Paul Graham, Founder of Y Combinator on Startups, Ambition, and Great Founders
Y Combinator
The Same Old Startup Advice 0:00
Paul Graham points out that this talk marks the 47th YC batch, the 21st year of Y Combinator, and yet he writes each talk from scratch every time. He says most of starting a startup stays the same no matter the era, whether the technology of the moment is microprocessors, the internal combustion engine, or AI.
Startups Getting More Serious 1:01
Graham pushes back on the idea that YC has fallen from some golden age, noting that critics have said this since around 2008. He argues today's startups are more ambitious than the old days, pointing to a company doing what he jokingly calls intercontinental ballistic cargo, meaning a rocket that lands and delivers goods instead of exploding.
Fighting Cancer With Many Approaches 2:00
Graham describes a current batch startup doing on demand research for cancer patients, which he compares to inflicting a death of a thousand cuts on the disease rather than seeking one grand cure. He notes this idea grew directly out of Sid from GitLab treating his own cancer like a startup, with the very people who helped Sid now building a company to do the same for everyone else.
Frighteningly Ambitious Ideas Realized 4:01
Referring to his 2012 essay on frighteningly ambitious startup ideas, Graham says building a new Google was one of them, and that it finally happened not through direct attack but because the underlying model became obsolete. He credits OpenAI, incubated at YC as one of Sam Altman's side projects, with replacing his own habit of searching Google with simply asking an AI for information.
Why Founders Need Ambition 5:02
Graham explains he already knew before starting YC, from his own startup experience, that the obstacles founders face are too fierce for mere talent alone, so something has to drive a person through them, and that something is ambition. He adds that day to day, founders are actually driven less by dreams of wealth and more by fear of failure, like a server crashing, and that many are genuinely surprised when they realize, often only when someone does the math for them, that they have become billionaires.
Formidable Founders and Hidden Ambition 7:00
Graham says cases of founders becoming more ambitious during a batch are rare because the quality is mostly inborn, citing Sam Altman as someone who was obviously formidable before YC even accepted him. Where ambition seems to be missing, he thinks it is usually just suppressed by upbringing, such as pushy parents demanding obedience, rather than actually absent. He also warns that treating a startup as a resume credential badly misjudges what it actually takes, since a failed startup is no badge at all and a successful one becomes your life's work rather than a line item.
What Formidable Really Means 9:31
Graham traces the word formidable to private language he and his wife Jessica used before YC existed, describing someone who reliably gets what they want. He explains that investors want formidable founders because their interests align, since if the founder gets what they want, the investor gets what they want too.
Lean Startups and Getting Cheaper 10:31
Responding to Patrick Collison's claim that lean startups may be dead, Graham says he has never fully understood the term but doubts it changes anything, since technology only gets cheaper and inference prices keep falling roughly thirtyfold a year despite current token shortages. He points to a rocket company called StarCloud that raised money on a white paper and a booked launch as proof you can still start big things with little money if you reach convincing milestones.
Fumbling Toward a Human AI 13:01
Graham recalls studying AI in the 1980s and expecting progress to move from a perfect fly's brain up through mice, cats, and monkeys toward humans. Instead, AI arrived as an imperfect full human, comparing early ChatGPT to a bullshitting undergraduate producing plausible but flawed answers, which nobody had predicted.
Defining AGI as a Smear 14:30
Graham says he had to look up the Turing test definition recently, which he takes as evidence AGI is close. He describes AGI not as a clean finish line but as a smear with width, where some capabilities are far advanced and others, like knowing restaurant hours, lag behind in what he calls the jagged frontier.
Shipping Speed Still Matters 16:01
Graham says the best predictor of startup success remains how fast a team ships, and AI tools have not changed that, since founders still need to come up with ideas worth shipping. The only new wrinkle is that AI bills, sometimes tens of thousands of dollars a day in tokens, have joined salaries as a major cost that startups did not used to face.
Advantages of Being in a Batch 17:01
Graham lists why batch participation helps founders, including escaping the usual loneliness of startup life by having colleagues, learning from others who have already solved the same technical problems, and what he calls the YC GDP, meaning the built in customer base of fellow startups willing to try your product as early adopters.
How YC Began by Accident 18:00
Graham explains YC started as an attempt to be an angel firm using standardized paperwork to invest small amounts early, unlike the large late stage rounds of traditional venture capital. The idea of funding many startups at once as a summer alternative to jobs at places like Microsoft was meant to teach them how to invest, and the batch format that resulted turned out to work so well by accident that it became permanent.
Scale Without Real Change 19:31
Graham argues YC has changed very little despite its growth, since the startups' problems stay the same and the organization functions as many small groups, similar in feel to the 2012 batches that had only seventy startups. He notes that people accused YC of having jumped the shark even back then, when it was far smaller.
Where the Next Giant Company Comes From 20:01
Asked where the next trillion dollar company will come from, Graham answers simply that it comes from the right founders, who he expects will look much like founders have for the past twenty years. He jokes about founders someday being robots before affirming that the pattern of formidable people building great companies has not changed and has no reason to change.
AI-generated summary. It can be wrong or incomplete - check anything that matters against the original.
