This video offers a clear answer to a dilemma every founder faces when starting a company: "Which idea should I choose?" Rather than wasting time hunting for the perfect idea, it emphasizes that you should pick one idea, go deep on it, and refine it through customer feedback. Through this process, the video explains, you can uncover an idea far better than the one you started with and build a successful startup.

1. Letting Go of the Obsession with the Perfect Idea

YC partner Jon Xu says founders tend to agonize between multiple ideas, or wait until they find the perfect one, and he points out that this is one of the most common ways startups fail in their earliest stage. 😩

1.1. There Is No Abstract Perfection

Many founders get trapped in the belief that they have to find the "perfect idea," but Jon Xu puts it this way:

"The problem with this approach is that it's impossible to figure out the abstract perfect idea. You only know what to do once you contact reality and get feedback from customers."

In other words, without real communication and feedback from actual customers, you simply cannot know which idea is the "perfect" one.

1.2. You Don't Have to Be the Perfect Founder

Founders also often get stuck on the question, "Am I the perfect founder for this idea?" Founder-market fit matters, of course, but you don't necessarily need decades of experience. Jon Xu stresses:

"The truth is that you don't have to be. If you pick an idea you're curious about, go incredibly deep, and most importantly talk to customers, it's possible to develop exceptional knowledge in a short amount of time." He cites Blake Scholl, CEO of Boom Supersonic, who went from working in ad tech to commercializing supersonic flight and built a multi-billion-dollar company — an example of how curiosity, immersion, and communication with customers are what truly matter.

2. The Problem with Pursuing Multiple Ideas at Once

You might think running several ideas in parallel is the best way to figure out which one will succeed, but Jon Xu says this approach has a serious flaw. 🤔

2.1. It Generates Bad Data

If you juggle multiple ideas without going deep on any of them, you won't get a clear signal on whether an idea actually works.

"If you don't actually go deep on an idea and instead juggle it with several different ideas, you won't get a good signal on whether what you're doing actually works. And if you don't get a good signal, you can either abandon a good idea prematurely, or convince yourself that a bad idea is worth continuing." This situation can make you miss a good idea, or chase a bad one in vain.

2.2. Focus on a Single Idea

Jon Xu offers "go in depth first" as the solution. Out of several appealing ideas, pick one and immerse yourself deeply in it.

3. How to "Go Deep" on an Idea

So what exactly does it mean to "go deep" on an idea? 🏊‍♀️

3.1. Burn the Other Boats

Jon Xu references the Spanish conquistador Hernán Cortés, who, during the conquest of Mexico, burned the ships they had arrived on to prevent his soldiers from retreating and force them to think only of going forward. In the same way, he says you must give up your other ideas and commit solely to the one you chose.

"You have to burn the other boats. That means explicitly giving up on your other startup idea options, stopping work on them, telling your customers you've pivoted, and focusing on the chosen idea with single-minded purpose."

3.2. Transform as if Putting On "New Skin"

"Going deep" is like making yourself into an entirely new person. You may have to change your company name, email address, website, and even the narrative behind why you started the startup. 😮 He cites GovDash, a startup that helps with bidding on government contracts; they pivoted five times to find their idea, changing the company name and mission each time. They became true domain experts in government procurement and grew successful enough to handle enormous demand.

3.3. Understand It Well Enough to Run the Customer's Business Yourself

The best way to check whether you're going deep on an idea is to ask yourself "Could I run the customer's business myself?"

"Let's say you want to build a voice customer service agent for cleaning services. The question isn't simply whether you talked to 20 owners. The question is, 'If I dropped you into the cleaning business tomorrow, would you know how to run it?'" You need to accurately understand what your customer's daily crises are, whether handling phone calls is one of their top problems, how much business they lose when a call goes unanswered, and how much they'd be willing to pay to solve it.

3.4. Become Knowledgeable Enough to Lecture on Solving the Problem

Going further, you should become one of the most informed people in the world on the problem you're trying to solve — informed enough to give a lecture on it. 🎓 This requires many conversations with customers and, sometimes, the experience of doing the work yourself.

