Dalton Caldwell on surviving startups, pivots, and avoiding tar pit ideas

Executive overview

Most startups that succeed do so because founders refuse to quit when quitting would have been rational. Dalton Caldwell, managing director at YC with 21 batches of experience, argues the single most important startup skill is sustaining irrational persistence through near-death moments that every startup encounters.

The corollary: knowing when it actually is time to stop. If you've stopped enjoying the work and run out of ideas, forcing it is just an opportunity cost on your life.

The founders who build big companies don't have a special personality type — they have an unshakeable internal belief that they will make it work, regardless of incoming evidence.

Just don't die

• Rational analysis would have told Airbnb founders to quit three or four times before YC.
• Near-death experiences are universal — 100% of startups hit a point where founders consider stopping.
• The most common cause of startup death is not running out of money; it's founders losing hope, fighting with co-founders, and running out of ideas.
• Founders who turn it around consistently love their product, their customers, and each other.
• Ask: am I still having fun? Do I love my co-founders and my customers? If yes, lean toward continuing.
• If the startup is actively damaging your life and relationships and you've stopped caring, stopping is legitimate.
• Shutting down with integrity is forgotten quickly; few will remember in 10–20 years.

What makes a good pivot

• Good pivots move warmer — closer to what you already know, not further away.
• Brex: pivoted from VR headsets to fintech because the founders had prior fintech experience in Brazil.
• Retool: pivoted to developer tooling because both founders had built internal dashboards at internships and for their prior startup.
• Segment: couldn't have started with their final idea — they discovered analytics insights only by grinding on earlier ideas for years.
• Zip: went through six pivots before finding procurement software; the co-founder's Airbnb experience gave him deep knowledge of broken procurement.
• Signs it's time to pivot: you've run out of ideas for growth on the current path; the few remaining ideas are low-quality grasps (e.g., "maybe we should pay influencers").
• Signs to stay: you still have a half-dozen untried growth ideas and genuine gasoline left in the tank.

Tar pit ideas

• A tar pit idea is not just a bad idea — it's one that seems unsolvable, generates positive validation, and has trapped founders since the 90s.
• The trap: you can get friends to say yes to it, early users to try it, and real signals that feel like traction — but the structural problems are intractable.
• Classic examples: apps to coordinate where friends go out; music discovery startups; consumer location-based apps (Foursquare clones).
• The danger is precisely that it draws you in and keeps you stuck, because it looks like progress.
• Avoid it by mining your own deep expertise instead of consuming the same information diet as every other founder.

Avoiding the same ideas as everyone else

• If you read the same blogs, follow the same Twitter accounts, and hang out with the same people, you will generate the same startup ideas.
• Go off the beaten path: dig into your personal experience, niche expertise, or underserved industries.
• YC's request for startups is an information-diet prompt, not a prescriptive list — it's designed to surface idea spaces founders filtered out (ERPs, open source, space, defense, manufacturing, enterprise glue, fine-tuned models).
• Find large incumbents with very low NPS and horrible software; that's a reliable prompt for a pivot or new idea.

Talking to customers

• Talking to customers is universally preached and almost universally under-practiced.
• Don't hide behind landing pages and Instagram ads — that is not talking to customers.
• The barrier is social anxiety and fear of looking stupid, not lack of time.
• Self-assessment: in the past month, how many in-person or live meetings have you had with potential customers?
• A healthy calendar has 20–30% of slots labeled as customer meetings or calls.
• Even after a customer says yes, you're not done — finish the last mile of implementation (the "Collison install": show up, sit down, take the keyboard, don't leave until it's live).
• Zip: ran hundreds of LinkedIn DMs asking about procurement pain before writing code; pre-sold the product to beta customers.

Why investors say no

• Investors make very few bets per year; saying no to something they like is normal — they're waiting for something that clears a higher bar.
• There is rarely a secret reason. The feedback is the no.
• Put yourself in the investor's shoes: if you could only do a few investments a year, you'd also pass on most things you liked.
• TAM matters much less at pre-seed/YC stage; early-stage investors need to believe in a future market, not a current one (Razorpay in 2015 India, Airbnb's tiny initial TAM).

Avoiding over-delegation

• Founders consistently over-delegate too early, especially after raising money and being pushed by investors to "build a world-class team."
• Hiring senior executives with shiny resumes too early causes founders to take their eye off product and customers.
• The founders who build great products stay deeply in the weeds on product and keep talking to customers regardless of company size.
• If you genuinely love your product and your customers, your intuitions on time allocation are correct — don't replace them with networking and investor relations.

Growth hacking is wrong for early-stage startups

• Most growth and A/B testing advice is written for companies that already have scale.
• Seed-stage founders who consume Series A/B growth frameworks get confused and distracted.
• When you have no customers, running split tests and analytics dashboards is actively harmful — it substitutes for the real work of getting a first customer.
• Zero-to-one strategy: look at how your archetype company got its first 1,000 users, not what they do today.

Patterns across founders who build big companies

• No single personality type — Tony Xu, Patrick Collison, Rajul Vohra, Grant from Whatnot are all very different.
• What they share: they deeply want it, believe they will make it work, and refuse internally to accept failure even when all external data points the other way.
• This conviction typically builds over time as customers reflect it back; it is not present on day one.
• Conviction grows with product-market fit; you build it, you don't start with it.
• Careers are long and multi-era. Reid Hoffman ran PayPal, LinkedIn, then became a VC. Reinvention is normal.

Before you write code

• Talk to potential customers and try to pre-sell before building anything.
• If you can line up even one person who's genuinely excited, that's a green light.
• Pre-selling beats PowerPoint decks and fundraising as a first step; it gives you conviction and early signal simultaneously.