John on How To Pick A Startup Idea

1. Key Themes

Overthinking Is the Enemy of Progress

The most dangerous startup failure mode isn't picking the wrong idea — it's failing to commit to any idea at all. John frames paralysis as an active harm, not a neutral waiting state.

"It's extremely hard to make meaningful progress on a startup without committing to a single idea." 00:00:09

The Perfect Idea Cannot Be Found in the Abstract

Founders believe they can think their way to the right idea before acting. John argues this is structurally impossible — reality and customers are the only valid testing ground.

"It's impossible to figure out the perfect idea in the abstract. You can only figure out what you should be working on by making contact with reality and getting feedback from customers." 00:01:06

Founder-Market Fit Is Real But Frequently Weaponized Against Yourself

Domain expertise matters, but founders — especially second-timers — use the concept as a reason never to start. Deep curiosity and rapid customer immersion can substitute for years of prior experience.

"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." 00:01:23

Parallel Exploration Produces Bad Data

Working on multiple ideas simultaneously feels like optionality but actually destroys signal. Shallow engagement with each idea means you can neither confirm a good one nor kill a bad one with confidence.

"If you don't actually go deep on an idea, but instead juggle it with several others, you won't get good signal about whether what you're doing actually works. And if you don't get good signal, then you could either prematurely talk yourself out of a good idea or convince yourself that a bad one is worth continuing." 00:02:37

"Burn the Other Boats" — Commitment as a Signal-Generation Mechanism

Commitment isn't just psychological resolve; it's an information strategy. Walking fast in one direction generates far more data per unit of time than cautious sampling in all directions.

"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." 00:10:25

The Benchmark for Going Deep: Could You Run the Customer's Business?

The standard for customer understanding isn't "talked to 20 people." It's whether you could be dropped into the customer's operation tomorrow and function. This is a much higher and more actionable bar.

"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?" 00:04:44

In the AI Era, Good Ideas Must Verticalize and Own the Outcome

As software cost approaches zero, the valuable assets become trust, licenses, regulatory permission, and outcome ownership — not the software layer itself.

"In the AI era, the cost of producing software is going to zero. So the things that actually become valuable aren't just software for X. They're customer trust, licenses, regulatory permission, and outcome ownership. So if you want to get into the insurance space, don't build software for insurance companies. Just be the insurer." 00:07:04

The Real Idea Is Usually Underneath the First Idea

Going deep on an idea is not primarily about validating that idea. It is a process for discovering the deeper structural problem that represents the actual opportunity — which almost always differs from the surface-level pain point you started with.

"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 almost always happens, especially if you're at the forefront of what models can do today. You'll notice the bottlenecks, the gaps, the dev tools nobody's built. And one of those could turn out to be the actual company." 00:09:54

2. Contrarian Perspectives

You Don't Need Years of Domain Experience — Curiosity Plus Customer Immersion Is Enough

Conventional wisdom says build what you know. John directly challenges this with a concrete example of an ad-tech founder successfully building a supersonic aviation company.

"If you pick an idea you're curious about, go extremely deep and most importantly, talk to customers. It's often possible to develop extraordinary knowledge in a short amount of time." 00:01:48

Pursue the Most Ambitious Version — It's Not Actually Harder Than the Modest Version

Most founders believe a smaller, more tractable idea reduces execution risk. John argues the costs are roughly equal, making ambition the dominant strategy on a risk-adjusted basis.

"The cost of pursuing a wildly ambitious startup idea and the cost of pursuing a modest one are roughly the same. They're both extremely hard. They both place extreme demands on your time. So aim at the version that, if it works, rewrites a sector of the economy." 00:08:29

Failure at a Startup Idea Is a Net Positive Starting Position

The standard view treats a failed startup as a sunk cost. John reframes it: going deep generates unambiguous customer data and a better-calibrated next idea — a genuinely superior position to where most founders begin.

"The good news is that you'll be in a dramatically better position than where you started. First, you have unambiguous customer data... you will often come away from the process with a new idea that will actually work." 00:08:57

Don't Wait to Fully Understand the Customer Before Writing Code — Do Both Simultaneously

The lean-startup orthodoxy often implies sequential steps: research, then build. John explicitly rejects this in favor of a tight parallel loop.

"Don't obsess over needing to talk to hundreds of customers before writing code. The goal is to do both at the same time, in a tight loop. Deep understanding of customer needs, then product delivery." 00:05:37

3. Companies Identified

Boom Supersonic

A commercial supersonic aviation company. Mentioned as proof that founder-market fit can be built rather than inherited — its CEO came from ad-tech with no aviation background.

"Take Blake Scholl, the CEO of Boom Supersonic. Blake spent his early career working on ad tech at companies like Amazon and Groupon before deciding to work on commercializing supersonic flight. Lots of people probably thought he was crazy, but now Boom is a billion dollar company." 00:01:48

GovDash

A startup that helps customers win government contracts. Mentioned as the canonical example of committing fully to each pivot — changing name, emails, and internal narrative each time — until finding a product-market fit so strong they couldn't keep up with demand.

