I spent weeks reading everything YC has ever published about what they look for. here are 9 things that would have changed how I thought about the application

45+ startups hiring in this thread: 1. @turbopuffer - remote, NYC, Sf 2. @convex - SF 3. @fal - SF / remote 4. @firecrawl - SF / remote 5. @attio - SF, NYC, London + remote 6. @Linear - North America & EU remote 7. @warpdotco - NYC 8. @speak - SF / Seoul / Tokyo 9. @BrightHarborCo - Austin 10. @Vizcom - SF & US remote 11. @PostHog - NA + EMEA 12. @sievedata - SF 13. @artie_labs - SF & remote 14. @concurrencehq - NYC / SF 15. @SpaceXAI - global 16. @EnclaveAI - NYC, SF, TLV 17. @tweetsbyport - US, TLV 18. @RevelHQ - LA / SF / NYC 19. @SnorkelAI - SF & NYC 20. @neatlogs - remote 21. @BabaHQ - NYC 22. Tremendous - remote (Americas) 23. @nucleussec - Florida / fully remote 24. Searchable - London / Salt Lake City 25. @flytbase - Pune, India 26. Indices - London / SF 27. Duna - across Europe 28. @aiunderwriting - San Francisco 29. @zenml_io - SF 30. @nexcade_ai - London 31. Social Fetch - remote 32. @gumloop - SF + Vancouver 33. @LulaConvenience - US remote 34. CLAR AI - Munich 35. @valkaicom - NYC / SF 36. @KeycardAI - SF 37. Latent Defense - NY 38. @DataUAcademy - Southeast Asia + global 39. @Nooqbook - remote 40. @saasflashstudio - remote 41. @SkydioHQ - US 42. @NotionHQ - global 43. @Replit - global 44. @forus - NYC 45. @harmonic_ai - NYC

YC PMF timeline data Founders talking to users 10+ times per week: → Find PMF on average in 6 months Founders talking to users 2 to 4 times per week: → Find PMF in 12 to 18 months Founders talking to users fewer than twice per week: → Often pivot multiple times or never find it User conversation frequency is the most predictive variable for PMF timeline, not market size or team credentials.

How to Talk to Customers YC says “talk to your users” so often it starts to sound like a slogan. It isn’t. Most startups die because the founders built something in their head and then went looking for people who would politely agree with them. Talking to customers is not pitching, It’s trying to find out how someone already deals with a problem when you’re not in the room. If you leave a call thinking “they loved the idea,” you probably ran the conversation wrong. What you’re actually trying to learn You want the messy version of their life. How they do the thing today. What sucks about it. How often it happens. What they already pay for. What they ignore. What they complain about to coworkers but never put in a survey. Paul Graham’s test is a good one: can you say what you learned? Not “users are excited.” What was wrong about your first picture of them? If you can’t answer that, you haven’t talked to enough people. Or you talked, but you were too busy selling. Who to talk to Talk to people who have the problem. Not your most encouraging friends, unless those friends actually live with the problem. Start with people you already know. Former coworkers. People one intro away. Then go where they hang out: LinkedIn, Reddit, Slack, Discord, industry groups, events. Warm intros get replies. Strangers are often more honest. Do the call live. Video, phone, in person. A five-minute conversation beats 2,000 survey answers because you can hear when someone is guessing, being nice, or describing something that actually happened last Tuesday. How not to blow the call Build a little rapport first. Then shut up about your product. This is the part founders hate. They want to explain the idea so the other person “gets it.” The second you explain it, the interview is contaminated. People start answering the version of reality that makes you feel good. Ask about their world. Let them talk. Take notes. Record if they’re okay with it. You’re trying to catch details you’ll forget ten minutes later. Ask about what they did, not what they would do Good questions sound almost boring: - How do you do X today? - What’s the hardest part? - Why is that hard? - How often does this happen? - Tell me about the last time it happened. - What did you do then? - What do you use now? - What happens if you don’t deal with it? Then keep poking: - What do you mean? - Can you say more about that? - Why does that matter? The gold is in the follow-up. First answers are usually clean and generic. The real stuff shows up when you make them slow down and tell the story. Don’t ask these - Would you use this? - Would you pay for this? - What features should we add? - What would the perfect product look like? - Anything that can be answered yes or no People are polite. They will invent a more disciplined future self who switches tools, pays annually, and onboards the whole team next month. That person does not exist. You want evidence from last week, not a speech about next year. How to tell if the conversation was real Look for - a specific recent example - an ugly workaround they already use - money, time, or internal pain attached to it - the same complaint from different people, in almost the same words - them getting sharper, not vaguer, as they talk One excited friend is not a market. Five unconnected people describing the same broken workflow starts to be something. If you later show a prototype, watch their face and their hands. Confusion is more useful than praise. After the calls Don’t jump straight to a feature list. Write down the boring synthesis: - who actually has this problem - what they’re trying to get done - why the current way sucks - what they already do about it - why they would bother switching This week, talk to 5 people who actually have the problem and write one sentence after each call: what surprised you. If you can’t do that by Friday, you’re still building in your head.

