This piece covers YC partner Michael Seibel's framework for evaluating founders and startups. He says don't obsess over the 'big idea'—most successful companies end up doing something else. Instead, focus on 'trajectory': how fast you've progressed given the time spent. He highlights Airbnb, noting its real customer is the host (not the traveler), who wants bookings daily and pays 12% fees. He also discusses Promise, a startup that started helping people pay fines and evolved into 'Stripe for government payments,' achieving zero default rates.
Michael Seibel (Y Combinator partner and CEO) shared insights from reviewing thousands of YC applications, interviewing founders, and observing company growth in an interview. Key points include: YC places the highest value on a founder's clear understanding of the problem and the frequency and inte
Michael Seibel is a partner and CEO of Y Combinator, having served as co-founder and CEO of Justin.tv (which later grew into Twitch) and Socialcam. This chapter centers on the founder evaluation framework and startup methodology he distilled from reviewing nearly 30,000 YC applications and interviewing thousands of entrepreneurs. Seibel makes the most counterintuitive judgment: when evaluating founders, one should "not be too smart"—set aside the bias toward "good ideas," because most successful companies end up solving a completely different problem from what they originally applied with; the only reliable indicator of founder potential is "trajectory," i.e., the rate of progress over a given time, not the starting point.
Seibel distilled three core screening criteria from reviewing nearly 30,000 YC applications (combined from the summer and winter batches of 2020), and stressed that evaluators need to "not be too smart" – set aside the bias for "good ideas," because most companies' paths to success are completely different from their original applications.
First, the team must be able to build and ship the initial version of the product. This is the most basic threshold, corresponding to "technical ability." Seibel does not care about the founder's background or education; he only cares about "can you get your hands dirty."
Second, given the time already invested, progress must be impressive. This is Seibel's most important "trajectory" metric. He explicitly distinguishes between the education system and startup evaluation: "Schools only measure absolute scores (SAT, GPA), not trajectory; while YC 'cares about only one thing – trajectory – how much forward momentum and velocity you've created, not where you started.'" Whether the founder has worked for two weeks or two years, Seibel wants to see "frightening progress."
Third, there must be a strong emotional bond between co-founders. Seibel explains with a vivid analogy: "You're signing up for a contract that says 'get slapped in the face every day, forever.' You get slapped in the face in the morning, and at night there's someone you can lean on and cry with – that person is your co-founder." He opposes mechanically looking at "complementary skills," arguing that emotional support is the key to an early-stage company's ability to withstand pressure.
In the interview, Seibel (trained by Paul Buchheit) uses two litmus tests:
Seibel also notes that in an interview, "whether the founders like each other" is hard to hide, and "physical cues" in body language are extremely obvious; internal team friction is one of the most lethal warning signs.
Seibel proposes a simple framework to help founders screen problems: frequency, intensity, and willingness to pay. He encourages founders to "stack rank" candidate problems along these three dimensions. The core purpose is not to select the absolute best, but to make founders realize, "Oh, this problem actually isn't that important."
Host Patrick uses Airbnb as an example: from a traveler's perspective, staying in a short-term rental has low frequency, medium intensity, and average willingness to pay — by this framework, Airbnb should have failed. Seibel points out that this is a carefully designed trap, with the key being a misdefinition of "who the customer is" .
Seibel argues that in two-sided markets, most people mistakenly assume they themselves are the customer. Airbnb's real customer is the host, not the traveler:
Seibel uses a car-buying website as an analogy: "An ordinary person buys a car once every seven years and thinks they're the car website's customer; in reality, the dealer who needs to sell cars every day is the real customer."
Seibel's early advice: in a two-sided market, first "cheat to acquire supply" — manually recruit valuable supply on a one-to-one basis, because "without supply, there is no demand to showcase." The critical supply for Airbnb in its early days was vacation rental properties in New York City, used by international travelers with children — this was not a model invented by Airbnb, as VRBO and Couchsurfing already existed. Seibel emphasizes: "Airbnb's uniqueness was not the idea, but finding a better tool to make something that was already happening easier. Its first investor, Greg McAdoo, already understood the vacation rental market."
Seibel observes a new trend: founders are increasingly focused on their company's long-term impact on society, daring to tackle "core challenges" — such as healthcare, mass incarceration, and similar issues. He categorizes "courage" into three types:
Seibel uses the startup Promise as a case study for in-depth analysis. The core goal of Promise's two founders is "to reduce mass incarceration in the United States." They went through two iterations:
First round: Helping people who cannot afford bail to appear in court on time. They developed an app that uses GPS tracking, SMS reminders, and one-click Lyft/Uber bookings to ensure people can attend court without being detained. They "sold to the government in a field where everyone said 'you can't sell to the government,'" and secured county-level contracts.
But a problem emerged: They found that the more efficient their system became, the more the police system tended to "arrest more people" — because "the system sees its role as punishing the bad guys, not caring about individuals." Contradicting the founders' mission, they voluntarily withdrew from the contracts that were already generating revenue.
Second round: Returning to the root cause — many people enter the judicial system because they cannot afford fines or fees. They created a front-end website that allows users to pay fines with credit cards or installment plans (governments usually do not accept credit cards or installment payments), while the back end pays the government on behalf of users. A few months later, zero default rate — people not only paid on time, but also proactively explained why they paid late. They later approached the municipal finance department, which said: "Your collection rate is higher than ours. We only care about how much money we collect, not how hard we make life for people." Seibel sums it up as "Stripe for government" — making every payment interaction between citizens and the government friendly. This is not charity, but a company that can succeed and "make America better."
