YC Application 2026: Get a W27 Interview by Nov 2

How to write a Winter 2027 YC application that earns the interview: every section, the one-minute founder video, the ten-minute call, and the thirteen days left before the deadline.

Submitted YC applications have grown 60% longer in three years, YC partner Tyler Bosmeny wrote on X in August, adding only "I wonder what could explain that..." A day earlier his colleague Pete Koomen reported that the share of applications containing the word "wedge" went from under 1% in spring 2025 to more than 20% in summer 2026. Nobody at YC had to spell out the cause. Founders are writing their YC application with the same AI assistants, and the assistants all reach for the same words.

That is the most useful fact you can know about the Winter 2027 cycle, because it tells you what no longer works. When anyone can generate fluent, confident, well-structured prose for free, fluency carries zero signal. The partners reading your application have seen thousands of answers that sound exactly like yours will sound if you let a model write them. What still separates an application in October 2026 is the same thing that separated one in 2007: evidence that is expensive to fake. Specific users, a real product, numbers that hold up when checked, and proof that the founders build fast.

The deadline is real and close. YC's apply page says the on-time deadline for Winter 2027 is November 2 at 8pm PT, that founders who apply before it "will get a decision by December 11," and that most interviews happen by video in November and December. If you are reading this on the day it was published, you have thirteen days.

This guide walks through the whole process from first principles: the W27 calendar, how a partner actually reads your application, what the 2026 batches tell you about who gets in, each section of the form, the founder video, the progress you can still make before November 2, the legal and equity questions, the ten-minute interview, the special cases (solo, non-technical, international, students, reapplicants), and what to do if the answer is no. Every rule here is sourced to YC's own pages wherever YC has published one, and where YC has not, we say so.

Contents

1. The W27 Calendar: What Happens Between Now and March
2. How a YC Partner Reads Your Application
3. The Odds and the Batch: What the 2026 Numbers Say
4. The Founder Section: Proving You Build Fast
5. The Founder Video: One Minute, Nothing but Founders
6. The Company Section: The 50-Character Line and "What Are You Making?"
7. Progress: The Thirteen Days Before November 2
8. Idea, Market and Money: Answering Like an Insider
9. Structure, Equity and the $500,000 Deal
10. The Interview: Ten Minutes on Zoom
11. Solo, Non-Technical, International, Student and Repeat Applicants
12. How Applications Fail, and Plan B If Yours Does

The Winter 2027 Field at a Glance

YC is the program this guide is about, but serious founders rarely apply to only one, and several programs with real money are open in the same weeks. Before the detail, here is the field on one page for a founder applying this winter. It is not a ranking of which accelerator is "best" in general (our US accelerator ranking does that). It answers a narrower question: if you are applying in late October 2026, which programs are worth an application right now, and on what terms?

Each program is scored from 0 to 10 on four criteria, and the final score is the weighted average. Capital and terms (30%) compares the money against the equity it costs. Open this window (25%) rewards programs you can apply to now for a start in early 2027, since a great program with a closed cycle does not help you this month. Program depth (25%) reflects cohort size, partner access and what the program bundles. Selectivity (20%) uses published acceptance rates as a proxy for the signal an acceptance sends; where a program publishes none, the cell shows a dash and the criterion is left out of that program's average. Every term below was checked on the program's own page on October 20, 2026.

YC tops the table mainly because it is open now, deep, and highly selective; its terms score below Neo's because a fixed 7% plus an MFN SAFE costs more equity than an uncapped note with participation rights. speedrun is the closest substitute on timing and signal, but its 10% upfront is the most expensive equity in the table and its deadline falls a day before YC's, so a founder applying to both should finish the shared material first. Neo and PearX tie on score for different reasons: Neo offers the most founder-friendly money, PearX the most flexible timing. The practical reading is that most founders applying to YC this month should also consider one program from rows two to four, chosen by whether money, timing or cohort size matters most to them. We return to these options in section 12.

1. The W27 Calendar: What Happens Between Now and March

Most applicants treat the deadline as the finish line. It is closer to the starting gun. YC's own apply page lays out a sequence that runs for five months, and each step changes what you should be doing in the weeks before it. The application closes November 2 at 8pm PT. Interviews run "by video conference in November and December." YC says it typically makes "decisions the same day as your interview," and everyone who applied on time hears back by December 11. The batch itself runs "from January to March in San Francisco," in person at YC's campus, starting with a three-day in-person kick-off.

Two details on that page matter more than they look. The first is that YC "invest [s] in companies as soon as they are accepted; we do not wait for the batch to start." An acceptance in mid-November means money in November, not January, which is why the time between your application and your interview is productive time rather than waiting time. The second is that YC explicitly tells founders, "We encourage you to submit your application as soon as you're ready to apply." YC does not promise that early applications are read first or scored more kindly, and you should not believe anyone who claims to know otherwise. But the logic of the calendar rewards submitting as soon as the application is genuinely good, then spending the remaining days making progress you can describe in the interview.

The calendar now repeats four times a year. YC moved to four batches per year (Winter, Spring, Summer and Fall) in 2025, and Dalton Caldwell's announcement of the first Spring batch gave two reasons: "with more frequent batches it's more likely the timing works for founders applying to YC," and "the rate of change currently happening as a result of AI is causing more people to create startups." The practical consequence is that a rejection is no longer a six-month wait. The 2026 Demo Day schedule shows the rhythm: March 24, June 16, September 10 and December 2. YC has not yet published the Winter 2027 Demo Day date, so treat "March 2027" as the batch window rather than a confirmed day.

The batch is physically in San Francisco, and that is not a formality. YC's FAQ says it "briefly did run YC remotely during Covid, but since 2022 it has been back in person," and that "YC works much better in person." After the three months, "you can go wherever you want." If you cannot be in San Francisco from January through March 2027, you should know that now, not after an acceptance.

The photo above is from the kick-off of the Summer 2025 batch, the same three-day in-person start that W27 will open with in January. It is a useful reminder of what you are applying for: not a cheque and a logo, but three months of working next to a couple of hundred other founders and the partners who picked them.

Why this matters: the calendar defines your real deadline, which is not November 2 but the day of your interview, because that is the last moment new progress can change the decision. How to apply it: submit as soon as the application is clear and complete, then treat every day until the interview as a sprint whose output you will describe on the call. If you are a student who cannot start in January, YC's Early Decision page lets you apply now for a later batch by selecting "A batch after Winter 2027" on the form.

2. How a YC Partner Reads Your Application

Before writing a single answer, it helps to picture the reader. Paul Graham's essay on how to apply, still hosted on YC's site and updated to open with "Four times a year Y Combinator takes applications," describes the arithmetic plainly: "If we get 1000 applications and have 10 days to read them, we have to read about 100 a day. That means a YC partner who reads your application will on average have already read 50 that day and have 50 more to go." That was written when YC received about a thousand applications a cycle.

