YC Partner Chaubard Tells PhD Founders Academic Rigor Kills Startups

Y Combinator

Academic spinouts routinely fail because incentives that drive citations and peer review penalize the speed, commercial aggression, and empirical customer validation required to survive.

AI-assisted summary of How To Go From Researcher to Founder | Startup School — original source. Check the original for full context.

Source published: 2026-11-04

Key ideas

Academic founders frequently fail because behaviors rewarded in research—chasing novelty, seeking perfection, and qualifying statements—directly conflict with building a commercial startup. (0:09)
Commercial execution beats technical novelty: fast followers like Instacart and Oracle routinely win massive markets by solving mundane operational problems faster than pioneers with patented inventions. (2:26)
AI founders must validate customer willingness to pay using human-in-the-loop tests before spending significant capital and time training models. (4:29)
Founders must embrace Jeff Bezos's 70% certainty rule, avoid outsourcing customer-facing operations, and adopt aggressive commercial instincts to survive competitor threats. (6:28)

Detailed notes

## Background Published on YouTube by Y Combinator on November 4, 2026, this Startup School video features YC visiting partner, PhD student, and former Focal CEO Francois Chaubard. Drawing on his experience transitioning from Fei-Fei Li's lab at Stanford to running a company, Chaubard breaks down the critical mindset shifts researchers must make to avoid startup failure and become formidable founders.

## TL;DR While academic researchers frequently possess exceptional intellect, their training in chasing citation metrics, theoretical novelty, and perfectionism actively undermines their success as entrepreneurs. To build a viable company, founders must abandon academic vanity and embrace extreme empiricism, prioritizing practical customer pain points and rapid execution over technical complexity. Founders cannot delegate sales or stay insulated in research labs; they must leave their desks, validate whether users will actually pay before burning capital, and adopt an aggressive capitalist mindset where speed consistently beats pure academic elegance.

## Atmosphere & style The presentation functions as a direct, tough-love masterclass that systematically dismantles the intellectual vanity and comfort zones of academia. Chaubard speaks with pragmatic, battle-tested conviction, using his own painful mistakes and sharp historical contrasts to deliver an urgent wake-up call. The delivery feels unpretentious yet intense, transforming from a sobering reality check into an energizing, high-stakes challenge that demands grit and commercial hunger over scientific pedigree.

## Bulletin - [00:09](https://www.youtube.com/watch?v=u-8KpXcPYCw&t=9s) Academic founders frequently fail because behaviors rewarded in research—chasing novelty, seeking perfection, and qualifying statements—directly conflict with building a commercial startup. - [02:26](https://www.youtube.com/watch?v=u-8KpXcPYCw&t=146s) Commercial execution beats technical novelty: fast followers like Instacart and Oracle routinely win massive markets by solving mundane operational problems faster than pioneers with patented inventions. - [04:29](https://www.youtube.com/watch?v=u-8KpXcPYCw&t=269s) AI founders must validate customer willingness to pay using human-in-the-loop tests before spending significant capital and time training models. - [06:28](https://www.youtube.com/watch?v=u-8KpXcPYCw&t=388s) Founders must embrace Jeff Bezos's 70% certainty rule, avoid outsourcing customer-facing operations, and adopt aggressive commercial instincts to survive competitor threats.

## Key takeaways - [00:09](https://www.youtube.com/watch?v=u-8KpXcPYCw&t=9s) Breakout companies like Google, Anthropic, and Databricks are rare exceptions among academic spinouts; most fail due to misaligned instincts. - [00:53](https://www.youtube.com/watch?v=u-8KpXcPYCw&t=53s) Researchers must kill their academic identity and abandon vanity metrics like citations, GPAs, and conference invites in favor of product-market fit and paid customer demand. - [01:45](https://www.youtube.com/watch?v=u-8KpXcPYCw&t=105s) Startups require prioritizing what is commercially important over what is intellectually interesting, accepting that code that 'just works' should ship without needing theoretical explanations. - [02:26](https://www.youtube.com/watch?v=u-8KpXcPYCw&t=146s) Academic peer review requires novelty, but startups succeed by unashamedly copying and executing established business models better than pioneers. - [03:54](https://www.youtube.com/watch?v=u-8KpXcPYCw&t=234s) Startups have no semester deadlines or finish lines, requiring relentless, single-minded focus on a single operational problem for years rather than hedging across multiple speculative projects. - [04:29](https://www.youtube.com/watch?v=u-8KpXcPYCw&t=269s) Product validation cannot occur in an academic office or through venture capitalists; founders must speak directly to prospective buyers. - [04:58](https://www.youtube.com/watch?v=u-8KpXcPYCw&t=298s) Focal wasted three months training retail tracking models that customers refused to pay for, an insight that could have been uncovered in a day using manual video review. - [05:53](https://www.youtube.com/watch?v=u-8KpXcPYCw&t=353s) Scientific pedantry and qualified claims ruin commercial messaging; effective marketing requires punchy, simplified statements rather than academic rigor. - [06:28](https://www.youtube.com/watch?v=u-8KpXcPYCw&t=388s) Under Jeff Bezos's 70% rule, founders must make decisions with incomplete information, as waiting for complete certainty allows faster competitors to capture the market. - [06:52](https://www.youtube.com/watch?v=u-8KpXcPYCw&t=412s) Larry Ellison's commercialization of SQL over Michael Stonebraker's technically superior QUEL demonstrates that market velocity consistently beats academic elegance. - [07:11](https://www.youtube.com/watch?v=u-8KpXcPYCw&t=431s) Founders cannot delegate messy tasks like sales and marketing to MBAs, as doing so prevents the direct customer empathy necessary to build a viable product. - [07:52](https://www.youtube.com/watch?v=u-8KpXcPYCw&t=472s) Academics must adopt an unapologetic capitalist mindset that embraces high gross margins, customer acquisition, and aggressive competition. - [08:16](https://www.youtube.com/watch?v=u-8KpXcPYCw&t=496s) Prospective clients and procurement officers do not care about a founder's raw intellect; successful sales require genuine curiosity about the customer's specific problems. - [08:52](https://www.youtube.com/watch?v=u-8KpXcPYCw&t=532s) Founders must physically travel into the field to meet users in their work environments, testing demand before competitors seize the opportunity. - [09:20](https://www.youtube.com/watch?v=u-8KpXcPYCw&t=560s) Raw brainpower is not the limiting factor for academic founders; the decisive bottleneck is personal willpower and the willingness to leave one's comfort zone.