The YC application is a mirror. It shows you exactly where your company is unclear, thin, and missing evidence. Use it as a diagnostic. Then fix what it reveals

The market will tell you everything. It just won't speak until you charge

Don’t apply to YC if you hate • $500K so you can build instead of hunt cash • partners who’ve seen your exact problem a thousand times • a batch of founders moving as fast as you • three months where shipping is the only plan • Demo Day instead of six months of cold emails • investors already looking for companies like yours • the first customers and hires actually replying • answers at midnight from people who just did this • help closing the round when every investor sounds the same • a network that keeps working after the batch ends You can skip all of that Just don’t pretend it isn’t a lot to skip.

Stripe growth data Stripe transactions processed in year 1 (2012): → $1M Stripe payment volume in 2022: → $817 billion Stripe valuation as of September 2026: → $159B From $1M to $817B in transaction volume in 10 years. 817,000x growth. One idea, done exceptionally well.

If Muse is the new default agent for people, Meta Enterprise Platform is Meta trying to become the default agent layer for companies.

Agents are useful only if people trust them. NVIDIA just released an open way to keep agents inside safe limits. Its time to "Make something people want an agent to do"

Steal this YC habit, charge early so the conversation gets serious • Price on the table • Objection in the open • Value named in dollars Free compliments are abundant. Budgets to run the startup are not.

The startup that ships 10 bad features and learns from each one: builds a better product than the startup that ships 1 perfect feature

Revenue is a lagging indicator. Retention is a leading indicator. Track the leading indicator. Fix it before the lagging one breaks

YC tells founders to make something people want then prove it with growth. • Revenue if you charge • Active users if you don’t • 5–7% a week during YC If that number moves, you’re not guessing

Rejected by YC? If you think a first rejection means you are not a real founder, you have the story backwards. Look at the last 20 years. How big this got - 2005: 8 or 9 companies. Whole batch. - Now: about 5,600–5,700 companies funded. - Alumni value often cited at $600B+. 100+ unicorns. - Peak year: 2021. - Biggest batch: Winter 2022, around 400 companies. - Now: 4 batches a year. Still 600–800 companies annually. - Applications: tens of thousands a cycle. - Acceptance: around 1%. Some batches closer to 0.6%. Garry Tan has said it is drifting under 1%. Who gets in - A lot of accepted teams were still an idea. - YC has said roughly 40% came in that early. - Only about 7% of recent batches were already at $50k+ MRR. - Most nos never become an interview. Reapplying is common - About 30–40% of accepted founders had already been rejected. - Some people in those rooms say it felt closer to half. - Airbnb got a no first. - Dropbox almost died as a solo app until PG told Drew to find a cofounder. - Replit took four rejections. - InsForge got in on the sixth try. YC remembers the last version of you. They watch whether the company moved. Judge the no - First no, vague pitch, thin proof: expected. Go build. - Repeat no, same numbers: they already answered. Don’t send the same form. - Repeat no, but users, revenue, a live product, a stronger cofounder: that’s the pattern that gets in. - Dont think, You don’t need YC. Use the deadline. Don’t make it your identity. If you apply again, change the company - Who hurts - What you shipped - What number moved - A 60-second demo of a real person doing the thing - Why this team doesn’t quit Test: can a partner repeat your progress in one sentence? Almost everyone gets rejected. The ones who get in later treat the no like a date, then come back with a steeper line. Stay in the game long enough that the next application is a different company.

YC acquisition data YC companies eventually acquired across all batches → 12.86% YC companies from mature batches (10+ years old) eventually acquired → 25% to 50% Notable YC acquisitions: → Twitch: $970M (Amazon), Segment: $3.2B (Twilio) Acquisition is not the consolation prize. For 12.86% of all YC companies it was the intended outcome.

YC Demo Day raise data Typical post-Demo Day seed round size: → $1.5M to $5M Median seed round for YC companies in 2025: → $3.1M Top decile seed rounds at Demo Day: → Cleared $10.5M The MRR you walk in with determines which of those tiers you land in.

I pulled the fastest growing startups on X by follower growth over last 90 days:

YC tells founders to build for a small desperate group, not for “everyone.” • 100 people who need it • Not 1 million who kind of like it • Depth first, width later Mild interest doesn’t make a company.