Seibel admits this is "the question he hates most yet is asked most often." He shares two of his own PMF experiences:
Justin.tv (Spring 2008): A user in Morocco pointed a camera at a TV so friends could watch a local soccer match. 3,000 Moroccans watched from around the world, with terrible quality, but it directly crashed the site. That year the company grew 1,200%, and the bottleneck was "whether it could stay online." Seibel's description: "PMF hits you like a sledgehammer to the face. You don't have time to sit and think, 'Is this PMF?' — everything explodes at once."
Socialcam (2012): Growth was so fast that they built their own servers, which took two weeks to arrive and install. One morning, he found he had to decide whether to place an order; by lunch, he realized the two-week window was too late; by dinner, he decided to "migrate to AWS that very night," completing the full migration by 6 a.m. Again, "no time for philosophizing."
Seibel's core warning: "The thing founders are best at lying to themselves about is whether they have PMF. Good founders limit how much they lie to themselves." He tells every YC batch: "You do not have PMF. Keep digging." Because what founders most crave is "company building" (hiring, building processes, company building), but without PMF, company building is wasting investor money — all capital should be poured into the product.
When asked whether “execution ability is trainable,” Seibel firmly answered “100% yes,” and revealed YC’s “secret weapon”:
> “If you take the fastest kid in high school and put him on a college team, he gets faster. Take the fastest college runner and put him on the Olympic team, he gets faster. YC’s secret is not advice, not Demo Day, not ‘ta-ta-ta’ — it’s the batch itself. The shareholders make YC, not us. If I moved all MIT students to a community college, they would still be one of the best engineering schools in the world. Our job is to ensure the batch is high quality, motivated, and everyone is chasing each other.”
Seibel points out that the core psychological mechanism for founders is “self-deception” — “Every founder must fundamentally lie to themselves, convincing themselves to go and do this startup thing.” From this perspective, entrepreneurship is more like sports than school: “If you’re a smart kid, staying average among smart kids, you can get into a good college and find a good job. But in a YC batch, average players lose. You have to be several standard deviations above those around you to achieve an extraordinary outcome.”
On the comparison between “education vs. entrepreneurial learning”: Seibel is blunt: “Learn more? Fuck that, it’s not even a question. You do nothing in school. The ‘best practice’ you learn is ‘fire people’ — but what you really learn is that you have an underperforming employee, six months in, and every time you see him you get a stomachache because you know you have to fire him but you don’t dare. That’s learning. School is not learning.”
| Ticker | Analyst View | Key Data |
|---|---|---|
| Airbnb | Bullish (Case Study) | Platform fee 12%; early supply from New York City vacation rentals, used by international travelers with children |
| Promise | Bullish (Deep Dive) | Starting from "reducing mass incarceration", went through two iterations; second iteration had zero default rate; now "Stripe for government" |
| Square | Neutral (Background Mention) | As a benchmark for "vertical SaaS", but Seibel believes "challengers can succeed" |
| Tesla | Cautious (Comparison Only) | Seibel points out the "government subsidy" background, believes the financing environment for non-software companies is far less favorable than for software |
| SpaceX | Cautious (Comparison Only) | Same as above |
1. "Don't be too clever" — Let go of the bias for "good ideas" (Michael Seibel)
Supporting practice: Most companies end up solving a completely different problem than the one they applied with; the predictive power that makes you "too clever" will cause you to miss the future X-shaped path.
2. "The only real metric for a founder is trajectory, not starting point" (Michael Seibel)
Supporting mechanism: YC only cares about "the speed of progress within a given time frame," rejecting absolute indicators such as educational background or pedigree. SAT/GPA are static; trajectory is dynamic.
3. "Frequency × Intensity × Willingness to Pay" framework, but pay attention to who the real customer is (Michael Seibel)
Key pitfall: Airbnb's customers are hosts (high frequency, high intensity, high willingness to pay), not travelers. Analysts must ask, 'In this market, whose need is the most urgent?'
4. "In a two-sided market, cheat to acquire supply first; manually recruit valuable supply-side participants one-on-one" (Michael Seibel)
Supporting mechanism: Without supply, there is no demand to show. Airbnb's early supply came from New York City vacation rentals — that was the "tinder."
5. "The feeling of product-market fit is the website crashing; you have no time to philosophize" (Michael Seibel)
Practical case: Justin.tv had 3,000 concurrent viewers causing the website to crash, and grew 1,200% that year; Socialcam migrated to AWS overnight. PMF is not "felt" — it is "overwhelming."
6. "The thing founders are best at lying to themselves about is whether they have achieved PMF" (Michael Seibel)
Falsification condition: If you start "company building" (hiring, processes, team building) rather than "product building," you have likely deceived yourself. Without PMF, 100% of funds should go into the product, not the organization.
7. "Startup is more like sports than school" — Average players lose (Michael Seibel)
Data support: In a YC batch, you must be several standard deviations above those around you to achieve an extraordinary outcome. "If you want to be normal, go to Yale Law School; if you want to win, be 'abnormal.'"
8. "The core psychological mechanism of a founder is self-deception" — but boundaries must be managed (Michael Seibel)
Supporting mechanism: The "lies" founders tell themselves are the fuel to start a company, but great founders limit the extent of their self-deception. "You signed up for a contract to be slapped in the face every day; the only thing that keeps you going is a willingness to lie to yourself a little."
9. "YC's secret is not advice — it's batch" — Let excellent people inspire each other (Michael Seibel)
Practical execution: Put high-quality, motivated founders together, and everyone has the drive to "catch up with each other"; advice itself is not important — the environment is.