摘要
Karri Saarinen 是 Linear(一家 2019 年成立、2025 年 6 月 $82M Series C 估值 $1.25B、2026 年 8 月以 $2.5B 估值完成 $99M tender offer 而不融 primary、ARR 已过 $100M、自 2021 年起持续盈利且至 2026 年仍保持"负累计 burn"的开发者工作流软件公司)的联合创始人兼 CEO。他是芬兰人,与另两位芬兰联合创始人 Jori Lallo(CDO)和 Tuomas Artman(CTO)于 2019 年 1 月共同创立 Linear——三人 2008 年在赫尔辛基(Helsinki)的初创圈相识,2012 年 Karri 与 Jori 一起被 YC S12 batch 接收(与 Coinbase / Brian Armstrong 同一届),2014 年 5 月他们的 bookmark 工具 Kippt 被 Coinbase 收购(Coinbase 的第一次收购)。
Karri 在 Linear 之前的轨迹是设计圈最被反复引用的"设计师创始人"模板之一:Coinbase 早期设计师(2014–2015,重新设计了 iOS+Android app)→ Airbnb 主设计师(2015–2018,2016 年作为 DLS / Design Language System 主推动者发表了 "Building a Visual Language",成为设计系统领域的奠基性文献)→ 在 Airbnb 内部写了一个 Chrome Extension 来减轻 Jira 的信息过载,被 100 个 Airbnb 同事内部采用——这就是 Linear 的种子。
他的核心信念可以浓缩成几句反复出现的语录。"Quality growth means growth based on something real... not artificially juiced. It's the growth coming from going to the gym. That's more sustainable than using steroids."(2024 年 20VC,定义了他对"硅谷式融资—烧钱—增长"的反命题。)"Small teams always delivered better quality, and faster."(2025 年 "The profitable startup" 散文。)"You shouldn't try to boil the ocean. You need to just boil the pot and get someone in the pot and cook with them."(聚焦的反空想格言。)"Chat is a very weak and generic form."(2025 年 4 月"Design for the AI age",定义了 Linear 在 AI 时代不跟随 chat-first 范式的赌注。)"Linear is the backbone... collecting signals, problems, decisions."(2026 年 3 月 "Issue tracking is dead" 重新定义 Linear 为 AI-era 的产品记忆层。)
他罕见地以"自负盈亏 + 增长缓慢 + 用户少而精 + 团队 99 人覆盖 15+ 国家 + 公司至今总营销支出 $35,000"这种数据点反复挑战硅谷正统。他对 AI 的态度同样反潮流——故意"等"到 2025 年才发布 AI 功能;公开说 chat 是 "weak and generic form";坚信 "you can't outsource the thinking purely to the AI agents";同时他也在 2026 年宣布 Linear 将转型为"AI 时代的 product memory / context infrastructure",覆盖了 77% 的 workspace 已经在用 cloud agents。2026 年 5–6 月,这个转型落地为产品:Code Intelligence(agent 读 codebase 回答问题)与 Coding Sessions(agent 直接从 issue 写代码开 PR——上月 Linear 内部近 700 个 PR 由 agent 合并)。
一、生平时间线
第一章:芬兰背景与设计的初识(~1985–2008)
Karri Saarinen 出生在芬兰,确切出生年未公开(基于职业轨迹推断为 1980 年代中期)。他在 Sequoia Spotlight [01] 与多次访谈里反复把自己的设计敏感性追溯到芬兰的设计传统:
"Back in the day, care for the craft was there and we would like to see more of it.""在过去的年代,对手艺的关心是存在的——我们希望今天也能看到更多。"Karri Saarinen, Sequoia Spotlight
"When I look at things, I could see, 'This could be better.""当我看东西的时候,我能看到——'这本可以更好。"Karri Saarinen, Sequoia Spotlight
Sequoia 在 Linear 的 Spotlight 文章里把他的设计 DNA 直接归到芬兰品牌 Iittala(成立 1881 年的玻璃器皿品牌)和 Artek(Alvar Aalto 1935 年创立的家具品牌)——这两个品牌共同代表的是芬兰极简主义、功能主义、与手工艺主义的合流。
Linear 的视觉语言直接承载了这一传统:UI 用近黑色背景配 Inter 字体;logo 是一个紫色 gradient sphere;色彩参考 Apple Mac Pro 的金属配色。这些不是事后被采编进来的"风格",而是 Karri 在 2010 年(远早于 Linear)就在自己的博客上明确表述的世界观。
"Your Product Needs a Soul"(2010 年 2 月 12 日)——Karri 最早的成型散文
他最早保存下来的散文 [06] 写于 2010 年 2 月——比 Kippt 早一年。这篇文章是理解他后来一切产品决策的基础文献。
核心论点:成功的产品需要刻意的"性格"和"愿景",而不仅仅是功能。强观点驱动的设计、由一个有品味的领导者引导——才能产生"灵魂"(soul)。
他用日本武士刀的隐喻:在古代日本,铸剑是一种复杂的工艺,几乎是宗教仪式。每一把剑都体现着铸剑师的个人风格。产品也应该如此——而不是像无主商品那样存在。
他引用 Paul Graham 的 "Taste for Makers":
"Taste is not mere preference but 'a sense of sophistication a person has.""品味不只是个人偏好——它是一个人所拥有的'精细度感知'。"Karri Saarinen, "Your Product Needs a Soul", 2010
最关键的一句是引自 John Gruber 的:
"The quality of any collaborative creative endeavor tends to approach the level of taste of whoever is in charge.""任何协作型创造性工作的质量,总是趋近于负责人的品味水平。"John Gruber, 转引自 Karri Saarinen, 2010
这句话后来被他在多次 Linear 决策中无声地兑现——他至今亲自审查每一个 Linear 功能的 GA 版本,即使在公司接近 100 人规模时也是如此。
他还在 2010 年这篇文章里就已经反对了"满足所有人"的产品观:
"Chrome vs Firefox example: strong opinions about simplicity and speed create soul. Attempting to please everyone via feature bloat produces hollow experiences.""Chrome 与 Firefox 的对比:对简洁和速度的强主张造就了'灵魂'。试图通过功能堆砌满足所有人,只会产生空洞的体验。"Karri Saarinen, "Your Product Needs a Soul", 2010
这是 Linear 2019 年立项时"highly opinionated"产品策略的祖先。
关键决策:在芬兰留下还是去硅谷?(~2008–2012)
- 背景:2008–2012 年的赫尔辛基已经是一个有活力的初创社区。Karri 此时在 Flowdock(一家芬兰协作工具公司)和 Kisko Labs(赫尔辛基设计/开发工作室)工作,已经接触到 startup-grade 的产品工作。他在这段时间还和 Linda Liukas 于 2010 年共同创立了 Rails Girls——一个用 Ruby on Rails 教女性编程的工作坊,后来成为国际运动。
- 关键假设:硅谷的市场(用户规模、付费意愿、distribution)远大于欧洲;芬兰的设计敏感性反而在硅谷"feature-bloated"的环境里能形成稀缺优势。
- 决策:和 Jori Lallo 一起把 Kippt 申请到 YC S12 batch,搬到旧金山。
- 结果:他从 2012 年起持续在硅谷生活,但保留了"芬兰人"的身份框架——三位 Linear 联合创始人 都是芬兰人,办公室也保留赫尔辛基的精神。
第二章:Helsinki 创业圈与三位芬兰人(2008–2010)
Sequoia Spotlight 把 Linear 的源头追溯到 2008 年的赫尔辛基初创社区——Karri、Jori Lallo、Tuomas Artman 三人在那个圈子里相识。这一点在多个访谈里被三人各自单独证实过。这是一个不到几百人的圈子,三人此后大约 10 年间分别走了不同的路径但保持联系——
- Karri Saarinen:Flowdock → Kisko Labs → 创立 Kippt(2011)→ Coinbase(2014)→ Airbnb(2015)→ Linear(2019)
- Jori Lallo:与 Karri 共同创立 Kippt(2011)→ Coinbase(2014)→ Linear(2019)
- Tuomas Artman:Groupon → Uber(早期工程师)→ Linear(2019)
这种"十年同伴 + 各自独立成长 + 然后再聚" 的模式——Karri 自己在多次访谈里把这视为 Linear 能在 2019 年第一天就开始执行的关键因素——三个人已经知道彼此的工作风格、判断标准、对"什么是好的产品"的内在感知。
Jori Lallo 在 Sequoia Spotlight 上对成立时的氛围有一句被反复引用的话:
"I don't think we ever questioned if there's a need for it, or if we could pull it off.""我觉得我们从来没有质疑过'是否真的需要它',或'我们能不能做出来'。"Jori Lallo, Sequoia Spotlight
Tuomas Artman 的话则点出了三人的共同野心:
"That's one of the reasons why we designed it so well. I hope that it inspires people to build better software.""这是我们为什么把它设计得这么好的原因之一。我希望它能启发人们做出更好的软件。"Tuomas Artman, Sequoia Spotlight
第三章:Kippt 与 YC S12(2011–2014)
Kippt 的诞生与"软启动"
2011 年,Karri 与 Jori Lallo 在赫尔辛基软启动 Kippt——一个"简单的网页书签工具",最初发布在 Hacker News 上。几个月内吸引了 10,000 用户。
2012 年夏,他们带着 Kippt 申请并被 Y Combinator S12 batch 接收——与 Brian Armstrong 的 Coinbase 同一届。这一届的 alumni 网络后来对 Karri 一生的事业都至关重要——Coinbase 的 acquisition、Airbnb 的人脉、最终 Linear 的早期客户都源自这个网络。
Kippt 在 YC 后期 pivot 成一个企业内部 link-sharing 工具叫 Inc(类似 Yammer 或 Convo),客户包括 Teeshan+Lax、Betaworks、Swiggle。但他们在 TechCrunch [11] 公开的复盘里坦白:
"Although our service has been loved by many, we never achieved the growth and the scale that would allow a sustainable future for Kippt.""虽然我们的服务被很多人喜爱,但我们从未达到能让 Kippt 持续下去的增长和规模。"Lallo & Saarinen, 关于 Kippt 关停时
关键决策:被 Coinbase 收购(2014 年 5 月 6 日)
- 背景:Kippt 在产品上小受欢迎但商业上不可持续。同届 YC 的 Coinbase 已经完成了 Andreessen Horowitz 与 Union Square Ventures 的 $30M+ 融资,并开始物色第一笔收购——他们需要的不只是产品工程,更是品牌和设计——一个加密货币公司天然存在的信任赤字。
- 关键假设:Kippt 团队的设计与开发能力可以直接转化为 Coinbase 的产品/品牌团队的核心;Coinbase 已经在 12 人左右的规模,加入早期能产生大的影响。
- 决策:Karri 与 Jori Lallo 同意被收购,并以现金 + 股票形式完成交易——这是 Coinbase 公司历史上的第一次收购。
- 结果:Karri 成为 Coinbase 的早期设计师(founding designer 角色),Jori Lallo 开发了 Coinbase 的 app gallery(开发者平台)。
Karri 自己在 20VC(2024-10)上对当时的判断是:
"With Coinbase, I was there much earlier... when I joined, it was only like 12 people... The main thing I saw there and learned there [was] how much you can do with very small teams... For example, Coinbase Exchange... initially was built by one guy. He was in a room... for maybe six months. And when it went live, it already transacted maybe millions or tens of millions of dollars.""Coinbase 时我加入得很早……我加入时只有大概 12 个人……我在那里看到和学到的最重要的事是——你用极小的团队能做出多少事……比如 Coinbase Exchange……最初是一个人做的。他在一个房间里大概待了六个月。它上线时已经在交易数百万甚至上千万美元。"Karri Saarinen, 20VC, 2024-10-30 [00:05:00–00:05:43]
这个"Coinbase Exchange 由一个工程师 6 个月做完"的故事——是他后来在 Linear 反复用来辩护"小团队优于大团队"的根本经验数据。
Kippt 的教训:"不想商业模式"的建议是错的
2025 年 6 月 First Round In Depth 深访(建档后回收)里,Karri 罕见地系统复盘了 Kippt 给他的教训——问题不在执行,而在从一开始就没想清楚这是一门什么生意:
"The advice on starting a company and not thinking about the business model too much is not good. I think you should think about it a little bit. It's like, what kind of business are you in? Who is going to pay for it?""那种'创业时别太想商业模式'的建议并不好。我觉得你应该稍微想一想:你做的是哪一种生意?谁来为它付钱?"Karri Saarinen, First Round In Depth, 2025-06-05 [00:10:44–00:10:58]
"But I still don't think it's a venture scale type of business. The market is not there to support it now or probably never.""但我至今仍不认为它(Kippt)是 venture-scale 的生意。支撑它的市场现在不存在——很可能永远不会存在。"Karri Saarinen, First Round In Depth, 2025-06-05 [00:11:14–00:11:17]
[事实] 他的具体归因:Kippt 的用户主要是个人,而那个年代人们不太愿意为 web 工具付费;且 Kippt 同时分散服务教师、作者、工程师等多类用户,没有聚焦一个愿意付钱的高价值人群。[推断] 这两条教训在 Linear 身上被逐条反转:Day 1 就定义付费客户(软件团队)、premium 优先不做免费版、"boil the pot" 聚焦——第七章的几个反直觉决策都能追溯到 Kippt 的镜像失败。
第四章:Coinbase 的早期设计师(2014–2015)
"Trust, humanness, modernity, simplicity"
Karri 自己在 [09] "Designing Coinbase" 里回忆 Coinbase 的状态:
"[The team] had been developing the product without any designers or much emphasis on design.""[团队] 之前在没有任何设计师、也没多少设计意识的情况下开发产品。"Karri Saarinen, "Designing Coinbase"
他给自己定义的 Coinbase 设计使命是四个词:Trust, humanness, modernity, simplicity(信任、人性、现代、简洁)。对于一个 2014 年在比特币领域工作的金融服务公司——信任是产品级别的核心特征,不是营销话术。
Twitter Bootstrap 的"hack project"美学问题
他在 YC Startup Podcast(2025-10)上对 Coinbase 加入时的视觉状态做了一次最坦白的回顾:
"For me at Coinbase, thinking about crypto back in 2014, it was a very different time. There's very little trust in the market. And also just because it's a new thing, you don't have that trust built in... we need to make the design really simple and explain it in a simple way so we can break out of this... The visuals, using this kind of standard library [Twitter Bootstrap], made it look like a hack project.""Coinbase 时,2014 年的加密货币世界——那是一个非常不同的时代。市场里几乎没有信任。而且因为它是一个新事物,你也没有内建的信任……我们需要把设计做得真的简单,把它用简单的方式解释清楚,这样才能跳出这个困境……视觉上,用了这种标准库(Twitter Bootstrap),让它看起来像个 hack 项目。"Karri Saarinen, Y Combinator Startup Podcast, 2025-10-14 [00:03:26–00:05:06]
这个决策不是"重做设计",是"重做美学"——Karri 没有重新设计功能,他重新设计了视觉层让 Coinbase 看起来值得托付钱。这种"先解决信任的视觉表达"思维方式后来在 Linear 也反复出现:Linear 的"high opinionated"美学不只是好看,是在向开发者发送一个信号——这个工具的制作者真的在意。
三个月 iOS + Android Apps(2015 年 1 月)
他在三个月内重新设计了 Coinbase 的 iOS 和 Android 应用,两个 app 在同一天上线,都在 app store 拿到 4 星评分。同时他还设计了 web 平台的账户视图、交易显示、Vault 安全系统。他的商业级商户工具吸引了 43,000 个商户,其中包括 Dell 和 Expedia。
但他对自己 Coinbase 工作的最终自评是它不只是 pixel work——而是建设了一个公司的"设计文化":
"His impact extended to 'building the company's design culture' — teaching stakeholder value, hiring designers, improving design consciousness across the bitcoin industry.""他的影响超出了 pixel 工作——'建设了公司的设计文化',向 stakeholder 传授设计的价值,招聘设计师,提升整个比特币行业的设计意识。"Karri Saarinen, "Designing Coinbase"
这是一种"设计师作为文化制造者"的自我定位——后来在 Airbnb DLS 和 Linear 上反复出现。
第五章:Airbnb 与 DLS(2015–2018)
为什么离开 Coinbase 去 Airbnb
Karri 在多次访谈里把这次转换的理由说得很直接——他想要一个更大的设计画布。Coinbase 已经有了品牌;Airbnb 是当时硅谷设计组织最大、最有规模建立 Design Language System 的公司。这是一次他把自己作为设计师的雄心放大的转换。
Brian Chesky 给他留下的最深印象
Karri 在 20VC 2024-10 上反复把 Brian Chesky 作为对他个人影响最深的领导者。当 Harry Stebbings 问他"你最钦佩的 CEO"时:
"If I really had to choose, I would really choose Brian Chesky. I think he's really this power of nature. He can thrive through anything. I think Brian Armstrong is as well. If you put Brian Chesky to do something, like he will really do it, then he will have all the energy to do it. So I would bet on him.""如果我真的要选,我会选 Brian Chesky。我觉得他真的是一种'自然之力'。他什么环境下都能 thrive。Brian Armstrong 也一样。如果你让 Brian Chesky 做一件事——他就会真的做它,他会把全部精力投进去。我会赌他。"Karri Saarinen, 20VC, 2024-10
更具体的影响是关于"经验导向的公司"这个观念:
"I think Brian Chesky is very driven or passionate about the experience, like how does this feel and how we want the company or the product to feel. That was it's really great to see that you can run a company that way, that you can actually focus on the experience and you don't have to focus on some of the other things like metrics.""我觉得 Brian Chesky 非常驱动或痴迷于体验——它感觉如何、我们希望公司或产品感觉如何。能看到原来可以这样运营一家公司——真正聚焦于体验、不必去专注那些其他东西比如 metrics——这件事本身对我影响很大。"Karri Saarinen, 20VC, 2024-10
这种"experience-first not metrics-first"的运营哲学——后来直接成为 Linear 的核心反命题之一。Linear 不做 A/B 测试,没有 OKR,没有传统的"data-driven product development"——根源就是 Karri 在 Brian Chesky 旁边看到这是可能的。
关键决策:DLS(Design Language System)作为 Airbnb 的内部基础设施(2015–2016)
- 背景:2015 年的 Airbnb 已经增长到几十名设计师、跨多个平台(iOS、Android、Web)、面对几十个国家。设计部门已经没法用临时的、个人化的方式协调——产品视觉到处是不一致。