YC's admissions team wrote in 2021 that the volume had already reached "the tens of thousands of applications we receive each year," and that review is "mostly, it's lots of people reading lots of applications! There's no secret machine learning model."

So the structural problem is attention. A human with limited time is trying to estimate one thing, the chance that this team builds something very large, from a few hundred words and a one-minute video. Every sentence you write either moves that estimate or wastes the reader's time. That framing explains nearly every piece of advice YC has ever given, and it is more useful than memorizing the advice, because it tells you what to do in the cases the advice does not cover.

From that starting point, three things follow. First, clarity is not style, it is throughput: an answer the reader has to decode costs attention that could have gone to your evidence. Graham's test is the sharpest version: "ask how close the reader is to reproducing it." After a vague sentence like "We are going to transform the relationship between individuals and information," he writes, "I'm no closer than I was before, so its content is effectively zero."

Second, rate of change beats level. A partner cannot verify that your product is good, but can see that you went from nothing to a working product and twelve users in six weeks, and speed is the trait that compounds over the next ten years.

Third, costly signals beat cheap ones. Anything that is free to produce (adjectives, market-size claims, polished prose) tells the reader nothing, while things that are expensive to produce (a shipped product, paying users, a hard technical result) tell them a lot.

Dalton Caldwell, who headed YC's admissions, gave the reader's actual checklist in his 2023 Startup School talk. Condensed, the questions in a reader's head are:

• Who are the founders, how did they come together, and who does what?
• What exactly exists today: "Is it built? Are there users?"
• Is there evidence people want this: customers, users, founders using their own product?
• Are the founders serious: quit a job, reapplied with more progress each time?
• What is impressive or unique about this application?

Read that list as a specification for your application, not as tips. Each item is a question the reader will try to answer whether or not you help, so the application that answers each one directly and early wins on attention alone. Caldwell's summary of the outcome is worth keeping on a sticky note: "If I can tell this kind of story about your company by reading your application, it means it was a good application. On the flip side, weak applications obfuscate everything." He is equally blunt about the opposite of clarity. "Our entire system rewards honest people acting in good faith," he says, and misrepresenting revenue, such as making "monthly revenue" look like annual, "is like automatic disqualification."

The diagram is our model, not a YC document, but every branch maps to a question YC has published. It is useful mainly as a test: for each section you write, ask which of the three branches it strengthens. An answer that strengthens none of them is a candidate for deletion, which is exactly what Graham recommends when he tells applicants to "print it out and take a red pen and cross out every word you don't need."

There is one more structural point, and it explains why good applications are rejected. YC's own page on why it does not explain rejections says: "paradoxical as it sounds, there often is no reason. The median application is usually pretty good. The reason it gets rejected is not that it seems particularly bad, but that there are a sufficient number of others that seem particularly good."

Graham's estimate is harsher still: "of groups good enough to make it to interviews, more than half blow the application." Why this matters: you are not graded against a bar, you are ranked against thousands of other good applications, so the goal is not to avoid mistakes but to be legibly excellent in the first few lines. How to apply it: for every answer, put the single most impressive, most specific fact in the first sentence, and cut everything that does not support it.

3. The Odds and the Batch: What the 2026 Numbers Say

Acceptance odds at YC have been under 1% for years. The last time YC published both numbers in one place was its post introducing the Winter 2024 batch: "260 companies joined YC for W24, as selected from over 27,000 applications. With an acceptance rate under 1%, this was one of the most selective cohorts in YC history." YC has not published application counts for the 2025 or 2026 batches, so any precise 2026 acceptance rate you see online is an estimate. What we can measure is the other side of the ratio: how many companies YC funds, and what kind.

To do that, we counted every company in YC's public startup directory, using the open YC companies dataset that mirrors the directory's search index daily (the snapshot we used is dated October 20, 2026). One caveat up front: the directory lists only companies with public pages, so its batch counts run slightly below YC's official ones. Winter 2024 shows 248 listed companies against the 260 YC announced, so treat our numbers as a floor.

Two things stand out. YC is funding more companies per year than at any point since 2022: 770 listed companies across the four 2026 batches (197 in Winter, 192 in Spring, 230 in Summer and 151 in Fall so far), against 494 in 2023. At the same time, each individual batch is far smaller than the giant 2022 cohorts, because the same annual volume is now split four ways. For an applicant, both facts cut in your favour. More total slots means more chances per year, and smaller batches mean each company gets a larger share of partner attention once inside. The counter-argument is real too: more YC companies per year means the YC name alone carries a little less weight with later investors than it did when there were two batches of a hundred. That is a reason to treat YC as an accelerant for a company that already moves fast, not as a substitute for one.

The second question is what kind of company gets in. YC's FAQ gives the official spread: "On average, 40% of the companies we fund in each batch are just an idea. Most don't have any revenue." At the other end, "7% of recent batches had more than $50k in monthly revenue when accepted," and "each YC batch has many companies who have already raised over $1M." The batch is not a single profile. It is a mix of pre-product teams with strong founders and companies that are already working.

What has shifted sharply in 2026 is the industry mix. Using the industry label YC assigns in its directory, we measured the share of each batch tagged as B2B software against the share tagged Industrials (manufacturing, robotics, defense, energy, logistics and similar physical-world categories).

Industrials went from roughly one company in ten through 2025 to nearly one in four in Summer and Fall 2026, while B2B software fell from about two thirds to about half. That is not noise. It matches the first line of YC's current Requests for Startups: "AI is moving into the physical world." The same page lists requests ranging from an adaptive AI tutor for young children to defense hardware requested by the sitting Secretary of the Army, and it is explicit that the list is "just a fraction of what we fund" and that "you don't need to work on these ideas to apply." We decoded every current request in our guide to what YC is funding in 2026.

YC's partners confirmed the same shift from the inside in September. In the Lightcone episode "The State of Startups in 2026", Diana Hu said the share of hard tech companies in the batch "has gone from eight percent to twenty percent," and that solo-founder companies went from "about five percent" of acceptances to "over eighteen, nineteen percent." Garry Tan summed it up as "almost one fifth of the batch." The partners also described how much faster companies now move once inside: "the median YC company when it gets accepted is at zero in revenue," Hu said, but by the end of the batch the median used to reach about $8,000 in monthly revenue and "now the companies in median are getting to 20,000." And the share of accepted companies that do "full stack end to end work" on a task, where "the agent does the job," has "gone from just 10% to over 25% of the batch."

