How To Pick A Startup Idea

Key Concepts

• Founder-Market Fit: The alignment between a founder's skills/experience and the startup idea.
• Going Deep: The process of fully committing to one idea, becoming a domain expert, and understanding the customer's business at a granular level.
• Burning the Boats: The act of abandoning all other startup ideas to focus exclusively on one, including rebranding and changing internal narratives.
• Verticalization: Moving beyond selling software to owning the entire outcome (e.g., becoming the insurer rather than selling software to insurers).
• High Watermark: A standard for validation where a founder understands a customer's business well enough to run it themselves.

---

1. Overcoming Overthinking

John, a partner at YC, identifies two primary "failure modes" that prevent founders from making progress:

• The Search for the "Perfect" Idea: Founders often wait for the ideal concept before committing. John argues this is impossible in the abstract; reality and customer feedback are the only ways to validate an idea.
• The "Perfect Founder" Trap: Founders often believe they need a decade of domain experience. John counters that curiosity and deep, rapid immersion are sufficient.
  • Example: Blake Scholl (CEO of Boom Supersonic) transitioned from adtech at Amazon/Groupon to commercial supersonic flight, proving that domain expertise can be developed quickly.

2. The Methodology of "Going Deep"

To avoid the "bad data" produced by juggling multiple ideas, founders must commit to a single path.

• Burn the Boats: Stop working on other ideas, pivot communications, and focus entirely on the chosen path.
• The "New Skin" Framework: Founders should fully immerse themselves in the new domain, changing company names, emails, and mission narratives to reflect their new focus.
  • Case Study: GovDash pivoted five times, changing their identity each time. By the fifth iteration, they became such deep experts in government procurement that they achieved significant scale and raised a Series B.

3. Validating Through Execution

John suggests a "high watermark" for validation: Could you run your customer's business?

• Actionable Knowledge: You must understand the customer's daily crises, top-five problems, and the financial impact of those problems.
• Tight Feedback Loops: Do not wait to talk to hundreds of customers before coding. Instead, use a loop: Deep understanding → Product delivery → Deeper understanding → Better delivery.

4. Qualities of Good Ideas in the AI Era

John outlines three criteria for evaluating startup ideas:

1. Edge of Model Capability: The idea should sit at the limit of what current AI models can do. If the product struggles due to a specific bottleneck, solving that bottleneck could become the business itself.
2. Verticalization (Outcome Ownership): In an era where software costs are trending toward zero, value lies in trust, licenses, and regulatory permission.
   • Case Study: Corgi Insurance (YC Summer '24) didn't just build software for brokers; they acquired an insurance carrier to own the entire stack, from underwriting to customer service.
3. Ambitious Scope: The effort required for a "modest" idea is the same as for a "world-changing" one. Aiming for the latter protects against competitors and attracts better talent.

5. The Value of Failure

John argues that even if an idea fails, the process of "going deep" is never wasted:

• Unambiguous Data: You gain concrete evidence of whether a "hair-on-fire" problem exists.
• Discovery of the "Real" Idea: Most founders start by solving surface-level pain points. By going deep, they uncover structural gaps or missing dev tools that lead to the actual, viable company.

---

Synthesis and Conclusion

The core takeaway is that speed and commitment generate more information than cautious, multi-directional exploration. In the early stages of a startup, the "fog" of uncertainty is best cleared by picking one direction and moving fast. The worst outcome is not being wrong, but rather "spinning your wheels" by dabbling in multiple ideas without ever reaching the depth required to learn anything meaningful. As John concludes: "Pick one and go deep."