We focused on lowering the 'activation energy' required for a developer to start accepting payments."

"我们聚焦于降低一个开发者开始接受付款所需的'activation energy'。"

Stanford GSB Industrialist's Dilemma, 2016

"Activation energy" 在化学里指的是触发反应需要的最小能量阈值。Patrick 把它移植到产品哲学：每个产品决策应该问"这把开始所需的能量降低了多少？"——这是 Stripe 整个发展史的统一线索：从把"几周开通银行账户"变成"五分钟集成 API"，到 Atlas 把"几个月办手续注册公司"变成"几小时网页流程"，到 Frontier 把"分散的小气候买家"聚合成"提前承诺的大需求"。

5.4 2016 年最年轻自手起家亿万富翁

2016 年 11 月，Stripe 估值 92 亿美元。Patrick 28 岁、John 26 岁，分别拥有约 11 亿美元身家——成为世界最年轻的自手起家亿万富翁。同年 Obama 任命 Patrick 为"Presidential Ambassador for Global Entrepreneurship"。

5.5 Greg Brockman 的离开（2015）

Brockman 2015 年 5 月离开 Stripe，5 月底参与了 Sam Altman 等人在 Sand Hill Road 的一系列会议，12 月联合创立 OpenAI。Patrick 失去了第一个 CTO，但保留了一段长期友谊——多年后他在自己的 Cheeky Pint podcast 上邀请 Brockman 回来谈 AI 基础设施。

第六章：思想家时期（2018–2021）

进入 2018 年，Stripe 的运营机器已经成熟，Patrick 开始把越来越多的时间投入到"非 Stripe"的思想性活动中。这并不是分心——按他的内部逻辑，Stripe 的天花板等于全球创业活力，而创业活力又取决于科学和文化的进步——所以这些"外围"活动其实都是 Stripe 的远景投资。

6.1 Stripe Press（2018）

"Stripe Press highlights ideas that can be broadly useful. Their publications include entirely new material, collections of existing work reimagined, and republications of previous works with renewed relevance."

"Stripe Press 聚焦那些有广泛用途的思想。他们出版的内容包括全新的作品、对既有作品的重新汇编，以及对在今天重新具有意义的旧作的重版。"

press.stripe.com

出版书目（部分）：

• The Dream Machine — M. Mitchell Waldrop（Patrick 多次推荐的"早期计算/互联网最佳概览"）
• The Art of Doing Science and Engineering — Richard W. Hamming
• Stubborn Attachments — Tyler Cowen（论证经济增长是道德义务的论文级著作）
• Poor Charlie's Almanack — Charlie Munger
• Pieces of the Action — Vannevar Bush
• Where Is My Flying Car? — J. Storrs Hall
• Scientific Freedom — Donald W. Braben
• Working in Public — Nadia Eghbal
• Scaling People — Claire Hughes Johnson
• The Scaling Era: An Oral History of AI, 2019-2025 — Dwarkesh Patel
• Boom: Bubbles and the End of Stagnation — Byrne Hobart and Tobias Huber
• Maintenance: Of Everything — Stewart Brand

逻辑很清楚：把人们通常不愿意出版（旧书太冷门、新书太冒险）但 Patrick 自己长期获益的书，重新做成精装。

6.2 Tim Ferriss Show #353（2018 年 12 月）

这期 2.5 小时的访谈是 Patrick 公开自传性质最强的一次。他第一次详细谈到童年、父母、古希腊语、阅读哲学、决策框架。几个核心观念：

关于阅读的偏见：

"I really think people should be kind of much more biased towards older books than they are. I think The Dream Machine is a good example of this."

"我真的觉得人们读书时应该更多地偏向旧书一些。我认为 The Dream Machine 就是个很好的例子。"

老书已经经过时间筛选，新书没有。这条原则贯穿了 Stripe Press 的选品逻辑。

关于决策：

他不问"How should I choose between option A and B?"，而是问 "How do I make sure that both options A and B are as good as possible, and there's also a C, D, and E, and that those options are great too?"

"you don't necessarily need to be that good at decision making if you get really good at remaking the decision when and as necessary... Instead, it's all about the constant feedback and course and error correction."

"如果你真的擅长在必要时重做决策，那你就不一定要在做决策本身上有多厉害……重要的是持续的反馈、纠偏和误差校正。"

Tim Ferriss Show #353

关于"Virtual Board of Directors"（虚拟董事会）：

"the idea that people can improve their own decision making. By simply choosing other people... imagining what they would say... and then averaging over the responses. But that's like a really amazing fact because of course, this is all occurring in your own head."

"这个想法是：人可以通过选定一些他人……想象他们会怎么说……然后对这些回答取平均，来改善自己的决策。这其实是个相当惊人的事实，因为这一切都发生在你自己的脑子里。"

他的 virtual board 包括已故的人，比如 Richard Feynman——遇到难题时在头脑里模拟"如果 Feynman 在这里，他会说什么？"

关于"wandering vs lost"：

"So not all, not all who wander are lost, right? So there's a difference between being lost, which is you have a destination and you have gone off track or you are preoccupied with where you are currently located and don't know where you're located versus exploring and wandering."

"并非所有漫游的人都在迷路。'迷路'指的是你有一个目的地但偏离了轨道，或是你被困在当前位置、不知道自己在哪儿；这与探索和漫游不同。"

为人生中的探索期辩护——你可以没有明确目的地但也不算迷路。

关于创业的心理真相：

"But the weird part is that even if things are going well and the effort or the company or whatever is succeeding, things will still often not feel great. And no one sort of told me that before I started... the actual reality is that you're always necessarily operating at the kind of outer edge of what you can handle because if it was, if you'd spare capacity, you just take on more."

"但奇怪的地方在于，就算事情进展顺利、努力或公司什么的在成功，感觉往往还是不太好。开始之前没人这样告诉过我……真正的现实是，你总是必然在自己所能承受的最外缘运转——因为一旦你有富余的余力，你就会再揽下更多。"

Tim Ferriss Show #353

"you have to be very optimistic... You also have to be very pessimistic... you exist in this superposition, this juxtaposition of pessimism and optimism."

"你必须非常乐观……也必须非常悲观……你处在这种叠加态、这种悲观与乐观的并置中。"

6.3 "We Need a New Science of Progress" — The Atlantic（2019 年 7 月 30 日）

与经济学家 Tyler Cowen 联合发表在 The Atlantic 上的这篇文章，是 Patrick 整个非 Stripe 工作的总纲领。

"Progress is amazing, influenceable, and understudied. @tylercowen and I decided to make the case for Progress Studies."

"进步是惊人的、可以被影响的、却被研究得太少。@tylercowen 和我决定为 Progress Studies 立论。"

Patrick's launching tweet

"Progress itself is understudied. By 'progress,' we mean the combination of economic, technological, scientific, cultural, and organizational advancement that has transformed our lives and raised standards of living over the past couple of centuries."

"进步本身被研究得太少。我们所说的'进步'，是指过去两个世纪里改变了我们生活并提升了生活水准的那种经济、技术、科学、文化和组织上进展的合力。"

"Plenty of existing scholarship touches on these topics, but it takes place in a highly fragmented fashion and fails to directly confront some of the most important practical questions."

"已有大量学术研究涉及这些主题，但都极为碎片化，未能直接面对一些最重要的实践问题。"

副标题："Humanity needs to get better at knowing how to get better."

这篇文章催生了一整个 "Progress Studies" 圈：Jason Crawford 的 Roots of Progress、Eli Dourado、Caleb Watney 和 Alec Stapp 的 Institute for Progress、Works in Progress 杂志（后来被 Stripe Press 收购）。

6.4 EconTalk（2019）— 进步在减速吗？

EconTalk 与 Russ Roberts 这一期里，他系统铺开了"科学减速论"的论证，关键点：

四类证据三角验证：

1. 宏观 TFP（全要素生产率）增长放缓
2. 领域专业指标（药物发现、半导体）
3. 结构指标（诺奖得主第一项重大工作年龄、团队规模）
4. 同行专家两两比较（科学家自己评估某发现的重要性）

关键效应量：

"Grant recipients who were funded differently were 98% more likely to produce work in the top 1% by citations. The fact that you can get such a large effect size just by changing one variable suggests it is possible to create other high-impact shifts."

"采用不同资助方式的受资助者，产出引用排名前 1% 工作的概率高出 98%。仅靠改变一个变量就能得到如此大的效应量，说明我们有可能创造出其他高影响的改变。"

98% 的提升来自仅仅一个变量改变（HHMI 给科学家个人长期资助 vs NIH 按项目资助）——对他来说这是"系统离最优非常远"的烟雾弹。

经济增长的道德分量：

"If you look at GDP per capita around the world and plot that against self-reported well-being, the correlation is 0.68. It is rare in human psychology and sociology to get a correlation that strong."

"如果你把世界各地的人均 GDP 与自报的幸福感画在一起，相关系数是 0.68。在人类心理学和社会学里，能拿到这么强的相关性是很罕见的。"

关于天赋分布：

"Talent is approximately evenly distributed, but opportunity is much more uneven. I conclude from my own upbringing that there must be many more people like me, but for whom the pachinko machine of outcomes did not fall the right way."

"天赋大致是均匀分布的，但机会的分布要不均匀得多。从我自己的成长经历看，我得出的结论是：一定还有许多和我类似的人，只是命运的弹珠机没有刚好掉到对的位置。"

"The rate of progress is declining rapidly on a per-person, per-hour, or per-dollar basis. Perhaps because of the enormous increase in our investment and efforts, we are sustaining aggregate progress, but our productivity per person is much lower."

"按人均、人时或人均美元计，进步的速率在迅速下降。或许是因为投入和努力大幅增加，我们才维持了总量上的进步；但每个人的生产率要低得多。"

"We have 50 times more researchers in these fields, but as judged by scientists themselves, we are producing significant discoveries at a nearly constant rate. This suggests that an individual researcher is becoming 50 times less likely to produce significant work."

"这些领域的研究者数量是过去的 50 倍，但按科学家自己的判断，重大发现的产出速度几乎保持不变。这意味着单个研究者做出重要工作的概率正在变成原来的 1/50。"

具体行动方案：

"carve out $100M-$1B/year and let proven high-impact scientists each unilaterally direct $1M to projects they personally believe in — bypass committee gatekeeping."