## Detailed walkthrough ### [00:09](https://www.youtube.com/watch?v=u-8KpXcPYCw&t=9s) Killing the Academic Identity: Vanity Metrics and User Importance

Some of the most valuable technology companies in modern history—including Google, Anthropic, Databricks, and Together AI—were founded by researchers with PhDs. Yet these breakout successes remain the exceptions. In reality, academic startups fail at an alarming rate because the exact behaviors that win accolades in research are often orthogonal, or directly antithetical, to what it takes to succeed as an entrepreneur. Francois Chaubard experienced this tension firsthand: before launching his startup Focal, he worked as a researcher in Fei-Fei Li’s lab at Stanford, only to realize later that the instincts drilled into him during those late, coffee-fueled nights in the lab actively held him back as a CEO. Bridging this chasm demands more than learning a few business terms; as Francois Chaubard argues, “To become a formidable founder, you will literally have to kill part of your academic identity, which is quite difficult to do.”

```mermaid flowchart TD accTitle: Academic Mindset vs. Founder Mindset accDescr: Contrasting academic research metrics with founder operational priorities. subgraph Acad["Academic Mindset"] direction TB A1["Vanity Metrics: Citations, talks, grades"] --> A2["Interesting Research: Uncover deep mechanisms"] A2 --> A3["Theoretical perfection before publication"] end subgraph Found["Founder Mindset"] direction TB F1["Commercial Value: What people pay for"] --> F2["Important Solutions: Solve real user pain"] F2 --> F3["Empiricism: 'It just works, ship it'"] end ```

The first major hurdle is abandoning vanity metrics in favor of genuine demand. In university environments, researchers spend years chasing proxies for prestige: GPAs, citation counts, accepted submissions, and invitations to give talks. Reflecting on this institutional conditioning, Francois Chaubard quotes Paul Graham: “The most damaging thing you learned in school wasn't something you learn in any specific class. It was learning to get good grades.” Founders must abandon these artificial scorecards and obsess over commercial reality: what do people actually want, what will they open their wallets to purchase today versus merely claiming they would buy, and will users genuinely adopt a new algorithm into their daily workflows?

This distinction flows directly into the second shift: prioritizing what is important over what is merely interesting. Researchers are trained to dissect mechanisms and probe theoretical nuances to earn publications, even when those findings have zero tangible impact on a single human life. Startups cannot afford intellectual indulgence; founders must care exclusively about product-market fit and working backward from user pain points. As Francois Chaubard points out, founders have to embrace pure empiricism: “In founder mode, you have to be okay with saying it just works. We don't know exactly why, but that's good enough. Ship it.”

- PhD-led giants like Google, Anthropic, Databricks, and Together AI are rare outliers among academic spinouts. - Academic success optimizes for proxy credentials—citations, grades, and prestige—rather than market adoption. - Marc Andreessen’s core principle holds that the only thing that matters is getting to product-market fit, meaning theoretical explanations matter far less than whether a product reliably solves a user's problem.

---

[02:26](https://www.youtube.com/watch?v=u-8KpXcPYCw&t=146s) Novelty, Plagiarism, and Working Without a Finish Line

In academic research, novelty is the sole currency. Peer reviewers reflexively reject any paper that lacks fundamentally original work. In entrepreneurship, however, an obsession with originality is often lethal, and shameless plagiarism of working models is a proven path to victory. Pioneers routinely end up with arrows in their backs while fast followers capture the market. Francois Chaubard references Joe Lonsdale's distinction between "MIT-style" startups—which revolve around an unprecedented, patented black-box invention that few customers actually want—and "Stanford-style" startups, which anchor on a simple, often mundane operational problem and iterate on it relentlessly. Companies like Instacart have spent fifteen years focusing on the unglamorous problem of grocery delivery, an idea attempted by dozens of companies before them; their victory came from execution, not pioneering the concept.

```mermaid flowchart LR accTitle: MIT-Style vs. Stanford-Style Startups accDescr: Comparing technology-push startups to customer-pull startups. subgraph MIT["MIT-Style: Technology Push"] M1["Patented Invention / Black Box"] --> M2["Search for Applications"] --> M3["Low Customer Demand"] end subgraph Stanford["Stanford-Style: Customer Pull"] S1["Boring Customer Problem"] --> S2["Iterative Execution"] --> S3["Massive Market Scale"] end ```

Copying and refining existing market dynamics is entirely legitimate: Lyft versus Uber, Anthropic versus OpenAI, and Rappi versus Instacart all demonstrate that solving an unaddressed customer friction matters far more than novelty. Channeling Steve Jobs, Francois Chaubard emphasizes that founders must always start with the customer experience and walk backward to the technology, rather than building a technical marvel and hunting for a use case.

Compounding this challenge is the fact that startups have no finish line. Academia operates on a structured cadence of finals, paper submission deadlines, and conference dates that offer cathartic moments of completion and rest. Startups offer no such relief. Academic training encourages a "breadth-first search" where researchers balance five speculative projects at once to hedge their bets. Applying this "smearing the peanut butter too thin" approach to a company divides team focus and leads to failure. As Francois Chaubard explains, “However, in founder mode, you have to wake up every single day and crank on that exact problem without getting sidetracked for decades.” Academics quickly grow bored of doing the same operational task every day, and that boredom frequently derails their ventures.