- 关键假设:(1) 视觉语言就像普通语言一样——共享性是必要条件,否则误解会传染。(2) 不能用"风格指南"这种被动文档解决——必须是一个主动的、有生命的组件系统。(3) 必须把它做成一个能被外部讨论的公开 artifact——这样它在内部就有"权威"。
- 决策:Karri 与一小组团队"清空日历、在外部工作室封闭工作",从 iOS+Android 原生平台入手,设计完整的 DLS。2016 年 4 月 17 日发布。
四个指导原则被他写进 [20] 的 Medium 文章(成为设计系统领域的奠基文献):
- Unified——每个组件对整个系统是正贡献的。
- Universal——全球可用、无障碍。
- Iconic——专注的设计与功能。
- Conversational——运动语言清楚地与用户沟通。
最被引用的一句关于"视觉语言就像语言":
"Visual language is like any other language. Misunderstandings arise if the language is not shared.""视觉语言就像任何其他语言一样。如果语言不共享,误解就会出现。"Karri Saarinen, "Building a Visual Language", 2016
而最重要的概念创新是把组件视为"living organism 的元素"——而不是被动的"原子"——每个组件都有自己的"个性"、自己的边界、自己独立的演化。
Chrome Extension 原型:Linear 的真正种子(2018 年前后)
在 Airbnb 工作期间,Karri 因为对 Jira 的不满在自己的浏览器里写了一个 Chrome Extension——专门用来"减少 Jira 的信息过载"。他自己在 [05] 的"Starting Linear"博文里讲:
"I wrote a Chrome Extension at Airbnb to 'reduce some of the information overload of JIRA.' ~100 internal Airbnb users — unexpected internal adoption.""我在 Airbnb 写了一个 Chrome Extension 来'减少 Jira 的信息过载'。大约 100 个 Airbnb 内部同事开始用——意外的内部采用。"Karri Saarinen, "Starting Linear", 2019-04-28
这是 Linear 的原型。一个真实的痛苦 → 一个真实的解决 → 一个真实的内部分布——这就是后来 YC "make something people want" 在他身上的具体兑现。
第六章:Chrome Extension 原型与 Linear 的诞生(2018–2019)
三位芬兰人共同的"Jira 痛苦"
Karri 在 [05] 里把 Linear 的成立动机定义为一种跨公司的共同抱怨:
"Teams start informally; eventually adopt JIRA. Then 'team motivation sank and how much time the team wasted arguing against it.""团队一开始非正式运作;最终采用 JIRA。然后'团队的动力下沉了——团队浪费多少时间在反对它'。"Karri Saarinen, "Starting Linear"
"I was surprised how badly they were designed. It felt like they were not designed for humans.""我惊讶于它们设计得有多糟糕。它们感觉不是为人类设计的。"Karri Saarinen, "Starting Linear"
"Often companies would overly customize the workflows to be unnecessarily complex.""很多公司会过度定制 workflow,把它弄得不必要地复杂。"Karri Saarinen, "Starting Linear"
Jori Lallo(Coinbase)和 Tuomas Artman(Uber)都报告了相同的体验。这种"在三家不同公司、由三个不同的工程团队、用同一个工具受同样痛苦"的事实——构成了 Linear 立项的最坚实理由。
关键决策:2019 年 1 月成立 Linear
- 背景:Jori Lallo 在 2018 年休假(sabbatical)期间,决定向 Karri 和 Tuomas 提出"我们做我们一直想要的开发工具"的想法。三人之前已经花了超过一年在做非正式的用户研究——和他们认识的工程师、PM、设计师聊天,了解每个公司是怎么"绕过 Jira"的。[事实] In Depth 2025 补充了这一年验证期的具体形态:三人每周三在附近一家酒吧碰头,带着电脑坐角落桌,过一遍"这周做了什么、学到什么、试了什么",再定下周的事 [00:41:27–00:41:45]。
- 关键假设:(1) Jira 在过去 20 年里基本没有变——这是一个稳定的、被锁死的市场,市场不是"快"的胜利,是"质量"的胜利。(2) 软件开发实践已经远离了纯 Agile/Scrum——但工具还停在那里。(3) 不需要 PM——一个设计师 + 2 个工程师的小团队可以驱动一个项目。
- 决策:2019 年 1 月正式成立 Linear。
- 结果:2019 年 4 月 18 日 Medium 上 "Announcing Linear" 帖子发布、4 月 28 日 Karri 的个人 essay 发布。
最关键的策略性句子被 Karri 写进了 essay:
"Most companies are no longer run by pure Agile or SCRUM practices, they have evolved their processes, but the tools haven't.""大多数公司已经不再用纯粹的 Agile 或 SCRUM 实践运作了——他们已经演化了自己的流程,但工具没有跟上。"Karri Saarinen, "Starting Linear", 2019-04-28
愿景一句话:
"Create the new standard for a tool that would support modern software development and issue tracking practices that actually in use in some of the top companies.""建立一个新的标准——一个能支持顶级公司实际使用的现代软件开发和 issue tracking 实践的工具。"Karri Saarinen, "Starting Linear", 2019-04-28
第七章:Linear 的"少即是多"路径(2019–2022)
Series Seed → Series A → 持续盈利
- 2019 年 11 月:Sequoia 领投 $4.2M Seed Round
- 2020 年 6 月 30 日:Linear 公开发布("open for all")
- 2020 年 12 月:Sequoia 领投 $13M Series A
- 2021 年 7 月:Linear 实现盈利——也就是公开发布后约 12 个月
至 2023 年初,Linear 已经达成一个硅谷少见的状态:
- 30 人 团队
- 2021 年起持续盈利
- 负累计 burn——账上的现金比总融资额还多
- 0.5% 的收入用于营销
- 4 人的销售团队
- 公司创立以来的总营销支出:$35,000——其中大部分是两个播客赞助(包括 20VC 的一次)
Linear 通过 Lenny Rachitsky 2023 年 10 月的播客访谈第一次完整公开地讲述了"how Linear builds product"——这是当时硅谷最被讨论的反 SaaS 增长指南之一。
"Premium 优先"的定价策略——刻意不做免费版
Linear 早期没有免费版。这是一个极其反直觉、几乎所有 PLG SaaS 都不会做的决定。Sequoia Spotlight [01] 里 Karri 解释这种"premium 优先"路径:他们要先把目标客户群(最有动机的工程师团队)定义清楚——这些人愿意付钱,因为他们对自己的工具有审美和效率要求。
后来他们才加了一个 250 issue 限额的免费 tier——但底层策略永远不是 "free first"。
[事实] 在敲定这套定价之前,Linear 有一段约一年的 private beta(他们用 waitlist 分小批放人进来、逐个 email 用户、专修用户抱怨的具体问题)。价格本身是用一个 "pay-what-you-want" 的滑块试出来的——让 beta 用户自己选择愿意付多少。据 Karri 在 2023 年 EO 访谈里的复盘,这个滑块给了他们"在做真正有价值的东西"的信心,是后来敢走"premium 优先、不靠免费引流"路线的经验依据。[推断] 这也解释了"premium 优先"为什么不是一个先验的意识形态选择,而是一次真实的价格发现——先用最低承诺的机制测出付费意愿,再据此定策略。
"Boil the Pot, Not the Ocean"——聚焦的语言
Karri 在 20VC 2024-10 上用了这个后来反复被引用的隐喻:
"You shouldn't try to boil the ocean. You need to just boil the pot and get someone in the pot and cook with them.""你不应该尝试把整个海洋煮沸。你应该先把锅里的水煮沸,把一个人放进锅里,和他一起煮。"Karri Saarinen, 20VC, 2024-10
这种"先把一个小但完整的群体真正服务好、再扩"的路径——是 Linear 不做"广告轰炸 + 增长黑客"路径的根本原因。
在更早的 2023 年 EO 访谈里,他把同一条"聚焦"纪律从产品范围延伸到了 founder 在最早期的时间分配——创业最初该几乎无视 legal、accounting、conference networking 这些次要事务,一切都为"做出人们真正想要的东西"让路:
"Whenever you think about doing something, ask yourself: do we really need to do this now? If it doesn't advance the product or get you more customers, it is something you can avoid or put off.""每当你想做某件事时,问自己:我们现在真的需要做这个吗?如果它不能推进产品、也不能带来更多客户,那它就是你可以避免或推迟的事。"Karri Saarinen, EO, 2023-11-18 [00:12:34]
[推断] "boil the pot" 是对产品的聚焦,这一句是对创始人自己时间的聚焦——两者是同一纪律的两个层面。这也是档案里最早期(2023)视角对"say no"哲学的一次落地:不是等公司大了才学会拒绝功能,而是从第一天起就把"次要但看起来该做的事"(合规、社交)主动推迟。
"We Don't Do A/B Tests"
Linear 的产品决策方式在 Lenny × Linear 2024 的 "How Linear Builds Product" 一文 [22] 里被完整披露——他们不做 A/B 测试,没有 OKR,没有 per-project metric:
"We don't do A/B tests. We validate ideas and assumptions that are driven by taste and opinions, rather than the other way around.""我们不做 A/B 测试。我们用品味和观点来验证 idea 和假设——而不是反过来。"Karri Saarinen, "How Linear Builds Product"
Karri 在 Every 的 AI & I(2026-04)上把这点讲得更哲学化:
"I never personally believed in this A/B testing and data-driven product development, which I think could work well for agents, but I think it doesn't work for all kinds of products. And I also think the best products are not necessarily being built that way. You still need the human touch of what is interesting or what would make this good.""我个人从来不相信 A/B 测试和数据驱动的产品开发。我觉得这对 agents 可能有用,但不适用于所有类型的产品。我也认为最好的产品并不是这样建造的。你仍然需要那种人的判断——什么是有意思的、什么会让这个东西更好。"Karri Saarinen, Every / AI & I (Dan Shipper), "Most SaaS Companies Got AI Wrong. Linear Waited.", 2026-04-01
2024 年 9 月 Product School 访谈补上了这个立场的正面版本——不做 A/B、不爱发问卷,那到底看什么:
"We don't love making surveys all the time. It feels kind of annoying and you don't always capture all the people in the surveys. I think a lot of times we rely on what we hear from customers and users.""我们不喜欢老是发问卷。那有点烦人,而且问卷也总会漏掉一部分人。很多时候我们依赖的是从客户和用户那里直接听到的东西。"Karri Saarinen, Product School, 2024-09-04 [00:19:26–00:19:43]
[事实] 他列出的实际"仪表盘":客户/用户发来的直接消息 + 在 Twitter 上搜 "Linear" 的自发声量;低 churn(长期满意度的代理指标);每个组织内的活跃用户数(防"买了不用");以及在特定阶段创业公司细分里的渗透率/饱和度——"each YC batch or each seed or Series A type of batch or segment"(他自己承认这类数据不总是干净,但他们想看到 Linear 在每个这样的 segment 里都被采用)[00:20:51–00:21:23]。[推断] 最后一项是 "quality growth" 的量化镜像:不追总量增长,追"下一代公司里被默认采用"的比例。
单一 Head of Product(不是单个 PM)
Linear 内部的组织结构在硅谷也极其少见——他们有一个 Head of Product(Nan Yu),但没有"per-feature PM":
"Instead of having a PM for every team or area of the product, we like to see that the broader team collectively thinks about the product and how features are delivered, not just the PMs.""不是让每个团队或产品领域都有一个 PM——我们更希望整个团队集体思考产品和功能如何交付,而不是只让 PM 想。"Karri Saarinen, "How Linear Builds Product"
"One of the unique aspects to Linear is that we expect the project team to be the PM.""Linear 独特的一面是——我们期望项目团队本身就是 PM。"Karri Saarinen, "How Linear Builds Product"
项目团队大约是 1 名设计师 + 2 名工程师,由工程师或设计师领头(绝不是 PM)。
工程师必须有"taste"
Karri 在 YC Startup Podcast(2025-10)上把 Linear 的招聘哲学讲得很直接:
"We are always looking for people who have opinions or they have some kind of product sensibilities or they have some kind of product taste. This is like explicit some kind of curiosity. It's like, I think there's a good way of doing things and there's a bad way of doing things. This feature doesn't feel good. This feature does feel good. So we're just looking for those people that can think and use their own judgment.""我们一直在找有观点的人,或者有某种产品敏感性、产品品味的人。这是一种非常明确的好奇心——'我觉得有一种做这件事的好方式,也有一种坏方式。这个 feature 感觉不好,那个 feature 感觉好'。我们就是在找那些能思考、能用自己判断的人。"Karri Saarinen, Y Combinator Startup Podcast, 2025-10-14
Linear 的招聘有几个反硅谷的特征:
- 不喜欢 10 年 Google 背景:他直接说"在 Google 工作 10 年的人,他们通常被锁在某个领域里、没机会思考其他事情"
- 付费 work trial:1-5 天的付费试用,让候选人加入真实项目
- Retention 数据:截至 2024 年文章发布时,Linear 历史上只有 2 个人离开过
在最早的 Lenny 2023-10 访谈里(该集 2026-07 才被 Podwise 处理完成、补入档案),Karri 把 work trial 讲成一个双向去风险装置——不只是公司考察候选人,更是候选人考察公司。候选人拿到真实 codebase 权限、构建一个真实版本、最后 present 自己的工作;而"codebase 本身的干净程度"在这个过程中成了 Linear 的人才磁铁:
"I think a lot of engineers also like that they see the code base and they're like, oh wow, this is like really clean and it's not like some kind of spaghetti code type of thing or situation. So I think it helps the candidates as well understand what are they signing up for, which I think can be like very risky sometimes with especially with startups — it's really hard to tell how the startup is operating just from the interviews.""很多工程师也喜欢这一点——他们看到 codebase 会想:哇,这真的很干净,不是那种 spaghetti code 的情况。所以我觉得这也帮候选人搞清楚自己到底在签约加入什么——尤其对 startup 来说这一点风险很大,光靠面试你其实很难判断一家 startup 究竟是怎么运转的。"Karri Saarinen, Lenny's Podcast "Inside Linear", 2023-10-08 [01:03:55]
[事实] 配套细节:work trial 可协商安排在周末或假期以尽量不干扰候选人的现职;据 Karri 说历史上几乎没有人拒绝过 work trial,且事后候选人普遍认为"对要加入的公司有了远更真实的感知"[01:04:55–01:05:21]。他最喜欢的面试问题是"你最自豪的是什么、为什么"——测的是候选人的 values 和思考方式,同时让候选人展示强项而非防守弱点 [01:34:24]。
[推断] "干净 codebase 吸引高质量工程师"把他的 craft 论点闭合成一个自增强循环:quality 不只通过产品吸引用户(quality growth),也通过代码库本身吸引维护 quality 的人——这是"质量作为增长策略"在人才市场上的镜像,档案此前只在用户/客户侧记录过这个机制。
第八章:"质量增长"与 AI 时代的克制(2022–2026)
Series B $35M(2023 年 9 月,Accel 领投,估值 $400M)
到 2023 年 9 月,Linear 已经盈利、负累计 burn、营销支出极低——这种业务状态使 Accel 的 Series B 成了一次罕见的"创始人有完全 leverage"的融资。Karri 在 20VC 上讲了他对这次 fundraise 的具体方式:
"I make a separate Google Doc for each investor and say like, hey, here's a memo. At the end of the memo, there's some questions for you that I would like you to think about. And then they should write them down and then we should talk about those answers. And to me, what it gives me, it's a little bit like simulating the relationship of working with them. I can see how they think and what their style is.""我为每个投资人做一个单独的 Google Doc,说——这是 memo,最后有一些问题给你思考。请你写下答案,然后我们再讨论这些答案。这对我来说像是在模拟和他们一起工作的关系——我能看到他们的思考方式和风格。"Karri Saarinen, 20VC, 2024-10
这种"给投资人布置作业"的做法——是把"我不需要这笔钱"这个 leverage 用到最满的具体战术。
"Quality Growth"——Karri 的核心反命题
20VC 2024-10 上 Karri 用了一个被反复引用的隐喻把"硅谷增长"重新分成两类:
"With Linear, we've been growing much more in this sustainable way, or I like to call it more like a quality growth way. What it means is that the growth is based on something real that is working in the company. And it's not something that is artificially juiced or it's not like the growth is on steroids. It's like the growth is coming from the fact that you go to the gym. That's more sustainable than using steroids.""Linear 一直以一种更可持续的方式增长——我更愿意叫它'质量增长'。意思是增长基于公司里真的有效的某种东西,不是人为推动的、不是'增长用了类固醇'。是那种'你去了健身房'式的增长——比用类固醇可持续得多。"Karri Saarinen, 20VC, 2024-10
"For us, like the quality growth has always been the product that we almost spent no money in marketing. We didn't even have sales for the first three years. We just wanted to see that the growth is coming from the fact that people find our product better than anything else out there.""对我们来说,质量增长一直是——产品本身。我们几乎没花钱营销。前三年甚至没有销售团队。我们想看到的是:增长来自于人们觉得我们的产品比其他一切都更好。"Karri Saarinen, 20VC, 2024-10
"What If You Don't Need to Raise?"——融资 leverage 的本质