"每年划出 1 亿到 10 亿美元，让那些已被证明有高影响力的科学家各自单方面把 100 万美元投向他们个人相信的项目——绕开委员会的把关。"

6.5 与 Mark Zuckerberg 的三方对谈（2019 年 11 月 22 日）

Atlantic 文章发表四个月后，Mark Zuckerberg 邀请 Patrick 和 Tyler Cowen 进行一次三方对谈。议题涵盖：进步研究的重要性、生物医学研究、新建大学和基金会、快速建造、住房和医疗负担能力、Chan Zuckerberg Initiative 的未来。这次对谈直接预示了两年后的 Arc Institute 成立。

6.6 Fast Grants（2020 年 4 月）

COVID-19 大流行爆发，传统科研资助流程慢得令人窒息。Patrick 与 Tyler Cowen 紧急联手成立 Fast Grants：审批从天计、每位科学家最高 50 万美元，专门资助 COVID 相关研究。最终发放了约 5,000 万美元给 260 个项目。

"Fast Grants was not this giant, impressive edifice, but it doesn't have to be giant, right, to have that kind of bigger impact?"

"Fast Grants 并不是某座庞大、壮观的建筑，但要产生那种更大的影响，本来就不必很庞大，对吧？"

Dwarkesh, 2024

"the most important thing, even in addition to these specific countermeasures trying to come up in advance, is when the thing is happening, having competent individuals who can synthesize and organize information, and also having these new initiatives and institutions to get the right thing done."

"除了那些试图提前准备好的具体应对手段之外，最重要的事是：当事情真正发生时，要有能够综合和组织信息的能干的人，并且要有这些新发起的项目和机构来把对的事做成。"

Fast Grants 的运营经验直接转化成了 Arc Institute 的设计——把"给科学家，不是给项目"的资助哲学制度化。

6.7 Paystack 收购（2020 年 10 月）

约 2 亿美元收购尼日利亚支付公司 Paystack——Stripe 在非洲扩张的关键一步。

6.8 California YIMBY（2018）

Stripe 向 California YIMBY（Yes In My Backyard 住房改革倡议组织）捐赠 100 万美元。这是 Patrick 把他的"湾区房价是对未来潜力的窒息"观点付诸行动：

"It is dispiriting because it feels like a suffocation of future potential. Silicon Valley is the greatest concentration of wealth creation in history, but its rise was enabled by cheap mobility and expansion — factors that no longer exist due to collective political decisions."

"这件事令人沮丧，因为它给人一种未来潜力被窒息的感觉。Silicon Valley 是历史上最集中的财富创造地，但它的崛起是建立在廉价的人口流动和扩张之上的——这些条件如今因为集体性的政治决定而不复存在。"

Knowledge Project #32

6.9 Arc Institute（2021 年 12 月 15 日）

Arc Institute 是 Patrick 与未婚妻 Silvana Konermann（斯坦福生化教授）和 Patrick Hsu（UC Berkeley 生物工程教授，CRISPR 先驱 Feng Zhang 的学生）共同创立的私营生物医学研究机构。

• 初始资金：6.5 亿美元
• 资助模式：科学家本人获得 8 年期可续约的资助，funded for whatever they want to pursue（自由探索）
• 共同捐助者：Vitalik Buterin、Patrick & John Collison、Dustin Moskovitz、Ron Conway、Crankstart、Jane Street 高管
• 研究方向：机器学习、基因组工程、细胞和动物疾病模型、多组学、bridge RNA、生物大模型（Evo 模型）

Patrick 在 Dwarkesh 访谈中清晰阐述了 Arc 与 NIH 的差异：

"what's different about Arc is one, scientists are funded themselves to pursue whatever they want. So it's curiosity-driven research, whereas NIH grants are given for projects. Second, we build a lot of in-house infrastructure so that scientists can draw upon other platforms and other capabilities that they don't have to go and build and maintain themselves."

"Arc 的不同之处有两点。第一，科学家本人获得资助，可以追求任何他们想做的方向，所以是好奇心驱动的研究，而 NIH 的资助是给项目的。第二，我们在内部建立了大量基础设施，让科学家可以使用各种平台和能力，不必自己去搭建和维护。"

Dwarkesh Podcast, 2024

注意时间线的隐藏深意：他与 Silvana 在 2004 年欧洲青年科学家大赛上相识，先一起创立了 Arc Institute（2021 年 12 月），几个月后才结婚（2022 年 4 月）——他们的合作关系先于婚姻关系。

6.10 Forbes "Stab City" 反驳（2021 年 4 月）

2021 年 4 月，Forbes 撰稿人 Stephen McBride 写了一篇题为《How Two Brothers Escaped 'Stab City' And Made $11 Billion》的文章，把利默里克描述为"刺城"、"欧洲谋杀之都"、"战区"。Patrick 公开反击：

"Not only mistaken about Limerick but the idea of 'overcoming' anything is crazy. We are who we are because we grew up where we did."

"不仅是对利默里克的误解，'克服'什么这种说法本身就很离谱。我们之所以是现在的我们，正是因为我们在那里长大。"

John 称那篇文章 "daft"。利默里克市长 Michael Collins 称之为"slurs"。#LimerickAndProud 在 Twitter 趋势 8.6 百万次曝光。Forbes 最终承认文章 "did not meet editorial standards"，删除了文章。

这段插曲是 Patrick 罕见的公开身份表态：爱尔兰性不是被克服的劣势，而是构成他的东西。他至今坚持使用 patrick@collison.ie 的邮箱地址。

第七章：婚姻、气候与 AI（2022–2024）

7.1 与 Silvana Konermann 结婚（2022 年 4 月）

7.2 Frontier — $925M Carbon Removal AMC（2022 年 4 月 12 日）

"We're excited to work with this coalition to help accelerate some of the technology development needed to mitigate climate change. Frontier will be one of the largest AMCs ever and ~20x what's been spent to date on permanent carbon removal."

"我们很期待与这个联盟合作，加速缓解气候变化所需的部分技术研发。Frontier 将是有史以来规模最大的 AMC 之一，约相当于此前在永久碳清除上所有支出的 20 倍。"

Patrick Collison

参与方：Stripe、Alphabet、Shopify、Meta、McKinsey，加上数千个使用 Stripe Climate 的企业用户。Frontier 用 AMC（Advance Market Commitment，预先承诺市场） 机制——这种机制最早被诺奖经济学家 Michael Kremer 用于资助为低收入国家研发疫苗。Frontier 的逻辑是：当一个新技术（永久碳清除）的市场没有可信的需求信号时，提前承诺购买可以激发供给端的爆发。

承诺期：9 年，9.25 亿美元，约相当于此前全球所有永久碳清除支出的 30 倍。

这就是 Patrick 在 Dwarkesh 访谈里总结的 "theory of Stripe"：

"It seems too self-aggrandizing to call it the theory of Stripe, but some vague hunch in Stripe is that that aggregation of demand can have important expansionary effects with respect to the ensuing supply. And yes, Stripe Climate is some version of this hypothesis applied on a much smaller scale than Stripe itself, but still real and well, we'll see, maybe important."

"把它叫作'Stripe 的理论'听起来太自我膨胀了，但 Stripe 内部有一个模糊的直觉——需求的聚合可以对随之而来的供给产生重要的扩张性效应。是的，Stripe Climate 就是把这个假设应用在比 Stripe 自身小得多的尺度上的某种版本，但它是真实的，至于是不是重要，我们走着瞧。"

7.3 Works in Progress 收购（2022）

Works in Progress 是英国进步研究圈的旗舰杂志，被 Stripe Press 整体收购，纳入"思想生态"的一部分。

7.4 2024 年度报告

Stripe 2024 年度报告（Patrick 和 John 联名发布）几个关键数据：

• 2024 TPV：1.4 万亿美元，38% 同比增长，约相当于全球 GDP 的 1.3%
• Stripe Billing：30 万家公司，2 亿活跃订阅
• Revenue and Finance Automation Suite：5 亿美元年化收入

"We're continually retraining dozens of machine learning models that optimize every part of the transaction flow over an economy-scale dataset. The resulting optimizations are big enough that businesses see them in their topline revenue figures. Businesses simply start making more money when they switch to Stripe."

"我们在持续地重新训练几十个机器学习模型，在一个经济规模的数据集上优化交易流程的每一个环节。最终带来的优化足够大，以至于商家在自己的营收数字里就能看到。商家切换到 Stripe 后，单纯就开始多赚钱了。"

"In each of the last six years, Stripe has reinvested a much higher proportion of our earnings in R&D than any comparable company."

"过去六年的每一年里，Stripe 把利润再投入研发的比例都远高于任何可比公司。"

"Stripe is profitable, and expects to remain so in the coming years."

"Stripe 是盈利的，并且预计在未来几年仍将保持盈利。"

Stripe 2024 Annual Letter

第八章：加密、AI 与新阶段（2024–2026）

8.1 Bridge 收购：Stripe 历史最大笔（2024 年 10 月 21 日宣布，2025 年 2 月 4 日完成）

11 亿美元收购 Bridge——一个由前 Coinbase 高管 Zach Abrams 和 Sean Yu 在 2022 年创立的稳定币 API 公司。

Patrick 当晚的推文成为标志性表述：

"Stablecoins are room-temperature superconductors for financial services. Thanks to stablecoins, businesses around the world will benefit from significant speed, coverage, and cost improvements in the coming years. Stripe is going to build the world's best stablecoin [infrastructure]."

"stablecoin 之于金融服务，就像 room-temperature superconductors。得益于 stablecoin，世界各地的企业将在未来几年里在速度、覆盖和成本上获得显著改善。Stripe 将打造世界上最好的 stablecoin 基础设施。"

这是 Stripe 历史上最大的单笔收购，也是加密行业历史上最大的并购。它完全反转了 Stripe 之前对加密的谨慎立场——时机选在 2024 年美国大选后预期监管放松。

8.2 国会作证（2025 年 3 月 11 日）

收购完成一个月后，Patrick 在众议院金融服务委员会作证，主题为《Navigating the Digital Payments Ecosystem》。他提出了稳定币监管的五原则：

1. 监管清晰
2. 灵活性与对创新的支持
3. 中立性、与既有系统的互操作性
4. 强消费者保护，重点在透明度
5. 以结果为导向的反洗钱与系统完整性框架

"Stablecoins represent a fundamental innovation in how money moves."

"stablecoin 代表了在'钱如何流动'这件事上的一个根本性创新。"

"Putting in place a shared framework with prudent rules will do a tremendous amount of good."