- Academic peer review mandates novel mechanisms, whereas commercial markets reward superior execution of established ideas. - Instacart did not invent online grocery delivery, just as Lyft, Anthropic, and Rappi did not invent their respective product categories. - Academics naturally default to spreading attention across multiple research threads, but building an enduring startup requires relentless, single-minded focus on one problem for years or decades.

---

[04:29](https://www.youtube.com/watch?v=u-8KpXcPYCw&t=269s) Customer Validation and Punchy Pitches over Academic Rigor

Discovering product-market fit cannot happen in the sterile isolation of an academic office; it is a brute-force exercise in talking to real people. Researchers frequently make the fatal mistake of staying behind closed doors, projecting their own assumptions onto user problems, and burning through years of runway building something nobody wants. Talking to venture capitalists does not solve this problem, because most investors have no idea whether a product will succeed. Founders must speak directly with the prospective buyers suffering from the pain point.

Francois Chaubard illustrates this through a costly mistake he made at his retail AI company, Focal. The team spent three months gathering computer vision data and training models to solve people-tracking for retail stores, accurately pinpointing shopper traffic. Retail executives had insisted this was a major problem, but when Focal delivered the finished models showing heavy foot traffic at checkout lanes and the deli counter during lunchtime, the retailers flatly refused to pay. The insight provided no actionable commercial value. Francois Chaubard realized he could have manually reviewed a few video feeds by hand in a single day, handed the retailers the exact same data, and discovered their unwillingness to pay without wasting three months of engineering time. This leads directly to his golden rule for AI founders: “Validate that your customers actually want what you're serving before you spend millions on compute training a model.”

```mermaid flowchart TD accTitle: Customer Validation Loop for AI Startups accDescr: Validating real willingness to pay before training expensive models. A["Identify Stated Customer Problem"] --> B["Build Human-in-the-Loop MVP"] B --> C["Present Results to Real Buyers"] C --> D{"Will they actually pay?"} D -- "No" --> E["Pivot / Discard without wasted runway"] D -- "Yes" --> F["Invest Capital in Scaled Model Training"] ```

Academic training also cripples founders when crafting commercial messaging. Researchers are trained to be pedantic, qualifying every claim with error bounds, ablation studies, and double-blind verification. Marketing, by contrast, relies on compelling, punchy statements. Consider Geico’s famous slogan: "15 minutes could save you 15% or more on your car insurance." The statement is scientifically inexact, unsupported by formal ablation studies, and utterly unverified by academic peers, yet it became one of the most effective advertising campaigns in history. While founders must never lie, attempting to bring peer-reviewed scientific precision into sales collateral will suffocate a startup's momentum: “If you obsess over rigor in your marketing, it will destroy you.”

- Validating customer willingness to pay must happen before writing complex code or buying GPU hours. - Focal lost three months training tracking models that retailers refused to buy, an outcome that could have been uncovered in twenty-four hours using manual video labeling. - Hyperbole and simplified marketing narratives are vital communication tools; burying prospects in scientific caveats destroys sales conversions.

---

[06:28](https://www.youtube.com/watch?v=u-8KpXcPYCw&t=388s) Speed, Responsibility, and the Capitalist Mindset

Perfectionism in decision-making is another classic academic pitfall. While submitting a paper to NeurIPS with only 70% confidence in the data would lead to swift rejection, operating with complete information in a startup is impossible. Jeff Bezos codified this reality by pointing out that most decisions should be made with around 70% of the information one wishes they had. Waiting for 100% certainty guarantees that competitors will beat you to market. During the early database wars, Michael Stonebraker developed QUEL at UC Berkeley, an engine widely considered technically superior to SQL. Yet Larry Ellison pushed SQL into commercial production at Oracle far faster. Larry Ellison famously declared: “QUEL might be better than SQL. I don't care. SQL is going to win.” Speed consistently triumphs over technical elegance.

Founders must also recognize that they cannot outsource the messy, unglamorous parts of building a business. Many academic founders nurture a fantasy where they raise venture capital, retreat to their desks to run fascinating algorithms on the investors' dime, and hire an MBA to handle sales, marketing, hiring, and firing. Francois Chaubard warns that this separation of duties is lethal: “If you abstract yourself away from sales and marketing, you will never develop empathy for your customers.” Without customer empathy, building a viable product is impossible.

Finally, researchers must shed their intellectual detachment and adopt what Francois Chaubard terms "outdoor animal psychology"—becoming unapologetic capitalists. Academics often feel squeamish about business realities like demanding 80% gross margins, charging customers aggressively, recruiting top performers, and fighting competitors. Paraphrasing Larry Ellison, Francois Chaubard reminds founders that competitors are actively trying to destroy their business and strip the food from their children's mouths.