The episode is the most current official view of who YC is accepting, and two of its claims change how you should write. The first is that YC is funding more deep expertise: Jared Friedman noted that "one in six of the founders" in the Summer 2026 batch "actually has a PhD." The second is the theme in the episode's own description, that "knowing what to build is becoming more important than simply knowing how to build it." That is the cheap-building argument again, from YC's side of the table, and it is the reason the idea and progress sections below matter as much as the founder section.

AI remains close to universal. Searching each listed company's own description for "AI," "LLM" or "agent," we found that 70% to 83% of every batch since Winter 2025 describes itself in AI terms, and the share mentioning agents rose from about a third in early 2025 to between 34% and 48% across the 2026 batches. A simple word count is a rough instrument, but the conclusion is robust: being "an AI company" differentiates you from nobody in the pile. The early Winter 2027 companies already listed in the directory (five as of October 20, including Rote, an "AI-native insurance department for auto body shops," and LinkLane, an "AI-powered freight brokerage") follow the same pattern of AI applied to a specific, unglamorous industry.

Why this matters: the numbers tell you where the competition is thickest (horizontal B2B AI tools) and where YC is actively leaning (physical-world, regulated and vertical companies), without telling you to change your idea. How to apply it: if you are building an AI product, assume every reader has seen ten like it this week, and make your first sentence about the specific customer and the specific job, not the technology. If you are building something physical or deeply vertical, you are swimming with the current, and your domain experience belongs at the top of the application.

4. The Founder Section: Proving You Build Fast

YC has said for twenty years that it bets on founders before ideas, and the reasoning is structural rather than sentimental. Graham puts it simply in How to Apply: "I care more about the founders than the idea, because most of the startups we fund will change their idea significantly." If the idea is likely to change, the only durable asset in the application is the team, so the founder section is where the reader spends real attention. It is also the section where AI-polished prose helps you least, because the questions ask for facts about your past that either exist or do not.

A note on wording before going further. The live W27 form sits behind a YC login, and YC does not publish the full question list on a public page, so the exact phrasing may differ from what you remember or what a template site shows. The questions below are drawn from YC's own pages and from the published wording of recent forms, and the substance has been stable for years even as the phrasing shifts. The founder questions have long asked who writes the code (and whether any was done by a non-founder), how long the founders have known each other and how they met, a project two or more of you built together, something impressive each founder has built or achieved, and, in the classic version, "the time you most successfully hacked some (non-computer) system to your advantage."

The "impressive thing" question

Graham calls the impressive-achievement question "the most important question on the application," and his guidance on it is precise. "It's not the type of achievement that matters so much as the magnitude." And: "You probably shouldn't list the startup itself as your most impressive achievement. We already know you've created that." The question is a proxy for a trait, the ability to make something happen that other people would not have, and a reader can only assess it if you give them something checkable. "Built a mobile app" is a category. "Built a mobile app that reached 40,000 downloads in a month with no ad spend" is a magnitude.

Write one or two sentences per founder, lead with the number or the outcome, and pick the thing that is most unusual rather than the thing that is most prestigious. A rank in a national olympiad, an open-source project with real users, a business you ran as a teenager, a research result, a sale nobody thought you could close: each works if it is specific and if it shows initiative. A job title at a famous company works less well on its own, because it tells the reader someone else selected you, not what you did.

Technical talent is the single biggest lever

Dalton Caldwell was unusually explicit about this in his 2023 talk on applying: "Probably the biggest variable for if someone is selected for interview is the quality and quantity of the technical talent on the founding team." Teams where at least one founder "has a skill set level to be hired into a technical role at a top YC company" have, in his words, "five times better odds than teams that don't." And at the other end: "teams where there aren't technical founders and they're working on an unlaunched tar pit idea, those folks have the lowest odds. That's a fact."

The reason is not prejudice against business founders. It is that YC's model depends on a team shipping and iterating during a three-month batch without hiring first, and YC's FAQ says so directly: "It's important for the founding team to have the skills to build their product themselves, rather than outsourcing it to someone else." The practical consequence for your application is to make the builder visible. Name who writes the code, state plainly that it was not outsourced, and point to the strongest technical evidence each builder has.

New in 2026: your coding session and the Paxel report

The biggest change to the founder section in years arrived quietly in February. On February 5, 2026, YC announced on X: "We want to see how you build with AI. To better capture that, we've added a new application question for the Spring 2026 batch," letting applicants "upload a markdown file or transcript from Claude Code (or your favorite coding tool). We want to see how you plan, design, debug, and ship your most important features." Applicants quoted the field on Hacker News as "Attach a coding agent session you're particularly proud of," and a developer-tools vendor that documented the export process describes it as an optional .md or .txt file of up to 10MB. Those format details come from a third party rather than YC, so check them on the form itself.

YC then added a second, separate signal. Its FAQ now asks "Do I need a Paxel builder report to apply?" and answers: "No, it's completely optional. Paxel analyzes your real coding sessions and produces a builder report that gives us extra signal about how you build. If you link a Paxel account to your YC account and opt in on your application, we'll consider your report alongside everything else; if you don't, your application is evaluated exactly as before." Paxel describes itself as "an experiment from YC" that reads local transcripts from tools including Claude Code, Codex CLI, Cursor, opencode and Gemini CLI, runs its analysis in Docker on your own machine in about 15 to 30 minutes, and scores you across "steering, execution, engineering, product instinct, and planning." Its page says your working tree, .env files and raw transcripts stay local, that short excerpts go to Claude or GPT for summarizing, and that YC receives "a small payload of scores, narratives, and redacted decisions."

Should you use either? Think about what they measure. A curated session is a demo of judgment: it shows how you break down a problem, catch the model's mistakes, and decide what to ship. A Paxel report is a sample of your habits across many sessions, which is harder to curate and therefore a more honest signal. If you build with coding agents daily, both are cheap ways to put a costly signal (months of real work) in front of the reader. If you rarely do, neither will help, and the FAQ is explicit that skipping them leaves your application "evaluated exactly as before."

• Pick a session with a real decision in it: a bug you diagnosed, a design you changed, a feature you cut
• Prefer a shipped feature over an exploratory chat that went nowhere
• Scrub secrets first: API keys, tokens, customer data and internal URLs
• Keep it readable: one coherent session beats a 10MB dump
• Link it to the product: say in the application what that session produced

The scrubbing point deserves emphasis this week in particular. On October 17, the team behind OpenCode disclosed on X a breach affecting 1.4% of users of its OpenCode Console, a subset "who had external providers linked," and TechCrunch reported that affected users reported unauthorized charges on their AI accounts, mostly in the hundreds of dollars, with a few describing totals above $1,000. Transcripts routinely contain pasted keys and environment variables. Before you upload any session anywhere, search it for anything that looks like a credential, and rotate any key that has ever appeared in one. Our pre-launch security checklist for AI-built apps covers the habits that keep secrets out of transcripts in the first place.