"建立一个有审慎规则的共同框架，会带来巨大的益处。"

立场要点：支持稳定币 + 反对美国央行数字货币（CBDC）。

8.3 Privy 与 Metronome 收购（2025）

• Privy：加密钱包基础设施
• Metronome：用量计费

8.4 加入 Meta 董事会（2025 年 4 月）

Mark Zuckerberg 邀请 Patrick 加入 Meta 董事会——同时还有 Nat Friedman、Tobi Lütke、Charlie Songhurst 加入"Meta Advisory Group"。

8.5 2026 年估值 1590 亿美元

2026 年 2 月，Stripe 估值升至 1590 亿美元（同比 +70%）。员工约 8,500 人。

8.6 2026 年 AI + a16z 访谈：v2 API 与 Conway's Law

最近这次访谈（2026 年初）是他对 Stripe 15 年的回望与对未来的展望。关键观点：

Conway's Law 反思：

"If I were to do everything at Stripe again, I would have spent even more time on APIs and data models. This is because of Conway's law effect of how both of those things end up shaping the organization. The strong version is that it substantially shapes your strategy and just your business outcomes. The right API design, the right abstractions, and the right data models can really endure. Stripe is now 15 years old, and there were lots of things we designed 15 years ago that are still in use today, which is kind of good and bad in the sense that they endured."

"如果我重新做一遍 Stripe 的所有事，我会花更多时间在 API 和数据模型上。这是因为 Conway's Law 的效应——这两样东西最终会塑造你的组织。强一点的版本是：它们会实质性地塑造你的战略和你的业务结果。正确的 API 设计、正确的抽象、正确的数据模型可以真正地持久。Stripe 现在已经 15 岁了，我们 15 年前设计的很多东西今天仍然在用——它们持久了，这件事既算好事也算坏事。"

v2 设计原则：

"The answer to that is to unify everything you can plausibly unify. Another lesson, maybe a bit tongue-in-cheek, would be to make anything that plausibly could be an N by M relationship to support that. If you only support 1-to-N or N-to-1 or whatever, and even if it's non-obvious how it could possibly be N-to-M, just inevitably, you'll end up needing that and you'll think, well, you could never have a company that's owned by two different companies or something, but it turns out that sort of every permutation in the space is, in fact, eventually explored."

"答案是：尽可能把一切你能合理统一的东西统一起来。另一条教训，多少带点玩笑——凡是有可能形成 N 对 M 关系的东西，就都把它做成支持 N 对 M。如果你只支持 1 对 N 或 N 对 1，即使现在看起来不明白怎么会变成 N 对 M，最终你一定会需要它。你会想'怎么可能有一家公司被两家不同的公司持有'之类的，结果这个空间里几乎每一种排列最终都会被探索到。"

对开发环境的批判：

"I think the basic idea of a development environment and not just a text editor is really the right idea. That's the thing that the Lisp machines had and Genera, that's the thing that to some extent Mathematica has, and that's the thing that Smalltalk has. I think it's just such a mistake that we have ended up with development environments where there is such a separation between the runtime and the text editing and the environment in which the code runs."

"我觉得'开发环境而不仅仅是文本编辑器'这个基本想法是非常正确的。Lisp Machines 和 Genera 有这种东西，Mathematica 在一定程度上有，Smalltalk 也有。我们今天竟然演变成把运行时、文本编辑、代码执行的环境分得这么开的开发环境，这真是个大错误。"

Stripe 的核心价值"craft and beauty"：

"we really care about what we call at Stripe craft and beauty, and we want our software to be well designed and pleasant to use, and pleasant to use not only in the superficial pixel sense, but also in the deep, it works very well sense, and is something you can set up and largely forget about and just trust"

"我们在 Stripe 内部非常在意我们所说的 craft and beauty——希望我们的软件被精心设计、用起来令人愉悦；这种愉悦不只是表层的像素美观，而是深层意义上'它就是工作得很好'，是那种你设置好就基本可以抛诸脑后、信任它的东西。"

对 AI 的复杂态度：

"AI could conceivably do a lot to help in the beautification and the refactoring of code bases. You can imagine producing all this ungainly, not quite correctly factored detritus at the front, and then nocturnally, this thing comes up behind you and makes it all beautifully factored."

"可以想象，AI 在代码库的美化和重构上能帮上很大的忙。你可以想象前期产出一堆笨拙、组织得不太对的杂物，然后夜里这个东西从你身后冒出来，把一切重新整理得漂漂亮亮的。"

"There's obviously a concern with AI that it leads to the creation of more slop and more kind of crappy things, but not more of the best things."

"对于 AI 显然有一个担忧——它会带来更多 slop、更多质量很差的东西，却不会带来更多最优秀的东西。"

Arc Institute 的最新进展：

"At ARQ, we're working on training foundation models for biology using DNA and things like that. We're working on a virtual cell and generally we're trying to better understand complex diseases. We've never cured a complex disease."

"在 Arc，我们正在用 DNA 之类的东西训练生物学的 foundation models，正在做一个 virtual cell，总体上是在尝试更好地理解复杂疾病。我们从未治愈过一种复杂疾病。"

"Over the last 10-ish years, we've gotten three new classes of technology in biology: much better sequencing technology, single cell sequencing, neural networks and deep learning and transformers, and huge improvements in functional genomics and CRISPR and bridge editing... If you put those together, you now have the ability to, again, at the kind of level of the individual cell, to read, think, and to write. And this starts to really feel like a new kind of Turing loop and to have its own sort of completeness."

"过去十来年里，我们在生物学里得到了三类新技术：好得多的测序技术，单细胞测序；神经网络、深度学习和 transformers；以及功能基因组学、CRISPR 和 bridge editing 上的巨大改进……把这些放在一起，你现在就有能力——同样是在单个细胞的层面——读、想、写。这开始真的像一种新的 Turing loop，自身具有某种完备性。"

对宏观经济的清醒判断：

"GDP growth in the U.S. looks well, over the last two years, it's been somewhat better than we expected. We certainly don't see any evidence for exponential takeoff. GDP growth outside of the US has not been that encouraging. We're not living in some massively accelerated period of economic growth for the world writ large."

"美国的 GDP 增长，过去两年其实比我们预期的略好一些。我们当然没有看到任何指数级起飞的证据。美国之外的 GDP 增长就没那么乐观。整体而言，我们并不是生活在某个全球经济大幅加速的时期。"

二、深度洞察集

按主题归档的 Patrick 关键观点。所有引用保留英文原文以避免翻译失真。

1. 关于创业 / Entrepreneurship

1.1 Dark Energy 检验

"We looked for it for months, and its absence was so strange that we initially worried there was some 'dark energy' or regulatory constraint preventing it."

"我们找了好几个月。它的缺席太奇怪了，以至于我们一开始担心是不是有某种'dark energy'或监管约束在阻止它出现。"

Knowledge Project #32

→ 当一件"显然应该存在"的事不存在时，先排查是否有隐藏的结构性力量阻止它，再决定是否值得投入。

1.2 创业的默认结局是不存在

"Doing anything of significance is hard. For a startup, the default outcome is near-term non-existence; to survive over the medium or long term is an unnatural act."

"做任何有分量的事都很难。对一家创业公司而言，默认结局就是短期内不复存在；中长期能活下来本身就是一件违背自然的事。"

"For a startup, the default outcome is near-term non-existence. To survive over the medium or long term is an unnatural act."

"对一家创业公司而言，默认结局就是短期内不复存在。中长期还能活着是一件违背自然的事。"

Knowledge Project #32, Knowledge Project 2022

1.3 缩放是 Tower Defense 游戏

"A startup feels like a tower defense game where you are constantly building defensive mechanisms against critters trying to break in."

"创业感觉像一场 tower defense 游戏：你在不停地建防御机制，去抵挡那些想冲进来的小怪。"

"Scaling a business is both straightforward and extremely hard. You do not control the rate of problem appearance; you only control the mechanisms you build to solve them."

"把一家公司做大，既直白又极其难。你无法控制问题出现的速率，你能控制的只有自己用来解决问题的那些机制。"

Knowledge Project, 2018 / 2022

1.4 同时持有极端乐观与极端悲观

"you have to be very optimistic... You also have to be very pessimistic... you exist in this superposition, this juxtaposition of pessimism and optimism."

"你必须非常乐观……也必须非常悲观……你处在这种叠加态、这种悲观与乐观的并置中。"

"Successfully building a company requires maintaining a simultaneous state of extreme optimism and pessimism, constantly anticipating problems while believing in long-term success."

"要成功地建一家公司，需要同时保持极端乐观和极端悲观——不停地预想问题，同时又相信长期会成功。"

"the actual reality is that you're always necessarily operating at the kind of outer edge of what you can handle because if it was, if you'd spare capacity, you just take on more."

"真正的现实是，你必然总是在自己所能承受的最外缘运转——因为只要还有富余的余力，你就会再揽下更多。"

Tim Ferriss Show #353

1.5 最深的成功之痛

"But the weird part is that even if things are going well and the effort or the company or whatever is succeeding, things will still often not feel great. And no one sort of told me that before I started."

"但奇怪的地方是，就算事情进展顺利、努力或公司什么的在成功，感觉往往还是不太好。开始之前没人这样告诉过我。"

Tim Ferriss Show #353

1.6 早期公司不靠委员会

"If we had relied on a committee to reach a consensus on our worthiness, Stripe would not exist."

"如果我们当年需要靠一个委员会就'我们值不值得'达成一致，Stripe 根本不会存在。"

EconTalk

"It is vividly clear how fragile, delicate, and implausible successful companies are at the larval stage. We see the potency of mechanisms to shepherd them, provide them with role models, and offer them cultural capital."

"你能非常清楚地看到，那些后来成功的公司在'幼虫'阶段是多么脆弱、精致、不像样。我们也看到那些托住它们、为它们提供 role models 和文化资本的机制有多重要。"

1.7 Silicon Valley 既是摇篮也是坟墓

"Silicon Valley is well known as a cradle of entrepreneurship, but it is just as successful in being a grave. Not only did these companies fail, but you haven't even heard of them."

"Silicon Valley 因为是创业的摇篮而闻名，但它做坟墓也同样成功。那些公司不仅失败了，你连听都没听说过。"

Stanford GSB, 2016

1.8 组织漂移现象

"I think most organizations, when they start out, are actually trying to achieve their stated goals. But then over time, that person and that set of people who initially populated the organization Depart, and some set of new people come to take their place... And so I think this phenomenon is kind of a fact of life."