When stepping into this arena, intellectual vanity becomes a massive liability. Academics frequently lecture prospects unprompted about finite automata or their theories on P versus NP, but customers, procurement officers, and investors do not care how smart a founder is. Building a business requires listening with genuine curiosity and understanding the user's specific world, echoing Dale Carnegie's maxim that one can win more friends in two months by showing interest in others than in two years of trying to get others interested in them.

- Jeff Bezos's 70% rule dictates moving forward without waiting for full certainty; waiting for complete data forfeits the market. - Oracle beat Michael Stonebraker's technically superior QUEL database purely on market velocity and aggressive distribution. - Delegating commercial operations to business hires isolates founders from customer friction and prevents product-market fit. - Founders must embrace commercial aggression, high margins, and active competition rather than relying on intellectual credentials to win respect.

[08:52](https://www.youtube.com/watch?v=u-8KpXcPYCw&t=532s) Get Out of the Office and Validate Demand

Understanding and caring about other people's problems is the fundamental engine behind both sales and product development. Too many academics fall into the trap of trying to impress others with their intelligence rather than focusing on the user. To succeed as a founder, you have to invert that dynamic entirely. The single most frequent piece of advice offered to researchers transitioning into the startup world is to physically leave the building, fly out, and spend real time in the environments where prospective customers live and work.

The primary objective of these visits is rigorous, unfiltered assumption testing. Founders must determine whether people actually have the problem being solved and, crucially, whether they are willing to pay cold, hard cash for a solution. Sitting inside a comfortable research lab or office leaves an open flank, because more aggressive competitors are already out in the field doing exactly this ground work. Facing prospective users directly can feel awkward and unnatural for those conditioned by the academy, but avoiding it is fatal. As Francois Chaubard puts it: “Get out of your office, get on a flight, and go meet with your users where they live and work.” He cautions that founders cannot afford to skip this step: “Because that is exactly what your competition is doing. And if you don't do it, they will beat you to it. It will be uncomfortable, but is entirely necessary.”

- **Field validation**: Travel directly to your prospective customers' workplaces to verify firsthand if they will pay for what you are building. - **Competitive reality**: If you refuse to engage directly with users in the real world, competing founders will do it and win the market before you.

[09:20](https://www.youtube.com/watch?v=u-8KpXcPYCw&t=560s) Rewiring the Mindset: Essential Reading and Willpower

For researchers who want to quiet their academic instincts and cultivate an operational founder mindset, structured reading outside the scientific literature is essential. Rather than reviewing academic journals, founders need frameworks for persuasion, people management, and personal efficacy. Francois Chaubard highlights a core reading list designed to dismantle academic hesitation and replace it with practical business instincts: Neil Rackham's *SPIN Selling*, Andy Grove's *High Output Management*, Stephen Covey's *The 7 Habits of Highly Effective People*, and Dale Carnegie's *How to Win Friends and Influence People*.

Ultimately, intellect is not the bottleneck for academic founders—determination is. Bringing the discussion to a close with a reflection attributed to Mahatma Gandhi—"If you want to change the world, start with yourself"—Francois emphasizes that academic founders already possess more than enough cognitive ability to succeed. The true test lies in whether they possess the grit and willingness to step out of their comfort zones and execute. As Francois Chaubard concludes: “You already have the intelligence. The only question left is, do you have the will?”

- **Recommended reading list**: - *SPIN Selling* by Neil Rackham - *High Output Management* by Andrew S. Grove - *The 7 Habits of Highly Effective People* by Stephen R. Covey - *How to Win Friends and Influence People* by Dale Carnegie - **The decisive factor**: Raw brainpower is already established; the deciding variable for whether a researcher succeeds as a founder is sheer personal will.

## Notable quotes > [00:09](https://www.youtube.com/watch?v=u-8KpXcPYCw&t=9s) To become a formidable founder, you will literally have to kill part of your academic identity, which is quite difficult to do. — Francois Chaubard

> [00:09](https://www.youtube.com/watch?v=u-8KpXcPYCw&t=9s) The most damaging thing you learned in school wasn't something you learn in any specific class. It was learning to get good grades. — Paul Graham

> [04:29](https://www.youtube.com/watch?v=u-8KpXcPYCw&t=269s) Validate that your customers actually want what you're serving before you spend millions on compute training a model. — Francois Chaubard

> [06:28](https://www.youtube.com/watch?v=u-8KpXcPYCw&t=388s) If you abstract yourself away from sales and marketing, you will never develop empathy for your customers. — Francois Chaubard

> [04:29](https://www.youtube.com/watch?v=u-8KpXcPYCw&t=269s) If you obsess over rigor in your marketing, it will destroy you. — Francois Chaubard

> [00:09](https://www.youtube.com/watch?v=u-8KpXcPYCw&t=9s) In founder mode, you have to be okay with saying it just works. We don't know exactly why, but that's good enough. Ship it. — Francois Chaubard

> [09:20](https://www.youtube.com/watch?v=u-8KpXcPYCw&t=560s) You already have the intelligence. The only question left is, do you have the will? — Francois Chaubard