Why this matters: the founder section is the part of the application that cannot be generated, so it is where your real advantage shows or does not. How to apply it: write it first, fill in every founder's history completely (Caldwell's first rule is "fill the whole thing out"), lead each answer with a magnitude, and attach a coding session only if it shows the kind of judgment you want the reader to see.

5. The Founder Video: One Minute, Nothing but Founders

The founder video is the one part of the application where the reader sees and hears you, and YC's instructions for it are short, specific and frequently ignored. YC's video page says: "The video should be 1 minute long and should contain nothing except the founders talking." It asks founders to "introduce yourselves, explain what you're doing and why, and tell us anything else you want to about the founders or the project." It says "This is not the place to submit a demo or promotional video," because "there is a separate place in the application for that. For this video, we want to hear how the founders communicate." All founders should appear, and "If you can't be in the same room at the same time, screen record a video call instead."

The most counter-intuitive instruction is about preparation. "Do not recite a written script: Use bullet points instead. Using bullet points, just talk spontaneously as you would to a friend. Reading a written script doesn't help your application or convey communication skills." That rule exists for the same reason as the interview guidance later in this guide: YC is sampling how you think and communicate under light pressure, and a script hides exactly the thing being sampled. A model can write you a flawless script in seconds. It cannot make you sound like someone who understands their customers.

Following the instructions is itself graded. Caldwell said so plainly in 2023: "our entire application process has these detailed instructions, and we definitely take into account if you follow them in our selection criteria. This is one of the ways to tell how detail-oriented someone is." In his 2018 version of the talk he described the failure modes he still saw: "We get the 20-minute, like, spirit-journey video," and "an unrelated video of a product demo that they made with one of those free tools where there's no humans in it ... So please follow the directions. I swear it matters."

The clearest way to understand the format is to watch one that worked, and YC chose the example itself. The video instructions page embeds Teespring's application video from the Winter 2013 batch, which runs 47 seconds. It is thirteen years old, which is exactly the point: the format has not changed because the purpose has not changed. Watch it for the shape (founders on camera, talking, under a minute) rather than for the specific product.

Alongside that embedded example, YC's video page plays application videos from Kernel (Summer 2025), Gumloop (Winter 2024), Medplum (Summer 2022) and Whatnot (Winter 2020), so you can compare how different teams use the same minute. Some YC companies also leave their original application video public on their directory page. By our count of the public directory data, 103 companies have done so, six of them from Summer 2026, including the AI infrastructure company Orchestra, whose page shows its application video today.

A practical recording routine makes the minute easy. Write four or five bullet points (who you are, what you are building, the one most impressive fact about your progress, why you are the right people, what you will do next), put them somewhere you can glance at, and record in one take with all founders in frame. If you go over a minute, cut a bullet rather than talking faster. Re-record until you sound like yourselves, not until you sound polished; the reader is listening for clarity and chemistry, not production values.

Why this matters: the video is a sixty-second sample of how the team communicates, and it is one of the few parts of the application the reader can use to check the written answers against real people. How to apply it: follow every instruction on YC's video page literally, use bullets instead of a script, keep the demo out of it, and put your single strongest fact in the first fifteen seconds.

6. The Company Section: The 50-Character Line and "What Are You Making?"

The company section opens with the two questions a reader uses to file your application in their head. Recent forms have asked founders to "describe what your company does in 50 characters or less," and then, "What is your company going to make? Please describe your product and what it does or will do." Graham explains why the second one comes first in his reading: "I look at it first because I need something to hang the application on in my mind. The best answers are the most matter of fact." Everything else in the application is interpreted through whatever picture these two answers create, so an unclear picture makes every later answer weaker.

The 50-character line is a compression test, and the best answers pass Graham's reproduction test: after reading it, the reader could roughly describe the product to someone else. Some YC companies have made their original applications public on their YC directory pages, which gives us real answers to study. The AI video company sync., accepted for Winter 2024, wrote "lipsync video to audio in any language in one-shot" on its application. Kamion, from Summer 2022, wrote "Uber for Trucks in Turkey" ( application). Propexo, from Winter 2023, wrote "Plaid for the Property Management Industry" ( application). None of them is clever. Each tells you exactly what the product does and for whom.

The "X for Y" format in two of those examples works only when both halves are concrete and the analogy carries real information. "Uber for trucks in Turkey" tells you the marketplace model, the supply side, the demand side and the market in five words. "Uber for productivity" tells you nothing, because productivity is not a market and Uber's mechanics do not map onto it. If your analogy needs a sentence of explanation, drop it and describe the product directly.

Answering "what are you making?"

The longer description should do three jobs in order: say what the product is, say who uses it and what it replaces, and give the strongest evidence that it is real. GitLab's answer from its Winter 2015 application, preserved on its YC page, is a model of the form: "We're making open source software to collaborate on code. It started as 'run your own GitHub' that most users deploy on their own server(s)." Then the details of what it does, then one line of proof: "Over 100k organizations use it including thousands of programmers at Apple." There is no adjective in the whole answer that the facts do not earn.

Graham names the most common mistake directly. "Don't begin your answer with something like 'We are going to transform the relationship between individuals and information.' That sounds impressive, but it conveys nothing. It could be a description of any technology company." Garry Tan, in his tips for applying, recommends the journalist's inverted pyramid: "figure out what that conclusion is and put it right at the beginning. Right after that, start adding proof." And on vocabulary: "Plain English is best because Y Combinator uses plain English to figure out whether or not you know your stuff."

This is where the AI-writing problem from the introduction bites hardest. Koomen's "wedge" statistic and Bosmeny's 60% length increase are symptoms of the same habit: founders asking a model to "make this sound better," and the model adding framing, strategy vocabulary and length. Each addition is cheap to produce, so it carries no information, and it dilutes the facts that do. Entrepreneur reported in August on data Koomen shared suggesting that more than half of applications used em dashes by summer 2025, and on Paul Graham's remark about AI-written founder emails: "It feels like being lied to." YC has published no rule against using AI to help write, and AI is genuinely useful for checking grammar or finding where an answer is unclear. The failure is letting it add words. Caldwell's rule applies with new force: "More words is not better."

The demo slot and the location question

YC's video page confirms that "there is a separate place in the application" for a product demo, and YC's February announcement called "the application demo" "key." If you have anything working, include a short demo: a two-minute screen recording that shows a real user flow beats a polished trailer. Make sure the link works without a login, and if your product requires an account, record the logged-in experience rather than asking the reader to sign up. Recent forms have also asked where the founders live now and where the company would be based after YC, which is a factual question, not a test; answer it plainly.