"我觉得大多数组织在创立之初，其实是在认真追求它们写明的目标。但随着时间推移，最初组成这个组织的那个人、那群人离开了，新的一批人来取代他们……所以我觉得这种现象差不多算是一种生活事实。"

Dwarkesh Podcast, 2024

1.9 真正的元主题：把目标当真

"I feel like maybe a kind of a theme through everything we've talked about is actually taking the goal seriously."

"我感觉，贯穿我们今天聊的所有内容的一条主线，可能就是真正地把目标当真。"

Dwarkesh Podcast, 2024

2. 关于产品与技术 / Product & Technology

2.1 Activation Energy（活化能）作为统一原则

"We focused on lowering the 'activation energy' required for a developer to start accepting payments."

"我们聚焦于降低一个开发者开始接受付款所需的'activation energy'。"

Stanford GSB, 2016

2.2 Stripe 不是替换金融基础设施，而是嵌入它

"we would much prefer that Stripe plugged into every existing system and rail and domestic infrastructure. And this has been Stripe's strategy very deliberately from the beginning...by enhancing the capabilities of an existing ecosystem, you create quite a bit more value."

"我们更希望 Stripe 嵌入每一个既有的系统、支付通道和本地基础设施里。这从一开始就是 Stripe 非常刻意的战略……通过增强一个既有生态的能力，你能创造出多得多的价值。"

Dwarkesh Podcast, 2024

2.3 Multi-Decadal Abstractions（跨数十年的抽象）

"we're all trying to impress upon people at Stripe the importance of multi-decadal abstractions. And I think people sometimes respond to that thinking that that's some insanely lofty, kind of implausibly ambitious, I don't know, hyperbole. But no, I think that's actually just what happens when you, when you get this stuff right."

"我们都在努力让 Stripe 的人理解 multi-decadal abstractions 的重要性。有时人们会觉得这是某种极其夸张、不切实际、自我吹嘘的高调说法。但其实不是——这就是把这些东西做对之后会自然发生的事情。"

Dwarkesh Podcast, 2024

2.4 Conway's Law

"If I were to do everything at Stripe again, I would have spent even more time on APIs and data models. This is because of Conway's law effect of how both of those things end up shaping the organization. The strong version is that it substantially shapes your strategy and just your business outcomes."

"如果我重新做一遍 Stripe 的所有事，我会在 API 和数据模型上花更多时间。这是因为 Conway's Law 的效应——这两样东西最终会塑造组织。强一点的版本是：它们会实质性地塑造你的战略和你的业务结果。"

AI + a16z, 2026

2.5 N-by-M 通用原则

"The answer to that is to unify everything you can plausibly unify. Another lesson, maybe a bit tongue-in-cheek, would be to make anything that plausibly could be an N by M relationship to support that."

"答案是：尽可能把一切你能合理统一的东西统一起来。另一条教训，多少带点玩笑——凡是有可能形成 N 对 M 关系的东西，就都把它做成支持 N 对 M。"

AI + a16z, 2026

2.6 Self-Serve in Hard-to-be-Self-Serve Domain

"We're a self-serve product for a fairly ubiquitous need. In a domain where it's actually hard to be self-serve."

"我们是一个面向相当普遍需求的自助式产品。而这个领域其实很难做成自助式的。"

"We're continually surprised when we build things that we conceive of as being for the low end, how frequently they get enthusiastically adopted at the high end."

"让我们持续感到惊讶的是，那些我们当初是为低端市场设计的东西，常常会在高端被热情采用。"

Invest Like the Best, 2023

2.7 Talk Up to the User

"We have another value at Stripe that we try to really inculcate is talking up to the user."

"Stripe 还有另一条我们努力灌输的价值观——以高于用户的姿态去尊重用户（不要把用户当成需要俯就的对象）。"

Invest Like the Best, 2023

2.8 Auteur Vision（作者主义）

"You really need the auteur. The whole thing is pointless unless someone has a vision."

"你真的需要那个 auteur（作者型的人）。除非有人有清晰的愿景，否则整件事都没意义。"

Invest Like the Best, 2023

2.9 软件不可被简单度量

"And part of it comes from the lack of legibility in what you're working with."

"其中一部分原因来自你所处理的东西本身缺乏可读性（难以被外部清晰看见）。"

Invest Like the Best, 2023

→ 软件工程的核心难题是"不可见性"——无法像 1950 年代制造业那样直接度量产出。这正是为什么搬来 1950 年代的管理理论是错的。

2.10 五分钟而非新的可能

"We weren't going from a world in which it was impossible to one in which you could, but rather a world where it took weeks of setup to a world where you could do it in five minutes."

"我们做的不是把一件原本不可能的事变得可能，而是把一件原本要花几周才能搞定的事变成五分钟就能做完。"

Gray Area, 2016

2.11 反计算环境的退步

"I think the basic idea of a development environment and not just a text editor is really the right idea. That's the thing that the Lisp machines had and Genera, that's the thing that to some extent Mathematica has, and that's the thing that Smalltalk has. I think it's just such a mistake that we have ended up with development environments where there is such a separation between the runtime and the text editing and the environment in which the code runs."

"我觉得'开发环境而不仅仅是文本编辑器'这个基本想法是非常正确的。Lisp Machines 和 Genera 有这种东西，Mathematica 在一定程度上有，Smalltalk 也有。我们今天竟然演变成把运行时、文本编辑、代码执行的环境分得这么开的开发环境，这真是个大错误。"

AI + a16z, 2026

"What I'd love to see, for example, is when I hover over a line of code, I would like to see profiling information about just the runtime characteristics of that code or that function or whatever. I would like to see logging and error information overlaid. When I hover over a variable, I would like to see the most common values that it takes on in production."

"比如我希望看到的是：当我把鼠标悬停在某一行代码上时，能看到这行代码或这个函数的运行时性能信息；能看到叠加在上面的日志和错误信息；悬停在某个变量上时，能看到它在生产环境里最常出现的取值。"

2.12 Craft & Beauty

"we really care about what we call at Stripe craft and beauty, and we want our software to be well designed and pleasant to use, and pleasant to use not only in the superficial pixel sense, but also in the deep, it works very well sense, and is something you can set up and largely forget about and just trust"

"我们在 Stripe 内部非常在意我们所说的 craft and beauty——希望我们的软件被精心设计、用起来令人愉悦；这种愉悦不只是表层的像素美观，而是深层意义上'它就是工作得很好'，是那种你设置好就基本可以抛诸脑后、信任它的东西。"

AI + a16z, 2026

"I think it's very hard to assemble groups of the best people if you don't take the practice of the work super seriously."

"如果你不把工作本身的实践极其认真地对待，是很难组建出由顶尖的人组成的团队的。"

Dwarkesh Podcast, 2024

2.13 互联网作为分发的重要变化

"And by efficient, I mean that it was becoming more likely that people would come to discover the kind of quote unquote correct thing that maybe kind of 30 years ago or something because there were sort of much slower or fewer kind of distribution channels that who won and what got used, we kind of determined more by marketing budgets or by sales tax or whatever. But that there's a new kind of sort of transparency and kind of efficacy and information dissemination that the internet now made possible that shifted the way that sort of a product could really get adopted."

"我说的'有效'是指：人们越来越有可能去发现所谓'正确'的那个东西。可能 30 年前由于分发渠道更慢、更少，谁胜出、谁的东西被使用，更多取决于营销预算或销售之类的因素。但互联网现在让信息传播获得了一种新的透明度和效力，这改变了一个产品真正被采用的方式。"

Tim Ferriss Show #353

2.14 巨头很难变成软件公司

"Once you attach a real sales and marketing engine to that, it's gonna be really freaking hard for a big company to effectively compete because again, this kind of organizational transformation into being good at software is just so profoundly hard."

"一旦你给它接上真正的销售和营销引擎，大公司想要有效地竞争就极其困难——因为再说一次，那种把组织转型成擅长做软件的转变，难度实在太深了。"

Tim Ferriss Show #353

3. 关于组织与文化 / Organization & Culture

3.1 三大招聘维度：Rigor / Willfulness / Warmth

"Many organizations prize smoothness of interactions and try to minimize ruffled feathers, inadvertently preferring cohesion over correctness. We seek people who prioritize correctness, who do not mind being wrong, and who are willing to contemplate ideas that seem improbable or divergent from the status quo."

"许多组织看重互动的顺滑、尽量不让人不悦，结果在不知不觉中把'团结一致'置于'正确'之上。我们要找的是把'正确'放在第一位、不介意自己被证明是错的、愿意认真考虑那些看似不可能或偏离现状的想法的人。"

"We look for people who not only tolerate that difficulty but actually enjoy the challenge of building in an undefined or broken space."

"我们要找的不仅仅是能忍受这种困难的人，而是真的乐于在一个未定义或破碎的空间里搭建东西的人。"

"Many organizations prefer cohesion over correctness, but we seek people who prioritize being right over minimizing ruffled feathers and who are willing to challenge the status quo."

"许多组织把'团结一致'看得比'正确'更重要，但我们要找的是把'正确'优先于'不让人不悦'的人，以及愿意挑战现状的人。"

Knowledge Project #32, Knowledge Project 2022

三个维度：

1. Rigor and clarity of thought（严谨与思维清晰）
2. Determination / willfulness（决心与意志力）
3. Interpersonal warmth（人际温度）

3.2 Cohesion vs Correctness

"Many organizations prize smoothness of interactions and try to minimize ruffled feathers, inadvertently preferring cohesion over correctness."

"许多组织看重互动的顺滑、尽量不让人不悦，结果在不知不觉中把'团结一致'置于'正确'之上。"

Knowledge Project #32

3.3 Pace Layering（节奏分层）

"We invest roughly 70% to 80% of our effort into optimizing what we already have... The remaining 20% is allocated to more speculative bets."

"我们大约把 70% 到 80% 的精力投入到优化我们已有的东西上……剩下的 20% 留给更具投机性的下注。"

"We devote 70-80% of our effort to optimizing what we already have... and roughly 20% to speculative bets."

