How To Pick A Startup Idea
YC partner John advises founders to stop overthinking startup ideas and instead commit fully to one idea, going extremely deep on customer understanding. He argues that meaningful progress requires single-minded focus, and that the process of going deep often reveals better ideas than the one you started with. He also outlines three qualities of strong startup ideas in the AI era.
Summary
The video opens by identifying a common founder failure mode: inability to commit to a single startup idea. John argues that overthinking takes two primary forms — searching for the 'perfect' idea in the abstract, and questioning whether you have sufficient domain expertise to pursue a given space. He counters both by pointing out that real validation only comes through customer contact, and that deep curiosity combined with rapid learning can substitute for years of prior experience. He cites Blake Scholl of Boom Supersonic as an example of someone who succeeded in a field far outside his background.
John then makes the case for committing fully to one idea at a time, warning that working on multiple ideas simultaneously produces corrupted data — you can't get meaningful signal without going deep. He describes 'going deep' as a process of completely burning other options, changing your identity around the new idea (including company name, emails, and internal narrative), and becoming a genuine domain expert. He illustrates this with GovDash, a startup that pivoted five times, fully reinventing itself each time, before finding a product that worked so well they couldn't keep up with demand.
To measure whether you've gone deep enough, John proposes a high watermark: could you actually run your customer's business? Using the example of a voice AI product for cleaning services, he argues that founders need to know the daily crises, the top five problems, and the exact dollar value of missed calls — not just surface-level familiarity. He also frames this as being able to teach a class on the problem, becoming one of the most informed people in the world on the subject.
John then outlines three qualities that define strong startup ideas in the AI era. First, the idea should sit at the edge of what current models can do — barely working today but clearly improving as models advance, with the founder understanding the specific bottlenecks. Second, the idea should verticalize — meaning it should sell an outcome rather than just software, since the cost of software is trending toward zero. He uses Corgi Insurance as an example: rather than building tools for insurers, they became a full-stack AI-powered insurer, acquiring a carrier during their YC batch. Third, the idea should be the most ambitious version of itself, since the cost of pursuing a modest idea versus a wildly ambitious one is roughly the same, and ambition creates moats, attracts talent, and produces outsized outcomes.
The video closes by reframing failure: if an idea doesn't work, you come away with unambiguous customer data, real conviction for a pivot, and often a better idea discovered in the process of going deep. John argues that the real opportunities are usually structural problems underneath the surface-level pain points founders initially identify. The worst outcome isn't being wrong — it's never committing deeply enough to learn anything.
About this episode
<p>Many founders get stuck trying to find the perfect startup idea before they commit. But the perfect idea doesn't exist in the abstract. The only way to find what works is to pick one, go deep, and get feedback from real customers.</p><p><br /></p><p>In this episode of Startup School, YC's Jon Xu breaks down how to choose what to build, "burn the other boats," and go deep enough to practically run your customer's business— and why that depth is what surfaces the better idea underneath.</p>
Key Insights
- John argues that working on multiple ideas simultaneously produces bad data — without going deep on any one idea, founders cannot get reliable signal about whether something is working or not, leading to premature abandonment of good ideas or false confidence in bad ones.
- John claims that the high watermark for customer understanding is not the number of conversations held, but whether you could actually run the customer's business — knowing their daily crises, their top priorities, and the precise economic cost of their problems.
- John contends that in the AI era, the cost of software is trending toward zero, making pure software products increasingly undefendable — the real value lies in owning outcomes, licenses, regulatory permissions, and customer trust, which points toward verticalization rather than tooling.
- John argues that the cost of pursuing a wildly ambitious startup idea and a modest one are roughly the same — both are extremely hard and demanding — making the ambitious version strictly superior because it creates moats, attracts better talent, and can rewrite entire sectors.
- John claims that going deep on an idea is not primarily a validation process for the original idea — it is more often a mechanism for discovering the better idea underneath, as founders at the frontier of what models can do today will naturally encounter bottlenecks and gaps that represent the actual company opportunity.
Topics
Transcript
Hi, I'm John, and I'm a partner at YC. I often meet founders who have lots of ideas about what to work on and can't decide between them. Sometimes they're working on multiple things. Often they'll say that they're waiting to find the best idea before fully committing. Often they'll say that they're waiting to find the best idea before fully committing. But it's extremely hard to make meaningful progress on a startup without committing to a single idea. So in this video, I'm going to give you a rubric for how to stop overthinking, pick an idea, commit to it, and then figure out fast whether it's actually working. The most important piece of advice I'd give to founders…
Full transcript available for MurmurCast members
Sign Up to AccessMore from Y Combinator Startup Podcast
Paul Graham On Startups, Ambition, and Great Founders
Paul Graham discusses YC's evolution over 21 years, arguing that startups funded today are more ambitious than in the past, and explores how founder ambition, shipping speed, and formidability drive success. He shares perspectives on AI's surprising development as a 'bullshitting' system rather than a perfect-to-complex progression, and explains how YC's batch model has remained fundamentally unchanged despite scaling.
The World’s Largest Electric Aircraft Just Flew
HART Aerospace has successfully flown the world's largest electric aircraft, a 100-foot wingspan hybrid-electric plane designed to reduce regional air travel costs. The company, founded by Anders Forslund, evolved from a 3D-printed model seven years ago to a 40-person team building a commercially viable airliner that swaps jet engines for electric motors.
Max Junestrand: You Need The Willingness To Learn Faster Than Anyone Else
Max Junestrand of Legora describes the company's rapid growth from zero to $100M ARR in 18 months by building an AI operating system for lawyers. He discusses the importance of learning the market deeply, maintaining competitive culture, hiring for growth potential rather than prestigious credentials, and betting on improving AI models without fine-tuning.
Susan Kare: Designing Icons & Graphics For the Original Mac
Susan Kare, the iconographer for the original Macintosh, discusses her journey designing system fonts, icons, and graphics under severe technical constraints (16x16 black and white pixels). She shares design principles learned from mentors like Paul Rand, her experiences across multiple tech companies, and the philosophy that simplicity, metaphor, and meaningful design create universal, memorable user interfaces.
Chelsea Finn: This is the State of the Art in Robotics
Chelsea Finn from Physical Intelligence discusses advancing robotics toward general-purpose models that achieve high reliability through reinforcement learning, memory systems, and diverse training data. She demonstrates how robots can perform complex real-world tasks autonomously and argues that physical AI has reached a ChatGPT-like era comparable to language models, with models now being deployed in real-world applications.