Pick One Idea and Go Deep

I’m John, a partner at YC. I often meet founders who have several startup ideas and can’t decide which to pursue. Some work on multiple things at once; others say they’re waiting to find the best idea before committing. But meaningful progress is extremely hard without committing to a single idea.

The way forward is to choose one, pursue it deeply, and learn quickly whether it works.

Stop looking for the perfect idea—or the perfect founder

The most important advice I give founders struggling to choose an idea is to stop overthinking. Two traps come up repeatedly.

The first is the search for the perfect idea. The impulse is understandable: startups are hard, so shouldn’t you choose the best possible idea before committing? The problem is that you can’t identify it in the abstract. You discover what you should work on through contact with reality and feedback from customers.

The second is asking whether you’re the perfect founder for the opportunity. Founder-market fit matters. A nontechnical founder, for example, probably isn’t the right person to devise a killer developer-tool startup. But that consideration can become a reason never to begin.

But often founders, especially second-time founders, weaponize this line against themselves. They convince themselves that they need a decade of domain experience before they can start. The truth is, you don't.

If you choose something you’re curious about, go extremely deep, and, most importantly, talk to customers, it’s often possible to develop extraordinary knowledge in a short time. We see remarkable examples at YC.

Blake Scholl, the CEO of Boom Supersonic, spent his early career in advertising technology at companies such as Amazon and Groupon before deciding to commercialize supersonic flight. Plenty of people probably thought he was crazy. Yet Boom is now a billion-dollar company. Don’t let doubts about whether you’re allowed to work on something stop you from starting.

Commit deeply enough to get useful evidence

Founders sometimes pursue several ideas simultaneously because they think that’s the best way to discover which will work. The most serious problem with this approach is the quality of the data it produces.

If you juggle several ideas without going deep on any of them, you won’t get a good signal about whether they work. You might prematurely abandon a good idea or convince yourself that a bad one deserves more time.

When several ideas look equally attractive, choose one and investigate it in depth.

The first thing is that you should burn the other boats. That is, you should explicitly foreclose your other startup idea options. Stop working on them. Tell any customers that you've pivoted and work with single-minded focus on the idea you've chosen.

That commitment should feel like putting on a new skin—becoming an almost unrecognizable version of yourself. It could involve changing your company’s name, email addresses, website, and even your internal account of why you’re building a startup.

I worked with GovDash, which helps customers win government contracts. They pivoted at least five times before finding this idea. Whenever they explored something new, they changed the company name and the way they described their mission.

At one point, I forgot how to get in touch with them because they changed their email addresses with each pivot.

By becoming genuine domain experts in government procurement, they made their fifth idea work so well that they could barely keep up with demand. They recently raised a Series B to scale the business and meet that demand.

Understand the customer well enough to do the job

How can you tell whether you’re going deep enough? The high standard I use is whether you could actually run your customer’s business.

Suppose you’re building voice customer-service agents for cleaning services. Having spoken to 20 owners isn’t, by itself, the test.

The question is if I dropped you into a cleaning business tomorrow, would you know how to run it? Do you know what their daily crises are? Do you know whether answering the phone is a top five problem? Do you know how much business they lose when a call goes unanswered and what they would actually pay to never lose another one?

You need to answer those questions with very high confidence. Another test is whether you could teach a class on the problem you’re solving. Are you among the most informed people in the world on the subject?

Reaching that level takes many customer conversations and sometimes doing the job yourself. But don’t become fixated on talking to hundreds of customers before writing code. Customer understanding and product development should happen together in a tight loop: understand the need, deliver a product, deepen your understanding, and deliver something better.

Real customers using your product generate concrete data that complements your abstract knowledge. That combination tells you whether what you’re building is actually working.

Look for ideas at the model frontier

Once you’re pursuing an idea deeply, customer pull is the most obvious sign that it deserves continued work. In the AI era, there are other qualities to watch for as well.

One is that the idea sits at the edge of what models can do today. Your product might barely work on current frontier models, yet clearly improve as those models get better.

You should understand its performance bottlenecks intimately. If a particular bottleneck doesn’t clear as you hoped, solving it might become the company. This is a version of Paul Graham’s advice to live in the future and build what’s missing.

Sell the outcome and own the whole solution

A second quality is that the idea should verticalize: ultimately, it should sell an outcome. That could mean providing insurance or medical care rather than selling software alone.

In the AI era, the cost of producing software is going to zero. What becomes valuable is customer trust, licenses, regulatory permission, and ownership of the outcome. If you want to enter insurance, become the insurer. If you’re considering back-office software for banks, become the bank.

Corgi Insurance, an AI-powered commercial insurance company from YC’s Summer 2024 batch, illustrates this ambition. The founders weren’t content to be a technology-enabled broker or even a managing general agent, because those roles owned only part of the solution. They wanted the entire commercial insurance stack, from underwriting to customer service.

During their YC batch, they took the unprecedented step of acquiring an insurance carrier to make that happen. Being a full-stack insurance company allows Corgi to underwrite any insurance line in any vertical with a fraction of the head count of traditional carriers. They can offer far better pricing, much faster turnaround, and own all the economics.

Pursue the most ambitious version

The third quality is ambition. An idea should become the most ambitious version of itself.

That may sound counterintuitive, but pursuing a wildly ambitious startup and pursuing a modest one cost roughly the same. Both are extremely hard, and both make extreme demands on your time. Aim for the version that, if it works, rewrites a sector of the economy. That is also the version that protects you from competitors, attracts the best talent, and creates a moat worth building.

It might mean building and selling into highly regulated industries such as legal services, healthcare, or financial services. It might mean taking on a $10 billion legacy SaaS company. Or it might mean building hard technology, such as robotics for space assembly.

Let depth reveal the better idea underneath

What happens if you commit, learn deeply, execute—and the idea fails?

You’ll be in a dramatically better position than when you started. You’ll have unambiguous customer data showing whether the space contains a truly urgent problem or whether you talked yourself into believing it did. You’ll have real conviction on which to base a pivot and a better understanding of how to execute next time.

More importantly, you’ll often leave the process with a new idea that will actually work. Most founders begin by addressing surface-level pain points. The real opportunities are almost always the deeper structural problems.

In other words, going deep isn't primarily a process for validating the idea you started with. It's a way to find the better idea underneath.

This discovery almost always happens, especially when you’re working at the frontier of model capabilities. You notice bottlenecks, gaps, and developer tools that nobody has built. One of those may turn out to be the actual company.

So choose an idea, close off the competing options, learn everything you can about the customer, and try to execute for them.

In the early idea fog, where you can only see 10 ft in front of you, the temptation is to take a few cautious steps in every direction. Sample a little here, a little there, stay close to home. The problem is that gives you almost no information. What actually works is to commit to one direction and walk fast. You're not guaranteed to end up in the right place, but you generate much more information per unit of time. And when you're walking, you might arrive at a better destination, one you couldn't have seen from the start.

The worst failure mode is indecision: spinning your wheels and dabbling among ideas without pursuing any of them deeply enough to learn. Pick one and go deep.