## Mentioned Francois Chaubard, Google, Anthropic, Databricks, Together AI, Focal, Fei-Fei Li, Stanford University, Paul Graham, Marc Andreessen, Joe Lonsdale, MIT, Instacart, Lyft, Uber, OpenAI, Rappi, Steve Jobs, Geico, Jeff Bezos, Michael Stonebraker, QUEL, SQL, Oracle, Larry Ellison, Dale Carnegie, Neil Rackham, Andy Grove, Stephen Covey, Mahatma Gandhi, SPIN Selling, High Output Management, The 7 Habits of Highly Effective People, How to Win Friends and Influence People

## Video description Google, Anthropic, and Databricks were all started by founders with PhDs. But they’re the exceptions. Most academic founders fail, because the instincts that get you published are the ones that can get you killed in a startup. Francois Chaubard knows both sides. He’s a visiting partner at YC and a current PhD student, and before that he was a researcher in Fei-Fei Li’s lab and the founder and CEO of Focal. In this episode of Startup School, Francois breaks down the ten mindset shifts you have to make to go from researcher to formidable founder. He'll cover how to thrive without deadlines or grades, the difference between an interesting idea and an important one, and why being the first to an idea can actually be a disadvantage.

Chapters: 00:00 - Why academics struggle as founders 00:53 - Vanity metrics vs. what people want 01:45 - Interesting vs. important 02:27 - Novelty vs. plagiarism 03:54 - There is no finish line 04:32 - Get on a plane and talk to people 05:53 - Hyperbole is okay 06:28 - The 70% rule 07:11 - You can't delegate the hard stuff 07:52 - Become a capitalist 08:16 - Nobody cares how smart you are 09:04 - Get out of your office

Apply to Y Combinator: https://www.ycombinator.com/apply Work at a startup: https://www.ycombinator.com/jobs

[Source](https://www.youtube.com/watch?v=u-8KpXcPYCw)

中文简报

YC合伙人Chaubard告诫博士创业者：学术严谨性会扼杀初创公司学术衍生项目屡屡失败，是因为驱动论文引用和同行评审的激励机制，往往会惩罚生存所必需的速度、商业进攻性以及基于经验的客户验证。学术研究人员虽然通常智力超群，但他们追求引用指标、理论新颖性和完美主义的思维训练，实际上严重阻碍了他们作为创业者的成功。要打造一家能生存的公司，创始人必须抛弃学术虚荣心并坚持彻底的经验主义，将切实解决客户痛点和快速执行置于技术复杂性之上。创始人绝不能把销售假手于人，也不能闭门自守在实验室中；他们必须走出办公室，在消耗资金之前验证用户是否真愿意付费，并建立一种速度始终胜过纯粹学术优美感的激进商业思维。学术型创始人经常失败，是因为学术界所奖励的行为（如追求新颖性、力求完美和措辞谨慎）与商业初创公司的运作要求存在直接冲突。 (0:09) 商业执行力重于技术新颖性：像Instacart和Oracle这样的快速跟随者，往往通过比拥有专利发明的先驱者更快解决日常运营问题，从而赢得庞大的市场。 (2:26) 人工智能领域的创始人必须在投入大量资金和时间训练模型之前，利用“人机协同”（human-in-the-loop）测试来验证客户的付费意愿。 (4:29) 创始人必须奉行杰夫·贝索斯（Jeff Bezos）的70%确定性法则，切忌将面向客户的业务外包，并保持敏锐激进的商业直觉以应对竞争对手的威胁。 (6:28)

Reading summary

Former Stanford researcher and Focal founder Francois Chaubard argues that successful academic founders must dismantle their scholarly instincts—such as prioritizing novelty, rigor, and theoretical perfection—in favor of rapid execution, commercial aggression, and direct customer validation. Abandoning Academic Vanity Metrics for Commercial Demand (0:09) While PhD-led companies like Google, Anthropic, Databricks, and Together AI represent massive successes, Chaubard notes that academic spinouts usually fail because research incentives conflict with business realities. Academia rewards proxy prestige metrics—such as citations, paper acceptances, talks, and grades—whereas startups require an exclusive obsession with product-market fit and commercial adoption. Quoting Paul Graham on the hazard of training students merely to earn good grades, Chaubard emphasizes that founders must abandon intellectual indulgence to focus on what users will actually buy. This shift demands adopting pure empiricism over deep theoretical explanations. Researchers are conditioned to probe underlying mechanisms and ensure perfection before publication. In contrast, startup founders must prioritize solving user pain points quickly and accept solutions that work practically, even without complete understanding of why, rather than waiting for exhaustive proof. Prioritizing Execution and Focus Over Novelty (2:26) Academic peer review strictly mandates original discovery, but commercial markets primarily reward the superior execution of existing ideas. Drawing on Joe Lonsdale's distinction, Chaubard contrasts patented, invention-driven 'MIT-style' startups that lack buyers with 'Stanford-style' startups that iterate persistently on mundane, operational problems. Companies such as Instacart, Lyft, Rappi, and Anthropic did not invent their product categories; instead, they won by focusing on user friction and working backward from the customer experience to the technology. Furthermore, building a startup lacks the structured milestones of academia, which usually provides regular closure through semesters, exam dates, and conference submissions. Academic researchers naturally default to spreading focus across several speculative threads at once. Chaubard warns that this approach dilutes organizational focus, whereas building an enduring enterprise requires unyielding commitment to a single problem for years or decades without getting bored. Direct Customer Validation Versus Premature Engineering (4:29) Chaubard cautions that founders cannot validate demand from an office or by consulting venture capitalists; they must talk directly to target buyers before writing code or spending capital. At his startup Focal, the team spent three months training computer vision models for retail tracking after executives stated it was a major problem, only for the retailers to decline paying for the resulting traffic insights. Chaubard realized the same lack of commercial interest could have been discovered in twenty-four hours by manually reviewing video footage, demonstrating the need to prove customer willingness to pay before committing heavy compute resources. Scholarly rigor can also cripple go-to-market communication. While research writing requires extensive qualifications, ablation studies, and error margins, commercial messaging demands punchy and simplified claims, exemplified by Geico's famous insurance tagline. Chaubard stresses that while