"What if you don't need to raise the round? Like what if you can sustain without raising a round or you raise some round, and then you just try to focus on making the business function work that actually try to make money and not spend that much money. And then you're no longer kind of tied to those metrics because you don't need to raise.""如果你不需要融下一轮呢?如果你可以不融下一轮地继续生存呢?或者你融一轮、然后专注于让业务真正赚钱、不花太多钱?这样你就不再被那些指标绑住——因为你不需要融。"Karri Saarinen, 20VC, 2024-10
"Being profitable gets you the highest leverage you can have in terms of when it comes to talking to investors because it gives you the leverage of not needing them. And so that's always the highest leverage if you can walk away from any deal.""盈利能给你和投资人对话时最高的杠杆——它给你'不需要他们'的杠杆。这永远是最高的杠杆:你可以从任何一笔交易里走开。"Karri Saarinen, 20VC, 2024-10
"The Profitable Startup"——具体的数据指标
[16] 2025 年 2 月发表的 "The profitable startup" 是 Karri 对硅谷"烧钱增长"教条最系统的反驳。最关键的数据点是他给出的人均收入指标:
"Target revenue per employee: $500K–$1M.""目标人均收入:$50 万–$100 万 美元。"Karri Saarinen, "The profitable startup", 2025-02-21
"Ten people before product-market fit should be your ceiling, not your target.""在 product-market fit 之前 10 人应该是你的上限,不是目标。"Karri Saarinen, "The profitable startup", 2025-02-21
"Small teams always delivered better quality, and faster.""小团队总是交付更好的质量、更快。"Karri Saarinen, "The profitable startup", 2025-02-21
公开数据点(来自 Karri 的 Twitter/Threads):
- 2023 年(Linear 4 周年):30 人、自 2021 年起盈利、负累计 burn
- 2024 年(Linear 5 周年):50 人、100% remote 覆盖 16 国家、为 tens of thousands 客户运营产品组织、Forbes top 50 AI 公司里有 66% 用 Linear、总融资 $52M、盈利、负累计 burn、0.5% 的收入用于营销、销售团队 4 人
- 2025 年(Linear 6 周年):70 人、10,000+ 客户、企业 ARR 增长 2000%、第一个 Fortune 100 客户、NRR 140%+、净运营收入增长 1400%、负累计 burn 仍在持续
关键决策:Series C $82M / $1.25B 估值(2025 年 6 月,Accel 领投)
- 背景:Linear 已经在持续盈利、不需要钱。但 AI 浪潮在 2024–2025 把整个 dev tools 市场推到一个新的阶段——Linear 需要资本来支持 AI 与 agent 平台的开发,以及推进企业级销售。
- 关键假设:(1) AI 不是 issue tracking 的对手,是它的赋能者——Linear 应该把自己变成 AI 时代的"产品记忆层"。(2) 这一波技术周期需要正确的人才布局、可能比之前的"小团队"路径更宽。(3) 不能因为有钱就放弃质量第一的原则。
- 决策:接受 Accel 领投的 Series C $82M,估值 $1.25B。
- 结果:[15] 2025-06-10 发布的 "Building our way" 公告——Karri 把这次融资同时定义为公司的 AI 时代承诺:发布 Linear AI + Linear for Agents,并明确声明"no A/B testing, no SEO, no metric chasing"。
首次 Tender Offer:给团队流动性(2025 年 8 月)
[29] Series C 两个多月后,Linear 完成了公司历史上第一次 tender offer——现任与前员工可以按 Series C 同一 $1.25B 估值出售一部分 vested options(只能卖一部分,保留多数持仓)。Karri 给出的理由不是"奖励",而是留人机制的设计:
"People leave not because they lose motivation but because their needs change.""人离开不是因为失去了动力,而是因为他们的需要变了。"Karri Saarinen, "Giving our team liquidity through Linear's first tender offer", 2025-08-27
配套的长期留任体系还包括:10 年期权行权窗口、early exercise、4 年后一个月带薪 sabbatical、equity refresher。公告未提及 IPO 计划或保持私有的意图。[推断] 在"持续盈利、不急于退出"的路径下,tender offer 是必要的压力阀——它把"等流动性"从离职理由清单上划掉,与他"小团队长期主义"的组织哲学自洽。
第二次 Tender Offer:$99M / $2.5B 估值,拒绝"融资成习惯"(2026 年 8 月)
[30] ⚠️PR(公司公告,Karri 署名,linear.app/now 与个人站同步发布)。Series C 一年零两个月后,Linear 完成第二次 tender offer:现任与前员工按 $2.5B 估值出售部分 vested equity,总额 $99M;老股东 Accel、01A 跟进,新进入者 Salesforce Ventures 与 S32。这是一次只有 secondary、没有 primary 的交易。
- 背景:2026 年初 ARR 已过 $100M;付费公司超过 40,000 家;NRR 177%(2025 年 1 月为 140%+)。Salesforce、Figma 均于 2026 年成为客户——Salesforce 是先 organically adopt、后 Salesforce Ventures 入股。agent 侧:95% 的付费 workspace 装了 agent,agent 创建的工作占比一年内从 3% 升到 50%,2026 年以来附带 PR 的 issue 数量增长 7 倍。
- 关键假设:公司 cashflow positive、账上现金多于历史融资总额,primary 融资没有用途、只会稀释;但估值标记与团队流动性仍需要更新。
- 决策:只做 $99M 二级交易,以当前估值给团队与早期投资人流动性,不接受新的 primary。
- 结果:估值从 $1.25B(2025-06 / 2025-08)到 $2.5B;同时首次公开承认"小团队"要扩编——30+ 个开放职位。
"Fundraising can become a habit. A company raises a round, then starts preparing for the next. Sometimes that is exactly what the business needs. This time, it wasn't what Linear needed. We are cashflow positive, and have more cash in the bank than everything we have raised to date.""融资会变成一种习惯。公司融一轮,然后就开始准备下一轮。有时那正是业务需要的。这一次,它不是 Linear 需要的。我们现金流为正,账上的现金比我们迄今融到的全部还多。"Karri Saarinen, "Sharing Linear's growth with the people building it", 2026-08-26
"We have resisted many of the usual company-building playbooks and still believe a small, focused team can do more than people expect. But the scope of Linear has expanded, as have the needs of the customers relying on it. Quality has always been central to how we operate, and maintaining that standard at scale means growing the team.""我们抵制了许多常见的公司建设 playbook,并且仍然相信一支小而专注的团队能做到超出人们预期的事。但 Linear 的范围扩大了,依赖它的客户的需求也扩大了。质量一直是我们运营的核心,而在这个规模上维持这个标准意味着扩大团队。"Karri Saarinen, "Sharing Linear's growth with the people building it", 2026-08-26
[事实] 配套的长期留任体系被命名为 Linear Infinity:equity refresher、满四年全薪休一个月、之后每两年再休一个月。[推断] 两点值得标记:(1)这是"盈利=leverage"策略的第二次兑现——leverage 这次不是用来"少融",而是用来"不融",只给流动性;(2)"maintaining that standard at scale means growing the team" 是档案里第一次见他把"扩编"说成质量的必要条件而非威胁——与 §二"Small Teams Always Deliver Better Quality, Faster"形成一个需要持续观察的张力(他给的调和是 "carefully expand",但 30+ 个职位对一家约 100 人的公司是 30% 级别的增量)。二手佐证:2026-09-03 的 20VC 新闻 roundup(episode 8815267)亦提及 Linear 的 $2.5B 标记。
"Why Is Quality So Rare?"——Config 2025 的关键演讲(2025 年 5 月)
Karri 在 Figma Config 2025 的演讲是他公开发表过最被引用的设计哲学文本之一 [17]。核心观点:
质量为什么稀有? 因为技术进步反复制造一个循环:速度和效率挤走手艺。最终人们怀念质量,重新追求它,复兴开始——这个循环每次重大创新都会重演。
"AI attempts to separate the maker's judgment and taste from creation itself.""AI 尝试把制作者的判断和品味从创造本身中分离出去。"Karri Saarinen, "Why is quality so rare?", 2025-05-27
这是空前的——外包的不只是劳动,还有思考与关心。
他给"质量"的三要素定义:
- Belief——相信质量重要
- Care——对用户深度关注
- Craft——技能与品味
"Quality creates gravity — it pulls people in.""质量创造引力——它把人吸引过来。"Karri Saarinen, "Why is quality so rare?", 2025-05-27
"Design for the AI Age"——Linear 的 AI 战略立场(2025 年 4 月)
[18] 2025 年 4 月 7 日的散文 "Design for the AI age" 是 Karri 对 AI 时代设计师角色的最系统性论述。最关键的一句:
"[Chat is] a very weak and generic form.""[聊天] 是一种很弱很泛化的形式。"Karri Saarinen, "Design for the AI age", 2025-04-07
Linear 不会变成另一个 ChatGPT-style 工具。他用了"workbench"(工作台)作为隐喻——AI 不是替代了工作环境,而是工作台上的一个新工具。木匠的工作台是结构化的环境,提升每件工具的效用——AI 在 Linear 里应该这样。
"Designers must work forward to understand capabilities, rather than backwards from user needs.""设计师必须向前工作——去理解能力,而不是从用户需求反推。"Karri Saarinen, "Design for the AI age", 2025-04-07
这种"从能力出发设计"的思路本质上颠倒了传统 UX 方法论。
"Most SaaS Companies Got AI Wrong. Linear Waited."(Every / AI & I,2026 年 4 月)
Karri 在 Every 的 AI & I(Dan Shipper,2026-04-01;此前误标为 "Lenny 2025 后期",本轮据 Podwise 元数据与他本人 /interviews/ 页校正)访谈里把 Linear 的 AI 策略更具体地讲了出来——注意这集晚于 2026-03 的 "issue tracking is dead" 宣言,是宣言之后的第一次长访谈:
"I think we shouldn't go fast in deciding things or just speed running the decisions or not even doing decisions. I think some people do it now where they just have an idea, then they build it. And now we're all looking at this idea that no one really knows why it exists. And should we even do it?""我觉得我们不应该在决策上跑得太快,或者用'速通'的方式做决策,甚至跳过决策这一步。现在有些人就是有个想法、然后就做了。然后我们都在看着这个东西——没人真的知道它为什么存在。我们到底应该做它吗?"Karri Saarinen, Every / AI & I (Dan Shipper), "Most SaaS Companies Got AI Wrong. Linear Waited.", 2026-04-01
"Once we commit on the thing or the fix or the project, then I want it to improve fast. I want the loop to be fast to actually work on the problem. But I don't want the problem finding to be fast. You should take the time to find the right problem and the right approach for the problem.""一旦我们承诺做某件事、某个修复、某个项目——我希望它能快速改进。我希望循环要快、真正在问题上工作。但我不希望'找问题'这件事快。你应该花时间找到正确的问题和正确的方法。"Karri Saarinen, Every / AI & I (Dan Shipper), "Most SaaS Companies Got AI Wrong. Linear Waited.", 2026-04-01
这是他对"AI 应不应该用慢策略"最完整的回答:决策要慢,执行要快。
"Self-driving SaaS"——三级自治框架(2025 年 10 月)
[28] 2025 年 10 月 22 日的散文 "Self-driving SaaS: When software runs itself" 是"issue tracking is dead"(2026 年 3 月)的思想前篇。核心命题:
"Real change comes when AI stops requiring constant input and instead proactively moves work forward.""真正的变化发生在 AI 不再需要持续输入、而是主动推进工作的时候。"Karri Saarinen, "Self-driving SaaS", 2025-10-22
他借自动驾驶分级给出三级自治框架:Level 1 Assistive(系统只建议,不动手——如给 issue 建议 assignee/label,类似车道偏离预警)→ Level 2 Moderate Autonomy(系统先动手、人审阅修正,可设条件规则选择性自动应用——"The car does most of the driving, but you can still nudge the steering wheel or tap the brake.")→ Level 3 Full Autonomy(系统独立处理完整 issue 工作流,人只评估最终结果)。但他同时给控制权画了边界:
"You, the user, are still in control—you decide which projects to pursue and where to allocate resources.""你——用户——仍然掌握控制权:你决定追求哪些项目、资源配到哪里。"Karri Saarinen, "Self-driving SaaS", 2025-10-22
第九章:Issue Tracking is Dead(2026)
关键决策:2026 年 3 月,宣布"issue tracking is dead"
- 背景:2026 年初,77% 的 Linear workspace 已经在使用 cloud-based AI coding agents(他在 TBPN 2026-03-27 上自己补的口径:统计的是 cloud agent,"we don't track the local agents" [02:32:20–02:32:30])。"修一个 bug"的工程师工作从几天/几小时压缩到几分钟。Karri 决定 Linear 应该公开重新定义自己。
- 关键假设:(1) 当 AI agent 5 分钟修一个 bug 但你的 bug reporting 流程需要一周时——bottleneck 已经从"build"转到"deciding what to build"。(2) Linear 的真正价值不在"track issues",而在作为决策层的 product memory / context infrastructure。(3) 围绕这点重新定义 Linear 比让它变成 "AI-first ticketing" 更有持久价值。
- 决策:3 月 26 日通过 TBPN Digest 公开宣布 [25] "Issue tracking is dead"——Linear 转型为AI 时代的 context infrastructure for coding agents。
- 结果:Linear 的 product 战略转向三个支柱: 1. Aggregating context:把 bug report、客户反馈、产品目标、系统状态聚合成一个统一的、agent 可访问的层 2. Building agency:自驱动系统能检测问题、自动 debug、只在必要时升级到人审核 3. Proprietary agent layer:把现有 agents(Devin、Cursor 等)包裹在更深的平台集成里
[事实] 这个判断的书面雏形早于宣言近三个月:2026-01-03 他在 X 上发的 "The disappearing middle of software work" [36] 已写下"软件工作的中段(把意图翻译成代码)正在变薄,重心移向前端的意图成形与末端的 review / test / release":
"In this era, directing and managing agent work becomes the craft. Writing code is less like constructing a solution and more like setting up the conditions for a good solution to emerge.""在这个时代,指挥和管理 agent 的工作成了那门手艺。写代码不再像是在构造一个解,更像是在搭建让好的解浮现出来的条件。"Karri Saarinen, "The disappearing middle of software work", 2026-01-03
最经常被引用的一句(本轮以 TBPN 2026-03-27 原始音频校正,此前据 TBPN Digest 转引;Digest 报道稿日期为 03-26、Podwise 节目集日期为 03-27,一天之差未进一步核实):
"If the agent can fix the bug in five minutes, but your bug reporting process or the process of running through that process takes a week, it starts to look pretty silly that the actual work takes a lot less time than actually running through the whole process.""如果 agent 五分钟就能修好 bug,但你的 bug 报告流程、或者走完那套流程要一周——那就显得挺蠢的:真正的工作花的时间远少于走完整个流程。"Karri Saarinen, TBPN, 2026-03-27 [02:22:27–02:22:40]
最深的战略洞见——⚠️转述校正:TBPN Digest 文中 "When execution becomes inexpensive, the bottleneck shifts from building to deciding what to build" 一句是编辑概括,原始音频里没有这句;他在节目里的原话是:
"Because I think the problem today starts to become if you can build anything fast, the problem is not building things anymore. The problem is deciding what to build and why, and even understanding what is the problem.""因为我觉得今天的问题开始变成:如果你能很快造出任何东西,问题就不再是造东西了。问题是决定造什么、为什么造,甚至是搞清楚问题到底是什么。"Karri Saarinen, TBPN, 2026-03-27 [02:26:03–02:26:10]
宣言背后的两个具体立场(TBPN 2026-03-27 原始音频,仅采用可归属 Karri 的 segment)
[事实] 他把这句口号的意图说成"让整个行业照镜子"——"we should all look in the mirror and say, what are we doing here? ... Do we still need to do these things? From our perspective, we don't" [02:23:13–02:23:23]。镜子照出来的两件事:
第一,拒绝把在位者的流程复杂度搬进来。他把问题的一半归到 legacy 厂商身上——"these legacy platforms vendors have kind of like created this kind of idea that it should be complex" [02:21:46]——并给了一个迁移谈判的实例:
"We have customers coming to us and like, hey, we want to switch to Linear. But then they said like, oh, we have this like bug tracking process, which is 40 different states. And the bug report needs to have this 30 different custom fields. And then they asked, like, can you do that? And our answer is like, I don't think we should or you shouldn't either. Like the point of moving to Linear is actually rethink some of this stuff.""有客户来找我们说,嘿,我们想换到 Linear。然后他们说,哦,我们有一套 bug 跟踪流程,有 40 个不同的状态,bug 报告还要填 30 个不同的自定义字段。然后他们问:你们能做到吗?我们的回答是:我不觉得我们应该做,你们也不应该。换到 Linear 的意义恰恰是重新想一想这些东西。"Karri Saarinen, TBPN, 2026-03-27 [02:22:00–02:22:25]