"我们把 70%–80% 的努力用来优化已有的东西……大约 20% 用于投机性的下注。"

Knowledge Project #32

→ 取自 Stewart Brand 的"pace layering"概念。Patrick 用它解决"优化 vs 探索"的张力——不是均匀分配，而是按团队分配，让不同节奏的团队"have dinner together and feel like they are on the same team"。

3.4 CEO 不应该是决策瓶颈

"If I am the one making a decision, it often suggests that something is broken organizationally or that the domain experts aren't empowered."

"如果是由我来做这个决策，那往往意味着组织上有什么东西出了问题，或者真正的领域专家没有被赋予应有的权力。"

Knowledge Project 2022

3.5 文化是集体认知探索，不是广播

"I don't view culture as a top-down broadcast that inevitably loses fidelity as it gets transmitted through more people. Instead, I see it as a collective exercise in learning what the platonic version of the company should be."

"我并不把文化看作一种自上而下的广播——经由越来越多的人传播之后必然失真。我把它看作一项集体的练习：共同摸索这家公司应有的'柏拉图式版本'是什么样的。"

"We aren't fighting a losing battle against entropy; we are chiseling away at our own identity through success and failure."

"我们并不是在打一场注定输给熵的仗，我们是在通过一次次成功和失败，凿刻出自己的身份。"

Gray Area, 2016

→ 反熵的视角：随着公司变大，文化变得更清晰而非稀释。目标是 "become more like ourselves over time"。

3.6 文化护城河的本质

"in as much as we have a moat, it's because we have a very good understanding of our domain and a set of people who actually care about solving the problems and who are continually paranoid at the prospect that we might be forgetting something important and sort of trying to figure out what the important thing that could supplant Stripe's approaches is and making sure that we build those first and so forth."

"如果说我们有什么护城河，那是因为我们对自己的领域有非常深入的理解，并且有一群真正在意解决问题的人——他们持续地为'我们可能漏掉了什么重要东西'感到焦虑，并不断试图想清楚什么东西可能会取代 Stripe 的做法，然后确保我们率先把这些东西做出来。"

Dwarkesh Podcast, 2024

3.7 强文化的对偶检验

"一个内部人能够 "expound on its merits for half an hour, but also describe all the things they dislike about it." "滔滔不绝地讲半个小时它的优点，同时也能说出他们所有不喜欢它的地方。"

Knowledge Project #32

→ 真正的强文化既能讲清楚为什么好，也能讲清楚自己不喜欢的地方——因为它有清晰的身份。

3.8 学到"真正的卓越"会永久改变标准

"That when they worked with X person or Y organization or in Z environment or whatever, that they learned what great actually is. And that just permanently changed their sense for what their own standard for their work ought to be."

"当他们和某某人合作、在某某组织或某某环境里工作时，他们学到了'真正的卓越'到底是什么样。这件事会永久改变他们对自己的工作标准应该是怎样的判断。"

Dwarkesh Podcast, 2024

3.9 大公司必须非理性地保守

"In order to endure, large companies have to be irrationally risk-averse. That is not a statement about coordination costs; it is a basic statement about survival."

"要长期存活，大公司必须非理性地厌恶风险。这不是在说协调成本，这是关于生存最基本的陈述。"

"Large companies must be irrationally risk-averse to survive: 'If an established company takes a bet that has a 50-50 chance of 10x growth but a risk of destroying the company, it's a bad bet for them because they have too much to lose."

"大公司要活下去就必须非理性地厌恶风险：'如果一家成熟公司去下一个有 50% 概率带来 10 倍增长、但同时有可能毁掉公司的赌注，对它们来说这就是一个坏赌注——它们要输掉的东西太多了。"

Stanford GSB, 2016

3.10 寻找不被周围世界观染色的人

"That is now, in this new context, in fact, in some instances, a competitive advantage because there sort of lack of absorbing worldviews from others or feeling that pressure enables them to come up with their own perspective and orientation. So if we filter those people out. Right, because they may be similarly difficult to imitate in a sense as a basketball player is to imitate in their height."

"在今天这种新环境下，这反而在某些情况下成了一种竞争优势——因为他们不太吸收别人的世界观，也不感受到那种压力，从而能形成自己的视角和方向。所以我们要把那些人筛选出来。某种意义上，他们和篮球运动员的身高一样难以被模仿。"

Tim Ferriss Show #353

→ 像篮球运动员的身高一样难以模仿。

3.11 反对组织实验的实用主义

"I think that most of the experiments have failed. I most efforts to do something that's kind of substantially different to the kind of best practice of the day, most of those efforts fail."

"我觉得大多数这种实验都失败了。绝大多数试图做出实质性不同于当下最佳实践的尝试，最后都以失败告终。"

Tim Ferriss Show #353

→ 对 holacracy 等花哨组织实验保持怀疑——多数都失败了。

3.12 好文化是反均值的

"Good cultures are often not regressed to the mean. They are differentiated. They have something that people strongly believe in a way they strongly act that is different from every other company."

"好的文化往往不是向均值回归的。它们是有差异化的。它们有一些被强烈相信的东西、有一种强烈的行为方式，和其他每一家公司都不一样。"

Invest Like the Best, 2023

4. 关于决策 / Decision-Making

4.1 速度优于精度

"I place more value on decision speed; making twice as many decisions at half the precision is often better."

"我更看重决策速度——以一半的精度做两倍数量的决策，往往是更好的选择。"

"If you can make twice as many decisions at half the precision, that is often better. Treat fast decisions as an asset and a capability in their own right."

"如果你能以一半的精度做两倍数量的决策，往往是更好的。要把'决策快'本身当作一种资产和能力来看待。"

Knowledge Project #32, Knowledge Project 2022

4.2 重做决策的能力 > 一次决策的精度

"you don't necessarily need to be that good at decision making if you get really good at remaking the decision when and as necessary... Instead, it's all about the constant feedback and course and error correction."

"如果你真的擅长在必要时重做决策，那你就不一定要在做决策本身上有多厉害……重要的是持续的反馈、纠偏和误差校正。"

Tim Ferriss Show #353

4.3 选项生成 > 选项选择

"How do I make sure that both options A and B are as good as possible, and there's also a C, D, and E, and that those options are great too?"

"我怎么才能确保选项 A 和 B 都尽可能好，而且还有 C、D、E，并且这些选项同样精彩？"

Tim Ferriss Show #353

4.4 按可逆性和量级分类

按 Bezos 的 "Type 1 / Type 2 doors" 类似：低可逆性、高影响的决策才需要长时间深思熟虑，其他的越快越好。

4.5 分歧时不在战术上吵，要回到底层模型

"Once we align on what we are optimizing for, disagreements over instrumental efficacy become much easier to resolve."

"只要我们就'到底在优化什么'达成一致，关于'哪种手段更有效'的分歧就会变得容易解决得多。"

Knowledge Project 2022

→ 当与同事意见不一致，去看双方对使命、关键指标、客户定义的不同认知，而不是争论具体选择。

4.6 Virtual Board of Directors

"the idea that people can improve their own decision making. By simply choosing other people... imagining what they would say... and then averaging over the responses. But that's like a really amazing fact because of course, this is all occurring in your own head."

"这个想法是：人可以通过选定一些他人……想象他们会怎么说……然后对这些回答取平均，来改善自己的决策。这其实是个相当惊人的事实，因为这一切都发生在你自己的脑子里。"

Tim Ferriss Show #353

→ 在头脑里模拟自己尊敬的人会怎么思考这个问题——包括已故的 Richard Feynman。

5. 关于阅读与学习 / Reading & Learning

5.1 偏向旧书

"I really think people should be kind of much more biased towards older books than they are. I think The Dream Machine is a good example of this."

"我真的觉得人们读书时应该更多地偏向旧书一些。我认为 The Dream Machine 就是个很好的例子。"

Tim Ferriss Show #353

5.2 Reader-Book Fit

"it really matters a lot when I read the book and the frame of mind that I'm in at the time when I stumble across it."

"我什么时候读这本书、以及我偶遇它时所处的心境，对我来说真的很重要。"

Tim Ferriss Show #353

→ 像 product-market fit 一样存在 reader-book fit。买一本看似有趣的书，让它在脑中"沉淀"，等到合适的时刻再读。

5.3 每一刻都该读你知道的最好的书

"At every moment, you should be reading the best book you know of in the world. As soon as you discover something more interesting or important, you should discard your current book in favor of that."

"任何时刻，你都应该在读你所知道的世界上最好的那本书。一旦你发现了更有趣或更重要的，就该放下手上这本，转而去读它。"

Knowledge Project 2022（引用 John Collison 的方法）

5.4 Basho 原则

"I follow the Basho principle: don't follow the people you admire, follow what they admired."

"我遵循 Basho 原则：不要追随你所敬仰的人，而要追随他们所敬仰的东西。"

Knowledge Project 2022

→ 取自日本俳句诗人松尾芭蕉。不是模仿你尊敬的人，而是追溯他们尊敬的源头。

5.5 主动、非线性的阅读

读书就是跳读、放弃、回头、跳到中间。不是顺序通读。

5.6 物理书制造"思想空间"

把书摊在房子里——厨房、卧室、床上——制造"productive collisions"（有生产力的碰撞）。

5.7 Pointer Graph（指针图）学习模型

"find generous, truth-seeking thinkers and follow their citation trails."

"找到那些慷慨、求真的思想家，跟随他们的引用线索。"

Gray Area, 2016

→ 找到几个慷慨的、真心追求真理的思想家，跟随他们的引用网络——这就是学习的引擎。Patrick 名单上：Tyler Cowen（Marginal Revolution）、Scott Alexander（Slate Star Codex）、David Deutsch、Hamming、Parfit。

5.8 关于不知道的焦虑

"I do not feel like I am weird; I feel like everyone else is weird. There is so much to know, and I feel stressed because it is obviously important, and I do not know it."

"我不觉得自己怪，我觉得是其他所有人怪。要知道的东西实在太多了，我会感到焦虑，因为这些显然很重要，而我不知道。"

Knowledge Project #32

5.9 Cold Email 习惯

"It is an asymmetric habit; it doesn't cost much when they don't respond, but half the time they do, and it often leads to a dialogue."

"这是一种不对称的习惯：对方不回应时你的代价并不大，但有一半的时候他们会回，而且常常因此开启一段对话。"

Knowledge Project 2022

→ 他与 Tyler Cowen 的友谊就是从一封冷邮件开始的。

5.10 Sussman 影响：修改既有系统的设计能力

"I took a class from Jerry Sussman on large-scale symbolic systems, but really what he was trying to focus on was the idea of creating code bases and environments and abstractions that were easy to modify. Every assignment was about modifying an existing system, thinking about how could you design things in such a way that those modifications, and there might be quite deep modifications, become straightforward."