Why this matters: the company section sets the frame for every other answer, so a vague first sentence costs you across the whole application, not just in one box. How to apply it: write the 50-character line until a stranger could repeat it, open "what are you making" with the plainest description of the product, end it with your best single piece of evidence, and cut every adjective the facts do not earn.

7. Progress: The Thirteen Days Before November 2

The progress section is where the application stops being about claims and starts being about evidence, and it is the one part you can still change materially before the deadline. Recent forms have asked how far along you are, how long each founder has worked on it and how much of that was full-time, how many active users or customers you have and how many are paying, "Who is paying you the most, and how much do they pay you?", and your revenue over recent months, specifically "not cumulative and not GMV." Each of those phrasings closes a loophole founders have used to inflate their numbers, which tells you how carefully the answers are read.

Start with the reassurance, because it is true. YC's FAQ says "it would be a mistake to wait to apply," that 40% of each batch is "just an idea," and that most companies have no revenue. Caldwell was even more direct in 2018: asked "What's the right amount of MRR to get into YC?", he answered, "Holy cow, the answer is zero." Zero revenue does not disqualify you. What hurts is zero movement, because the progress section is really asking about rate of change, and a team that has existed for six months with nothing to show for it tells the reader something about speed.

What changed when building got cheap

Here the 2026 context matters more than in any previous cycle. Building software is radically cheaper than it was even a year ago. Anthropic wrote this month that "Claude writes about 90 percent of Claude Code's own code," and that from June to September merged pull requests to Claude Code's codebase "more than doubled while the share that had to be reverted fell by more than 40 percent." Vercel's AI Gateway Production Index for October found that across its traffic "the average price per token fell 47%" between December and September. The tools that let a small team ship a real product in days are now cheap and widely available.

The first-principles consequence for your application is uncomfortable. When building a working prototype costs a weekend and a few dollars of inference, having one stops being evidence of much. Every serious applicant can show a product. The scarce, costly signals move downstream: people using it, people paying for it, a measured growth rate, and insight you could only have gained by talking to customers. A prototype is now table stakes for many software ideas, and the progress section is judged on what happened after the prototype existed.

That also means the next thirteen days are enough to change your answers, if you spend them on the right thing. If you have no product, there is still time to build a usable first version and put it in front of real people. If the building is the obstacle, the tooling is not: AI coding agents like Claude Code and Codex, app builders, and platforms that build and run a complete product from a plain-language description all exist now (Founden, which publishes this guide, is one of the latter, and just over 1,000 companies have been built on it by its own measured count). Our guide on how to build an app with AI and our breakdown of what it costs to build an app with AI cover the options. But the tool is never the point. The point is what real users do with the result.

YC's own recent Startup School material makes the same argument about where those first users come from. In a June 2026 episode on getting your first ten customers, YC Visiting Partner Max Kolysh argues that they "almost never come from a tool. They come from you, working your warm network and doing the unscalable things most people won't." The episode's description adds the two habits that matter most for an applicant on a deadline: finding "where your buyer actually spends their time" and "why showing up in person wins."

That emphasis on doing it yourself is directly usable for the next two weeks. Ten conversations with the exact people you are building for will improve your progress answer, your idea answer and your interview more than any amount of rewriting. For more on turning early users into word of mouth, see our guide on how to get people to talk about your product.

A thirteen-day plan

With thirteen days, sequencing matters more than effort. The order below front-loads the parts of the application that cannot change (your history) so the remaining days go to the parts that can (your progress). It assumes you submit early and keep working, since YC's FAQ says the review process starts when you submit and "we don't allow continuous editing" except for material changes such as a co-founder joining or leaving.

The plan also assumes the honest constraint of most applicants: you have a job, classes or a current product to run, so the days are not all equal. If you can only find a few hours a day, keep the order and stretch the first three steps, but do not let writing eat the days meant for users. A finished application on day six plus a week of real conversations beats a perfect application on day thirteen with nothing new to say in the interview, because the interview is where the extra week of evidence gets heard. If you are a student, the same logic applies to Early Decision: the progress you make this autumn is what the application for a later batch will be judged on.

1. Days 1 to 2: write the founder section, the 50-character line and "what are you making," then cut them by a third
2. Days 3 to 4: record the founder video from bullet points, re-record once, stop
3. Days 5 to 6: fill every remaining field, attach the demo, check every link logged out, submit
4. Days 7 to 13: talk to users, ship, measure, and write down every number with its date
5. Before the interview: one page of metrics and what changed since you applied

The plan deliberately puts submission at day six rather than day thirteen. Nothing in YC's published guidance rewards submitting at the last minute, its apply page encourages submitting "as soon as you're ready," and its interview guide says the best way to improve your chances is "to have your startup improve between the time that you applied and the time that you interview." A week of measurable progress after submission is worth more than a week of polishing sentences, and the polishing, as the AI statistics show, is more likely to make the application worse.

Why this matters: progress is the most honest signal in the application, and it is the only one you can still move this month. How to apply it: report monthly numbers exactly as asked (not cumulative, not GMV), state user counts with the date you measured them, and spend the days after submitting on progress you can describe in the interview. If you state a number, be ready to prove it; YC's interview guide warns, "if you state numbers in the interview we may ask for verification of them afterwards."

8. Idea, Market and Money: Answering Like an Insider

The idea section is where YC tests insight, and the questions are designed to separate founders who know their market from founders who have read about it. Recent forms have asked why you picked this idea and whether you have domain expertise, how you know people need it, "What's new about what you're making? What substitutes do people resort to because it doesn't exist yet," who your competitors are and "What do you understand about your business that they don't?", how you will make money and how much, and how users find the product. Graham's summary of what the reader wants: "what we look for in ideas is not the type of idea but the level of insight you have about it."

Insight has a recognizable texture. It is specific, slightly surprising, and checkable, and it usually comes from direct contact with the problem. Jessica Livingston, one of YC's founders, described it in her tips for YC interviews, and YC's current interview guide repeats the line: "We love learning new things. And a good startup idea usually teaches you something when you encounter it. Don't worry if the new things about your idea are things only someone in your field would care about. We like that. We'd rather have interesting details than boring generalizations."

"Why now" is where the current market helps you

The strongest idea answers often contain a "why now" that is specific to this year, and October 2026 has handed founders several. Consider how a founder building on cheap open models might answer "what do you understand that competitors don't." Vercel's October index found that open-weight models now process "more than seven in ten gateway tokens" for "nearly one in four dollars," and that "DeepSeek's models processed 58% of all gateway tokens for 6.6% of spend." A founder who has priced their product around those economics, and can explain the gross margin it produces, is showing exactly the kind of insight the question asks for. Our guide to pricing an AI product to beat token costs works through that math.