founders must not lie, bringing scientific pedantry into marketing collateral suffocates sales conversion. Embracing Speed, Capitalist Realities, and Commercial Empathy (6:28) Decision-making in startups cannot wait for complete data. Invoking Jeff Bezos's principle of making decisions with roughly 70 percent of desired information, Chaubard points to the historical database rivalry where Larry Ellison scaled Oracle using SQL despite Michael Stonebraker's QUEL at UC Berkeley being considered technically superior. Ellison prioritized market velocity over technical elegance, proving that shipping fast beats academic perfection. Chaubard cautions founders against retreating into pure technical work while delegating sales, marketing, and operations to business hires, as this separation prevents founders from developing customer empathy. Academic founders must shed intellectual detachment, drop the habit of lecturing prospects on technical arcana, and adopt an aggressive capitalist mindset focused on strong margins, competitive defense, and active listening. Chaubard concludes that raw cognitive ability is not the limiting factor for academic founders transitioning into business. Success depends instead on personal determination, the willingness to leave the lab to engage uncomfortable commercial environments, and studying foundational frameworks on management and influence from authors like Neil Rackham, Andy Grove, Stephen Covey, and Dale Carnegie.

精炼阅读摘要

Focal 创始人 Francois Chaubard 在 Y Combinator 的演讲中指出，学者转型为创业者必须打破原有的学术身份与心智模式，从追求理论原创与虚荣指标转向追求商业验证、执行速度与客户真实痛点。虽然 Google、Anthropic 和 Databricks 等巨头由学者创立，但大多数学术背景的创业公司因无法克服学术训练带来的本能偏差而面临极高失败率。摒弃学术虚荣与原创执念 (0:09) Chaubard 指出，学术界长期通过引文数量、演讲邀请与论文接收等虚荣指标评估成功，但创业的核心完全取决于市场采纳与商业变现——即用户是否真正愿意付费并将其融入日常工作流。学者往往受训于解构深层机制，探究纯粹有趣但对实际生活缺乏直接影响的理论细节；初创企业则必须保持纯粹的经验主义，以解决客户痛点为唯一导向，奉行“只要有效就尽快交付上线（It just works, ship it）”的实用主义标准。在商业市场中，对纯粹原创性的执念往往十分危险，对成熟商业模式的借鉴与改良才是常态。Chaubard 引用 Joe Lonsdale 对“MIT 式”与“斯坦福式”创业的划分：前者执着于开发少有人买单的专利黑盒发明并四处搜寻应用场景，而后者如 Instacart 则深耕杂货配送这一看似乏味平凡的运营问题，通过持续迭代执行做大规模。此外，学术研究依托学期、答辩和论文截稿日等明确周期，习惯同时推进多项课题以分散风险；初创企业则没有休止符，创始人必须克服对日常重复性事务的厌倦，数年甚至数十年如一日地专注攻克单一核心问题。先行验证付费意愿并精简表达 (4:29) 寻找产品与市场契合点（PMF）必须通过与真实潜在买家直接对话来完成，而不能在实验室中闭门造车。学者型创始人常犯的致命错误是将自身主观假设强加于用户，在无人买单的方向上耗尽资金。Chaubard 以自身零售 AI 初创公司 Focal 的经历为例：团队曾耗费三个月收集计算机视觉数据并训练追踪店内人流的模型，尽管零售商高管此前坚称需要该功能，但当模型准确呈现收银台与熟食区的高客流时，零售商却因缺乏可操作的商业价值而拒绝付款。实际上，团队原本只需通过人工抽检一天监控录像呈现相同数据，即可零成本验证付费意愿。AI 创始人必须在消耗大量算力训练昂贵模型之前，切实验证客户的真实采购意向。学术严谨性同样会削弱商业传播的有效性。研究人员习惯在陈述中附加误差边界、消融实验与双盲检验等限定条件，而商业营销依赖简洁明快的主张。以 Geico 著名的广告语“15 分钟可能为你节省 15% 或更多的汽车保险”为例，该表述在科学上缺乏严密的学术证明，却成为了商业史上极具转化效力的推广口号。Chaubard 强调，尽管创始人绝不能欺骗用户，但在营销和销售环节过度追求学术严密性将会扼杀初创企业的商业增长势头。决策速度优先并直接承担商业职能 (6:28) 在学术界，数据只有 70% 置信度会导致论文被顶会拒稿，但在初创企业中追求完全信息是不可能的。Chaubard 引用 Jeff Bezos 的 70% 决策准则，强调绝大多数商业决策应在掌握所需信息的约 70% 时迅速做出，等待 100% 的确定性只会让竞争对手抢占市场。在早期数据库竞争中，加州大学伯克利分校 Michael Stonebraker 开发的 QUEL 在技术机制上被公认为优于 SQL，但 Larry Ellison 带领 Oracle 以极快的商业化速度将 SQL 推向市场，正如 Ellison 所言，即便对手技术更佳，SQL 也能凭借执行速度与激进分发赢得最终胜利。此外，学者型创始人常抱有不切实际的幻想，试图躲在办公桌后专注于前沿算法，将销售、营销、招聘与解雇等商业事务外包给 MBA 人才。Chaubard 警告，创始人一旦脱离销售与营销一线，就永远无法建立对客户痛点的同理心，从而无法打磨出合格的产品。创始人必须放下智识优越感，培养“野外动物心理学”般的商业进取心，坦然追求高毛利与激进竞争，停止向客户宣讲有限状态机或 P 对 NP 等理论课题，转而以倾听与好奇心深入理解客户的具体业务现实。深入一线调研与重塑心智模式 (8:52) 理解并重视他人的现实痛点是驱动产品开发与销售转化的根本动力。Chaubard 建议学术背景的创始人必须走出舒适的实验室与办公室，亲自飞往客户日常工作和生活的实际场景中，进行未经滤网过滤的假设检验。如果创始人因心理不适而回避与潜在用户的直接接触，行动更积极的竞争对手就会抢先占领市场。为了剥离学术本能并建立商业直觉，学者必须阅读科学文献之外的实用管理与人际沟通书籍以重塑思维方式，包括 Neil Rackham 的《SPIN 销售高收法》（SPIN Selling）、Andy Grove 的《格鲁夫给经理人的第一课》（High Output Management）、Stephen Covey 的《高效能人士的七个习惯》（The 7 Habits of Highly Effective People）以及 Dale Carnegie 的《人性的弱点》（How to Win Friends and Influence People）。Chaubard 总结道，科研人员并不缺乏智力储备，向创始人转型的关键瓶颈在于是否具备走出舒适区并日复一日坚决执行的个人意志力。 Chaubard 的核心论点在于，学术卓越与商业成功所要求的思维框架存在根本冲突；学者型创始人成功的关键不在于技术深度或智力高低，而在于能否主动打破原有的学术身份包袱，亲自扎根于用户验证、快速决策与激烈的市场竞争之中。

Source quotes

[0:09] Francois Chaubard: To become a formidable founder, you will literally have to kill part of your academic identity, which is quite difficult to do. [0:09] Paul Graham: The most damaging thing you learned in school wasn't something you learn in any specific class. It was learning to get good grades. [4:29] Francois Chaubard: Validate that your customers actually want what you're serving before you spend millions on compute training a model. [6:28] Francois Chaubard: If you abstract yourself away from sales and marketing, you will never develop empathy for your customers. [4:29] Francois Chaubard: If you obsess over rigor in your marketing, it will destroy you. [0:09] Francois Chaubard: In founder mode, you have to be okay with saying it just works. We don't know exactly why, but that's good enough. Ship it. [9:20] Francois Chaubard: You already have the intelligence. The only question left is, do you have the will?

Original transcript

[0:01] [music] [0:07] [music] [0:09] Some of the most successful companies in [0:10] the world were created by founders with [0:12] PhDs, including Google, Anthropic, Data [0:14] Bricks, Together AI, and more. But these [0:17] are exceptions to the rule. It's [0:18] actually more common for academic [0:20] startups to fail because many of the [0:21] skills you're rewarded for as a [0:23] researcher can be antithetical or at [0:25] least orthogonal to succeed as a [0:27] founder. I know this in part because [0:28] before I was a founder, I was a [0:30] researcher in Feel Lee's lab and many of [0:32] the skills I learned there in retrospect [0:34] held me back as a startup CEO. To make [0:36] it as a founder, academics need to [0:38] unlearn many instincts that have been [0:40] beaten into them by their principal [0:42] investigators over years of blurry eyed [0:45] coffeefueled nights solo in a lab. To [0:47] become a formidable founder, you will [0:49] literally have to kill part of your [0:50] academic identity, which is quite [0:52] difficult to do. So today I want to [0:54] break down