"我上过 Jerry Sussman 关于大规模符号系统的课，但他真正想关注的是：如何创造易于修改的代码库、环境和抽象。每一道作业都是在修改一个既有系统，让你思考——你怎么设计才能让这种修改（而且可能是相当深层的修改）变得直截了当。"

AI + a16z, 2026

5.11 用 LLM 阅读

"Even when I'm reading a book, I'll sometimes activate Grok's voice mode, and I'll just passively ask questions while I'm reading, and Grok is just listening in the background, and the answers are very helpful."

"即使我在读书时，我有时也会打开 Grok 的语音模式，一边读一边随口问问题，Grok 在后台听着，给出的回答非常有用。"

"I use them mainly for answering factual or empirical questions I'm curious about. So for deep research style questions. I wish they were useful for writing, but I usually end up dissatisfied with the writing that they produce, so I don't really use them very much for that."

"我主要用它们来回答我好奇的事实性或经验性问题，也就是 deep research 风格的问题。我希望它们对写作有用，但我通常对它们写出来的东西不满意，所以在写作上其实用得不多。"

AI + a16z, 2026

6. 关于科学与进步 / Science & Progress

6.1 进步的核心命题

"Progress is amazing, influenceable, and understudied."

"进步是惊人的、可以被影响的、却被研究得太少。"

Atlantic, 2019

6.2 进步是可以影响的

"Humanity needs to get better at knowing how to get better."

"人类需要变得更擅长'知道如何变得更好'。"

Atlantic essay subtitle

6.3 科学减速论的证据三角

四类独立证据：

• 宏观 TFP
• 领域专业指标（药物发现、半导体）
• 结构指标（诺奖年龄、团队规模）
• 同行专家两两比较

"The rate of progress is declining rapidly on a per-person, per-hour, or per-dollar basis. Perhaps because of the enormous increase in our investment and efforts, we are sustaining aggregate progress, but our productivity per person is much lower."

"按人均、人时或人均美元计，进步的速率在迅速下降。或许是因为投入和努力大幅增加，我们才维持了总量上的进步；但每个人的生产率要低得多。"

"We have 50 times more researchers in these fields, but as judged by scientists themselves, we are producing significant discoveries at a nearly constant rate. This suggests that an individual researcher is becoming 50 times less likely to produce significant work."

"这些领域的研究者数量是过去的 50 倍，但按科学家自己的判断，重大发现的产出速度几乎保持不变。这意味着单个研究者做出重要工作的概率正在变成原来的 1/50。"

EconTalk, 2019

6.4 HHMI 98% 效应量

"Grant recipients who were funded differently were 98% more likely to produce work in the top 1% by citations. The fact that you can get such a large effect size just by changing one variable suggests it is possible to create other high-impact shifts."

"采用不同资助方式的受资助者，产出引用排名前 1% 工作的概率高出 98%。仅靠改变一个变量就能得到如此大的效应量，说明我们有可能创造出其他高影响的改变。"

EconTalk, 2019

6.5 经济增长的道德分量

"If you look at GDP per capita around the world and plot that against self-reported well-being, the correlation is 0.68. It is rare in human psychology and sociology to get a correlation that strong."

"如果你把世界各地的人均 GDP 与自报幸福感画在一起，相关系数是 0.68。在人类心理学和社会学里，能拿到这么强的相关性是很罕见的。"

EconTalk, 2019

"Well, I guess the first thing I'd say is I think that question you just asked is the big question I think we should all be obsessed with... It is the, it, it's, I, it is an issue of true moral importance."

"嗯，我先说的第一件事是：你刚才问的这个问题，我觉得正是我们所有人都应该痴迷其中的大问题……这是一件具有真正道德重要性的事。"

Tim Ferriss Show #353

6.6 Talent vs Opportunity

"Talent is approximately evenly distributed, but opportunity is much more uneven. I conclude from my own upbringing that there must be many more people like me, but for whom the pachinko machine of outcomes did not fall the right way."

"天赋大致是均匀分布的，但机会的分布要不均匀得多。从我自己的成长经历看，我得出的结论是：一定还有许多和我类似的人，只是命运的弹珠机没有刚好掉到对的位置。"

EconTalk, 2019

6.7 Arc Institute 的设计哲学

"what's different about Arc is one, scientists are funded themselves to pursue whatever they want. So it's curiosity-driven research, whereas NIH grants are given for projects. Second, we build a lot of in-house infrastructure so that scientists can draw upon other platforms and other capabilities that they don't have to go and build and maintain themselves."

"Arc 不一样的地方有两点。第一，科学家本人获得资助，可以追求任何他们想做的方向，所以是好奇心驱动的研究，而 NIH 的资助是给项目的。第二，我们在内部建立了大量基础设施，让科学家可以使用各种平台和能力，不必自己去搭建和维护。"

Dwarkesh Podcast, 2024

6.8 "We've never cured a complex disease"

"At ARQ, we're working on training foundation models for biology using DNA and things like that. We're working on a virtual cell and generally we're trying to better understand complex diseases. We've never cured a complex disease."

"在 Arc，我们正在用 DNA 之类的东西训练生物学的 foundation models，正在做一个 virtual cell，总体上是在尝试更好地理解复杂疾病。我们从未治愈过一种复杂疾病。"

"the pleiotropy of the genes in terms of all the different parts of the body and the systems and the mechanisms inside the cell that they affect is so, you know, there's so much combinatoric complexity there."

"基因的多效性体现在它们影响身体的众多不同部位、系统和细胞内部的机制——这里头组合上的复杂性实在太大了。"

AI + a16z, 2026

6.9 生物学的新 Turing Loop

"Over the last 10-ish years, we've gotten three new classes of technology in biology: much better sequencing technology, single cell sequencing, neural networks and deep learning and transformers, and huge improvements in functional genomics and CRISPR and bridge editing... If you put those together, you now have the ability to, again, at the kind of level of the individual cell, to read, think, and to write. And this starts to really feel like a new kind of Turing loop and to have its own sort of completeness."

"过去十来年里，我们在生物学里得到了三类新技术：好得多的测序技术，单细胞测序；神经网络、深度学习和 transformers；以及功能基因组学、CRISPR 和 bridge editing 上的巨大改进……把这些放在一起，你现在就有能力——同样是在单个细胞的层面——读、想、写。这开始真的像一种新的 Turing loop，自身具有某种完备性。"

AI + a16z, 2026

6.10 Fast Grants 的小却有力

"Fast Grants was not this giant, impressive edifice, but it doesn't have to be giant, right, to have that kind of bigger impact?"

"Fast Grants 并不是某座庞大、壮观的建筑，但要产生那种更大的影响，本来就不必很庞大，对吧？"

Dwarkesh Podcast, 2024

6.11 经济学之外的约束

"By just talking about these things in funding and dollar terms, you're making the implicit assumption that the only relevant constraint is the financial one. In practice, maybe it's permits or it's labor shortages or it's other things."

"仅仅用资金和美元的口径来谈这些事，你就在隐性地假设唯一相关的约束是财务上的。但在实践中，约束可能是审批，可能是劳动力短缺，也可能是其他东西。"

Dwarkesh Podcast, 2024

6.12 mannose 例子：激励机制的盲点

"But mannose is like, it's a generic sugar. It's been understood for I guess more than a century. And you couldn't patent it, importantly. It's not clear who has the incentive to even fund the work to test whether or not this would actually work in practice."

"但 mannose 就是一种普通的糖，被认识得我猜超过一个世纪了。重要的是它无法被申请专利。所以不清楚谁会有激励去资助哪怕是验证它在实际中是否真的有效的研究。"

Dwarkesh Podcast, 2024

6.13 进步研究比五年前更需要

"I think the need for progress studies has increased because I think the degrees of freedom are increasing. I think the kinds of questions that progress studies tries to answer are now more pressing and urgent because I think the degrees of freedom are increasing."

"我认为对 progress studies 的需要在增加，因为自由度在增加。progress studies 想回答的那些问题如今变得更紧迫、更急切，原因也在于自由度在增加。"

AI + a16z, 2026

"Maybe a critique you could have leveled at Progress Studies or Progress Studies style of thinking five years ago is these are all nice questions, but the world is on a kind of it's a fore-ordained escalator path to some kind of teleological outcome. And I don't think the world feels that way today, or certainly feels much less that way today than it did."

"五年前你或许可以这样批评 Progress Studies 或那种思考方式：这些问题都不错，但世界本来就走在一条注定通往某种目的论结果的自动扶梯上。我不认为今天的世界还是这种感觉，或者至少远没有当时那么强烈。"

AI + a16z, 2026

7. 关于行业与战略 / Industry & Strategy

7.1 时间视野作为护城河

"There is something deep about the notion of using time horizons as a competitive advantage. You are simply willing to wait longer than other people."

"把时间视野作为竞争优势这个想法里有某种很深的东西。你只是单纯地比别人愿意等更久而已。"

Gray Area, 2016（论 Bezos）

7.2 拒绝 IPO

"An IPO would be a solution in search of a problem."

"IPO 对我们来说是一个在寻找问题的解决方案。"

Patrick Collison, 多次访谈

"Stripe is profitable, and expects to remain so in the coming years."

"Stripe 是盈利的，并且预计在未来几年仍将保持盈利。"

2024 Annual Letter

7.3 Stablecoin 是金融服务的"室温超导"

"Stablecoins are room-temperature superconductors for financial services. Thanks to stablecoins, businesses around the world will benefit from significant speed, coverage, and cost improvements in the coming years. Stripe is going to build the world's best stablecoin [infrastructure]."

"stablecoin 之于金融服务，就像 room-temperature superconductors。得益于 stablecoin，世界各地的企业将在未来几年里在速度、覆盖和成本上获得显著改善。Stripe 将打造世界上最好的 stablecoin 基础设施。"

Patrick Collison tweet, 2024-10-21

"Improvements to the basic usability of money make economies more prosperous... Stablecoins are a new branch of the money tree."

"对'钱的基本可用性'的每一次改进都会让经济更繁荣……stablecoin 是这棵货币之树上的一个新分支。"

2024 Annual Letter

7.4 信息网络的垄断陷阱

"Every other information network that has been started has started out with lots of innovation and the healthy ecosystem. Low barriers to entry, and has ended up as a monoculture regulation captured oligopoly with two or three big players that get to extract rents and that has happened with every other information network to date."