The same week shows why a good answer also names its risks. On October 20 the Federal Register published a Department of Homeland Security order excluding DeepSeek products from federal civilian procurement and removing "all DeepSeek and DeepSeek-affiliate products, and third-party products and services developed using DeepSeek implementations, from civilian agency information technology." For most startups this changes nothing. For a startup selling to federal agencies on top of DeepSeek models, it changes the whole go-to-market. Graham's guidance applies directly: "we want to see that you're aware of the obstacles, and have at least a theory about how to overcome them." A founder who writes "we run on cheap open models, and we can swap providers in a day because the government channel now requires it" sounds like an insider. A founder who never mentions supply-chain risk while selling to government sounds like they have not looked.

Competitors, money and the alternates question

On competitors, the instinct to claim you have none is always wrong, because it signals either a market nobody wants or a founder who has not looked. YC's interview guide is explicit: "You should be intimately familiar with the existing products in your market, and what, specifically, is wrong with them. It's not enough to say that you're going to make something that's more powerful, or easier to use. You should be able to explain how." It also helps to know that YC does not penalize overlap with its own portfolio. Its FAQ says "we don't consider it a factor whether we've already funded a company working on something similar."

On money, the reader wants a plausible path, not a spreadsheet. Garry Tan's framing in his application tips is a useful sanity check: "in the next five to ten years, is there some real way for you to get to a hundred million dollars a year in revenue?" Answer with who pays, roughly how much, and how many such customers exist, using numbers you can defend. A top-down "1% of a $50 billion market" calculation is precisely the kind of cheap claim the reader skips.

Finally, do not skip the question about other ideas. Graham writes that when "the founders seem promising but the idea doesn't, I check the question near the end that asks what other ideas the founders had. It's quite common for us to fund groups to work on ideas they listed as alternates." Recent forms phrase it as an invitation: "One may be something we've been waiting for." Two or three crisp alternates, each in a sentence, give a reader who likes your team a second way to say yes. For inspiration grounded in what YC is actually funding, our analysis of what software is left to build in 2026 and our guide to boring businesses AI can transform map the territory.

Why this matters: the idea section is where a reader decides whether you know something the other several thousand applicants do not. How to apply it: answer every question with a fact you learned from customers or direct experience, name your real competitors and what specifically is wrong with them, include a "why now" grounded in something that changed this year, acknowledge the biggest risk with a theory for handling it, and list your alternates.

9. Structure, Equity and the $500,000 Deal

The equity section of the application looks administrative, and most of it is. Recent forms have asked whether you have formed any legal entity, where each one was formed, the ownership breakdown among founders, employees and other stockholders, whether you have taken investment, how much you spend per month, how much is in the bank and how long your runway is. The reader is not grading your paperwork. They are checking for problems that would make a fast investment impossible (a messy cap table, a co-founder with no equity, a foreign entity that needs restructuring) and for one signal that matters a great deal: whether the founders have split ownership in a way that suggests a durable partnership.

The single most common worry is unnecessary. YC's FAQ answers "Do we need to incorporate before applying?" with one word: "Nope." Founders outside the US do not need to do anything first either: "If we accept your application, we'll help you get visas and incorporate." What YC requires is that the company you eventually take money into is in the right place. Its deal page says "We invest in US, Canada, Cayman, and Singapore corporations," and Canada is a recent return: Garry Tan announced on February 5, 2026 that "YC will once again invest in US, Canada, Cayman, and Singapore corporations," after a period in which Canada was dropped because top Canadian companies tended to reincorporate in the US anyway.

LLCs, Delaware and the "flip"

If you already have a company somewhere else, YC's deal page explains that "you will need to 'flip' your corporate structure to have a parent company in one of those four countries," usually by making the existing entity a subsidiary of a new parent that owns the intellectual property. The FAQ is candid that "many companies in each YC batch do this process," and that while YC will connect you with lawyers, "it will require a significant effort on your part." That effort is a reason to know your structure before the interview, not a reason to restructure before applying.

The question founders search most often, LLC versus C corporation, has a clear answer for anyone planning to raise venture capital. In YC's startup legal mechanics talk, YC partner Carolynn Levy describes founders who formed an LLC because it was better for taxes: "for our purposes, this is usually a mistake because even though LLCs are better for optimizing taxes, the vast majority of angels and VC firms will not invest in an LLC." She adds, "We see a lot of LLC-to-corporate conversions, and it's just kind of a headache you don't need," and recommends Delaware because the process is fast and "some investors actually will require that you're in Delaware before they'll fund you." YC does not list Delaware as a hard requirement on its deal page, but the recommendation is close to universal. If you incorporate before applying, the same talk names Clerky and Stripe Atlas as the tools YC considers great. Our guide to starting a company in 2026 walks through the steps for first-time founders.

Splitting equity and vesting

How you split ownership tells the reader how you think about your co-founders. Michael Seibel's YC post on splitting equity is blunt: "what we almost always recommend at YC: equal equity splits among co-founders," and "If you aren't willing to give your partner an equal share, then perhaps you are choosing the wrong partner." The same post describes the standard protection against a co-founder leaving: "four years of vesting with a one year 'cliff,'" so that someone who leaves within the first year walks away with nothing and everyone earns the rest monthly. YC's legal talk calls skipping vesting because "our founding team has been working on this idea for years" "a big mistake," and says the same of solo founders who think they do not need it.

There is one deadline in this area that cannot be fixed later. When founders buy their stock subject to vesting, the US tax code lets them file an 83(b) election to be taxed on its value at purchase rather than as it vests. The IRS instructions for Form 15620 say "An 83(b) election must be filed no later than 30 days after the date the property was transferred." YC's lawyers put it more memorably: "it includes calendar days, it's not a business day period. And if you miss it, there's really not much you can do about it." Jon Levy's one instruction to every founder in that talk is "Never forget your 83(b)."

What the deal actually costs

YC's deal page describes the standard investment exactly. "We invest $500,000," split into "$125,000 on a post-money SAFE in return for 7% of your company" and "$375,000 on an uncapped SAFE with a Most Favored Nation ('MFN') provision." The MFN SAFE has no valuation cap of its own; it converts on the terms of the lowest-cap SAFE (or a more favourable term such as a discount) that the company issues between around the start of the batch and its priced round. YC's own example: if you "raise your next SAFEs at a $15M post-money valuation cap, the $375,000 MFN SAFE would convert into $375,000 / $15,000,000 = 2.5% of the company." YC also says it charges no fees and keeps a pro rata right and a participation right in later rounds.

That means YC's total stake depends on the price of your next round, and the arithmetic is simple enough to chart. The figures below apply YC's published formula; they show YC's ownership after its two SAFEs convert and before a priced round and option pool dilute everyone.