the 10 most important [0:55] perspective shifts academics need to [0:58] make in order to build a successful [1:00] startup. Number one, focusing on vanity [1:02] metrics versus what people want. In [1:05] academia, you're optimizing for what I [1:07] call vanity metrics. Grades, the number [1:09] of citations, the number of talks you [1:11] give, etc. Paul Graham has a great quote [1:13] about this. He said, "The most damaging [1:15] thing you learned in school wasn't [1:16] something you learn in any specific [1:18] class. It was learning to get good [1:19] grades." That is an academic mindset. [1:22] The founder mindset is entirely [1:24] different. It asks, "What is something [1:26] people actually want? What will they [1:29] actually pay money for versus what they [1:31] say they would pay money for?" You might [1:33] think they want your new shiny [1:35] algorithm, but will they actually adopt [1:37] it? So often the answer is no. And you [1:40] don't realize this until you try to [1:41] start talking to people and asking the [1:44] tough questions. This leads directly [1:46] into number two, interesting versus [1:48] important. Academics often optimize for [1:51] what is interesting research and that is [1:53] a good thing. You should double click [1:55] into the things to understand the [1:56] mechanisms behind something that gets [1:58] you published. But it often has [2:00] absolutely no real ramification on [2:02] anyone's daily life. Founders must focus [2:04] on what is important to their users. In [2:07] founder mode, you have to be okay with [2:09] saying it just works. We don't know [2:10] exactly why, but that's good enough. [2:12] Ship it. Empiricism is completely fine. [2:15] As Mark Andre says, the only thing that [2:18] matters is getting to product market [2:19] fit. That means caring only about what [2:22] is important to your users and working [2:23] backwards from there. Number three, [2:26] novelty and plagiarism. In academia, [2:28] novelty is everything. When you submit a [2:31] paper, the reviewer judges the original [2:33] nature of your work. If it's not [2:35] completely novel, it's rejected. In the [2:37] startup world, plagiarism is often a [2:39] really good strategy. Pioneers often get [2:41] errors in the back, as they say. Joe [2:43] Longdale makes a brilliant distinction [2:44] between MIT style startups and Stanford [2:47] style startups. MIT startups often [2:49] revolve around a brand new invention, a [2:51] brilliant black box that we patented [2:53] that does some unprecedented magic. But [2:56] they rarely turn into massive companies [2:57] because frankly no one really wants your [2:59] blackbox. A Stanford style startup on [3:01] the other hand starts with a single [3:03] boring customer problem and iterates on [3:04] it forever. Something simple like [3:07] listening to music online or getting [3:09] people a local hotel room or getting [3:11] people their groceries. Instacart has [3:13] been around for 15 years and is still [3:15] iterating on that single court problem [3:17] today. And Instacart didn't invent [3:19] delivering groceries from an online [3:21] order. Countless startups tried it [3:23] before them. Instacart was just the most [3:25] successful at executing it. Plagiarism [3:27] is completely okay. See Lyft versus [3:29] Uber, Enthropic versus OpenAI, WPPY [3:31] versus Instacart. As long as you are [3:33] solving a real problem that your [3:34] customers have, one that the market has [3:36] not adequately addressed, you're in [3:38] business. Novelty is not required. [3:41] Often, it's not even preferred. As Steve [3:43] Jobs famously said, you have to start [3:45] with the customer experience and walk [3:46] backwards to the technology, not the [3:48] other way around. Number four, there is [3:50] no finish line. In academia, there is [3:53] always some cathartic submission date or [3:56] final, a moment where the paper or test [3:58] is sent off and you can finally rest. [4:01] Startups don't have that. You have to [4:02] develop a long horizon vision of where [4:04] things will go and not waver constantly. [4:07] Researchers are quite bad at this. They [4:09] work on five different ideas all at [4:10] once, smearing the peanut butter too [4:12] thin. That's a good idea for academic [4:14] research. Breath first search. However, [4:16] in founder mode, you have to wake up [4:18] every single day and crank on that exact [4:20] problem without getting sidetracked for [4:22] decades. Many academics just get bored [4:25] and this distracts their team from the [4:27] core mission. Number five, get on a [4:29] plane and talk to some people. Finding [4:31] product market fit is a brute force [4:33] exercise. It requires iterating quickly [4:36] and talking to customers constantly. [4:38] Academics like doing this in a vacuum. [4:41] They presume they know what their [4:42] customer problem is without talking to [4:44] anyone, without leaving their office. [4:46] But very often, if you spend years [4:48] solving a problem, it turns out that [4:51] problem isn't big enough for anyone to [4:52] care and you just wasted years of [4:54] runway. And when I say talk to people, I [4:56] do not mean investors. Most investors [4:58] have no clue what is going to be a great [5:00] product or not. This is why they so [5:01] often misinventure. Talk to the people [5:04] who have the problem you are trying to [5:05] solve and ask them if you did X, would [5:08] they pay money for it? We made this [5:09] mistake at Focal. We spent a lot of time [5:11] collecting training data, training [5:13] models to solve people tracking for [5:15] retailers to know where people are and [5:17] when in retail stores. Every retailer [5:20] told us this was a major problem. 