别死掉：YC合伙人Dalton Caldwell谈如何让创业公司活下去

Lenny Rachitsky notes that founders who worked with Dalton Caldwell kept citing his simple, pragmatic advice: sell, make money, don't run out of cash. Caldwell compares it to a mic'd-up basketball coach or Tiger Woods talking to his caddy, where even the best athletes hear mundane fundamentals. Being reminded of the basics, he says, puts people in the right mindset.

His own mantra is "just don't die", and he says it is often what founders call his most impactful advice. The reasoning is that across YC's portfolio, founders rationally should have given up at some point. Airbnb, he says, probably should have shut down three or four times before YC, yet its founders kept working through near-death experiences before they looked like an overnight success.

Asked when giving up is the right call, Caldwell says it is nuanced but offers a test: are you still having fun, and do you enjoy your co-founders? If so, lean toward continuing; if the work is hurting you and your relationships, lean the other way. Founders who turn things around, he says, often truly love their customers and product, as the Airbnb founders did with their first hosts.

He adds that shutting down is no big deal: if you act with integrity, people will remember you fondly in 10 or 20 years. Spending your limited career on something miserable just to save face is, in his view, a heavy opportunity cost. Rachitsky agrees that life is short and the question of enjoyment is the key.

To reassure a struggling batch, Caldwell tells the story of Winter 2017. In his group, a VR headset startup from Stanford dropouts was ashamed and ready to quit, and another, a UK peer-to-peer payments app, wasn't growing. After pivoting, the first became Brex and the second became Retool, the most successful companies of the roughly 35 to 40 in his group.

Caldwell's pattern for successful pivots is that they get warmer rather than colder to the founders' expertise and build on what they learned. Brex's founders had earlier worked on fintech in Brazil, and Retool's had built internal dashboards for their payments app. Zip's founder knew Airbnb's procurement process from working there. He likens a good pivot to going home.

Segment shows how expertise can come from the failed idea itself. It began as a tool for students to tell professors they were confused, then became a Mixpanel competitor, then a JavaScript snippet sending events to multiple endpoints. The founders could not have started with the final idea because they only learned how analytics works by grinding through the earlier ones.

On when to pivot, Caldwell looks at how many untried growth ideas a founder has. Airbnb kept trying zany tactics like cereal and conventions; if the ideas left are as weak as "maybe we should pay influencers", pivoting makes sense. For Zip, he suggested finding large, publicly traded or private-equity-owned incumbents that customers hate, which led them to procurement software.

Caldwell says that seeing all YC applications shows founders pitching the same ideas, because they read the same blogs and follow the same people on Twitter. His advice is to go to the metaphorical mountains rather than the middle of the city, using personal experience as Brex, Retool and Whatnot did. Nobody else, he notes, was building a marketplace for Funko Pops. Trends such as trucking startups turn fashionable fast once someone succeeds in an unfashionable space.

A tarpit idea, as Caldwell defines it, is not an obviously bad idea but one that seems like an unsolved problem, collects positive feedback, and has plausible arguments in its favor. Examples include apps for coordinating where friends meet at night and music discovery, which he worked on himself in his first startup. People have been starting these since the '90s, and the early validation is what draws founders in.

Rachitsky shares his own tarpit, Localmind, an app for asking people checked in nearby how a place is, which users loved and then never reused. Caldwell recalls how Foursquare clones dominated for years and says those that did not pivot out didn't work. A tarpit is defined by being appealing, popular and easy to validate.

To understand why investors decline, Caldwell advises founders to put themselves in an investor's shoes. Most make few investments, so even a pitch they like gets a no if it isn't the one they are most excited about. He thinks founders imagine a secret reason or crave feedback, when the real feedback is simply that they didn't want to invest.

On market size, Caldwell says it matters more the later the stage and valuation, and much less at pre-seed. The TAM of Uber or Airbnb would have looked tiny, and Razorpay's market was small because few people in India used credit cards in 2015. For early YC companies he worries instead about how they get users and whether they make something people want.

Rachitsky points out that many other investors do focus on TAM, so founders may get turned down for that. Caldwell agrees and adds that investors dislike being asked to take a crazy leap of faith when they have less risky alternatives. Debating the TAM with them, even with proof points, rarely works.

Asked for his top advice for product-minded founders, Caldwell says it is not to over-delegate and to stay close to the product. A trap he sees is hiring very senior people with impressive big-tech resumes early. Founders say they get it, and then two years later find they over-delegated and need to clean it up.

The pressure often comes from investors and from the feeling that once you raise money, you must spend it and show you are building a world-class organization. The result is nice people with shiny resumes, and a founder who took their eye off the ball. Caldwell says the best product founders stay in the weeds and keep talking to customers at any stage, because caring about users can't be delegated.

On finding time, he says that if you care about customers and product, your instincts about priorities are good. He would cut tons of time spent with investors and networking before anything else. The rule is to stay obsessed with the product.