"I worked with a startup called GovDash that helps customers win government contracts. They pivoted at least five times before finding this idea. And each time they explored something new, they changed their company name and how they talked about their mission... By truly becoming domain experts in government procurement, their fifth idea worked so well that they could barely keep up with demand. They recently raised the Series B to scale the business." 00:03:49

Corgi Insurance

An AI-powered commercial insurance company from YC's Summer 2024 batch. Mentioned as the exemplar of full-stack verticalization — refusing to be a tech-enabled broker and instead acquiring an insurance carrier during their YC batch to own the entire stack.

"They set an ambitious goal of owning everything from underwriting to providing customer service, the entire commercial insurance stack, and even took the unprecedented step of acquiring an insurance carrier during their YC batch to make it happen. Being the full-stack insurance company allows Corgi to underwrite any insurance line in any vertical with a fraction of the headcount of traditional carriers." 00:07:32

Amazon

Referenced as one of the companies where Boom Supersonic CEO Blake Scholl worked before pivoting to aviation. 00:01:48

Groupon

Referenced alongside Amazon as part of Blake Scholl's ad-tech background before founding Boom Supersonic. 00:01:48

4. People Identified

Blake Scholl

CEO and founder of Boom Supersonic. Mentioned as a compelling counter-example to the "you must have domain expertise" rule — he came from ad-tech and built a billion-dollar supersonic aviation company.

"Blake spent his early career working on ad tech at companies like Amazon and Groupon before deciding to work on commercializing supersonic flight. Lots of people probably thought he was crazy, but now Boom is a billion dollar company." 00:02:10

Paul Graham

Co-founder of Y Combinator. Cited for his "live in the future and build what's missing" framework, which John applies to the AI-era product development context.

"This is a version of Paul Graham's well-known quote, that you should live in the future and then build what's missing." 00:06:34

5. Operating Insights

Make Each Pivot Total — Change the Name, Email, and Internal Narrative

Most teams pivot the product but preserve the old identity, which means the old mental model subtly persists and contaminates the new direction. GovDash's practice of changing everything — name, emails, stated mission — is a forcing function that ensures genuine cognitive reset.

"This could mean changing your company's name, your emails, your website, and even your internal narrative about why you're building a startup in the first place." 00:03:49

Use "Could I Run This Business Tomorrow?" as Your Customer Discovery Benchmark

Rather than measuring discovery by conversations completed, measure it by operational readiness. This reframe converts a fuzzy qualitative process into a concrete, testable standard and prevents premature product-building on shallow understanding.

"The question isn't just whether you've talked to 20 owners. The question is, if I dropped you into a cleaning business tomorrow, would you know how to run it?" 00:04:44

Target AI Ideas That Sit at the Edge of What Models Can Do Today

Products that barely work on current frontier models but have a clear improvement trajectory as models advance are strategically superior to products that either work comfortably today or are hopelessly out of reach. Understanding your bottlenecks at a technical level is itself a product strategy.

"The idea sits at the edge of what models can do today. This might mean that your product barely works on today's frontier models, but will clearly improve as they get better. You should understand the bottlenecks impeding your product's performance intimately. If a particular bottleneck doesn't clear the way you hoped, solving that might become the company." 00:06:05

6. Overlooked Insights

Acquiring a Regulated Asset During a Seed-Stage Batch Is Now a Viable Playbook

John mentions almost in passing that Corgi Insurance acquired an insurance carrier during their YC batch — not after raising a Series A, not after proving the model, but while still in the seed program. This is a non-obvious signal that AI-native companies can compress the timeline to owning regulated infrastructure dramatically, and that YC is actively encouraging this. For investors, this suggests the moat-building phase in regulated verticals (insurance, banking, healthcare) is arriving far earlier in a company's life than traditional venture pattern-matching would suggest — and deals may be missed if investors wait for conventional proof points.

"They even took the unprecedented step of acquiring an insurance carrier during their YC batch to make it happen. Being the full-stack insurance company allows Corgi to underwrite any insurance line in any vertical with a fraction of the headcount of traditional carriers." 00:07:32

The Dev-Tools Gap Inside AI Applications Is an Underappreciated Investment Category

John makes a single throwaway observation that founders working at the frontier of AI will inevitably notice missing infrastructure — and one of those gaps "could turn out to be the actual company." This is not a generic encouragement; it is a specific prediction that the next wave of high-value companies will emerge as internal tools built to solve bottlenecks within AI applications get spun out. Investors and operators should be watching what serious AI application builders are building for themselves.

"You'll notice the bottlenecks, the gaps, the dev tools nobody's built. And one of those could turn out to be the actual company." 00:09:54