[推断] 这是第六章 "Starting Linear" 里那句 "companies would overly customize the workflows to be unnecessarily complex"(2019)在 2026 年的销售现场版——七年里立场没有变,变的是现在有 agent 作为论据:当 agent 把 40 个状态压成"人只负责 verify"的一个状态 [02:25:09],保留那 40 个状态就失去了最后的理由。
第二,对"把 SaaS 全扔掉、agent 原生重来"的叙事保持怀疑。
"Everyone's running their agent and now they have 10 agents running. Then they put them on a Kanban board, which Linear has, and then it's like, oh, now I invented the agent orchestration. No, you invented the Kanban board, which has been around for 30 years, but you just put the agents on a Kanban board. Kudos to connecting those two topics, but it's not like that. Some things don't need to be reinvented.""每个人都在跑自己的 agent,现在有 10 个 agent 在跑。然后他们把这些 agent 放到一块 Kanban 板上——Linear 就有——然后就说,哦,我发明了 agent orchestration。不,你发明的是 Kanban 板,它已经存在 30 年了,你只是把 agent 放到了 Kanban 板上。把这两件事连起来值得称赞,但事情不是那样的。有些东西不需要被重新发明。"Karri Saarinen, TBPN, 2026-03-27 [02:30:31–02:30:55]
[事实] 同一段里他对"大家都在做同一个东西"的感受给了一个历史类比:很多人同时注意到同一组 primitives,就像数据库刚被发明时 "why is everyone making databases";共享 primitives 不意味着相同结果 [02:29:10–02:30:06]。[推断] 这与 §二"几乎与所有在位者反着做"互为镜像:他反对在位者的流程复杂度,也反对挑战者的"从零重来"浪漫——两边都被同一把尺子量:这个状态、这个抽象是否仍然必要。
Linear 的"workbench"模式真正落地
[相关播客访谈] Karri 在 "Most SaaS Companies Got AI Wrong" 上演示了具体的工作流:
- 用户在 Linear 创建一个 issue。
- Linear AI 把它转交给一个 coding agent(通过 Linear 的 agent platform 集成)。
- Agent session 对整个团队可见——这是 Linear 的核心 UX 决定,因为"multiplayer context"是他们的根本属性。
- Agent 工作几分钟后给出 PR diff。
- 人类做决策性 review,但不需要参与执行。
最后他说:
"I don't think we will solve all the different coding needs, but we don't necessarily also have to. I think what we see the value is just sitting upstream where the work is coming from. There's really good leverage there that we can offer to companies where it's like work comes in or bugs come in, they automatically get spawned into agents, even delegated to the agents. Engineers never even see them.""我不觉得我们会解决所有不同的编码需求——但我们也不一定要解决。我们看到的价值是'坐在工作流的上游'。在那里有非常好的杠杆——我们可以给公司提供:工作或 bug 进来时自动 spawn 成 agent、甚至自动派给 agent。工程师甚至看不到这些 bug。"Karri Saarinen, Every / AI & I (Dan Shipper), "Most SaaS Companies Got AI Wrong. Linear Waited.", 2026-04-01
从宣言到产品:Code Intelligence 与 Coding Sessions(2026 年 5–6 月)
"issue tracking is dead" 宣言在两个多月内落地为两次产品发布。
Code Intelligence(2026 年 5 月 14 日) [26]:Linear Agent 可以读取并分析 codebase,直接从源代码回答问题——bug triage 自动查 commit 找根因、support 一线自查不必升级到工程、PM 写 spec 时评估技术工作量。论证支点是"代码是客观的 context":
"Code is one of the best forms of product context because it is objective.""代码是最好的产品 context 形式之一,因为它是客观的。"Karri Saarinen, "Code Intelligence for Linear Agent", 2026-05-14
Linear 内部查询量从 2026 年 2 月的 1,055 次涨到 4 月的 4,267 次,使用方扩散到 bug triage、support 和 product 团队。
Coding Sessions(2026 年 6 月 11 日) [27]:Linear Agent 开始直接写代码——从 issue 读 context、调查 codebase、提方案、写代码、开 PR;配套 Automated Triage(新 issue 由 agent 自动先行调查,接 Sentry/Datadog)和 Diffs(Linear 内原生 code review 界面,change 与原 issue 保持关联)。Karri 把此前的缺口讲得很直白:
"The loop wasn't complete. The coding work itself still happened somewhere else.""这个环没有闭合。写代码这件事本身仍然发生在别的地方。"Karri Saarinen, "Now Linear writes the code, too", 2026-06-11
数据点:上月 Linear 内部 "almost 700 PRs were merged by the agent";客户侧 Ramp 的 Inspect 通过 Linear API 写出超过 60% 的 merged PR。定位语也从 "product memory" 进一步收敛为 "the context layer that product development runs on"。
[事实] 一个更早的中间读数来自 2026-04-25 的 "Some Notes on AI" [39]:仅凭自家 cloud coding agent,Linear 当时每月修复超过 1,000 个 issue,且"增长很快";但他同时给了与叙事相反的提醒:
"But hard problems remain hard. Agentic coding does not speed those up as much as the narrative often suggests. It helps around the edges, with scaffolding, refactoring, investigation, and smaller fixes.""但难题仍然是难题。Agentic coding 对它们的加速远不如叙事里常说的那样多。它在边缘处帮忙:脚手架、重构、调查,还有小修小补。"Karri Saarinen, "Some Notes on AI", 2026-04-25
他把最大的收益定义为 bandwidth——"We can tackle more smaller problems than before"——而不是难题被解决得更快。
[推断] 至此第九章的战略弧线可以被验证:2025 年 4 月 workbench 隐喻(AI 是工作台上的工具)→ 2025 年 10 月三级自治框架 → 2026 年 3 月宣言 → 2026 年 5–6 月产品落地。节奏仍是他式的"等到能做对再做";但方向上 Linear 已从"上游决策层"伸进了执行层本身——与他此前"我不觉得我们会解决所有编码需求"的表态构成一个值得持续跟踪的张力。
团队规模:99 人覆盖 15+ 国家,10 个时区
[19] 2025 年 10 月的 "Designing remote work at Linear" 给出了当时最具体的组织数据:
- 99 人
- 15+ 国家、10 时区
- No permanent HQ
- 整公司每年共聚一次(一周长的 offsite)。近期 offsite 地点:哥本哈根、里斯本、赫尔辛基、墨西哥城
- 可选的 coworking hubs:柏林、纽约、SF
核心运营哲学:三种自由
- Freedom of place——你在哪都行
- Freedom of rhythm——你的节奏由你定
- Freedom of deep focus——深度工作时间被保护
而质量需要 intent:
"Quality doesn't come automatically, it needs to be intentional.""质量不会自动到来——它需要刻意。"Karri Saarinen, "Designing remote work at Linear", 2025-10-29
具体实践:
- Zero-bugs policy + SLA
- Quality Wednesdays——每人每周找出并修掉一个小质量缺陷、会上互相分享(机制与初衷见 §二"Quality Wednesday 与 Feature Roast";此前此条被概括为"团队培训",本轮按 2026-09 Summit keynote 校正)
- 每个 feature 上线前都有 review
- "The company says no frequently to maintain high standards."
二、深度洞察
关于产品:Quality 不等于 Perfection
Karri 在 YC Startup Podcast(2025-10)上对"质量"做了一个非常具体的辨析——它不是 polish 到完美:
"Quality doesn't mean perfection, that you can't have anything rough or anything. You need to polish everything for years. It just means that you have this direction and idea in the mind that we are about quality. We want the customers to have a great experience.""质量不意味着完美——它不意味着你不能有任何粗糙的东西、或必须把所有东西打磨好几年。它只意味着你脑子里有一个方向和理念:我们关心质量。我们希望客户拥有出色的体验。"Karri Saarinen, Y Combinator Startup Podcast, 2025-10-14
这种"质量是一个方向,不是一个状态"的定义——是他与"完美主义"做的明确切割。Linear 内部用 feature flag 做的"内部粗糙 → 上线 polished"的两段式流程,就是这个原则的具体兑现。
关于产品:Design Is Finding Things That Fit and Feel Good
Karri 在 YC Startup Podcast(2025-10)上对设计的本体性定义:
"I like to think like design is finding things that fit and kind of feel good. And I think that in a way, when you're building products or companies, that's kind of like what you're doing ideally.""我喜欢把设计想成——找到那些 fit 而且感觉好的东西。我觉得某种意义上,当你在做产品或公司时,理想状态下你也在做这件事。"Karri Saarinen, Y Combinator Startup Podcast, 2025-10-14
"As a designer, it's not always easy to articulate that, but I can somehow sense or feel it. This is the way to do it. I can see that this will work. And with designs or other things, I can visualize how they will look or how they will work or maybe even how the users might react to it.""作为一个设计师,要把它清楚说出来不总容易,但我能某种方式感觉到——'就是这样做。我能看到这会奏效'。在设计或其他事情上,我能想象它会怎样、它会怎么工作、用户甚至会怎么反应。"Karri Saarinen, Y Combinator Startup Podcast, 2025-10-14
这是 tacit knowledge 的明确承认——他不假装自己是用 data 或 framework 在做决定。设计师的核心能力就是对"fit"的内在感知——这种感知必须靠经验培养,不能完全外包。
Conversations on Quality(2024-10,Linear 自办播客)里他把这种 tacit 感知进一步祛魅——intuition 不是天赋,是可训练、且必须持续校准的模式识别:
"Intuition is not a magical force that just exists somewhere in the world and you either have it or you don't have it. I think the intuition is just your brain or your consciousness telling you something based on the experiences you had in the past.""Intuition 不是某种存在于世界某处、你要么有要么没有的神秘力量。我认为 intuition 只是你的大脑或意识,基于你过去的经验在告诉你一些东西。"Karri Saarinen, Conversations on Quality, 2024-10-12 [00:04:23–00:04:33]
"Talking to the customer and the interaction with customer as a way to train and keep your intuition in check.""与客户交谈、与客户的互动——是训练你的 intuition、并让它保持被校准的方式。"Karri Saarinen, Conversations on Quality, 2024-10-12 [00:05:53]
[事实] 他明确警告了漂移风险:只凭自己的观点做东西,有时会做对,但有时"你可能就飘走了——不再匹配客户的现实" [00:06:00–00:06:10]。[推断] 这补全了他的 taste 论证三件套:taste 是感知 fit 的能力(本节上文)、过往经验是它的训练集、客户交互是它的校准项——防止 taste 过拟合于自我。
关于产品:Local-first —— 为什么自建 sync engine,让工具"感觉像专业软件而非网站"
档案此前充分记录了 Linear 的"美学 / 品味"是它的差异化来源,但从未触及它"快"背后那个具体的工程决策。在更早的 2023 年 EO 访谈里,Karri 第一次(在档案覆盖范围内)把 Linear 的"快"落到一个明确的架构选择上——自建 local-first sync engine,而不是复用现成框架。[事实] 据他在这期访谈的复盘:他们不想要一个"感觉像迟钝网站"的工具,而要一个"感觉像高性能原生应用"的工具;做法是让用户的每一个操作先在本地设备上瞬时发生、再同步到云端,从而消除那种打断工作流的 "lag"。
驱动这个决策的判断,是他对目标用户的一句很硬的定性——工程师是 "source of truth":
"Engineers are the source of truth. If they don't use the tool, the tool isn't helping the company.""工程师就是 source of truth。如果他们不用这个工具,那这个工具就没有在帮公司。"Karri Saarinen, EO, 2023-11-18 [00:08:40]
[推断] 这条补上了档案里"quality growth = 产品本身即营销"论证缺失的一块工程地基:Linear 的"人们觉得它比别的都好"并不只来自视觉品味,也来自一个大多数竞品没有付出的底层工程成本(local-first 同步)。它和第五章 Coinbase 时期"重做美学以传递信任"是同一种思路的两种介质——在别人偷懒的地方多做一层,向专业用户发出"制作者真的在意"的信号:在 Coinbase 是视觉信任,在 Linear 是零延迟的手感。
关于产品:Suffering as a Clarification Tool
Karri 在 YC Startup Podcast(2025-10)上对 AI 时代的产品设计提了一个反直觉的论点——做得太容易反而让你不思考:
"The danger is that when something is easy to do, you stop paying attention to it. So if it's easy to generate designs and ship them, you maybe didn't think about this feature at all. I think sometimes when something is really hard, you're constantly thinking about it, like, is this worth it?... So the suffering helps to maybe clarify the design.""危险是当某件事变容易时,你就不再注意它了。如果生成设计和上线很容易,你可能根本没思考过这个 feature。我觉得有时候当某件事真的很难时,你会一直在想——这值得做吗?……所以痛苦帮助你澄清设计。"Karri Saarinen, Y Combinator Startup Podcast, 2025-10-14
这是一种"摩擦作为思考激励"的反 frictionless 设计观——和 Stewart Butterfield 的"comprehension over friction"形成一个有趣的对照。
关于产品:AI 降低地板,但天花板继续升高
"Usually technology will happen regardless. I think we can't go back not having AI anymore. It's not going to disappear from the world. So now the next question is, well, what should I think about it? Or how should I use it? I think it will lower the floor on things like design.""技术通常无论如何都会发生。我觉得我们已经回不到没有 AI 的世界了——它不会从世界上消失。那么下一个问题就是:我应该怎么想它?怎么用它?我觉得它会降低设计等领域的'地板'。"Karri Saarinen, Y Combinator Startup Podcast, 2025-10-14
"I do think that the top, the ceiling, it will just keep getting higher. And so if you really want to be the best design company or the best company, you have to still keep pushing the boundaries and you can't just expect that the AI will solve it for you.""我确实认为'天花板'会继续被推高。所以如果你真的想做最好的设计公司或最好的公司——你必须继续推动边界,不能指望 AI 会替你解决。"Karri Saarinen, Y Combinator Startup Podcast, 2025-10-14
2026 年 5 月 Sequoia × Linear fireside 上,他给"AI 降低地板"加了一个重要的自我修正——如果你没有评估输出的能力,地板其实并没有为你升高:
"People have this idea where AI is increasing the floor of the design, the skill. But what we're seeing now, it's not always the case. Because if you have people who don't really understand design, they might be using the tools wrong and... they don't have a tools to evaluate the output.""人们有一种想法:AI 在抬高设计这项技能的地板。但我们现在看到的是——并不总是这样。因为如果一个人并不真正理解设计,他可能把工具用错,而且他没有评估输出的工具。"Karri Saarinen, Startup Grind fireside(Sequoia × Linear), 2026-05-05 [00:25:04–00:25:25]
"It's a little bit like me doing my own legal work. I don't understand legal enough, so I'm not going to do it with AI, even though AI sounds very confident... I'm not going to trust it with the thing that I have no way of evaluating with.""这有点像我自己给自己做法律工作。我对法律理解不够,所以我不会用 AI 去做——即便 AI 听起来非常自信……我不会把一件我完全没办法评估的事交给它。"Karri Saarinen, Startup Grind fireside(Sequoia × Linear), 2026-05-05 [00:25:30–00:25:43]
[推断] "AI 降低地板"与"没有评估能力则地板不动"并不矛盾——前者说的是产出的地板,后者说的是质量判断的地板。这条修正让他的 taste 论证闭环:AI 时代真正稀缺的不是生成能力,而是评估能力,这几乎是 taste 的操作性定义。
关于组织:Small Teams Always Deliver Better Quality, Faster
这是他在"The profitable startup"和 20VC 上反复出现的核心组织信念:
"One is that the team should be smaller and more focused. If I look back to different jobs in different companies, I always felt that the smallest teams made the best results. Not all the time, but usually when you put this special team together that are really talented, maybe it's only three people or five people or something, they could make really big impact in a company very fast.""团队应该更小、更聚焦。回头看不同公司的工作经历,我总觉得最小的团队做出了最好的结果。不是所有时候——但通常当你把这种特殊的、真正有才华的小团队聚到一起(也许只是 3 人或 5 人),他们能很快在公司里产生巨大影响。"Karri Saarinen, 20VC, 2024-10