Rachitsky relays Gustav's view that startups fail because they don't talk to customers. Caldwell agrees, but frames the cause differently: founders lose hope. He can tell when founders have resigned themselves to failure versus when they say they have one more move, and as long as you don't accept failure there are usually many more moves, such as getting profitable or launching a new product.

He argues it is rare for a company with lots of firepower and a positive outlook to simply run out of money. More often there is some cash left, but the idea isn't working, the founders fight, they can't agree on what to build and they shut down. If you are truly done, he adds, there is no need to go through the motions.

On how common near-death experiences are, he says that if defined as founders seriously wondering whether it was over, essentially everyone goes through it. For truly dire situations the share is still high, perhaps 50 percent. Many founders come close to the end and keep going through sheer will.

Caldwell says when he asks aspiring founders how many customers they talk to, they get quiet. People hide behind a keyboard, building landing pages and buying Instagram ads, largely because they are shy. His advice is to go out in the physical world and show people a product, and to push through the awkwardness, as the Airbnb founders had to.

As a heuristic, 20 to 30 percent of your calendar should say customer call or meeting. He notes that some founders say they will raise a pre-seed round before talking to customers, which he sees as social anxiety. Brex and Retool sold within their YC batch, Zip relentlessly got procurement leaders on calls and pre-sold, and PostHog launched open source and filled its calendar with eager users.

Caldwell also describes the Collison Install. When a customer said yes to Stripe but never implemented it, the Collison brothers would drop by, ask to drive the keyboard and install it themselves. The lesson is that even after a yes, you are not done with sales until the last mile of implementation is complete. He adds that Patrick Collison checked in with him weekly when he was an early customer.

Asked what successful founders have in common, Caldwell says it isn't personality: he has seen quiet and extroverted ones, and names Tony of DoorDash, Rujul, Grant of Whatnot, Patrick and Ryan of Flexport as very different people. What they share is a deep belief that they will make it work and won't accept it failing. That internal force, he says, can bend the world and convince employees and others.

Rachitsky asks how much of it is internal. Caldwell says internal, with a big caveat: nobody feels it before they have a good idea and customers. Conviction builds in a virtuous cycle as the product works and customers reflect it back; he recalls Patrick Collison saying he wasn't sure Stripe would work until a year or two in. Rachitsky ties this to Scott Belsky's test of whether conviction is rising or falling.

Caldwell explains that YC's request for startups, with 20 categories, isn't a list of the only ideas it will fund. It is meant to broaden founders' information diet toward ideas they don't consider. He wrote one on ERP software because he receives few applications there, and expects more after the post.

Other items include open source companies and space companies, which he says have done well and which founders may think too bold. Rachitsky lists more, including cancer, spatial computing, defense technology and bringing manufacturing back to America. Caldwell adds better enterprise glue, since the software that connects business systems is brittle and LLMs may improve it, and Rachitsky mentions small fine-tuned models as an alternative to giant generic ones.

Looking back at the early 2000s, Caldwell says the startup scene was tiny, like the Homebrew Computer Club: he cold emailed Reid Hoffman when LinkedIn had 12 employees and got a lunch. He didn't notice personality traits that predicted success, only staying power. Sam Altman was then a young Stanford dropout building Loot, and Caldwell once drove Sean Parker to the Oakland airport.

Caldwell also says these people reinvent themselves across eras, as Hoffman moved from PayPal to LinkedIn to venture capital. Rachitsky draws the lessons that careers are long, introverts and extroverts can both succeed, and that enjoying the work drives you forward. Caldwell, for his part, sold his music startup imeem to MySpace, where the new CEO was soon fired.

He then started PicPlz, a mobile photo-sharing app, with Andreessen Horowitz as a major investor and Marc Andreessen on its board. It reached around half a million to a million users, but another portfolio company pivoted into a similar product by combining photo filters with a social graph and took off; that company was Instagram. Because a16z was on his board, Caldwell says, the conflict meant it did not invest in Instagram, and he doesn't believe they held it against him.

For his contrarian view, Caldwell says growth hacking, analytics and A/B testing are a waste of time for very early startups. Much startup advice is written for series A or B founders, and seed-stage founders absorb it and get confused. When you have no customers, you should get one and talk to that person; imagine the Airbnb founders running split tests with four users whose names they knew.

Asked what is acceptable, Caldwell says to understand the metrics for your type of business, citing Whatnot's focus on buyers in a marketplace and Facebook saturating Harvard first. He advises looking at what comparable companies did from zero to one and ignoring what they do today. On failure, he says he has failed often as founder and investor, and the lesson is not to let it get to you and to keep going.

His closing advice is to start with customer validation rather than a pitch deck or fundraising: line up people who are excited, try to pre-sell something before writing code, and take that as a green light. Build once you believe at least one person would use it. In the lightning round he recommends basic sales books like Getting to Yes, and his life motto is to check that you're having fun.