3.5. Talk to Customers and Write Code at the Same Time

You don't need to be so fixated on talking to hundreds of customers before writing any code. It's important to pursue deep customer understanding and product development simultaneously.

"Deep understanding of customer needs, then ship product, then deeper understanding of customer needs, then ship a better product." Having real customers use the product provides concrete data that complements abstract knowledge, letting you know whether what you're building actually works.

4. How to Find Good Ideas in the AI Era

In the AI era, there are a few additional traits a good idea should have. 🤖

4.1. Push Against the Limits of AI Models

An idea should sit at the "edge" of what AI models can do today. Even if it barely works with current models, you should look for an idea that will clearly improve as the models advance.

"It can mean your product barely works on today's frontier models, but it will obviously get better as the models improve." Deeply understanding the bottlenecks that hold back model performance — and solving those bottlenecks — can sometimes become the company's core business. This aligns with Paul Graham's famous line about living in the future and building what's missing.

4.2. You Have to "Verticalize"

In the AI era, the cost of producing software is approaching zero. So what's valuable isn't simply "software for X," but things like customer trust, licenses, regulatory approvals, and ownership of outcomes.

"For example, providing insurance or healthcare itself, not just providing the software." If you want to break into the insurance industry, don't build software for insurance companies — become the insurer yourself. Instead of selling back-office software to banks, become the bank. Corgi Insurance made the bold decision to acquire an insurance carrier during its YC batch in order to own the entire commercial insurance stack, which resulted in much lower prices, faster processing, and ownership of all the economic upside.

4.3. Aim for the Most Ambitious Version

Jon Xu advises that you should aim for the most ambitious version of an idea. 🚀 The cost of pursuing a wildly ambitious startup idea and the cost of pursuing a mediocre one are nearly the same. Both are extremely hard and demand enormous amounts of time.

"Aim for the version that, if it works, could rewrite a sector of the economy. Because that's the version that protects you from competitors, attracts the best talent, and gives you a moat worth building." Good examples include diving into the most heavily regulated industries — like legal, healthcare, and financial services — taking on entrenched incumbents, or building hardware technology such as robotics for assembly in space.

5. What You Gain Even If the Idea Fails

What if you give it everything and the idea still fails? Jon Xu says you'll be in a far better position than when you started. 🌟

5.1. Clear Customer Data and Execution Ability

First, you gain unambiguous customer data. You'll know whether there's a genuinely "hair-on-fire" problem in the problem space, or whether you just thought there was. With this, you'll have firm conviction to pivot from and a much better understanding of how to move forward.

5.2. Discovering a Better, New Idea

Second, and more importantly, this process makes it highly likely you'll discover a new idea that actually works.

"Most founders, when they start, are solving a surface-level pain point. The real opportunity is almost always a deeper structural problem. In other words, going deep is mostly not a process of validating the idea you started with. It's a way to find the better idea beneath it." Along the way, you'll discover bottlenecks, gaps, and dev tools no one has built — and one of these could become an actual company.

6. Conclusion: Walk Fast in One Direction

The core message of this video is as follows:

1. Stop trying to find the perfect idea and just pick one.
2. Burn all the other "boats," learn everything you possibly can about your customers, and execute for them.

   "In the fog of an early idea, when you can only see 10 feet in front of you, the temptation is to take a few cautious steps in every direction. Tasting a bit of this and a bit of that, staying close to home. The problem is that gives you almost no information. What actually works is committing to one direction and walking fast." Jon Xu emphasizes that if you move quickly in one direction, you might not reach the right destination, but you'll generate far more information per unit of time. And while you're walking, you might arrive at a better destination you couldn't see at the start. The biggest failure isn't being wrong — it's failing to make a decision, wasting time, hopping across multiple ideas, never going deep enough on any of them, and learning nothing. So, pick one and go deep! 💪