Paul Graham On Startups, Ambition, and Great Founders
In a Nutshell
Paul Graham's core message is that startup success hinges on ambitious, formidable founders who relentlessly ship and overcome daily crises, not on credentials, prestige, or even the specific idea. Ambition is mostly inborn and essential because the day-to-day grind is driven by fear of failure, not the distant goal of wealth. The next trillion-dollar company will come from these formidable founders, and while AI has shifted costs from salaries to GPUs, the fundamentals of building startups remain unchanged.
These notes were generated by AI and may contain inaccuracies.
The conversation takes place at the original Y Combinator offices in Mountain View. This is the 47th YC batch and the 21st year of YC. Paul Graham creates the talk from scratch every time, though he tends to repeat the same core ideas.
Most aspects of starting a startup remain fundamentally the same regardless of the industry, whether it's microprocessors, AI, or internal combustion engines. The core challenges and processes don't fundamentally change.
Paul Graham recently noted that startups YC funds now are much more serious than in the "supposedly good old days." People have been claiming YC has "jumped the shark" since around 2008, arguing it was once great but has fallen. The old days included companies like Reddit, which Graham describes as not exactly "intercontinental ballistic cargo."
Current YC batch includes highly ambitious projects like intercontinental ballistic cargo systems and multiple approaches to curing cancer. One cancer-focused startup takes an "on-demand research" approach for people with cancer, inflicting "the death of a thousand cuts" on the disease rather than seeking a single cure. This startup helped Sid from GitLab deal with his cancer using a startup-like approach. Sid's experience inspired the concept of doing for everyone what he did for himself.
Graham wrote an essay in 2012 about "frighteningly ambitious" ideas, many of which have since been realized. OpenAI represents the realization of creating "a new Google" - not by attacking Google head-on, but by waiting for their model to become obsolete. The shift from wanting web pages to wanting information directly from AI made Google search less necessary.
Graham knew before starting YC that founders needed to be really ambitious. Starting a startup is grueling, and "mere beautifulness" isn't enough to overcome the obstacles. Ambition provides the necessary drive.
While becoming a billionaire motivates founders in theory, day-to-day motivation comes from the fear of failure - disaster, looking like a fool, server crashes. Founders focus on immediate problems like "the engine of my model train set is falling off the edge of the table" rather than long-term wealth. Billionaire status often takes founders by surprise when they finally do the math on their shares.
Stories of founders becoming ambitious during YC are rare because ambition is largely inborn. Sam Altman was "extremely formidable" the first time Graham met him. Sometimes founders have been "trained not to show" their ambition due to environments emphasizing obedience, but the underlying ambition was always present.
Some people view startups as resume credentials, but this misunderstands their nature. If a startup fails, it's "not much of a badge." Unlike Harvard where you can choose easy majors, startups have "no easy major" - it's like being forced to study theoretical physics. People applying to YC for prestige don't realize what they're asking for.
The word "formidable" comes from Maya and Jessica's private language at YC. It describes someone who "gets what they want" in any situation. This quality aligns investor and founder interests because if founders get what they want, investors get what they want through their equity stake.
Graham isn't clear on what "lean startup" specifically refers to, having never read the book. He argues that starting with little money remains viable because technology always gets cheaper. Token prices will decrease dramatically as GPU shortages resolve. Even rocket startups can begin with minimal funding by creating designs and simulations, then raising more money based on milestones achieved.
Graham studied AI in the 1980s expecting a different trajectory: starting with perfect simple systems (like a fly's brain) and working up to human-level intelligence. Instead, AI development went in the opposite direction - starting with flawed human-like systems ("a bullshitting undergraduate") and working toward perfection. The first ChatGPT versions were "full human but terrible."
Graham suggests the Turing test remains a reasonable definition of AGI. The concept of AGI has proven to be a "smear" rather than a clear finish line - some AI capabilities are way across it, others are in the middle. We're currently "on the smear."
The best predictor of startup success remains the pace at which they ship new stuff, even with AI tools. Many YC batch companies still aren't shipping fast enough. AI has introduced a new major cost: GPU and token expenses that can reach tens of thousands of dollars daily, replacing salaries as the primary startup expense.
YC provides colleagues in an otherwise lonely endeavor, allows learning from other startups' solutions to similar problems, and offers the "YC GDP" - the ability to sell products to other batch companies who serve as ideal early adopters.
YC began as an angel firm concept - standardized early-stage investing with smaller amounts than traditional VCs. The batch model emerged from wanting to learn investing by funding multiple startups simultaneously. The summer timing was designed as an alternative to college students getting summer jobs at places like Microsoft. The batch concept proved so effective that it became the permanent model.
Despite growth, YC's core remains unchanged - the same types of problems, just more of them. The batch structure maintains intimacy through pods of around 70 startups, similar to 2012 batch sizes. Claims that YC has become too large have persisted since the 40-startup era.
The next trillion-dollar company will come from formidable founders, not from any particular idea. These founders will have good ideas, but their formidability is the key factor. Graham doesn't expect founder characteristics to change significantly in the future, based on 20 years of consistent data.
Keep Y Combinator in your library
Save the videos and channels worth coming back to, and find them again in one place.
