--- author: EO date: '2026-10-08' guest: '' layout: post.njk source: https://www.youtube.com/watch?v=REsrgNeuZkU speaker: EO tags: - physical-ai - autonomous-driving - product-market-fit - business-model - engineering-mindset title: 从底特律工程师到150亿独角兽：Applied Intuition创始人谈极端实用主义与物理AI商业化 summary: Applied Intuition创始人Qasar Younis分享了打造估值150亿美元物理AI公司的核心心得。他提出创业如同手工艺需长期锤炼，主张将商业模式深度融入底层技术，以横向工具平台赋能全行业，并阐述了工程思维、动态PMF以及面对恐惧的心力管理。 insight: '' draft: true series: '' category: ai-application area: tech-engineering project: [] people: - Qasar Younis companies_orgs: - Applied Intuition products_models: [] media_books: [] status: evergreen --- ### 匠人视角与物理AI：将智能赋予十亿台机器 初次创业者往往将打造一家公司视为一次性任务，但真正的创业更像是一名手艺人打磨工艺。如果建立起长期投入的心理视角，明白自己可能会经历多次创业，对创业的理解就会截然不同。许多人认为**物理AI**（Physical AI: 具备感知、推理并在物理世界执行动作的智能系统）主要指人形机器人，这固然是一种引人注目的形态，但汽车、卡车以及重型工业设备能够更早地被赋予智能，并更快产生实质性社会价值。 以日本道路上实际自主运行的自动驾驶卡车为例，核心使命是将智能赋予全球数十亿台机器。当前大多数机械设备依然相对迟钝，高度依赖人工操作；一旦注入智能使其自主运转，所创造的生产力价值将呈指数级提升。目前，Applied Intuition 已成长为拥有近1500名员工、估值达150亿美元的企业，产品落地于全球数十个国家，涵盖广泛的制造厂商与垂直行业。
Original English Source First time founders tend to think that building a company is a one time thing and really actually building a company is like being an artisan or holding a craft. I think if you have the mental perspective that you're going to do this for many, many years, and you might do it two or three or four times, then you actually approach starting a company slightly differently. And it's great, if first company works, then often times it doesn't work. Often company, I spent three years on. It was successful, some of the biggest lessons I learned that make Applied Intuition, you know, much more successful came from that company twenty years ago. So what makes a company succeed versus company fail? The biggest mistake I've made in terms of starting companies is, but it was Applied Intuition. I built Applied Intuition, before that I was COO at Y Combinator, and before that, I founded a company which was acquired by Google. And at the beginning of my career, I worked as an engineer at General Motors and Bosch. Applied Intuition is a physical AI company. A lot of times people think physical AI is just humanoids, and humanoids is a very interesting form factor of physical AI. But I actually think cars and trucks and industrial equipment is something that can be made more intelligent and have a bigger impact sooner. This is the gig truck. It has some self-driving hardware on there. These trucks are running live on Japanese roads right now autonomously. Our mission is to put intelligence on a billion machines. The only way you do that is you take the same platform and you put it on lots of different machines. And why do you want to put intelligence in those machines? Because those machines right now are pretty dumb, they require human beings to operate them. If you put intelligence in those machines, they can function themselves. And if they can function themselves, they provide a lot more value. The company is over a thousand engineers or almost 1500 people, the company is valued at 15 billion dollars. Our products are deployed in dozens of countries around the world in lots of manufacturers in many verticals.
### 横向平台战略：化解垂直全栈的高昂试错风险 在汽车工业衰退的阴影下步入行业，能清晰感知到依附于萎缩产业的职业风险，这也是选择自主创业以掌控命运的起点。2017年创立 Applied Intuition 时，自动驾驶行业的前景并不明朗，当时的**Transformer**（Transformer: 基于自注意力机制的深度学习模型架构）等核心基础技术尚未在自动驾驶中普及，行业面临巨大的研发不确定性。当时许多公司采取垂直整合策略，试图自行搭建从数据引擎、专用工具乃至车辆本体与 Robotaxi 车队的全栈系统，导致资本消耗极为高昂。 在各垂直细分市场胜负未定之时，垂直自动驾驶公司往往将大量资源耗费在终端客户并不在意的边缘环节上。相较之下，横向基础设施平台的抗风险能力更为优越。例如在数据标注环节，每家车企都能自研标注工具，但购买自动驾驶系统的终端买家根本不在意标注软件出自何处。与其承担巨大的资本与市场双重风险去做封闭的垂直方案，不如横向提供核心软件与工程工具平台，让主机厂与自动驾驶团队专注于解决其最具价值的核心难题。
Original English Source Now it's more famous as the place where the industry has already started to die. I grew up very much in the shadow of the American car industry losing its position as the best in the world. So I'm entering as an engineer in the industry which was in a very volatile situation. My reflection on that was if I stay in this industry and industry shrinks that can be very bad for me as an engineer. So part of my ambition to become a founder was to control my own destiny. I left really great jobs at General Motors and Bosch, really great companies, to make my own company. It seems like a pretty risky thing, but twenty years later, it was the right thing as cars can drive themselves. It still sounds like science fiction, but guess what, it is here. The industry was very different when we founded Applied in 2017. In self-driving, people were not clear how self-driving would be built. It wasn't clear self-driving would happen. It seemed like a very research-oriented field, a lot of the AI breakthroughs hadn't happened. And then this is really important: the Transformer and this new architecture, which the large language models are really benefiting from, also has a lot of impact in self-driving. All of these building blocks didn't exist. And maybe the biggest point was, we weren't sure if self-driving would ever happen. I think at that time, lots