"We took the different approach where let's just try to find highest quality people we can and then not grow the team too much because it can cause additional problems. It can dilute the culture. You spend more time managing things and you have to onboard people more.""我们采取了不同的路径——找到我们能找到的最高质量的人,然后不要把团队扩太大,因为那会带来额外问题——稀释文化、花更多时间管理、需要更多 onboarding。"Karri Saarinen, 20VC, 2024-10
Coinbase Exchange 由一个工程师在 6 个月内做完——这是他反复举的数据点,是他"大公司迷信"的具体反例。
关于组织:Bike-Shedding 和 Design by Committee 的成本
"One of the lessons I learned in some of these other companies was that having more people working on things doesn't necessarily make it better. It often can make it worse. You start to lose the thread of what is this feature even about, because everyone has a little bit different opinions. You start into this bike-shedding or design by committee.""我在其他公司学到的一个教训是——更多的人参与不一定让事情变好,反而经常变差。你开始失去这个 feature 到底是关于什么的主线,因为每个人都有一点不同的观点。你陷入 bike-shedding 或 design by committee。"Karri Saarinen, Y Combinator Startup Podcast, 2025-10-14
"From the beginning, we wanted that the people work in small teams and like there's often like two, three people. It's like an engineer, maybe a couple of engineers, one designer. And we, for a long time, we didn't have any product managers.""从一开始我们就想要人们在小团队里工作——通常 2-3 人。一个工程师、可能几个工程师、一个设计师。很长一段时间我们根本没有 PM。"Karri Saarinen, Y Combinator Startup Podcast, 2025-10-14
关于决策:你不能 Spec 出质量
"It's hard to spec quality or the right solution. Like you can maybe spec the solution to some degree, but you can't really spec the quality execution of it. And I think that's where I think if someone is driving the project and building it, they have a lot of opportunities to tweak it to make it better.""你无法 spec 出质量或正确的解法。你也许能 spec 出某种程度的解法,但你无法 spec 出'执行的质量'。如果某个人在驱动这个项目并构建它——他们有很多机会去调整、把它做得更好。"Karri Saarinen, Y Combinator Startup Podcast, 2025-10-14
这是 Linear 为什么"项目团队 = PM"的根本理由——质量的微观决策只能在执行者手中产生,不能从外部 spec 出来。
关于商业:Profitability Is Leverage, Not Anti-Investor
"Being profitable gets you the highest leverage you can have in terms of when it comes to talking to investors because it gives you the leverage of not needing them. And so that's always the highest leverage if you can walk away from any deal.""盈利给你和投资人对话时最高的杠杆——它给你'不需要他们'的杠杆。如果你能从任何一笔交易里走开,那永远是最高的杠杆。"Karri Saarinen, 20VC, 2024-10
这种"盈利不是反对融资,是融资 leverage"的反正统观点——是他和 Paul Graham "ramen profitability" 概念的直接呼应。
关于商业:Default-Alive 却仍融多轮的理由——Signaling 与"别融太多"的双刃
Grit 2026-04 的访谈里 Joubin 直接把"既然已经 default-alive 又盈利,为什么还要融多轮"这个问题摆在他面前。Karri 的回答补上了前面 20VC 那套"盈利是 leverage"叙事的另一面——融资本身有一个正向价值(signaling),而 leverage 恰恰用在"不融太多"上:
"It's a mix of signaling and strategic positioning. While the seed round was necessary for survival and early hiring, subsequent rounds were about signaling to the market that we are a serious, long-term player. Larger enterprise customers often look at your funding stage to determine if they can trust you with their business.""这是 signaling 和战略定位的混合。种子轮是为了生存和早期招聘所必需的,但后续的几轮是为了向市场发出信号——我们是一个认真的、长期的玩家。更大的企业客户经常看你的融资阶段,来判断能不能把业务托付给你。"Karri Saarinen, Grit (Joubin Mirzadegan), 2026-04-20 [00:24:48 前后]
而 leverage 的用处不在"要不要融",而在"融多少、估值定多高":
"We are careful not to raise too much or increase our valuation too quickly, as that reduces the upside for future employees and can box us into a corner if we hit a market snag. We want to maintain control of our own destiny without needing to constantly chase the next fundraiser.""我们很小心不要融太多、不要把估值抬得太快——因为那会减少未来员工的 upside,而且一旦碰到市场逆风会把我们逼到墙角。我们想掌握自己的命运,而不必一直追着下一轮融资跑。"Karri Saarinen, Grit, 2026-04-20 [Q2]
对 VC 关系的默认姿态也在这里被讲明白——默认是 No,但保持一个慢慢养起来的短名单:
"By default, my answer is no. I'm not focusing on fundraising every week... I might take a meeting every now and then with a firm or partner I find interesting, building those relationships slowly over time so that when we are ready to fundraise, I already have a shortlist of people I trust... The most important thing is alignment; I want investors who understand our long-term view and don't try to meddle in our operations or push us to hire faster than we want to.""默认情况下我的回答是 No。我不会每周都在想融资……我偶尔会跟一家我觉得有意思的机构或合伙人见一面,慢慢地把关系养起来——这样当我们真要融资时,我手上已经有一份我信任的人的短名单……最重要的是 alignment;我想要那种理解我们长期视角、不会插手我们运营、也不会逼我们比我们想要的更快招人的投资人。"Karri Saarinen, Grit, 2026-04-20 [Q3]
配套的"好投资人"定义(highlight):
"The best investor is one you can trust who stays out of the way. When they try to do too much, they become a distraction.""最好的投资人是那种你能信任、又不挡道的人。当他们想做太多时,他们就变成了一种干扰。"Karri Saarinen, Grit, 2026-04-20 [00:24:48]
[推断] 这条把档案里原有的"盈利=最高 leverage"从一个静态论断补成一个可操作的融资策略:leverage 不是用来拒绝所有钱,而是用来(a)保留 signaling 带来的企业信任,同时(b)压住估值/融资额、保护未来员工的 upside 与自身回旋余地。这与他"小团队长期主义 + tender offer 保护员工流动性"的组织哲学(第八章)是同一套内在逻辑。
[事实] 2026-08 的第二次 tender offer 把这条策略推到下一格:公司 cashflow positive、现金多于历史融资总额,于是只做 $99M secondary、不融 primary——"Fundraising can become a habit... This time, it wasn't what Linear needed."(详见第八章)。
关于商业:Quality Is the Only Differentiator in a Feature-Parity Market
"I think it's just that the quality and doing things well still matter. It seems to be when I look at the market or me personally buying things that today it's really hard to find quality things. A lot of things or products are marketed really well and they can seem like they're high quality, but then when they actually use them, you're disappointed. There is some kind of power that if you actually do things really well, it is rare and people will notice and people will talk about it.""我觉得质量和把事情做好仍然很重要。当我看市场或我个人买东西时,今天真的很难找到高质量的东西。很多产品营销得很好、看起来好像很高质量——但真用起来你会失望。如果你真的把事情做好,是有一种力量的——它稀有、人们会注意到、会谈论它。"Karri Saarinen, 20VC, 2024-10
关于战略:几乎与所有在位者反着做
In Depth 2025 上他给出了自己最明确的定位启发式——先看市场里什么被服务不足,然后:
"I find that startups sometimes or founders make the mistake of trying to emulate the other companies and not play into their own strengths or where they're coming from or the context they're in... you should almost do the opposite of what all the other players in your domain are doing.""我发现创业公司或创始人有时会犯的错误是:试图模仿其他公司,而不是发挥自己的长处、自己的出身和所处的 context……你几乎应该与你所在领域里所有其他玩家反着做。"Karri Saarinen, First Round In Depth, 2025-06-05 [01:13:49–01:14:07]
"So if you are like everyone else, then no one is probably [paying attention]... or you are worse version of or a simpler version of something. It's probably not that interesting.""如果你和所有人一样,那大概没人会注意你——或者你只是某个东西的更差版本、更简化版本。那大概不会有意思。"Karri Saarinen, First Round In Depth, 2025-06-05 [01:13:09–01:13:23]
[事实] 他给的操作化版本:如果所有玩家都很怪(weird),你也许该非常不怪;如果都很僵硬(rigid),你该反过来;也可以从其他领域借范式——"Are you the Apple of this industry? Or are you the Amazon of this industry?" [01:13:37–01:14:23]。[推断] 这是"Quality Growth 反命题"的方法论泛化:反着做不是气质,是注意力稀缺下的结构性策略——Linear 的各个反直觉决策(无免费版、不做 A/B、慢招人、AI 上刻意等)都能回溯到这一条。
关于商业:与 IC 结盟,而非 Buyer——B2B 工具的"代理人问题"反转
2024 年 9 月 Product School 访谈里,他把 Linear 的 GTM 赌注表述为一个明确的结盟选择:
"They may be loved by the buyers or the management, but I don't think it really matters if the ICs are not really using the tool. So that's kind of like the bet we are taking is that we want to align ourselves with the engineers and people who build things.""它们(多数同类工具)也许被 buyer 或管理层喜爱,但如果 IC 们并不真正使用这个工具,我认为那并不重要。所以我们下的赌注就是:把自己和工程师、和真正做东西的人结盟。"Karri Saarinen, Product School, 2024-09-04 [00:14:17–00:14:32]
[事实] 配套的销售时序:等公司内部出现自发使用的"activity"之后,才去接触管理层/buyer 谈 pilot——卖给已经在内部要这个工具的人,远比冷启动卖给没听过你的人容易;他也直白承认 buyer/管理层仍是 gatekeeper,第一次谈不成就等——"often it happens that way, that next year they will come to us again and the timing is better then" [00:16:29–00:17:52]。[推断] 这与 quality growth 构成因果链:产品质量赢得 IC → IC 持续使用让工具里的数据保持新鲜 → 工具对管理层才有价值。多数 PM 工具走的是反向路径(先赢 buyer,IC 逃离,数据陈旧)——这正是他说"loved by the buyers 但 ICs 不用就没意义"的结构性依据。
关于 AI:你不能把 Thinking 外包给 Agent
"You can't just outsource the thinking purely to the AI agents. The more you can clarify your own thinking and the strategy or something, the better it's for your team, but also the better it is for the agents too.""你不能把思考完全外包给 AI agent。你越能澄清自己的思考、战略——对你的团队越好,对 agents 也越好。"Karri Saarinen, Every / AI & I (Dan Shipper), "Most SaaS Companies Got AI Wrong. Linear Waited.", 2026-04-01
"I don't see the future in a way that we are replacing humans and I don't quite believe in it. Maybe I don't want to believe in it, but I think things will change. The roles will change... I still don't see how the agents, how the AI actually does all the thinking and the choices or decisions. I think product building is still a craft or an art.""我不觉得未来是'我们替换人类'的样子——我不太相信。也许我不想相信。但我觉得事情会变。角色会变。我仍然看不到 agent / AI 怎么做所有的思考、选择或决策。我觉得做产品仍然是一种手艺或艺术。"Karri Saarinen, Every / AI & I (Dan Shipper), "Most SaaS Companies Got AI Wrong. Linear Waited.", 2026-04-01
关于 AI:Agent 是团队能力,不是个人生产力工具
2026 年 6 月 Coding Sessions 发布文里,他给出了与"个人工程师 + agent = 10x"叙事相反的立场——agent 的工作必须发生在团队共享的 context 上:
"Building a product is something a team does together.""做产品是一个团队一起做的事。"Karri Saarinen, "Now Linear writes the code, too", 2026-06-11
"The loop happens on the context everyone can see and add to, so it belongs to the team.""这个循环发生在每个人都能看到、都能补充的 context 上——所以它属于团队。"Karri Saarinen, "Now Linear writes the code, too", 2026-06-11
这与第九章里"agent session 对整个团队可见"的 UX 决定同源:multiplayer 是 Linear 的根本属性,AI 时代他把它升级为对抗"agent 工具个人化"趋势的差异化立场。
关于 AI:"让 AI 为你工作,别让 AI 工作你"——反谄媚的 Devil's Advocate 用法
2026 年 5 月 Startup Grind Global 的 Sequoia × Linear fireside(Karri × Stephanie Zhan)是档案里他亲身示范怎么用 agent 最具体的一次。他先把 AI 归入一类新的危险工具:
"It's a dangerous tool because it also does thinking. It's not just a mechanical tool, but it's also a thinking tool. So it can make decisions for you if you're not careful.""它是一种危险的工具,因为它也做思考。它不只是机械工具,还是思考工具。所以如果你不小心,它会替你做决定。"Karri Saarinen, Startup Grind fireside(Sequoia × Linear), 2026-05-05 [00:20:49–00:20:55]
"In your head, always, like, I'm making the decisions. You're not making any decisions here. You're just telling me what I could do... And then make the AI work for you, not let the AI work you.""在你脑子里要一直保持:做决定的是我。你(AI)在这里不做任何决定,你只是在告诉我我可以做什么……让 AI 为你工作,别让 AI 工作你。"Karri Saarinen, Startup Grind fireside(Sequoia × Linear), 2026-05-05 [00:20:58–00:21:16]
他自己的用法有两层。第一层,[事实] 把 Linear agent 当"思考伙伴"用在组织共享的 context 上——系统里累积了约 40,000 条客户请求,他和单个客户聊完后,让 agent 把这条请求对照全部历史反馈找 pattern,而不是就事论事 [00:06:40]。第二层是反谄媚纪律——主动要求 agent 论证"为什么不该做":
"And then I often, like, reverse it and ask, like, why we shouldn't build this. Like, can you explain the reasons we haven't built it so far or we shouldn't even build it? Because the AI LLMs have a tendency to please you.""然后我经常把问题反过来问:为什么我们不该做这个?你能解释一下我们至今没做它、或者根本不该做它的理由吗?因为 AI LLM 有讨好你的倾向。"Karri Saarinen, Startup Grind fireside(Sequoia × Linear), 2026-05-05 [00:08:31–00:08:41]
[推断] 这是"你不能把 Thinking 外包给 Agent"从立场落到日常操作的样子:检索与论证生成可以外包,但"决定权在我"必须作为显式心智状态被维持。[事实] 同场他还给了两个新数据点:77% 的企业客户至少装了一个 coding agent;过去一个季度经由 Linear 完成的 agent 工作量增长约 5 倍 [00:06:08–00:06:19]。(主持人转述:前一周 Sequoia AI Ascent 上 Figma CEO Dylan Field 在讲设计角色时引用了 Karri 的 "Output isn't design" 一文——与本档案库 dylan-field 档案互见。)
关于个人:Designer-Founder 的自我感知
这条洞察来自 Sequoia Spotlight——一句把他设计师身份的"职业病"具象化的话:
"When I look at things, I could see, 'This could be better.""当我看东西的时候,我能看到——'这本可以更好。"Karri Saarinen, Sequoia Spotlight
"For me, work is serious. If I'm building a house, I don't want my tools to be fun. I want them to be good. I want them to be professional.""对我来说工作是严肃的。如果我在盖房子,我不希望我的工具是'好玩的'。我希望它们是好的、专业的。"Karri Saarinen, Sequoia Spotlight
第二句是 Linear 整个产品哲学的浓缩——"don't make tools fun, make them good"。这一句话在 Sequoia 文章后被无数次引用,几乎成了 Linear 的非官方 mission statement。
In Depth 2025 上他还回答了一个更结构性的问题——为什么设计师创始人这么少。他的解释不是能力差异,而是验证路径的不对称:
"So there's a little more like clear, like quicker path to some kind of validation, I would say. Versus like designers, you might have this idea, something could be better. But so you design it, but you still just have designs.""(工程师)有一条更清晰、更快的验证路径。而设计师——你可能有一个'这可以更好'的想法,你把它设计出来了,但你手里仍然只有设计稿。"Karri Saarinen, First Round In Depth, 2025-06-05 [01:26:03–01:26:22]
[事实] 他的完整论证:工程师有 "side project 路径"——做出能用的 functional prototype,朋友能直接上手用,很快获得某种真实验证;设计师往往止步于设计稿。他还点出设计师的第二个短板:过窄地只关注 interface,而当创始人需要对更广的商业 context(公司为什么存在、谁付钱、激励如何运转)有兴趣。[推断] 这也解释了他自己的路径选择:Kippt 和 Linear 他都与工程师联创绑定(Jori、Tuomas)——用信任网补上设计师缺的那半条验证路径。
关于个人:Founder 必须仍然亲自做质量审查
"Before we launch anything to everyone, I or the other founders will do some kind of check on it. It's like, does this feel good? Does it feel right? Are the animations correct? And like those little details are correct. I'm not sure if we can keep doing that or it's fully scales. And even now it's hard.""在我们把任何东西公开发布之前,我或其他联合创始人会做某种 check——它感觉好吗?感觉对吗?动画对吗?这些小细节对吗?我不确定我们能不能一直这样、能不能完全 scale。即使现在也很难。"Karri Saarinen, 20VC, 2024-10
这是 John Gruber 2010 年说的"质量趋近负责人品味水平"在 100 人公司中的实际兑现——Karri 自己至今在每个 GA release 前做亲自审查。
关于组织:更快 ≠ 更多人;速度来自 Focus 与 Clarity
档案里已经充分记录了他"小团队优于大团队"的信念(第七章、"Small Teams Always Deliver…")。Grit 2026-04 上他把这个信念收成一个更锋利的因果句——"更快"的反义词不是"人少",而是"焦点模糊",而多招人反而会主动制造不该做的工作:
"I don't believe the way to go faster is hiring more people. You end up launching things you didn't need to launch just to keep people busy.""我不相信'更快'的办法是招更多人。你最终会为了让人有事做,而发布一些你本不需要发布的东西。"Karri Saarinen, Grit, 2026-04-20 [00:00:00]
"The way to go faster is having better focus and clarity. Without that, you don't get speed; you just get mistakes.""更快的办法是有更好的焦点和清晰度。没有这个,你得不到速度——你只会得到错误。"Karri Saarinen, Grit, 2026-04-20 [00:32:45]
他把这条逻辑直接接到 AI 上——"build 变快"不等于"应该 ship 更多"(这一点在"SaaS Got AI Wrong"里有相邻表述,但此处是最紧的一句):
"If AI makes building faster, it doesn't mean you should ship more. Shipping too many features just leads to a bloated product.""如果 AI 让 build 变快,这不意味着你应该 ship 更多。发布太多 feature 只会导致一个臃肿的产品。"Karri Saarinen, Grit, 2026-04-20 [00:33:16]
[推断] 这补上了档案里"小团队"论证的因果机制:不是"人少所以慢地做好",而是"人多 → 为填满产能而造工作 → 产品臃肿 → 反而更慢"。这也是他把 AI 效率红利主动拒绝兑现成"更多产出"的组织级理由。
关于组织:Roadmap 只看六个月 + 年度主题——远程公司的"方向稳定器"
"Generally, how the roadmapping goes is we usually look at six months ahead.""一般来说我们做 roadmap 的方式是:通常只往前看六个月。"Karri Saarinen, Product School, 2024-09-04 [00:24:40]
[事实] 机制的完整形态:六个月滚动视野 + 有时定一个年度主题(例如从 issue tracking 的执行侧转向 planning 侧);由他和 leadership 小组坐下来定,不搞纯 metrics 驱动;大版本刻意排在春/秋发布,避开美欧客户的暑假。理由与远程结构直接相关——Linear 从第一天起就是横跨两大洲的远程公司,remote 意味着方向必须更深思熟虑、更稳定,否则团队每天都在被动转向 [00:24:40–00:25:41]。[推断] 这是"低协调成本"哲学在时间维度的投影:小团队降低空间协调成本,六个月主题制降低时间协调成本。
[事实] 2026-04-25 的 "Some Notes on AI" [39] 补了这套机制背后的理由层——在"AI 时代还要不要 planning"的争论里,他把 plan 这个产出和 planning 这个动作分开:
"I often think planning is not about the output (the plans) but exercise on an alignment and commitment. It forces the organization to come together and debate what actually matters, decide priorities, create meaning and then share that direction with everyone.""我常常觉得 planning 不在于产出(那些计划),而是一次关于对齐与承诺的练习。它迫使组织聚到一起,辩论什么真正重要、决定优先级、创造意义,然后把这个方向分享给每个人。"Karri Saarinen, "Some Notes on AI", 2026-04-25
"When you don't do the planning exercise, you also never fail your thinking, commitments, plans, or timelines because you never make them.""当你不做 planning 这个练习,你也永远不会在你的思考、承诺、计划或时间表上失败——因为你从来没有做出过它们。"Karri Saarinen, "Some Notes on AI", 2026-04-25
[推断] 这解释了为什么 Linear 的 roadmap 只看六个月却仍坚持做:六个月是对"市场在变"的让步,"坚持做"是对"没有 plan 就没有可被证伪的判断"的坚持——执行变便宜之后,"做错东西"的成本反而上升(同文:"If it becomes easier to make more things, it also becomes easier to make the wrong things.")。
关于组织:Values 不能被发明,只能被"发现"
⚠️ 本条来源为 Linear 自办播客 Conversations on Quality(2024-10),有自我叙事成分;但内容是可核查的操作机制而非成就叙事,按机制折入:
"I think especially in startups, when you first get started, it's hard to just invent them. They have to like come from somewhere... the execs go in the room and they talk about things and they like try to like pick the values. But I think a lot of times the values end up being kind of not really connected to anything.""尤其在创业公司,刚起步时你很难凭空发明 values。它们必须有来处……(很多公司是)高管进会议室讨论一番、试着挑出几条 values。但很多时候这些 values 最后和任何东西都没有真实连接。"Karri Saarinen, Conversations on Quality, 2024-10-12 [00:00:04–00:00:18]
[事实] 他的替代做法:先写出公司的历史——"the story we tell ourselves is important"——从"哪些东西真实促成了我们的成功"里让 values 自己浮现;写完分享给团队时没有任何意外,因为大家本来就在按这些原则做事 [00:00:45–00:01:06]。[推断] 与第七章 hiring 的 work trial、"你最自豪的是什么"面试题同构:他对"文化制品"的一贯态度是只承认已被行为验证过的东西,拒绝先验宣告。
关于个人:40 小时工作制——创意工作不能"被磨出来"
档案此前没有触及他的工作节奏观。Grit 上他确认 Linear 维持 40 小时工作制,理由不是福利,而是对"creative work"本性的判断——磨(grind)本身会杀死反思:
"We value the quality of the output over the volume of the input. I've found that creative work—which is what we do—cannot be 'ground out' for 20 hours a day. If you are constantly hammering away, you don't stop to reflect on whether you are working on the right things or if there is a better way to solve a problem.""我们看重产出的质量,而不是投入的量。我发现创意工作——也就是我们做的事——没法靠一天磨 20 小时'磨'出来。如果你一直在锤,你就不会停下来反思:你在做的是不是对的事、有没有更好的方法解决这个问题。"Karri Saarinen, Grit, 2026-04-20 [Q5]
"When I have the right vision or inspiration, I can accomplish in a day what might take others a week of 'grinding.""当我有正确的愿景或灵感时,我一天能完成别人磨一周才能完成的事。"Karri Saarinen, Grit, 2026-04-20 [Q5]
这与他"deciding slow / executing fast"(第九章)以及"suffering as a clarification tool"是同一套底层信念的不同切面——保护反思的空间,比堆投入时间更重要。
关于个人:Sisu 与"不 brute-force"——把 grit 用在长期,不用在日常
主持人以节目名 "Grit" 收尾提问,Karri 的回答给了档案一个新的自我框架——他把 grit 分成两层,并明确拒绝在日常里 brute-force:
"It reminds me of the Finnish concept of 'Sisu,' which is a form of grit used as a last resort... in my day-to-day, I don't like to just 'brute force' things. If something feels hard, I don't just hammer at it; I stop and reflect on why it's hard and look for a way to make it easier or more efficient. True grit is about the long-term commitment to see things through, but the day-to-day should be about finding leverage and making smart choices.""它让我想起芬兰的概念 'Sisu'——一种作为最后手段的 grit……但在我的日常里,我不喜欢单纯地 brute-force。如果某件事感觉难,我不会硬锤;我会停下来反思它为什么难,找一个让它更容易、更高效的办法。真正的 grit 是那种把事情坚持做完的长期承诺,但日常应该是关于找 leverage、做聪明的选择。"Karri Saarinen, Grit, 2026-04-20 [Q8]
配套的 highlight 把这条日常操作原则讲得最干净:
"When something feels hard, I don't just grind it out or brute force it. I look for an easier way to achieve the same result.""当某件事感觉难时,我不会硬磨或蛮干。我会去找一个更容易的方法来达到同样的结果。"Karri Saarinen, Grit, 2026-04-20 [01:15:08]
[推断] 这把他芬兰设计 DNA(第一章 Iittala/Artek)从"美学"延伸到"工作伦理"层:Sisu 是芬兰文化里的国民概念,但他刻意把它降格为"last resort",日常反而推崇"找更省力的路"——这与"更快来自 focus/clarity 而非 brute force"(上一节)在同一逻辑面上。
关于个人:成为父亲之后——领导力就是"以身作则"
档案此前没有任何 Karri 的家庭/个人生活维度。Grit 上被问及为人父后的变化时,他把它接到了一个组织洞察上——员工模仿领导者,就像孩子模仿父母:
"It forced me to set tighter boundaries, such as stopping work in the evening to be present with my family. More importantly, it made me realize that leadership is about setting an example. Just as children observe and mimic their parents, employees observe their leaders' actions and behaviors. I became much more mindful about what I do and say because I realize that the organization is constantly looking to me to define the culture.""它逼我设立更紧的边界,比如晚上停止工作、陪伴家人。更重要的是,它让我意识到——领导力就是以身作则。就像孩子观察并模仿父母,员工也观察领导者的行为和举止。我变得对自己做什么、说什么谨慎得多,因为我意识到整个组织时时刻刻在看着我来定义文化。"Karri Saarinen, Grit, 2026-04-20 [Q6]
[推断] 这与"founder 至今亲自做质量审查"(上文)以及 John Gruber "质量趋近负责人品味水平"的引用,构成同一条"CEO 作为文化传导者"的线——只是这次是从家庭生活反推出来的,而非从产品审查推出来的。
关于哲学:技术演进中"care"会被反复挤走、然后被再次寻找
[17] 在 Config 2025 的演讲里,Karri 把"质量稀有"放进了人类技术史的循环框架:
- 工业化时代:手工艺被机器挤走
- 软件时代:制作变快、care 被 cycle time 挤走
- AI 时代:思考和品味本身被外包
每个时代之后都会有人怀念质量、重新追寻它——而 Linear 的赌注是 AI 时代会再次出现一个"质量复兴",他们想在这个复兴里成为参考。
关于组织:不要成为 Software Factory——产品工作产出两样东西,产品与学习
这是他 2026 年下半年最集中的论题,有三份一手材料:Lenny & Friends Summit 2026 的 keynote(2026-09-29,17 分钟独讲,YouTube 版后改题 "Why I took the summer off from AI (and what I learned)")、同月的 X 长帖 "Tools Are Getting Better. Are We?"(2026-09-25)[31]、以及 "Someone behind the software"(2026-10-15)[33]。
[事实] 演讲的开场是一个个人实验:他 2026 年夏天刻意停止关注 AI——不看新模型、新 agent、新技巧——度假前担心回来会"落后",回来后的结论是 "nothing has changed much":
"There's new models, there's new techniques, there's new agents. But I think in the end, making great products and building increasing revenue or building the business, it's still hard. So it made me wonder how much attention we're really paying to these tools... It's clear that these tools are getting better... But I think the more important question is, are we actually making better things with these tools?""有新模型、新技巧、新 agent。但我觉得到最后,做出好产品、把收入做上去、把生意建起来,仍然很难。所以这让我想:我们到底把多少注意力花在了这些工具上……这些工具显然在变好……但我觉得更重要的问题是:我们真的在用这些工具做出更好的东西吗?"Karri Saarinen, Lenny & Friends Summit 2026 keynote(Lenny's Podcast feed), 2026-09-29 [00:00:44–00:01:12]
核心命题是一个组织层面的禁令:
"You can build the software factories and you can automate things and that all makes sense. But just as an organization, don't become one. Don't become a software factory where you outsource all the work and thinking to somewhere else and just focus on how efficient can the output be. Because the output is not the product. Customers are not buying the lines of code. They're not buying the experiments.""你可以建软件工厂、可以自动化很多事,这些都说得通。但作为一个组织,别变成一座工厂。别变成一座把所有工作和思考都外包到别处、只盯着产出能有多高效的软件工厂。因为产出不是产品。客户买的不是代码行数,也不是实验。"Karri Saarinen, Lenny & Friends Summit 2026 keynote(Lenny's Podcast feed), 2026-09-29 [00:02:48–00:03:15]
论证的支点是"做产品有两个产出":
"Making products produces two things, the product and the learning... The danger really now with the product organization is that the more we automate stuff to AI to do and more we automate it or tell teams to use AI, you kind of create this separation from execution and from the learning. It's not necessarily a bad thing... Eventually, if you don't learn from the work you do, I think eventually you would lose the advantage of what you have currently in your space or in your company.""做产品产出两样东西:产品,和学习……现在产品组织真正的危险是,我们把越多的事自动化给 AI 做、越多地让团队用 AI,你就在执行与学习之间制造了一种分离。这不一定是坏事……最终,如果你不从你做的工作里学习,我想你最终会失去你目前在这个领域或这家公司里拥有的优势。"Karri Saarinen, Lenny & Friends Summit 2026 keynote(Lenny's Podcast feed), 2026-09-29 [00:03:27–00:06:12]
由此得出的自动化边界——不是"能自动化的都自动化",而是按"能不能从中学到东西"划线:
"We should automate more of the known things, like the things that are repeatable and maybe are not that impactful. It's not something we can learn a lot from.""我们应该把更多已知的事自动化——那些可重复的、也许影响不大的事。那不是我们能从中学到很多的东西。"Karri Saarinen, Lenny & Friends Summit 2026 keynote(Lenny's Podcast feed), 2026-09-29 [00:14:56–00:15:04]
[事实] 他给的正例正是 Linear 自己的 bug 修复 loop(接 Datadog / Sentry、读 codebase、写修复,见第九章 Coding Sessions);而省下的时间该去的地方是 "the customers, the explorations, the critiques, the quality, or even the reflection of what we do" [00:13:27–00:13:33]。标题句 "context becomes the product" 的解释是把视线从产出掉头(注意与第九章 Linear 的定位语 "the context layer that product development runs on" 不是一回事:那里 context 是 Linear 这一产品层,这里 context 是做决定的人所持有的理解):
"We have this obsession to look at what is the output, the software or the code. But I think in the end, I look the other direction. It's like, what is actually creating the software? Who is making the decisions? And that's the people are doing that. And the better context those people have and better understanding they have about what they're supposed to do, I think the products are better.""我们有一种执念,总盯着产出——软件、代码。但我觉得到最后,我看的是另一个方向:到底是什么在创造软件?谁在做决定?是人在做。这些人拥有的 context 越好、对自己该做什么理解越好,我认为产品就越好。"Karri Saarinen, Lenny & Friends Summit 2026 keynote(Lenny's Podcast feed), 2026-09-29 [00:16:03–00:16:26]