Where on that chart a YC company lands is set by the market after Demo Day. TechCrunch reported from Winter 2026 Demo Day that "the 'default' valuation this quarter seems to be around $30 million, which investors told me is roughly twofold the current seed market average," which on YC's formula puts its total stake near 8.25%. Those are figures investors gave a reporter, not filings, but they show why the MFN structure is cheaper for a company that raises well than the headline "7% plus" suggests.

Is it worth it? The honest case against is real. Fast Company summarized critics' complaints in 2025 as "increased batches, diminished seed rounds, more duplicate companies, less specialized training," and Forbes noted that the deal "isn't cheap" when YC's educational material is free online. High Demo Day prices have a cost of their own: Work-Bench general partner Jonathan Lehr told TechCrunch that higher seed valuations mean "less room for experimentation, less tolerance for pivots, and more scrutiny if progress doesn't match the capital raised." The case for is structural rather than sentimental: about 8% to 10% of the company buys $500,000 committed on acceptance, three months of focused partner attention, the "$12M+ in free credits and deals" listed on the apply page, and access to a Demo Day audience that consistently prices companies above the seed market average. For a pre-product team, that trade is usually good. For a company already growing fast with investors calling, it is worth doing the math rather than applying on reflex.

Why this matters: structure questions rarely decide an application, but a messy one can slow an acceptance, and equity splits are read as a signal about the team. How to apply it: do not incorporate just to apply, do not form an LLC if you plan to raise venture money, split equity close to equally with vesting, file any 83(b) within 30 calendar days, and answer the runway and spend questions with exact current numbers.

10. The Interview: Ten Minutes on Zoom

If your application works, you will get an email inviting you to interview, and YC has published unusually clear guidance on what happens next. Its interview guide states the basics: "YC interviews are 10-minute conversations over Zoom," "All founders should be present on the call," "Expect 2-3 YC Partners to be on the call," and "Interviewers will have read your application and have it open during the call." It also says what the ten minutes contain: "We only do two things at interviews: we ask you questions, and we look at what you've built so far."

The structure follows directly from the problem YC is solving. Ten minutes is far too short to evaluate a business, but it is long enough to sample how founders think when asked something they did not prepare for, and that sample is the scarce information the written application could not provide. That is why the guide's most counter-intuitive instruction is also its most important one: "Don't rehearse." YC writes that beyond basic preparation, practice "is not useful and is often counter-productive," because "we just want to have a conversation, and that works better when you are talking spontaneously," and practice sessions often produce "feedback from friends (including YC alumni) that is well-meaning but counter-productive."

What to prepare instead

The guide's replacement for rehearsal is progress: "the best way to get an edge is to work hard and have your startup improve between the time that you applied and the time that you interview. This may mean you launched, improved your product, increased revenue, etc. Demonstrating you can move fast and make quick progress is the most surefire way to impress your interviewers." The specific preparation it does recommend is narrow and practical.

That ordering is worth taking literally. A team that spends the ten days before its interview practising answers will arrive with polished sentences and the same metrics it had when it applied, while a team that spends them shipping will arrive with a new number, a new customer or a new insight to report. The second team has something to say that the partners have not already read, and the conversation naturally starts there. The preparation YC does ask for takes an evening, and it is mostly about having facts at hand rather than rehearsing how to say them:

• Answer "what are you working on?" in a few plain, jargon-free sentences; YC says it is "typically the first question"
• Know your users and metrics: where users come from, growth, retention, unit economics
• Write key numbers down so you can glance at them, since numbers "may" be verified afterwards
• Have a demo ready to screenshare, or the hardware in the room
• Let every founder answer at least one question

Each item on that list maps to something the partners are testing. The opening question tests clarity under time pressure. The metrics questions test whether you know your business or only your pitch, and the guide lists the ones it often asks of launched companies: "Where do new users come from? What is your growth like? How much are your users using the product, and do they stick around? What are your unit economics? What makes new users try you? Why do the reluctant ones hold back? What are the top things users want? What has surprised you about user behavior?" The demo tests whether the thing exists. And asking each founder to speak tests whether this is a real team or one founder with passengers. Dalton Caldwell's 2023 description of a good interview ties it together: "the founders can demonstrate a mastery of their own business. They understand their own numbers. They understand the risks. They have a crisp idea on what to work on next."

YC's interview guide itself points applicants to a video, and it is worth watching for the same reason. It embeds Caldwell's Startup School talk on applying and succeeding at YC, starting at the 18:13 mark, where he moves from the written application to what to expect in the interview and what sets successful ones apart.

The talk dates from 2023, but YC still embeds it on its current interview guide, which is the best evidence that the advice in it has not changed. Two pieces of guidance from YC's written material complement it well. On candour: "we tend to be more convinced by candid discussion of difficulties than glib dismissal of them," and "if we ask a question you don't know the answer to, don't panic. A smart person trying sincerely to answer an unexpected question can lead to a great discussion." On tone, Jessica Livingston's tips for YC interviews are as true now as in 2015: "Don't use buzzwords. Speak in plain language," "Have a conversation with us. Don't deliver a prewritten 'pitch,'" and "please don't take it personally if we jump right into things."

After the call

YC's apply page says it typically decides the same day. If the answer is no, the interview guide promises feedback by email, "genuinely useful advice that will make you more likely to succeed," and adds the sentence every rejected interviewee should reread: "It's very common for teams to take our feedback, re-apply the following batch and get accepted." With four batches a year, "the following batch" is now a few months away rather than half a year. The guide's last instruction is about character rather than tactics: "Be earnest. The YC folks in your interview are likely the same people you'll work closely with should you be accepted."

Why this matters: the interview is the decisive ten minutes, and it rewards exactly the things that cannot be crammed: mastery of your own numbers, honest thinking about risk, and visible momentum. How to apply it: do not rehearse a pitch; instead keep shipping between submission and the call, keep one page of dated metrics in front of you, have the demo open, decide in advance which founder takes which kind of question, and answer the first question in two plain sentences.

11. Solo, Non-Technical, International, Student and Repeat Applicants

Most of the anxiety around YC applications comes from founders who think their situation disqualifies them. YC's published answers are more permissive than the folklore, but each special case comes with a specific thing the reader will look for, and knowing it lets you answer the unasked question before it costs you.

Solo founders

YC's FAQ is unambiguous: "We regularly accept solo founders. That said, our advice remains that one-person startups are tough and you're more likely to succeed with a co-founder." The 2026 data shows the first sentence is getting truer. In the Lightcone episode cited earlier, Diana Hu said solo founders have gone from about five percent of accepted companies to "over eighteen, nineteen percent," and Garry Tan said, "it turns out YC is the YC for solo founders." The structural reason is the same cheap-building shift: one strong builder with coding agents can now ship what used to take a small team, so the cost of not having a co-founder has fallen. The risk YC worries about (one person carrying everything) has not. Our analysis of the rise of the solopreneur covers the broader trend.