of self-driving car companies, many that don't exist anymore, were doing actually what some of the AI companies today are doing. One thing is they would try to go vertical. They'd try to build everything: the tools, the data engine, sometimes even the vehicles themselves. They're going to do robotaxis, they're going to do trucks. That becomes very, very expensive. At that time, we felt pretty strongly it's not obvious which vertical is going to work, and it's not obvious that being a vertical self-driving car company is actually valuable. A lot of times companies will spend a lot of their resources and energies on things that actually the end customer doesn't really care about. This is the inside of a car. This is kind of a classic inside of a vehicle, and this is using some of our platform. The big difference between these two is there's a lot fewer components here. As you make a vehicle intelligent, you also want to change the platform that it's on. Traditional platforms are not really made for autonomy or intelligence; each component is separate. In newer vehicles, we can reduce a lot of that onto individual compute, put intelligence on there, and handle non-intelligent tasks like rolling up windows or playing music. The downstream impact is just a lot fewer components. The most exciting part is taking the same platform and putting it on so many different vehicles. If you're building this type of platform just for one vertical, it might never happen because it's too expensive. In most industries, horizontal companies actually do really well. Take data labeling: every company can make a data labeling product for themselves, but the end customer who buys the self-driving system doesn't care who made the data labeler. Fast forward to today, there's a lot more horizontalization, allowing you to buy parts of the stack and focus on the most valuable core problem. The 2017 vertical approach required massive capital and carried both capital and market risk, which is why the vast majority of those companies did not survive. We chose to provide the tools and let everyone else figure out self-driving, which proved to be the right move, especially since vehicle manufacturers always intended to build much of that technology internally.
### 极端实用主义：将商业模型熔铸进技术研发 当前不少 AI 企业投入巨额资金并不直接等同于商业成功，优秀的创业者必须兼顾技术攻关与商业变现。汽车原始设备制造商（OEM）身处极度严酷的竞争环境，每一笔开支都经过精打细算，因此技术方案必须具备不可替代的明确价值。最关键的反共识观点在于：商业模式必须内生于技术架构之中，绝不能在技术开发完成后再行追加——这如同房屋建好后再铺设管道一样困难。 这种业务哲学的核心被概括为**极端实用主义**（Radical Pragmatism: 以客观度量与真实商业价值为唯一导向的决策哲学），亦即高度审慎与目的明确（Intentionality）。从工厂流水线的管理经验出发，一切工序、质量与产出皆可被量化度量。与主观的艺术不同，物理学与工程学遵循绝对标准，非黑即白： * **量化与归因**: 无论是评估产品战略、人员招聘还是设立分支机构，均需明确书面记录需求痛点、竞品优劣、采购动机及支付意愿。 * **客观复盘**: 详尽的逻辑推演能够确保在决策失误时，团队能够迅速定位逻辑断点并纠偏。
Original English Source One of my earliest engineering managers had a sign in his office that said: "No problem can withstand the constant onslaught of thought." You see AI companies today spending a lot of money; that doesn't necessarily mean that they will be successful. It doesn't mean they are automatically going to be unsuccessful, but it's more risk. The best founders are thinking about the whole problem: part of it is engineering and technology, but the other part is business. You want to make sure both those things are working together. Having a fair amount of focus on commercialization earlier than you think you should be thinking about it is important. Automotive OEMs work in an extremely competitive industry. Growing up around that taught me that when you're in a competitive market, you are very thoughtful about how you spend money, so there has to be very clear value. There's no company on the planet that does physical AI in the breadth that we do as a horizontal company providing that intelligence layer. My biggest contrarian view is that you have to build a business model into the technology, and the technology built around a business model; you can't add it on afterwards. It's very hard to put in plumbing after the house is built. Working as an engineer and spending time in factories reinforced that even a company is a system. A factory is heavily process-driven: throughput and efficiency can all be measured. Physics and material science are absolute—a product either functions as designed or it doesn't. One door design is objectively better than another based on hinges, knobs, and heat transfer. Bringing that engineering mindset to business means objectively answering whether something is a better product strategy, a better hire, or a better investor. You can boil our values down to two words: radical pragmatism. Another way of saying that is being intentional. Who becomes an investor, what product we enter, whether we open an office in Korea—we don't just open an office and hope for business. We write down what customers need, what they're good or weak at, why they'd buy from us, and what they'd pay. Writing things down and intentionality are the same thing; even if you make a wrong decision, you can understand where the mistake was in your logic.