"过去启动过的每一个信息网络，最初都有大量创新和健康的生态、进入门槛低，最终都演变成一个被监管俘获的单一文化寡头格局——只有两三个大玩家能够攫取租金。迄今为止每一个信息网络都是这样。"

Invest Like the Best, 2023

7.5 监管是不可见的成本

"We get to observe when regulations fail to prevent something harmful, but we almost never get to observe when regulations prevent something that would be really good."

"当监管未能阻止某件有害的事时我们看得到，但当监管阻止了某件本可以非常美好的事时，我们几乎从来看不到。"

Gray Area, 2016

7.6 中国是唯一真竞争对手

"But in terms of technology versus China, I'm not sure that's necessarily true. I worry more about the United States falling behind China or at least, China getting parity with us...Basically, if I'm thinking about competition, I don't see any reason to think about any other country except China to compete with us."

"但在与中国的技术竞争上，我不太确定这一定成立。我更担心的是美国落后于中国，或至少中国在某些领域追平我们……基本上，如果我在思考竞争，我看不出有任何理由去考虑除中国以外的国家作为我们的对手。"

Econ 102, 2025

7.7 Aggregation of Demand 理论

"It seems too self-aggrandizing to call it the theory of Stripe, but some vague hunch in Stripe is that that aggregation of demand can have important expansionary effects with respect to the ensuing supply."

"把它叫作'Stripe 的理论'听起来太自我膨胀了，但 Stripe 内部有一个模糊的直觉——需求的聚合可以对随之而来的供给产生重要的扩张性效应。"

Dwarkesh Podcast, 2024

→ Stripe Climate / Frontier 是这套理论的小规模应用。

7.8 GDP 增长的恒定性是个谜

"I think the constancy of US GDP growth is, I think, just like one of the weirdest things."

"我觉得美国 GDP 增长的恒定性，是最古怪的事情之一。"

Dwarkesh Podcast, 2024

7.9 大公司被低估

"I think big business is underrated. And if you look at the survey data, people tend to have very positive sentiments, not only towards startups, but towards small business as a class."

"我觉得大企业被低估了。如果你看调查数据，人们对创业公司、对作为一个类别的小企业，普遍都怀有非常积极的情感。"

Dwarkesh Podcast, 2024

7.10 美国劳动生产率的独特性

"Approximately post-GFC, labor productivity in pretty much all OECD countries, with the exception of the U.S., falls significantly off-trend. The U.S. is the only country that remains approximately on its 1990 through 2010 trend line."

"大致从全球金融危机之后开始，几乎所有 OECD 国家的劳动生产率都明显偏离趋势——美国是唯一的例外。美国基本上仍处在它 1990 到 2010 年的趋势线上。"

Econ 102（与 Noah Smith 对话），2025

7.11 关于移民的 40% 数据点

"Roughly 40% of successful firms in the U.S. have at least one immigrant founder, despite immigrants making up a much smaller fraction of the workforce."

"美国大约 40% 的成功企业至少有一位移民创始人——尽管移民在劳动人口中占的比例要小得多。"

Gray Area, 2016

7.12 关于 SF 房价对国家的代价

"It is dispiriting because it feels like a suffocation of future potential. Silicon Valley is the greatest concentration of wealth creation in history, but its rise was enabled by cheap mobility and expansion — factors that no longer exist due to collective political decisions."

"这件事令人沮丧，因为它给人一种未来潜力被窒息的感觉。Silicon Valley 是历史上最集中的财富创造地，但它的崛起是建立在廉价的人口流动和扩张之上的——这些条件如今因为集体性的政治决定而不复存在。"

Knowledge Project #32

7.13 SF 不鼓励深度技术专精

"There is a set of career paths that require accumulating a lot of expertise and really studying a domain in tremendous depth. San Francisco doesn't encourage the pursuit of really deep technical depth."

"有一类职业路径需要积累大量专业知识、对某一领域进行极深入的钻研。San Francisco 并不鼓励人们去追求真正的深度技术专精。"

Dwarkesh Podcast, 2024

7.14 Stripe 顾客增长超过互联网整体

"Well, the customers that Stripe serves are outgrowing the internet, the internet economy as a whole, like an aggregate. So at some point those have to converge for kind of obvious mathematical reasons but you know we're 14 years in and they haven't converged yet."

"Stripe 服务的客户群整体上长得比互联网经济还快。出于显而易见的数学原因，这两个增速早晚会收敛——但我们做了 14 年，到现在还没有。"

Dwarkesh Podcast, 2024

7.15 智能手机是真实世界的魔杖

"Stripe is essentially a bet on human convenience."

"Stripe 本质上是对人类'图方便'的一次下注。"

"智能手机 "is a magic wand for the real world." "（智能手机）是真实世界的一根魔杖。"

Stanford GSB, 2016

7.16 卫生医疗与教育的停滞

"Healthcare and education stagnation comes from cost-management framing rather than progress-rate framing."

"医疗和教育的停滞，源于我们用'控制成本'的框架去看它们，而不是用'提高进步速率'的框架。"

Gray Area, 2016

医疗被 Certificate of Need 法律掐住。Baumol cost disease 解释为什么劳动密集型部门成本不断上涨。

8. 关于哲学与认识论 / Philosophy & Epistemology

8.1 Argumentative Theory of Reason

"When we are smart, we don't necessarily become better at finding the truth; we just become better at generating sophisticated rationalizations to defend our social standing or group identity."

"当我们变得更聪明时，我们不一定更善于找到真相，而是更善于产生精致的合理化论证，去捍卫我们的社会地位或群体身份。"

"There is nothing about your brain that made an effective truth-seeking machine. There is a disconnected moment in history that gave you tools, but you can lose those tools or misuse them."

"你的大脑本身并没有什么东西让它成为一台有效的求真机器。历史上有一个非连续的时刻给了你一些工具，但你可能会失去这些工具，也可能会误用它们。"

Gray Area, 2016（引用 Dan Sperber 的论证理论）

8.2 启蒙运动作为历史异常

"The Romans, the Mughals, and the Chinese dynasties all had the necessary technology to build steam engines, yet they didn't. Something changed about 250 years ago in how we think, how we generate ideas, and how we value explanations."

"罗马人、莫卧儿人、中国的历代王朝都拥有制造蒸汽机所需的技术，但他们都没有这么做。大约 250 年前，我们思考的方式、产生想法的方式、对解释的看重发生了某种改变。"

"We could easily lose the tools of the Enlightenment if we forget how fragile they are."

"如果我们忘了 Enlightenment 的那些工具有多脆弱，我们很容易就会失去它们。"

Gray Area, 2016

8.3 Press Secretaries vs Scientists

"They have enough stature to be credible, but they aren't in positions that force them to censor their thoughts. They are essentially scientists in a world of press secretaries."

"他们的地位高到足以让人信服，但他们所处的位置并不逼着他们对自己的想法做自我审查。他们本质上是 press secretaries 的世界里的科学家。"

Gray Area, 2016

→ 对 Tyler Cowen、Scott Alexander 这类思想者的最高赞誉——他们既有可信度又能不受迫害地讲真话。

8.4 Washington vs Silicon Valley

"In Washington, people are forged by seeing problems that should be solvable prove impossible to solve, leading to a focus on limits and compromise. In Silicon Valley, the culture is forged by seeing 'impossible' things happen routinely."

"在 Washington，人是被这种经历塑造的：看着那些本该可以解决的问题被证明无法解决，于是把注意力放在边界和妥协上。而在 Silicon Valley，文化是在不断目睹'不可能的'事情成为家常便饭中被塑造的。"

Gray Area, 2016

8.5 你是你五个最亲密朋友的均值

"You end up the average of your five closest friends. We do choose those people, though we may not think of it that way; you are choosing who you are."

"你最后会变成你五位最亲密朋友的平均。我们其实是在选择这些人——虽然我们可能没那样想；你是在选择你成为谁。"

"You end up the average of your five closest friends. There is a deep truth to that; choosing your five closest friends is choosing who you are."

"你最后会变成你五位最亲密朋友的平均。这里有一个深刻的真相：选择你的五位最亲密的朋友，就是在选择你自己是谁。"

Knowledge Project #32, Knowledge Project 2022

8.6 Wandering ≠ Lost

"So not all, not all who wander are lost, right? So there's a difference between being lost, which is you have a destination and you have gone off track or you are preoccupied with where you are currently located and don't know where you're located versus exploring and wandering."

"并非所有漫游的人都在迷路。'迷路'指的是你有一个目的地但偏离了轨道，或是被困在当前位置、不知道自己在哪儿；这与探索和漫游不同。"

Tim Ferriss Show #353

8.7 Applied Epistemology 是长寿机构的共性

"I am struck by how much these successful institutions prioritize applied epistemology. They are obsessed with structuring doubt, accounting for biases, and creating mechanisms for clarity of thinking."

"让我印象深刻的是，这些成功的长寿机构把 applied epistemology 看得多么重要。它们痴迷于把怀疑结构化、考虑偏见、为思维的清晰建立机制。"

Knowledge Project 2022

→ 他研究的长寿机构：Sequoia Capital、Koch Industries、The New Yorker、Apple——它们的共性是关于自己思考过程的严苛。

8.8 思考独立性

"nobody is going to teach you to think for yourself"

"没有人会教你独立思考。"

"don't make the mistake of judging your success based on your current peer group. Status lags by a generation or more."

"不要犯这个错误——以你当前的同辈群体作为衡量自己成功的标准。地位的反映滞后整整一代人甚至更久。"

Tim Ferriss Show #353 / patrickcollison.com/advice

9. 关于身份与个人 / Identity & Personal

9.1 爱尔兰是构成性的，不是劣势

"Not only mistaken about Limerick but the idea of 'overcoming' anything is crazy. We are who we are because we grew up where we did."

"不仅是对利默里克的误解，'克服'什么这种说法本身就很离谱。我们之所以是现在的我们，正是因为我们在那里长大。"

2021 Forbes pushback tweet

9.2 创业是日常

"starting a company was really not some strange, foreign thing. It was pretty normal."

"创业并不是什么奇怪、陌生的事。它挺平常的。"

Tim Ferriss Show #353

9.3 Patrick 的"Kindest Things"

"The two major ones are this one of someone noticed something about me and did something customized and shaped me in a certain way or bet on me when they didn't need to."

