> ## Content Index
> Fetch the complete content index at: https://www.signalful.com/llms.txt
> Use this file to discover other available public pages before exploring further.

# Most of Starting a Startup Is Always the Same
- URL: https://www.signalful.com/most-of-starting-a-startup-is-always-the-same/
- Published: 2026-10-02T15:49:00.000Z
- Updated: 2026-10-09T19:40:14.000Z
- Description: After 21 years of YC, Paul Graham says almost nothing about winning founders has changed. The one new expense is tokens.
- Author: Signalful Editorial
- Tags: Builders, Y Combinator, Paul Graham, Startups, Founders, #sf-0012, #yt-5bxp78i96S8, #Import 2026-10-06 15:50

*Paul Graham, cofounder of Y Combinator, interviewed by Vivian Shen (Y Combinator)*

**Summary:** Paul Graham argues that startups, and the people who win at them, have barely changed in 21 years of YC, even as AI changes the tools and the ideas get far more ambitious. The founder decides the outcome: ambitious from birth, driven day to day by fear of failure, and formidable, meaning they get what they want. Take that seriously and you bet on people and shipping speed. You also stop believing that AI, cheap capital, or YC's size has rewritten the rules.

1. **The Unchanging Startup.** Most of starting a startup is the same whether the technology is microprocessors, AI, or the internal combustion engine. Graham has given the YC batch talk 47 times, writes it from scratch each time, and suspects he says the same things anyway. Asked what AI has changed, he says "so far almost everything is exactly the same," and pace of shipping is still the best predictor of success. The one new thing is the cost structure: startups used to spend almost everything on salaries, and now some pay tens of thousands of dollars a day in tokens.
2. **Serious Startups.** Graham says the companies YC funds today are much more serious than the ones from the "supposedly good old days." People have claimed YC jumped the shark since about 2008, and to say it has fallen they have to admit it was once great. The old days produced Reddit, which he says is "not exactly intercontinental ballistic cargo" (a recent office-hours idea: an ICBM that lands and drops stuff off). The latest batch also includes a startup going after cancer.
3. **Death of a Thousand Cuts.** Graham wants YC to fund many approaches to curing cancer, because any one of them working would be good. Vaccines sit at one end; at the other is a startup in this batch doing on-demand research for individual patients. The same team helped Sid from GitLab go "founder mode" on his own cancer, acting as cofounders in what was in effect a startup. Graham had once thought someone should do for everyone what Sid did for himself, and now the people who did it for Sid are doing exactly that.
4. **The New Google.** Graham's 2012 essay "Frighteningly Ambitious Ideas" predicted that nobody would build a new Google by attacking it head-on. You have to wait until things change so much that Google's model is obsolete. OpenAI is that company: once Graham started using it, he stopped using Google, because what he wanted from search was information and the AI gives it to him directly. OpenAI was incubated at YC as one of Sam Altman's many side projects.
5. **Inborn Ambition.** Graham knew founders had to be very ambitious before YC started, because running his own startup had shown him how grueling the work is. Stories of founders who become ambitious during the batch are rare, since the quality is mostly inborn. Sam Altman was "extremely formidable" the first time Graham met him, and that is the rule. The exception is a young founder trained to be obedient, maybe by pushy parents, who has never let go; Graham says that person was ambitious all along and had learned to hide it.
6. **Fear Over Fortune.** Fear of failure drives founders from moment to moment far more than the hope of becoming a billionaire. When the server crashes, a founder thinks "oh no, this will be a disaster," and the billion dollars never comes into it. Graham compares it to saving your model train set as the engine slides off the edge of the table. Founders keep their heads down for 10 years, and sometimes Graham is the first to do the math and tell them that at their last round's valuation they are billionaires.
7. **The Worst Credential.** People started applying to YC for prestige early in its history, and Graham says they had no idea what they were asking for. Harvard has easy majors anyone can slide through and still collect the degree; a startup is like getting into Harvard and being forced to study theoretical physics. A startup that works becomes your life's work, so the only badge you could collect is a failed startup, and that isn't much of a badge. If you want to seem cool, starting a startup is about the least efficient route, with years of brutal work before anyone is impressed.
8. **Formidable Founders.** "Formidable" came from the private vocabulary Graham and Jessica shared before they started YC. The test is whether a person gets what they want in any situation, since you can't be very formidable if you don't. Investors want formidable founders because both sides own stock in the same company. When the founder gets what they want, the investor gets a share of it.
9. **Cheap Starts Still Work.** Graham hasn't read The Lean Startup ("Tolkien didn't read all those other fantasy books"), and he disagrees with Patrick Collison's suggestion that starting on little money is dead. Tokens are expensive right now because of a GPU shortage, and inference at any given quality level gets something like 30x cheaper each year. Even a rocket company can start small: design the rocket, simulate it, show experts, and raise enough for the next milestone. Philip from StarCloud raised money with a white paper and a booked launch, though Graham notes that a famous expert has credibility a new college graduate doesn't.
10. **Backwards AI.** When Graham studied AI in the 1980s, the plan was to build a perfect fly and work up through mice, cats, and monkeys to humans. What arrived was a full human that was terrible: early ChatGPT read like an undergrad bullshitting through a paper. From the 1980s, AGI looked like a thin finish line; standing on it, Graham finds it has width, and "we're on the smear." Some parts of AI are far across (maybe even the Riemann hypothesis), while Graham still can't get it to tell him when restaurants are open.
11. **The Accidental Batch.** YC started as an angel firm: invest small amounts early, with standardized paperwork. To learn how to invest, Graham and his partners funded a group of undergrads all at once, as an alternative to summer jobs at Microsoft. The investors and founders both became real very quickly. The batch gives founders colleagues in a lonely job, peers who already solved their technical problems, and the "YC GDP," a set of early adopters who decide fast and have to hear your pitch. At today's size, each startup sits in a pod of 70, the same as a whole batch in 2012, so YC is "basically like the same thing but more of it."
12. **Trillion-Dollar Founders.** The next trillion-dollar company will come from formidable founders, and Graham says that is the whole answer to where it comes from. It probably won't be dog walking, but startup ideas are very mutable, and the right founders will likely be working on something promising. YC now has 20 years of data, and founders look exactly like they did 20 years ago. Graham sees no reason they'd look different 20 years from now.

Watch the full video at [https://www.youtube.com/watch?v=5bxp78i96S8](https://www.youtube.com/watch?v=5bxp78i96S8&ref=signalful.com). Read the full transcript at [https://www.usetranscribe.io/yt/5bxp78i96S8/paul-graham-startups](https://www.usetranscribe.io/yt/5bxp78i96S8/paul-graham-startups?ref=signalful.com).