Paul Graham Just Published 'How to Earn a Billion Dollars.' Here's the One Number It Says Your YC F26 Application Must Show

Paul Graham published a new essay this week, How to Earn a Billion Dollars, based on a talk he gave at the Oxford Union. It hit the front page of Hacker News with more than 500 points within hours. If you're writing a YC Fall 2026 (F26) application right now, it is the most useful thing PG has published in months, because it tells you, in his own words, the first question he asks every founder he meets.

This post pulls out what actually matters for your application, and what to do about it before you hit submit.

What is Paul Graham's 'How to Earn a Billion Dollars' essay about?

The essay is a response to a politician who claimed it's impossible to earn a billion dollars without cheating. PG's counter is pure arithmetic. A startup's outcome comes down to just two numbers: its growth rate, and how long that growth continues. Get both, and a billion-dollar outcome isn't a moral failing, it's compounding.

His examples are worth memorizing because they are the numbers YC thinks in:

• A founder he'd funded was growing 93% month-over-month. Starting from $2M in value, it takes only 9.45 months of that growth to reach $1B (he literally walks the Oxford audience through typing log(500, 1.93) into Google).
• A more conservative 15% per month, which he says he "constantly" encounters, compounds to 4,384x over five years (1.15^60). A company making $10K/month becomes one making ~$526M/year.

The punchline for applicants: "Now you see why, when I meet a founder, the first thing I ask about is their growth rate."

What growth rate does YC actually want to see?

YC has said for years that a good weekly growth rate is 5-7%, and a great one is 10%. PG's new essay reframes the same idea monthly: 15% month-over-month is the floor he treats as genuinely promising, and he doesn't consider it rare. If you are pre-launch or pre-revenue, the number can be users, signups, waitlist conversions, or activated accounts, but it has to be something real that is compounding.

Here's the mistake F26 applicants make. They write "we're growing fast" or "we 3x'd." YC partners read thousands of applications and those phrases are noise. PG's essay tells you exactly what registers instead: a clean monthly percentage, measured against a defined denominator, over a defined period. "We went from 40 to 62 weekly active users over the last four weeks, ~15%/week" is a sentence that makes a partner lean in. "Strong early traction" is a sentence they skim past.

The second number: are you in a market that can grow 4,000x?

PG's second variable is how long growth can continue, which is just market size in disguise. To grow 4,000x, there has to be at least 4,000x more demand than you're capturing today. This is the part of the YC application most founders underweight.

The application asks why your market is big, or will be big. PG's framing gives you the answer structure: you don't need the market to be big today, you need a credible path where the thing you and your friends already want becomes the thing everyone wants in ten years. He points out that the way you legitimately expand that second number is by predicting future demand, then expanding into adjacent markets from a beachhead. So your market answer should name the beachhead and the first adjacent expansion, not just cite a $50B TAM slide.

How do you get an idea that grows like this?

The most quotable part of the essay is also the most relevant to founders who feel like they don't have a defensible idea yet. PG's claim: the best startup ideas don't come from looking for startup ideas. They come from building projects with your friends that you think would be cool, because your own needs predict future demand.

His evidence is the YC canon, and you should use these same names in your own thinking:

• Justin.tv was one guy, Justin Kan, walking around with a camera strapped to his head, livestreaming his life. It sounded ridiculous. It became Twitch.
• Airbnb sounded so bad that YC funded it almost in spite of the idea, on the strength of the founders. "Who's going to pay to sleep on an airbed on someone's floor?"
• Facebook, Apple, and Google all started as projects, not companies.

If your F26 idea sounds a little lame when you describe it at a party, that's not disqualifying. What matters is whether you and a few people like you genuinely use it and tell friends about it. That telling-friends loop is the source of the 15% number. Growth isn't a marketing tactic you bolt on later; it's the signal that you built the right thing.

What to fix in your F26 application this week

Read through your draft with PG's two numbers in hand and check three things.

First, find your growth claim. If it isn't a specific monthly or weekly percentage with a clear denominator and time window, rewrite it until it is. Vague traction language is the single most common reason a promising application reads as weak.

Second, make the telling-friends loop explicit. Show one concrete instance of a user pulling in another user organically. That is the closest thing to proof that your growth rate is real and not bought.

Third, name your beachhead and your first expansion. Don't claim a huge market; claim a small one you dominate and the adjacent one you grow into. That's how PG describes the second number actually working.

Get a second read before you submit

The hard part is that you can't see your own application the way a YC partner will. You wrote it, so the growth claim feels obviously strong to you and the idea feels obviously good. A founder who has actually been through YC reads it differently, and will tell you in thirty seconds whether your growth number lands or reads as fluff.

That's exactly what YC Roaster is for: it connects F26 applicants with YC alumni who give brutally honest feedback on your application before you submit. If PG's essay convinced you that your growth rate is the number that matters, the next step is making sure someone who has sat on the other side of that question agrees it's stated clearly enough to register.

The two numbers haven't changed in 21 years and 6,500 funded companies. The applicants who get in are the ones who can show them.