3 [5:22] months of work later, we gave it to the [5:23] stores and it told them that people [5:25] spend a lot of time in the checkout [5:27] counter and the deli counter at [5:28] lunchtime. I asked them, "Will you pay [5:30] for this?" And they said, "Absolutely [5:32] not." Because there was no value [5:34] delivered. I could have manually labeled [5:35] a few videos a day and given them the [5:37] same data and gotten the same conclusion [5:39] and that would have saved me 3 months. [5:41] So that's why I advise if you are an AI [5:43] company, start with a human in the loop. [5:46] Validate that your customers actually [5:47] want what you're serving before you [5:49] spend millions on compute training a [5:51] model. Number six, hyperbole is okay in [5:54] a sales pitch versus being [5:55] scientifically exact. Think about one of [5:57] the most famous advertising slogans of [5:59] all time. 15 minutes could save you 15% [6:02] or more on your car insurance. What a [6:04] beautifully inexact statement. What does [6:06] that even mean? Where was that measured? [6:08] Was there some ablation study somewhere? [6:10] Was it independently reviewed in some [6:12] double blind conference submission? [6:14] Absolutely not. Yet, this was one of the [6:16] most successful slogans of our lifetime. [6:18] If you obsess over rigor in your [6:19] marketing, it will destroy you. By no [6:21] means should you lie, but using punchier [6:23] language is necessary to get people [6:25] excited. Number seven, the 70% rule. [6:28] Jeff Bezos famously said that most [6:30] decisions should probably be made with [6:31] somewhere around 70% of the information [6:34] you wish you had. Publishing with 70% [6:37] confidence in your results isn't exactly [6:39] a great strategy to get into nurups. So, [6:41] it's natural for academics to hate this. [6:43] In startups, if you wait for 100% of the [6:45] data, you are way too late. By that [6:48] point, your competitor has already [6:49] launched and you've lost your first [6:50] mover advantage. In the great database [6:53] wars, Michael Stonebreaker had a [6:54] technically superior product, Quill. But [6:57] Larry Ellison, founder of Oracle, pushed [6:59] SQL into the market faster. Ellison [7:01] famously said, "Quill might be better [7:03] than SQL. I don't care. SQL is going to [7:05] win." And he was right. Speed beats [7:08] perfect information. Number eight, you [7:10] can't delegate the hard stuff. [7:12] Researchers love doing research, but as [7:14] CE of a startup, you can't be doing [7:16] research, likely ever. In a startup, [7:18] you're responsible for everything. Team [7:20] management, cash runway, hiring, firing, [7:22] marketing, sales, product development. A [7:25] lot of academics love the idea of a [7:26] startup, but they don't actually want to [7:29] do these messy jobs. They think, "Oh, [7:31] I'll just hire an MBA to do the business [7:33] stuff, and I will do the cool research [7:35] on VC's dime." This is a terrible idea. [7:37] If you abstract yourself away from sales [7:39] and marketing, you will never develop [7:41] empathy for your customers. And without [7:43] customer empathy, you cannot build a [7:45] product they really want. Number nine, [7:47] outdoor animal psychology. To be a [7:49] successful founder, you must adopt a [7:51] more aggressive psychology. You must [7:53] become a capitalist. Starting a company [7:55] is the most capitalistic thing you can [7:57] possibly do. And frankly, a lot of [7:59] academics are deeply uncomfortable with [8:01] that. You have to be comfortable [8:02] charging people money for things, [8:04] winning talent, making 80% margins, [8:06] raising capital, and most importantly, [8:08] beating your competition. To paraphrase [8:10] Larry Ellison, your competitors want to [8:11] kill your company and take the food out [8:13] of your child's mouth. This is the world [8:15] you're walking into. Number 10, nobody [8:17] cares how smart you are. Academics love [8:20] explaining how smart they are to the [8:21] world, completely unprompted. They will [8:23] talk your ear off about finite automa or [8:25] their point of view on P versus NP. But [8:28] this does not work in sales. It doesn't [8:30] work in product development and it [8:32] certainly doesn't work when pitching [8:33] VCs. You need an innate curiosity about [8:35] people. You need to think from your [8:37] users's perspective constantly and you [8:39] only get that perspective by listening. [8:41] Whether you're selling to an IT [8:42] department, an intern or a procurement [8:44] officer or a consumer, you must [8:46] empathize with them. As Dale Carnegie [8:48] once wrote, you can make more friends in [8:50] two months by becoming interested in [8:52] other people than you can in two years [8:54] by trying to get others interested in [8:56] you. This is exactly how sales and [8:58] product development work. To summarize, [9:00] the most common piece of feedback I give [9:01] to academics turn founders is this. Get [9:04] out of your office, get on a flight, and [9:06] go meet with your users where they live [9:08] and work. Validate your assumptions. [9:10] Will people actually pay you for this? [9:12] Because that is exactly what your [9:13] competition is doing. And if you don't [9:15] do it, they will beat you to it. It will [9:17] be uncomfortable, but is entirely [9:19] necessary. If you want to quiet your [9:21] academic brain and start growing your [9:22] founder brain, I highly recommend [9:24] reading spinelling, high output [9:26] management, the seven habits of highly [9:28] effective people, and how to win friends [9:30] and influence people. I'll leave you [9:32] with a quote from Mahatma Gandhi. If you [9:35] want to change the world, start with [9:36] yourself. You already have the [9:38] intelligence. The only question left is, [9:40] do you have the will? [music] [9:47] [music]