他自己怎么维持这个 learning loop:[事实] 除了"让 AI 为你工作"一节已记的"agent 当 40,000 条客户请求上的思考伙伴",keynote 里新给了一个机制——在汇集的客户 context 上放 "watcher" agent,让它只在"有我感兴趣的东西"时来找他;他设了一个关于客户 AI 工作流的 daily briefing,每天几条 bullet [00:09:13–00:10:03]。理由是产品组织一大,人就会 "away from the customer problems",因为没人有时间看每一封邮件、每一条请求 [00:09:00]。
书面版 [31] 里多了一层 CEO 的自我观察,这在他的访谈里很少见:
"I feel this myself as a CEO at times. Whenever I stop spending time with the product organization, things start to drift. The product becomes hazy. It becomes harder to imagine what we should do, what matters, or what should stay untouched.""我作为 CEO 有时自己也能感觉到。每当我不再花时间在产品组织上,事情就开始漂移。产品变得模糊。更难想象我们该做什么、什么重要、什么该保持不动。"Karri Saarinen, "Tools Are Getting Better. Are We?", 2026-09-25
三周后的 "Someone behind the software" [33] 把"工厂"从组织层面推到个人层面——每个人带着自己的 agent 各自成为一座小工厂:
"Individuals work with agents to produce changes independently, each operating a small software factory. Each change might make sense on its own, but together they start to resemble Frankenstein's monster rather than a carefully shaped sculpture.""个人与 agent 一起独立地产出改动,每个人都在运转一座小型软件工厂。每个改动单独看都可能说得通,但合在一起,它们开始像弗兰肯斯坦的怪物,而不是一件被仔细塑形的雕塑。"Karri Saarinen, "Someone behind the software", 2026-10-15
"I want to feel the creator behind the software, not the factories.""我想在软件背后感觉到创造者,而不是工厂。"Karri Saarinen, "Someone behind the software", 2026-10-15
[推断] 这篇文章里他重新用了 "the soul of a product" 这个词——与第一章 2010 年 "Your Product Needs a Soul" 隔了十六年。当年的敌人是"讨好所有人的功能膨胀",现在的敌人是"每个人各自的 agent 工厂叠加出来的平均值";论点结构没变(产品要有一个背后的人 / 视角),变的是那个把视角稀释掉的机制。这一节与"你不能把 Thinking 外包给 Agent"的区别在层级:那一节讲个人判断,这一节讲组织的学习回路——以及下一节两个具体的仪式;"思考与品味被外包"作为时代诊断,Config 2025 已有(见 §二"关于哲学:care 被反复挤走"),这里新增的是"产品 + 学习"两个产出与按"能否学到东西"划的自动化边界。
关于组织:Quality Wednesday 与 Feature Roast——训练"集体的眼睛",而非清 backlog
第八章的远程工作一节里 "Quality Wednesdays" 只是一条 bullet;Summit 2026 keynote 是他第一次公开讲这两个仪式的机制与初衷。[事实] 起因是招聘带来的标准漂移,以及 agent 时代新出现的"各自与 agent 独处"的孤岛化("a lot of people are working alone with their agents... we're not learning from each other anymore" [00:10:12–00:10:26]):
"This happened because as we keep hiring people, we've noticed that not everyone has the same standard or understanding what the quality means. And so we task everyone to go every week, spend some time, look into the product, try to find one quality defect in the product, and then fix it. It can be very small. Weird hover state like janky animation. It can be a copy error.""这件事的起因是:随着我们持续招人,我们注意到不是每个人都有同样的标准、或对质量意味着什么有同样的理解。所以我们让每个人每周花点时间,去看产品,试着找出一个质量缺陷,然后修掉它。可以非常小——奇怪的 hover 状态、卡顿的动画、一个文案错误。"Karri Saarinen, Lenny & Friends Summit 2026 keynote(Lenny's Podcast feed), 2026-09-29 [00:10:38–00:11:04]
"The bigger impact with this is not really that we do these fixes. We do like that too. The bigger impact is that the team trains, everyone trains themselves to look for this, like they train their eye to notice the small mistakes. And then because we do this in a meeting, everyone shares their findings and their fixes. Everyone can learn from that and see how other people are finding these problems.""这件事更大的影响并不真的在于我们修了这些东西——我们也喜欢修。更大的影响是团队在训练——每个人都在训练自己去找这些东西,训练自己的眼睛去注意小错误。而且因为我们是在一个会议里做这件事,每个人都分享自己的发现和修复。每个人都能从中学习,看到别人是怎么找到这些问题的。"Karri Saarinen, Lenny & Friends Summit 2026 keynote(Lenny's Podcast feed), 2026-09-29 [00:11:12–00:11:32]
第二个仪式针对新功能:
"The second thing we do for new features is this feature roast... It's this optional meeting, anyone in the company can join. The team who is building the feature hosts this. And then what they ask people to do is just critique the whole feature. Like you can be as nitpicky or confused or whatever you want. Put your raw feedback and it's not personal.""我们为新功能做的第二件事是 feature roast……这是一个可选的会议,公司里任何人都可以加入。正在做这个功能的团队来主持。他们请大家做的就是把整个功能批一遍——你可以尽管挑剔、尽管困惑、随便什么都行。把你最原始的反馈放出来,这不是针对个人。"Karri Saarinen, Lenny & Friends Summit 2026 keynote(Lenny's Podcast feed), 2026-09-29 [00:11:36–00:11:58]
"Because a lot of people in a company, this is the first time for them seeing the feature. So if they are confused, it's likely that the users are confused too.""因为公司里很多人是第一次看到这个功能。所以如果他们困惑,用户很可能也会困惑。"Karri Saarinen, Lenny & Friends Summit 2026 keynote(Lenny's Podcast feed), 2026-09-29 [00:12:22–00:12:26]
[事实] 流程收尾:lead 把反馈归类、合成为可执行的 issue。他特别点出的二阶效应是记忆:下次你自己做功能时,会想起"那个人总是在意 onboarding、那个人在意这个动画",于是提前检查,"so I don't get that feedback" [00:12:51–00:13:05]。
[推断] 把这两个仪式放回他的 taste 理论里看:Conversations on Quality(2024)说 intuition 是过往经验训练出来的模式识别、要靠客户交互校准;Quality Wednesday 和 roast 就是组织级的训练集与校准器——前者训练"看见小错误"的眼睛,后者用"公司内第一次看到它的人"模拟用户的困惑。它们都刻意是"一起做"的(会议、分享),对应的敌人正是 agent 时代的孤岛化。这也让 "values 只能被发现"那一节有了操作面:标准不是写在文档里,是每周三被一群人一起练出来的。
关于产品:Design Is a Search, Not a Production Pipeline——保护探索期,别让约束过早定义空间
2025 年 12 月的"设计师应该写代码"论战里,他连发两篇 X 长帖澄清立场:"Design is a search"(2025-12-12)[34] 与 "Part 2: Design is a search for the opinions"(2025-12-15)[35]——两者是 "Design is more than code" [24] 的姊妹篇(档案此前只有源 10 的第三方转载,本轮补回一手)。核心是把设计定义为搜索而非生产:
"I tend to think about design as a search, not a production pipeline. You start with a messy problem. Early on, you do not know the answer. This is why I never fully buy the idea that design is about output. I agree that design is useless without shipping, but the process of designing is not.""我倾向于把设计想成一次搜索,而不是一条生产流水线。你从一个凌乱的问题开始。一开始,你不知道答案。这就是为什么我从来不完全接受'设计就是产出'这个说法。我同意设计不上线就没用,但设计的过程不是没用。"Karri Saarinen, "Design is a search", 2025-12-12
关键区分是约束到来的时机——约束不是敌人,但过早到场会"塑造你的想象力":
"If you let constraints define the space too early, you do not just get a worse outcome. You lose outcomes that never get discovered.""如果你让约束过早地定义了空间,你得到的不只是一个更差的结果。你失去的是那些永远不会被发现的结果。"Karri Saarinen, "Design is a search", 2025-12-12
"Code is a medium of commitment. Designing inside an existing system means inheriting its past decisions. You gravitate toward what is already supported. You make smaller bets because the cost of a big swing becomes obvious immediately. The result is often work that is elegant inside the current system, but less likely to change the system.""代码是一种承诺的媒介。在一个既有系统里做设计,意味着继承它过去的决定。你会被已经支持的东西吸引。你会下更小的注,因为大幅挥棒的代价立刻就显形了。结果常常是在当前系统内部很优雅、但不太可能改变这个系统的工作。"Karri Saarinen, "Design is a search", 2025-12-12
对"统一工具链"愿景的保留意见,是本档案"AI 降低地板、抬高天花板"论点的一个反向补充——工具也可能压低天花板:
"Tools raise the floor, but they can also lower the ceiling if they quietly define what is worth attempting. If the easiest path is always the most conventional path, convention becomes the product.""工具抬高了地板,但如果它们悄悄定义了什么值得尝试,它们也可能压低天花板。如果最容易的路总是最常规的路,常规就成了产品。"Karri Saarinen, "Design is a search", 2025-12-12
"Use whatever tools you want, but be deliberate about what mode you are in. Protect exploration from premature constraint. Invite constraints when you are ready to learn from them. Use code as feedback, not as a cage.""用你想用的任何工具,但要清楚自己处在哪种模式。保护探索不被过早的约束侵入。当你准备好从约束中学习时再邀请它们进来。把代码当作反馈,而不是牢笼。"Karri Saarinen, "Design is a search", 2025-12-12
Part 2 回应的是"用通用 primitives 让每个人自己拼工具"的世界观,他的反驳落在手艺人的需求方向上:
"People and businesses often buy opinionated systems for the same reason majority of the world buys cars, iPhones, or sandwiches. Someone already made a thousand decisions. You get the benefits quickly and move on to what matters, the work, the goal, getting from A to B.""人和企业买 opinionated 系统的理由,和世界上大多数人买车、买 iPhone、买三明治是同一个:有人已经替你做了一千个决定。你很快拿到好处,然后去忙真正重要的事——工作、目标、从 A 到 B。"Karri Saarinen, "Part 2: Design is a search for the opinions", 2025-12-15
"As someone goes deeper into a craft, their needs get more specific, not more generic.""一个人在一门手艺里走得越深,需求就越具体,而不是越通用。"Karri Saarinen, "Part 2: Design is a search for the opinions", 2025-12-15
[推断] 这两篇把档案里散见的几条线接到了一起:第六章 Linear 立项时的 "highly opinionated"(为什么不做可无限配置的工具)、"Suffering as a Clarification Tool"(为什么设计过程中的挣扎本身有价值)、以及 "Output isn't design"(为什么产出不是设计)。新增的是一个可操作的判据——不是"该不该用代码设计",而是"你现在处在探索模式还是承诺模式"。他对 "great unification" 的不信任("often it might be driven by a corporate need to dominate many industries")也顺手解释了 Linear 为什么始终拒绝做 all-in-one。
关于哲学:Taste 是在工作内部行使的判断,不是末端的筛选
2026-02-18 的 "Against 'Against Taste'" [37] 是一篇论战文——回应 Will Manidis 在 X 上"taste 只是 patronage 消失后剩下的消费 / 选择"的论点。它给了档案里此前一直缺的一个定义:
"Taste did not suddenly arrive. Every few years the word comes back into fashion and starts carrying more weight than it should. But taste itself is simple. It is knowing what is good.""Taste 不是突然冒出来的。每隔几年这个词就重新流行一次,开始承载超出它应有的重量。但 taste 本身很简单:它是知道什么是好的。"Karri Saarinen, "Against 'Against Taste'", 2026-02-18
他用来检验这个定义的是一个自测:
"Could I serve a better sushi dinner than Jiro? I am almost certain I could not. No amount of AI, no amount of reading, no perfectly optimized workflow would close that gap... The gap isn't information. It's judgment. Taste is judgment built through experience under constraint. You cannot shortcut it by having access to more tools.""我能做出一顿比小野二郎更好的寿司晚餐吗?我几乎确定我不能。再多的 AI、再多的阅读、再完美优化的工作流都填不上那个差距……差距不在信息,而在判断。Taste 是在约束下经由经验建立起来的判断。你没法靠拿到更多工具来走捷径。"Karri Saarinen, "Against 'Against Taste'", 2026-02-18
两条结构性区分。第一,艺术与设计不同——设计有外部的裁判:
"Product design operates differently since it has external, more or less objective, validators: users or customers. A car has to work. A hospital has to function. Software has to perform reliably and scale.""产品设计的运作方式不同,因为它有外部的、或多或少客观的验证者:用户或客户。车必须能开。医院必须能运转。软件必须可靠地运行并能扩展。"Karri Saarinen, "Against 'Against Taste'", 2026-02-18
第二,taste 不是过程末端的选择,而是过程中的判断——这直接针对"AI 时代人类沦为 discriminator"的恐惧:
"Taste is exercised inside the work, not curated at the end.""Taste 是在工作内部被行使的,不是在末端被策展的。"Karri Saarinen, "Against 'Against Taste'", 2026-02-18
"Humans after AI are not just critics. They are definers of quality. They decide what is worth building in the first place and when it is good enough to ship.""AI 之后的人类不只是批评者。他们是质量的定义者。他们决定什么一开始就值得做,以及什么时候好到可以上线。"Karri Saarinen, "Against 'Against Taste'", 2026-02-18
[事实] 文中他把一部分批评接了下来:纯消费是空的——"Filling an apartment with perfectly 'tasteful' furniture is not necessary taste, it might be just vanity";并顺带批评硅谷"把 taste 当口号、同时为速度 / 增长 / 融资优化"。[推断] 这篇与 "Fit and Feel Good" 节、Conversations on Quality 的 intuition 祛魅构成一个闭合的三角:taste = 知道什么是好(定义);它通过约束下的经验建立(来源);它在工作内部行使并有外部验证者(位置)。对本档案"AI 降低地板"一节的 floor-caveat(没有评估能力则地板不动)来说,这里补的是另一半:有评估能力的人不该只被用来评估。
关于产品:Build 变便宜了,Carry 没有——"Every Feature Should Earn Its Place"
"build 变快 ≠ 该 ship 更多、否则产品臃肿"这个结论档案已有(Grit 2026-04-20,见 §二"更快 ≠ 更多人")。2026-04-12 的 X 长帖 [38] 新增的是背后的成本机制——针对"执行便宜了,为什么不多做、看哪个能活下来",他把成本拆成两段:
"But while the cost of building a feature is going down. The cost of carrying a feature is not.""但是,虽然造一个功能的成本在下降,承载一个功能的成本没有。"Karri Saarinen, "Every feature should earn its place", 2026-04-12
真正的风险不是失败的功能,而是勉强活下来的功能:
"Many teams worry that a feature might fail but failure is usually clear. You can remove it, learn from it, and move on. The harder case is a feature that succeeds just enough to survive. Some users adopt it. A few workflows start to depend on it. The team now owns it. It may not have made the product meaningfully better, but it has made the product larger.""很多团队担心功能会失败,但失败通常很清楚。你可以把它拿掉、从中学习、继续前进。更难的情况是一个功能刚好成功到能活下来。一些用户用上了。几个工作流开始依赖它。团队现在拥有了它。它可能没有让产品变得更好,但让产品变得更大了。"Karri Saarinen, "Every feature should earn its place", 2026-04-12
他对臃肿成因的诊断因此是一个门槛错位:"the threshold for adding something falls below the threshold for living with it"——加一样东西的门槛降到了与它共处的门槛之下(臃肿这一结论本身同 §二"更快 ≠ 更多人")。
[推断] 这是 "Suffering as a Clarification Tool" 的经济学版本:那一节说"做得太容易让你不思考",这一节把"不思考"的代价定价——restraint 过去是被昂贵的执行免费附带的,agent 把它拿走了,于是 restraint 必须被显式地重新装回流程。"Someone behind the software"(2026-10)里的一句可以当它的注脚:"A feature can work perfectly well on its own and still make the product worse."