### 合伙人动力学与PMF动态区间 在 Y Combinator 担任 COO 的经历提供了观察数千家初创企业的宏观视角。合伙人关系的构建类似于伴侣关系：两人或三人的小团队在技能互补、管理领导风格、反馈机制及野心规模上高度契合，往往最为稳固；四人及以上的团队则会因决策沉重与沟通摩擦显著增加解体概率，而单打独斗则极难涵盖初创期所需的全部技能。此外，选择正确的市场容量至关重要——正如在烈日炎炎下卖冰块远比在严寒地区容易，选错市场往往让卓越的团队徒劳无功。 初创团队常将**产品市场匹配度**（Product-Market Fit: 产品功能精准契合目标市场真实需求的平衡状态）误认为是一劳永逸的终点，但它本质上是一种处于动态变化中的“状态区间”。即使年营收达到千万美元量级，面对技术更迭与竞品涌现，PMF 也必须时刻重新验证： * **警惕虚假指标**: 单纯的应用下载量绝不代表 PMF，若缺乏持续活跃度与高粘性留存，庞大数字毫无意义。 * **常态化穿透**: 如同可口可乐必须持续进行品牌投入以维系消费习惯，真正的 PMF 判定依赖于持续深入客户一线：产品是否切实接入生产流程、是否真正改变了客户的项目成效，以及客户是否愿意为此持续投入时间与预算。
Original English Source Y Combinator taught me a lot of things I couldn't have learned just by being a founder. I saw thousands of companies. If you're in a relationship, you only have that experience, but a couple's therapist sees many relationships. YC is kind of like couples therapy: you start seeing the patterns of functioning co-founder relationships. A good co-founder relationship is where two or three people complement each other really well. I would never recommend starting a company with four people or more—there are too many people likely to leave, and logistics make decision-making hard. I also wouldn't recommend doing it alone, because multiple founders increase the likelihood of covering all needed skills. Emotional compatibility—how you lead, give feedback, and talk to each other—and aligned ambition are critical. You can't have one founder wanting a generational company while the other just wants to get rich quickly. Picking the right market is equally vital. I saw hard-working, smart teams fail simply because they picked the wrong market. Selling ice in a cold area won't yield many buyers; selling ice on a hot day will. The mistake first-time founders make about product-market fit is thinking it's a permanent destination. PMF is really a state that can slip away. Proxies include whether people give you time and money, but you should remain cynical. Even at $10M in revenue, I wasn't complacent, because your product sits in a dynamic market where substitutes constantly emerge. Coca-Cola spends heavily on marketing because consumer preference must be earned every day. Instead of declaring static PMF, we ask: are we in the PMF zone? Are customers using the product in production and seeing real impact? Millions of app downloads mean nothing if users never return. Founders must be brutally honest about whether they remain in that zone.
### 心理防线与职业定力：在恐惧与沮丧之间做抉择 持续的高强度阅读是构建底层认知与宏观视野的重要基石，深入研读长篇经典著作能够培养系统性深度专注力。从印度哲学经典《I Am That》关于超然观察的智慧，到山姆·沃尔顿自传《富甲美国》、甘地自传《我体验真理的经历》以及曼德拉的《漫漫自由路》，深刻的人生记录远比快餐式碎片信息更具精神滋养。 这种追寻自我命运掌控权的执念，根植于从巴基斯坦移民至美国的劳工阶层成长背景。长期身处对生计账单的不确定性中，锤炼出对恐惧的极高耐受阈值。评估一个人适宜成为大企业骨干还是创业者，本质上取决于其对负面情绪的承受类型： * **大企业的代价是挫败感（Frustration）**: 在大型机构中，稳定的薪资消除了生存恐惧，但日常面对的是工作未被赏识、推进受阻与体制约束带来的挫败感。 * **创业者的代价是恐惧（Fear）**: 创业者摆脱了层级汇报的挫败感，但必须直面公司随时可能倾覆的存亡恐惧。 易被体制与官僚激怒、但具备极高恐惧抗性的人，更适合成为创业者；而面对不确定性极易惊慌失措的人，则应选择深耕大型组织。
Original English Source I don't believe you can be very successful and not read a lot. I'm reading a book called "I Am That," an Indian spirituality book about Hindu philosophy on detaching from the world around you to see things more clearly. Reading hundreds of pages requires continuous focus. My book list includes Sam Walton's "Made in America," Gandhi's autobiography "My Experiments with Truth," and Nelson Mandela's "Long Walk to Freedom." Reading those makes other contemporary books look like disposable chocolate. The root of my ambition is to control my own destiny. Coming from Pakistan to America as a child in a working-class family meant fear was part of our everyday existence—worrying about bills and making ends meet. That gave me a high fear threshold. One way to find out if you're suited to be a founder or a company person is to ask how much having a boss bothers you. At a big company like Bosch, the day-to-day cost is frustration: does my boss acknowledge my work? Can I work on what I want? Is the company heading where I want? The downside of being a founder isn't frustration—you pay with fear, because you don't know what will happen to the company. Big companies remove fear by guaranteeing salaries, but bring frustration. If you get frustrated easily, maybe consider being a founder; if you get scared easily, staying at an established company is likely the better path.