How To Pick A Startup Idea
Jon Xu, a YC General Partner, argues against the "perfect idea" trap and juggling multiple startup concepts. He advocates for founders to commit deeply to a single idea, immersing themselves in customer problems to the point of understanding their business intimately. This approach, especially vital in the AI era, involves building at the frontier of technology, verticalizing to own outcomes, and aiming for the most ambitious version of a solution, ultimately yielding invaluable insights and potential for pivot, even if the initial idea doesn't succeed.
Jon Xu, a partner at Y Combinator, advises aspiring founders to overcome indecision and the “perfect idea” trap by committing to a single startup concept and pursuing it with intense focus. He outlines a rubric for rapidly validating an idea and generating meaningful data.
The Fallacy of the “Perfect Idea”
Founders often fall into two traps when beginning their startup journey:
- Seeking the perfect idea: This is an understandable impulse given the difficulty of building a startup, but it’s fundamentally flawed. Xu states, “You can only figure out what you should be working on by making contact with reality and getting feedback from customers.” The “perfect idea” cannot be conceived in a vacuum.
- Questioning founder-market fit: Founders, especially experienced ones, often doubt their suitability, believing they need decades of domain expertise. However, genuine curiosity combined with deep investigation and constant customer interaction can rapidly build “extraordinary knowledge.” As an example, Blake Scho of Boom Supersonic transitioned from adtech to supersonic flight, demonstrating that commitment and deep learning can outweigh prior direct domain experience.
Why Juggling Ideas Produces Bad Data
Working on multiple ideas simultaneously is a common but detrimental approach. It leads to “bad data” because founders cannot go deep enough on any single concept to get a clear signal. This lack of depth can cause founders to prematurely abandon a good idea or, conversely, to mistakenly pursue a bad one.
The “Go Deep” Strategy
To counter this, founders must commit fully to one idea. Xu details what “going deep” entails:
- Burn the other boats: Explicitly abandon all other startup ideas. This involves stopping work on them, communicating pivots to any involved parties, and channeling all focus into the chosen idea.
- Embrace a “new skin”: This means a comprehensive transformation of the company’s identity to align with the new focus. Changing the company name, emails, website, and even the internal narrative ensures singular commitment. GovDash, a YC-backed company, exemplified this by pivoting five times, each time adopting a new identity and becoming true domain experts in government procurement, which ultimately led to their success.
Understanding Your Customer’s Business
To truly go deep, founders must achieve an intimate understanding of their customers’ world:
- Run their business: The ultimate test is whether you could, hypothetically, run your customer’s business. For example, if building voice agents for cleaning services, it’s not enough to talk to 20 owners; you need to understand their daily crises, top problems (e.g., unanswered calls), business losses, and willingness to pay for solutions.
- Teach a class: You should be among the most informed people globally on the problem you’re solving.
- Tight feedback loop: This deep understanding comes from a continuous cycle of customer conversations, product delivery, and further refinement based on real-world usage. “Real customers using your product produces concrete data that will complement your abstract knowledge.”
Qualities of Good Ideas in the AI Era
Beyond customer pull, good ideas in the AI era possess specific characteristics:
- Building at the edge of what models can do: Your product should “barely works on today’s frontier models, but will clearly improve as they get better.” Intimately understanding current bottlenecks is crucial; solving these bottlenecks might even become the company itself. This aligns with Paul Graham’s advice to “live in the future and then build what’s missing.”
- Verticalization: Selling an outcome, not just software: As the cost of software production approaches zero with AI, true value lies in owning the entire outcome. Instead of building software for insurance companies, be the insurer. Corgi Insurance, for instance, acquired an insurance carrier to own the full commercial insurance stack, from underwriting to customer service, allowing them to offer better pricing and efficiency than traditional players.
- Aim for the most ambitious version: The effort required for a modest idea is often similar to that for an ambitious one, both being extremely challenging. Ambitious ideas—such as those in highly regulated industries (legal, healthcare), competing with large incumbents, or developing hardtech (e.g., robotics for space assembly)—create defensible moats, attract top talent, and have the potential to rewrite entire sectors of the economy.
The Value of Failure
Even if an initial idea fails after going deep, the founder is in a significantly better position:
- Unambiguous data: You gain clear customer data, understanding whether a “hair on fire” problem truly exists. This provides solid conviction for future pivots and improved execution.
- Discovery of new ideas: Going deep often reveals deeper, structural problems that represent greater opportunities than the initial surface-level pain points. Especially in the AI space, exploring the limits of current models highlights bottlenecks or missing development tools that can evolve into entirely new, viable company concepts.
Conclusion: Walk Fast in One Direction
The core takeaway is to stop searching for the “perfect” idea. Instead, pick one, commit fully by “burning the other boats,” become an expert on your customer, and execute rapidly. While the path ahead might be unclear, taking decisive, fast steps in one direction generates significantly more information than cautiously dabbling across multiple options. The worst failure mode is not being wrong, but rather indecision and a lack of deep commitment that prevents genuine learning and progress. By going deep, founders often find a better destination than they could have imagined at the start.