"最重要的两类是：有人注意到我身上的某些东西，并为我做了一件特别定制的事，从而以某种方式塑造了我；或者在并不需要这么做的情况下，押注在我身上。"

Invest Like the Best, 2023

→ 谈到他高中校长——校长允许他在课上读书、休假去做编程语言项目、找到方法让他申请美国大学而不是参加爱尔兰统一考试。

9.4 关心多过自信

"I wasn't confident in myself, but I was confident in my ideas."（这句出自 Evan Spiegel，但 Patrick 在多场访谈中表达过类似的"对想法自信、对自己谦卑"的状态。）
"我对我自己没什么自信，但我对我的想法有自信。"

"How much people care about what they do and the ability to move quickly is the predictor of success."

"人对自己所做之事在意到什么程度，加上移动速度的快慢，是预测成功的指标。"

类似精神，多次访谈

9.5 情绪自我管理是核心运营技能

"You have to realize that the constellation of problems often looks so implausible that it feels like someone is secretly screwing with you. You have to practice emotional self-management and, at times, simply smile at the unreasonableness of it all."

"你必须意识到，问题的组合常常看起来荒诞到像是有人在背后偷偷玩你。你必须练习情绪上的自我管理，有时候就得对着这种不可理喻笑一笑。"

Knowledge Project 2022

三、Patrick 的 23 个开放问题（保留英文原文）

发布在 patrickcollison.com/questions 上的 23 个他长期琢磨的问题，是他思想脉络最直接的索引：

1. Why are certain things getting so much more expensive?
2. Why do there seem to be more examples of rapidly-completed major projects in the past than the present?
3. Why is GDP growth so weirdly constant?
4. How do you ensure an adequate replacement rate in systems that have no natural way to die?
5. How do we help more experimental cities get started?
6. How do people decide to make major life changes?
7. Why are there so many successful startups in Stockholm?
8. Is Bloom's 'Two Sigma' phenomenon real? If so, what do we do about it?
9. How can we better understand the dynamics of progress in science?
10. Will end-user applications ever be truly programmable? If so, how?
11. What's the successor to the book? And how could books be improved?
12. What's the successor to the scientific paper and the scientific journal?
13. What's the right way to understand and model personality?
14. Could there be more good blogs?
15. Why are programming environments still so primitive?
16. What does religion cause?
17. Why is there no canon for life's most important questions?
18. Why are so many things so much nicer in Switzerland and Japan?
19. Why isn't China (yet) producing a lot of top-tier research?
20. Why don't we build nice neighborhoods any more?
21. What influences when people act in accordance with their self-interest and when they don't?
22. What's going on with infrastructure?
23. Why did climatic variability suddenly decline in the Holocene period?

四、Patrick 的 /fast 页面：他对"快速建造"的执念

发布在 patrickcollison.com/fast 上，列举 24 个"小团队短时间完成大事"的历史例子。完整列表（保留原文表述）：

1. BankAmericard: Dee Hock launched the card in 90 days, signing over 100,000 customers from scratch.
2. P-80 Shooting Star: Kelly Johnson's team designed and delivered the first USAF jet fighter in 143 days.
3. Marinship: Shipyard constructed in 197 days after telegram March 2, 1942; first ship completed September 15, 1942.
4. The Spirit of St. Louis: Donald Hall and Charles Lindbergh designed and built the aircraft in 60 days in 1927.
5. The Eiffel Tower: Built in 793 days (2 years, 2 months); completed 1889.
6. Treasure Island: 400-acre man-made island, 1935-1937, for the Golden Gate International Exposition.
7. Apollo 8: NASA decided on moon mission Aug 9, 1968; launch Dec 21, 1968 (134 days later).
8. The Alaska Highway: 1,700 miles military roadway built over 234 days starting 1942.
9. Disneyland: Built in 366 days.
10. The Empire State Building: Constructed in 410 days.
11. The Berlin Airlift: 463-day operation; Tegel airport built in 92 days (Aug 5–Nov 5, 1948).
12. The Pentagon: Constructed in 491 days, Sep 11, 1941 → Jan 15, 1943.
13. Boeing 747: Started March 1966; first 747 completed September 30, 1968 (~930 days).
14. The New York Subway: First contract Feb 21, 1900; 28 stations opened Oct 27, 1904 (4.7 years). Second Avenue Subway phase 1 opened Jan 1, 2017 (~17 years after 2000 approval).
15. TGV: Approved April 30, 1976; opened September 26, 1981 (1,975 days). California's HSR approved 1996, est. completion 2033 (~37 years).
16. USS Nautilus: Decided July 1951; entered service September 30, 1954 (1,173 days).
17. JavaScript: Brendan Eich implemented prototype in 10 days in May 1995; shipped in beta September 1995.
18. Unix: Ken Thompson wrote the first version in three weeks.
19. Xerox Alto: First GUI computer built in three months, Nov 1972 → March 1, 1973.
20. Shenzhen: Added 1 million residents (22% increase) between 1998-1999.
21. iPod: Hired late January 2001; greenlit March 2001; announced October 2001; shipped November 2001 (~290 days).
22. Amazon Prime: Implemented late 2004, announced February 2, 2005 (six weeks).
23. Git: Work began April 3, 2005; self-hosting in 4 days; public release with Linux April 20, 2005 (17 days).
24. COVID-19 Vaccines: Genome published January 10, 2020; Moderna finalized vaccine sequence January 13; first batch manufactured February 7; shipped to NIH February 24 (45 days from publication).

这个名单不是怀旧——它是 Patrick 用来反衬当代停滞的工具。他的核心问题是："Why do there seem to be more examples of rapidly-completed major projects in the past than the present?" Stripe Climate、Frontier、Fast Grants、Arc Institute 都是他对这个问题的实践回应。

五、几个关键交叉主题（贯穿整个档案）

主题 A：从 Lisp/Croma 到 API 设计的连贯线

• 16 岁用 Lisp 写 Croma
• Mindstorms（Logo 是 Lisp 方言）在他书架上
• 选 Smalltalk 做 Auctomatic
• 选 Ruby 做 Stripe（"主流化"决定）
• 把 Stripe 比作 Mathematica
• 反复说"程序设计环境是 Lisp Machine 时代退步了"
• v2 API 设计哲学：unify everything、N-by-M

这条线是他作为技术人的身份内核——他不是普通的 CEO，他是一个对 metaprogramming、self-modifying systems、abstraction quality 有近乎美学执念的工程师。

主题 B：从 Shuppa 到 Stripe Press 的"启用工具"线

• Shuppa（2007）：让 eBay 卖家更容易卖货
• Stripe（2010）：让开发者更容易收钱
• Atlas（2016）：让全球创业者更容易注册公司
• Stripe Press（2018）：让重要的旧思想更容易被发现
• Frontier（2022）：让碳清除技术更容易获得资金
• Arc Institute（2021）：让科学家更容易做好奇驱动研究

每一个都是同一种模式：找到一个高摩擦的 activation energy，把它降下来，让被压抑的需求/供给爆发出来。

主题 C：从青年科学家到 Arc Institute 的 18 年伏笔

• 2004 年欧洲青年科学家大赛认识 Silvana
• 2005 年 Croma 让他明白"竞赛能改变人生"
• 2019 年 Atlantic 文章呼吁"new science of progress"
• 2020 年 Fast Grants 验证"小团队也能加速科学"
• 2021 年与 Silvana 共同创立 Arc（结婚前几个月）
• 2022 年与 Silvana 结婚
• 2026 年 Arc 发布生物 foundation models（Evo）

这条线特别动人：他人生的最重要的两段关系（妻子 + 共同事业）都源于一场青少年科学竞赛的偶遇。

主题 D：拒绝伪标志的一致性

• 拒绝 IPO（"a solution in search of a problem"）
• 拒绝 Forbes 的"escape from stab city"叙事（"the idea of 'overcoming' anything is crazy"）
• 拒绝把困境用财务术语简化（"the only relevant constraint is the financial one"）
• 拒绝 holacracy 等组织实验（"most of the experiments have failed"）
• 拒绝委员会决策（"if we had relied on a committee... Stripe would not exist"）
• 拒绝把 SF 当成必须去的地方（"don't move to SF"，给 20-30 岁的建议）
• 拒绝"Stripe 是颠覆者"的叙事（"plug into existing rails"）

他几乎每个公开姿态都是对某个流行叙事的拒绝。这不是 contrarian 表演——这是一种深刻的 "the goal seriously" 思维：先把目标想清楚，然后选择不被路径依赖、不被流行词、不被短期社会奖励所污染的最直路径。

六、数据来源

6.1 网络资料（23 篇，全部英文）

6.2 播客访谈（10 期，全部英文）

6.3 已扫描但被排除的内容

• TBPN 节目（多嘉宾）：6454538、7023581、7293425 — 多人 panel 不符合一对一深度访谈标准
• CNBC TV 短片：多个新闻短片 — 时长 <15 分钟，PR 性质
• C-SPAN 听证会：政府发言不算访谈
• 业绩电话会：标准化披露，不是思想性内容

6.4 处理失败需要重试

• Greylock — Micro Pessimism, Macro Optimism (2023) [ID: 6808224] — Podwise 转录两次失败，音频源问题。值得后续手动再试或寻找替代来源（如 YouTube 视频）。

6.5 V2 待补充的高优先级源（master index 列出的）

• The Atlantic 进步研究文章全文（需翻越付费墙）
• Conversations with Tyler 2017 年 4 月一期
• Reid Hoffman Masters of Scale 访谈
• Stripe Sessions keynotes（多年的）
• Patrick 作为采访者：与 Daniel Ek、Brian Chesky、Jony Ive 的对谈
• New Yorker / Bloomberg 长篇人物特写

七、迭代记录

见 changelog.md。

八、反馈与备忘

见 feedback/ 目录。如需补充或修改，可：

1. 直接在本文件中插入 <!-- COMMENT: ... -->、<!-- TODO: ... -->、<!-- EXPAND: ... --> 标注，运行 /refine Patrick Collison
2. 在 feedback/ 文件夹创建 markdown 文件
3. 在对话中直接说

本档案由网络公开英文资料和高质量英文播客访谈综合整理，所有原话引用保留英文以避免翻译失真。叙述部分使用中文。Patrick Collison 本人的言论以引文形式保留原貌；推断或解读已明确标注。