How To Pick A Startup Idea

This video advises founders to stop overthinking their startup ideas, since waiting for a perfect concept in the abstract is a failure mode—real insight comes from customer conversations and real-world feedback. Instead, pick one idea you're genuinely curious about, commit fully by burning other options, and go deep enough to know your customer's business intimately. In the AI era, founders should verticalize to own the full stack and sell outcomes rather than just software, and they should choose the most ambitious version of the idea because the effort is the same but the payoff is rewriting a sector. Going deep often uncovers entirely better new ideas through discovery, not just validation of the original. Ultimately, the worst failure is indecision; dabbling across ideas teaches nothing, while committing to one path and executing for the customer is how real learning and progress happen.

• ▶ 0:44 The most important advice is simply: don’t overthink your startup idea; overthinking in the early days mainly shows up as two failure modes.
• ▶ 0:59 Failure mode one is waiting for the "perfect" idea before starting—this doesn’t work because you can’t find the perfect idea in the abstract; you need to talk to customers and get real-world feedback.
• ▶ 1:51 Instead of obsessing over whether you’re the perfect founder, pick an idea you’re curious about, go deep quickly, and talk to customers—don’t let perceived lack of experience stop you (e.g., Blake Scholl from adtech founded Boom Supersonic).
• ▶ 3:12 Pick one idea and go deep: burn other boats, stop working on alternatives, and focus single-mindedly on the chosen idea.
• ▶ 4:41 Know you’re going deep enough if you could run your customer’s business—knowing their daily crises, top problems, and what they’d pay to solve them.
• ▶ 5:43 Get depth through a tight loop: combine customer conversations and product delivery so real usage data sharpens your understanding and your product iteratively.
• ▶ 6:23 A good idea should sit at the edge of current model capabilities: it barely works today but clearly improves as models get better, and you must understand its bottlenecks intimately—solving a bottleneck can itself become the company.
• ▶ 6:53 In the AI era, verticalize to sell an outcome, not just software: own the full stack (e.g., be the insurer, be the bank) to capture trust, licenses, and economics—like Corgi Insurance owning the entire commercial insurance stack.
• ▶ 8:23 Choose the most ambitious version of the idea, because effort is the same and the ambitious version rewrites a sector; if it fails, you still gain unambiguous customer data, real conviction, and sharper execution for the pivot.
• ▶ 9:42 Going deep into a problem often uncovers a better, new idea rather than just validating the original one — it’s a discovery process, not just validation.
• ▶ 10:19 Stop trying to find the perfect idea: pick one, commit fully, and focus on learning about the customer and executing for them.
• ▶ 11:06 The worst failure is not being wrong — it’s not making a decision; dabbling across ideas without going deep means you learn nothing.

• ▶ 0:09 Intro and Overthinking Failure Modes (0:09 - 2:31) - - John explains why founders overthink and covers the two common failure modes: waiting for the perfect idea and questioning founder fit.
• ▶ 2:31 Commit to One Idea and Go Deep (2:31 - 6:06) - - Commit fully to one idea, go deep, and use the "run your customer's business" test as the high watermark of execution.
• ▶ 6:06 Validating Good Ideas and Going Deep Through Failure (6:06 - 9:42) - - Covers the three qualities of good AI ideas and what happens when the idea fails while you’re fully committed.
• ▶ 9:42 Going Deep Uncovers Better Ideas and Takeaway (9:42 - 11:30) - - Going deep often reveals a better idea underneath; the key takeaway is to stop hunting for the perfect idea and just commit.