关于品牌:Brand 是"你是谁",不是 logo——values 不必独特,差别在它多大程度改变决策
2026-10-13 的 "Brand is who you are" [32] 是他第一篇专门谈 brand 的个人文章(此前的品牌材料来自 YC 的 "Brand Design Tips" 访谈与 Airbnb 时期)。
"People talk about it as the logo, the website, the colors. Those are the things you can see, so it's natural to start there. But they are expressions of the brand. The underlying thing is who you are and what you stand for.""人们把它说成 logo、网站、颜色。那些是你能看见的东西,所以从那里开始很自然。但它们是品牌的表达。底下那个东西是你是谁、你主张什么。"Karri Saarinen, "Brand is who you are", 2026-10-13
"The values themselves don't need to be unique. Many companies say they care about quality. The difference is how much that belief affects their decisions. Does it change what they ship or who they hire? Do they put time into something because they believe it should be better, even when nobody is asking?""价值观本身不需要独特。很多公司都说自己在意质量。差别在于这个信念在多大程度上影响他们的决策。它会改变他们上线什么、雇谁吗?他们会因为相信某样东西应该更好就投入时间——哪怕没有人要求?"Karri Saarinen, "Brand is who you are", 2026-10-13
对早期公司的具体忠告,附带了 Linear 自己的例子:
"A small startup looks at an established company and copies its website. The result might look more mature, but it can make the actual company harder to understand. You lose the specific problem, the people, and the reason they care about solving it.""一家小创业公司看着一家成熟公司,照抄它的网站。结果可能看起来更成熟了,但会让这家公司本身更难被理解。你丢掉了那个具体的问题、那些人、以及他们在意解决它的理由。"Karri Saarinen, "Brand is who you are", 2026-10-13
"Early at Linear, we described the product as issue tracking. It gave people something concrete to understand. We could explain what we cared about through how we approached that problem.""Linear 早期,我们把产品描述为 issue tracking。这给了人们一个具体的东西去理解。我们可以通过我们处理这个问题的方式,来解释我们在意什么。"Karri Saarinen, "Brand is who you are", 2026-10-13
[推断] 两处值得交叉引用。其一,他形容"没有回答这些问题就请人做一套漂亮 identity"的结果是 "hollow"——与第一章 2010 年 "Your Product Needs a Soul" 里对功能膨胀产品的用词一模一样,十六年间"有没有一个真实的东西在后面"始终是他判断产品与品牌的同一把尺。其二,"values 不必独特、差别在它是否改变决策"与 Conversations on Quality 的"values 只能被发现"互补:前者说 values 的真伪看决策而非措辞,后者说 values 的来源是公司历史而非会议室。"Early at Linear, we described the product as issue tracking" 也为第九章提供了一个对照——2026 年宣布 "issue tracking is dead" 的公司,正是靠 "issue tracking" 这个具体标签让早期用户理解自己的。
三、数据来源
网络研究(39 篇)
- Linear: Designing for the Developers — Sequoia — Sequoia spotlight,三位创始人的相识与公司起源
- Inside Linear: Why craft and focus still win — First Round Review — opinionated design 与聚焦
- Linear Business Breakdown — Contrary Research — 权威的财务数据与时间线
- Karri's Medium articles index — 个人 Medium 索引
- Starting Linear (karrisaarinen.com) — 2019-04-28 Linear 立项之日的个人散文
- Your Product Needs a Soul (karrisaarinen.com) — 2010-02-12 他最早的世界观成型散文
- Lessons from Karri Saarinen of Linear — Antoine Buteau — 跨访谈高密度引言综述
- The Linear Method — Linear 的公开 manifesto,8 条原则
- Designing Coinbase (karrisaarinen.com) — Coinbase 设计师时期的第一人称回顾
- Design Is Search — AI Creators Media — AI 时代的设计搜索框架
- Coinbase Acquires Kippt — TechCrunch — 2014-05-06 收购报道
- Linear's Path to Product-Market Fit — First Round Review — 10K waitlist、weekly cohorts of ~10 等具体机制
- Rails Girls 与芬兰早期职业(综合 Wikipedia 与多源)— 赫尔辛基初创圈
- Linear blog by Karri index — 完整 essays 列表
- Building our way: Series C $82M / $1.25B — 2025-06-10 Series C 公告 + AI 战略
- The profitable startup — 2025-02-21 资本效率论
- Why is quality so rare? — 2025-05-27 Config 2025 演讲
- Design for the AI age — 2025-04-07 workbench 模型
- Designing remote work at Linear — 2025-10-29 远程组织设计
- Building a Visual Language (Airbnb DLS) — 2016 DLS 文献
- Lenny's Podcast: Inside Linear (Oct 2023) — paid work trials、no A/B testing 等
- How Linear Builds Product — Lenny × Linear — 内部运营机制的最完整公开账户
- Output isn't design — Linear blog — 2026-04-17 Christopher Alexander 框架
- Design is more than code — Linear blog — 2025-12-19 反"设计师写代码"的两阶段过程
- Linear CEO Karri Saarinen declares 'issue tracking is dead' — TBPN Digest — 2026-03-26 AI-era pivot 公告
- Code Intelligence for Linear Agent — Linear blog — 2026-05-14 agent 读 codebase;"code is objective context"
- Now Linear writes the code, too — Linear blog — 2026-06-11 Coding Sessions / Automated Triage / Diffs;近 700 PR 由 agent 合并
- Self-driving SaaS: When software runs itself — Linear blog — 2025-10-22 三级自治框架(上次刷新遗漏的补抓)
- Giving our team liquidity through Linear's first tender offer — Linear blog — 2025-08-27 首次 tender offer,$1.25B 定价(上次刷新遗漏的补抓)
- Sharing Linear's growth with the people building it (karrisaarinen.com / linear.app/now) — 2026-08-26 ⚠️PR 第二次 tender offer $99M / $2.5B、不融 primary、ARR > $100M、NRR 177%、agents 95% 付费 workspace、"will require us to grow"
- Tools Are Getting Better. Are We? (karrisaarinen.com) — 2026-09-25 Summit keynote 的书面孪生:产品 + 学习两个产出;CEO 自述离开产品组织后"产品变模糊"
- Brand is who you are (karrisaarinen.com) — 2026-10-13 brand = 信念的表达;values 不必独特、差别在是否改变决策;早期自称 issue tracking
- Someone behind the software (karrisaarinen.com) — 2026-10-15 个人 + agent 各成小型 software factory → Frankenstein's monster;"the soul of a product" 再现
- Design is a search (karrisaarinen.com) — 2025-12-12 一手版(源 10 为第三方转载):探索期 vs 约束期、code is a medium of commitment、tools raise the floor / lower the ceiling
- Part 2: Design is a search for the opinions (karrisaarinen.com) — 2025-12-15 opinionated = "someone already made a thousand decisions";craft 越深需求越具体
- The disappearing middle of software work (karrisaarinen.com) — 2026-01-03 "issue tracking is dead" 论点的书面雏形;"directing and managing agent work becomes the craft"
- Against "Against Taste" (karrisaarinen.com) — 2026-02-18 回应 Will Manidis;taste = knowing what is good;Jiro 测试;design 有外部验证者;taste exercised inside the work
- Every feature should earn its place (karrisaarinen.com) — 2026-04-12 building cost ↓ 而 carrying cost 不变;"succeeds just enough to survive"
- Some Notes on AI (karrisaarinen.com) — 2026-04-25 planning 是对齐与承诺的练习;"hard problems remain hard";自家 cloud agent 每月修 > 1,000 issue(expertise paradox 与"AI 降低地板"floor-caveat 重叠未折)
播客访谈(15 期)
| 节目 | 主持人 | 集数标题 | 日期 | 重点引用 |
|---|---|---|---|---|
| Lenny's Podcast | Lenny Rachitsky | Inside Linear: Building with taste, craft, and focus | 2023-10-08 | work trial 的双向去风险机制、干净 codebase 作为人才磁铁、"最自豪的是什么"面试题(2026-07 Podwise 处理完成后补入;main-quest/side-quest、go-slow-to-go-fast 等与既有内容重叠部分未重复折入) |
| Grit | Joubin Mirzadegan | From Airbnb to Linear: How Karri Saarinen Redefined Product Design | 2026-04-20 | 更快=focus/clarity 非人多、VC 融资的 signaling 双刃、40 小时工作制/反 grind、Sisu 与不 brute-force、成为父亲→领导以身作则、好投资人=不挡道 |
| EO | EO Studio | Future of Product Management From the Era of AI | 2023-11-18 | local-first sync engine(工具"感觉像原生 app 而非迟钝网站")、engineers = source of truth、最早期 founder 时间纪律(无视 legal/accounting/networking,"do we really need to do this now")、private beta 的 pay-what-you-want 价格滑块 |
| 20VC | Harry Stebbings | How to Grow Capital Efficiently in a World of BS Growth | 2024-10-30 | quality growth 隐喻、investor "homework"、Brian Chesky 影响、$30K 营销总支出、Coinbase Exchange 一个工程师故事 |
| Y Combinator Startup Podcast(此前误标 Lenny's Podcast;dup feed 6429404 "The Vault") | YC | How To Design Products That Truly Stand Out | 2025-10-14 | quality ≠ perfection、design as "finding fit"、bike-shedding、small teams、Coinbase Bootstrap 故事、AI 降低地板/抬高天花板、craft 的 suffering |
| Every / AI & I(此前误标 Lenny's Podcast;dup feed 7673574 "If SaaS Is Dead, Linear Didn't Get the Memo") | Dan Shipper | Most SaaS Companies Got AI Wrong. Linear Waited. | 2026-04-01 | self-driving SaaS、kitchen-ticket 隐喻、deciding slow / executing fast、Linear 作为 product memory、skills、agent session 可见性 |
| Y Combinator(YouTube 视频;含重复 feed 4799984 / 4963808;此前标为"Brand Design 播客") | YC | Brand Design Tips From Linear Founder Karri Saarinen | 2025 | 品牌 V1→V2→V3 演化、Stripe 比较、Airbnb 品牌教训 |
| 开发者工具播客(Tom Moor 主持) | Tom Moor | Why Linear Built an API For Agents | 2025 | Linear agent platform 技术细节 + Karri 的战略定位 |
| TBPN Live (Karri 片段) | TBPN Live(Andreessen / Sutskever 嘉宾主持) | TBPN Live (Karri segment) | 2025 | self-driving SaaS 的独特论述 |
| In Depth (First Round) | Brett Berson | Inside Linear: Why craft and focus still win in product building | 2025-06-05 | "与在位者反着做"定位启发式、Kippt 商业模式教训、设计师创始人的 validation gap、每周三酒吧验证会(craft/focus/small-teams 主干与 web 源 02 同集文章重叠,未重复折入) |
| Startup Grind(Sequoia × Linear fireside,YouTube 导入) | Stephanie Zhan | AI Agents and the Future of Software Teams | 2026-05-05 | "dangerous tool because it also does thinking"、make the AI work for you、devil's advocate 反谄媚用法、agent 当 40K 客户请求上的思考伙伴、"AI 抬地板不总成立"(法律工作类比)、agent 工作量季度 5 倍 |
| Product School | Carlos Gonzalez de Villaumbrosia | The tool product managers love and is disrupting Jira (E234) | 2024-09-04 | 与 IC 结盟而非 buyer、销售时序(先内部 activity 再谈 pilot)、无问卷的正面度量(churn/YC batch 渗透率)、六个月 roadmap + 年度主题 |
| Conversations on Quality(⚠️ Linear 自办) | Linear | Karri Saarinen – Conversations on Quality Ep 02 | 2024-10-12 | values 不能被发明只能被发现(从公司历史里浮现)、intuition = 可训练的模式识别 + 客户交互校准(仅折机制类内容,自我叙事已 PR 折扣) |
| Lenny's Podcast(Lenny & Friends Summit 2026 keynote) | —(17 分钟独讲,无主持;形态例外收录) | Context is now the product: Product leadership when software can build itself(YouTube 版后改题 "Why I took the summer off from AI (and what I learned)") | 2026-09-29 | software factory 禁令、产品工作的两个产出(产品 + 学习)、自动化边界按"能否学到东西"划、context becomes the product、Quality Wednesday 与 feature roast 机制、watcher agent daily briefing、夏天停看 AI 的实验(intuition 祛魅、taste 招聘与既有内容重叠未重复折入) |
| TBPN(多嘉宾日播,仅采用可归属 Karri 的 segment [02:19:55–02:33:30]) | John Coogan / Jordi Hays | Arm Pumps CPUs, Social Media Addiction, Data Center Ban ... Karri Saarinen ... | 2026-03-27 | 第九章两条 TBPN Digest 转引以原始音频校正("bottleneck shifts" 一句系 Digest 编辑概括)、40 状态 / 30 字段迁移请求的拒绝、"你发明的是 Kanban 板"、77% 为 cloud agent 口径(kitchen 隐喻、agent 修 bug 流程、coding agent 为 wrapper 与既有内容重叠未重复折入) |
未使用但已扫描的(质量过滤掉的 / Podwise 未处理的)
- "How Linear's Karri Saarinen is Redefining What Scale Looks Like" (Accel spotlight, 2024-11-05, episode 5313910/7525175) — Podwise 未处理;与 20VC 内容重叠,不主动追
- "Creating magical software" (Changelog Founders Talk #94, 2023-03-11, episode 117697) — 嘉宾是联创 Jori Lallo 而非 Karri,排除
- "TBPN Live Karri Saarinen 多嘉宾" (2025-04-26, episode 3741427) — 多嘉宾节目;Karri segment 已通过 episode 5745282 单独捕获
- "Marc Andreessen & Ilya Sutskever Host TBPN" (TBPN, 2025-11-01, episode 7023644) — 疑似即已折入的 5745282(Karri segment,档案表注明 Andreessen/Sutskever 嘉宾主持)所属完整集的另一 feed;多嘉宾长节目,未重复折入
- "Config 2025: Crafting quality that endures" (Figma, 2025-05-13, episode 4140634) — 会议演讲录音;内容已由 "Why is quality so rare?" 文字稿 [17] 覆盖
- Pending(2026-10-18 提交 Podwise 处理,下轮步骤 0 回收):"Building for Quality in a World of AI Slop with Linear's Karri Saarinen"(Founders in Arms,Immad Akhund & Rajat Suri,2026-05-22,54 分钟,YouTube Nuzt8o46mFQ → episode 9150166);"Hacking design and not dying"(Startup Sauna / Slush 2012,Karri 与 Jori Lallo,2012-11-25,13 分钟演讲,YouTube 54v3kqcmDGQ → episode 9150170;Kippt 时期唯一同期一手材料,用于对冲第三章的后视镜偏见)
- "If SaaS Is Dead, Linear Didn't Get the Memo"(AI and I feed,episode 7673574 / 8627717)— 与已折入 7685615(Every,2026-04-01)为同一集 Dan Shipper 访谈的另一 feed
- "How To Design Products That Truly Stand Out"(Y Combinator: The Vault feed,episode 6429404,2025-10-13)— 已折入 5431244 的重复 feed
- TBPN 多嘉宾集:2025-12-17(episode 6464138 / 7023579,Karri 段为 design-vs-code,与 [24] 重叠)、2025-06-11(episode 4288696 / 4311840,Series C 公告日 PR 段,已由 [15] 覆盖)、2025-04-26(episode 3941409,为已排除 3741427 的重复 feed)— 未处理、不提交
- "How Linear Builds AI Agents"(MLOps.community,episode 6737070,2025-11-30)— 15 分钟短片、嘉宾两轮未能确认为 Karri;按短片排除,结束此前两轮的悬置
- 中文转述节目:野格知识贩子 047(episode 9024401,2026-09-30)与 2025-07-28 集(4809482)、跨国串门儿计划 #484(7703128)— 翻译转述,排除
- "Définir ses priorités Produit..."(Clef de voûte,episode 5428903,2025-10-13)— 法语多创始人合集,Linear 发言人很可能是 Adrien Griveau 而非 Karri
- 同公司噪音:Cristina Cordova(Product Podcast E299,8202594)、Tom Moor(5617481)、Nan Yu(Lenny,3024573)、Jori Lallo(Founders Talk feed 182439 = 已排除 117697 的重复)
- 20VC 2026-10-12 Atlassian CEO Mike Cannon-Brookes(episode 9107467 / 9109251)— Karri 仅被提及;20VC 2026-09-03 新闻 roundup(8815267)— 非嘉宾出镜,仅作 $2.5B 标记的二手佐证
- YouTube 短片:Founders in Arms 6 分钟剪辑(ntqqV-Jb2yo,源自上述 54 分钟集)、Startup Archive 2 分钟(F9EVetIQcP0)— 剪辑排除
- 可选 backfill(非 Podwise 索引、< 30 分钟演讲):Builders' Studio by Slush #21 "Product Design"(2022-02-13,27 分钟,6KpWP8WC71w)、Nordic.design 2017 "Scaling design with systems"(25 分钟,TuLY1cYM57g)— 预计与第五章 DLS / 第七章产品设计 canon 重叠,仅在 deep refresh 时考虑
- Bryce Roberts "At Home with Karri Saarinen"(p3tHGu84vOg)— 2026-08-04 判定观感向、价值低,维持
- 网络侧已见未折:"This website is written and built with Linear"(karrisaarinen.com,2026-10-16)— 个人站点迁入 Linear 的 dogfooding 记录,无新角度;印刷访谈待跟进:Entrepreneur "Rejecting Silicon Valley's Hustle Culture"(2025,§40 小时工作制目前仅引 Grit)、Runtime "How Linear became the new Jira"、Figma "10 rules for crafting products that stand out"、Business of San Diego substack 2023-10-25(迁居 San Diego;疑付费墙)
- 标准发现源(本轮起):https://karrisaarinen.com/interviews/(他本人维护的全部访谈 / 演讲清单)与 https://karrisaarinen.com/rss.xml(essay 权威索引;linear.app/now 为 JS 渲染且 sitemap lastmod 过期)