If you want a co-founder before November 2, YC runs a free tool for exactly that. YC Co-Founder Matching describes itself as "the largest network of its kind," with "over 100K matches made," and lists active profiles by city, led by San Francisco (3,200), New York (3,000) and London (2,900).

Be careful with timing, though. A co-founder found in the last week before a deadline is a risk the reader will notice, because recent forms ask how long the founders have known each other and how they met. A solo application with a strong builder and real progress is usually better than a rushed pairing. If you do add someone after submitting, YC's FAQ says a "co-founder addition/removal" is exactly the kind of material change you should submit as an update.

Non-technical founders

This is the hardest case, and YC does not soften it. The FAQ says "it's important for the founding team to have the skills to build their product themselves, rather than outsourcing it to someone else. For most businesses, that usually means you need a technical co-founder." Caldwell's 2023 talk put numbers on it, with technical teams having "five times better odds" of an interview and non-technical teams on "an unlaunched tar pit idea" having "the lowest odds." In his 2018 talk he described the other side of the rule: founders without known software credentials are "A-okay. We still fund folks like this. But the burden of proof is on them to prove they can do it."

That burden of proof is where 2026 genuinely changes the picture. The new coding-session upload and the Paxel report are, in effect, ways to prove you can build regardless of your job history. A founder without a computer science degree who has shipped a working product with coding agents, and can show the sessions where they planned it, debugged it and decided what to cut, has evidence that did not exist as a category two years ago. What does not work is outsourcing: a product built by an agency, or by a tool you cannot explain, fails the FAQ's test. If you are in this position, our guides to building a live app with Claude Code and to the top AI app builders are the place to start, and the honest goal is to become the person who can build it, not to find a way around being that person.

International founders, students and people with jobs

YC says its process is "open to any startup, anywhere in the world," and international founders need no US entity or visa to apply. If accepted, YC says it "works with excellent immigration attorneys that have helped hundreds of our founders get visas to move to or stay in the US," and founders "can also use the investment to cover their travel expenses to San Francisco." The batch itself is in person, so the practical question is whether you can be in San Francisco from January to March. The early Winter 2027 companies in YC's directory already include Lucebox, which lists Milan, Italy as its base.

Students and employees can apply, with one condition: YC "expect [s] the founders to commit to working full-time on their company during the batch and afterwards if accepted." Students who cannot start in January can use Early Decision to apply now for a later batch; the most common path, YC says, is "students applying in the fall of their final year and joining the summer batch after graduating." Founders who have already done another accelerator can still apply, but the FAQ warns that "we may expect that you've reached a higher level of progress."

Reapplicants

If you have applied before, you are in the majority of YC's eventual companies. The FAQ says "in a typical YC batch, about half the companies applied multiple times before being accepted," and "Having made progress since your last application is a strong signal to us." YC's admissions team checked this in its own data and reported that repeat applications do not hurt: "Teams that re-apply and have made progress since their first app are much more likely to get into YC." You must fill out a new application each time, but you can reference old ones, and recommendations carry over.

One founder's account shows how this works in practice. The team behind Metorial published both of its applications in August, writing that "We interviewed for S25 and were rejected. A few months later, we applied again and got into F25." Their rejection email, which they quote, said "it wasn't clear to us how you'd differentiate among the many MCP projects," and the second application answered that question with progress. Their framing of why reapplying works is the best summary we have seen: "A first application gives YC a point ... A second application gives them another point. Now they can calculate a slope." This is a single self-reported case, but it matches what YC itself says.

Why this matters: each special case has a specific doubt attached, and an application that leaves the doubt unanswered loses to one that addresses it. How to apply it: solo founders should show the work of a team, non-technical founders should show they built the product themselves, international founders should confirm they can be in San Francisco from January, students should consider Early Decision, and reapplicants should lead with exactly what changed since last time.

12. How Applications Fail, and Plan B If Yours Does

Almost every rejected application fails in one of a small number of ways, and almost none of them are about the idea. YC's own descriptions of failure are consistent across a decade of talks. In 2018 Caldwell said, "you would be shocked how many people don't actually fill out the dang thing properly ... There's like whole areas where they don't answer the question whatsoever, they don't put in URLs, they don't fill out their bio."

In 2023 he added that "if there's some hint somewhere in your application or elsewhere that you are being misleading in your answers, your application will be rejected." And Graham's estimate that "more than half" of interview-worthy teams "blow the application" is a reminder that the most common failure is simply not conveying how good you are.

Reasoning from the reader's position, the failures fall into two kinds. The first kind wastes the reader's attention: missing fields, broken links, vague first sentences, long answers, buzzwords, a scripted video. The second kind destroys trust: inflated or ambiguous numbers, cumulative revenue presented as monthly, a "demo" made by someone outside the team. The first kind costs you the ranking; the second kind costs you the application.

• Incomplete or sloppy: empty fields, no bios, dead or login-gated links
• Unclear: a first sentence the reader could not repeat to someone else
• Padded: AI-lengthened answers, strategy vocabulary, "wedge"
• Unproven builder: no one on the team visibly able to build the product
• Untrustworthy numbers: cumulative, GMV, or annual figures dressed as monthly

The common thread is that every item is cheap to fix before you submit and impossible to fix after. YC's FAQ says the application "enters our review process once you submit it, so we don't allow continuous editing," and the why-not page explains that if you are not invited to interview you will not get feedback, because "there often is no reason" beyond others looking stronger.

So the last thing to do before you click submit is a hostile read: give the application to someone who does not know your company, ask them to explain back what you do and why you will win, and fix every point where they hesitate. Then check every link while logged out.

Plan B: what to do with a no

A rejection from YC in 2026 is far less final than it used to be, for two reasons. The calendar now runs four times a year, so the next cycle is a few months away, and YC's own data says reapplicants with progress do well.

If you interviewed, you will have written feedback to act on. If you did not, YC's guidance is to keep building and apply again with a measurably better application, and its apply page notes you can apply to future batches now.

Meanwhile, other programs are open in the same window, and the table at the top of this guide scores the ones whose current terms we could verify. The most time-sensitive is a16z's speedrun, whose SR008 applications close on November 1 at 11:59pm PT, one day before YC's, with a standard deal of "$500K for 10% upfront on a SAFE, and up to an additional $500K in your next qualifying round."

PearX takes applications "year round" for a winter cohort that runs from January to April, and Techstars has its Spring 2027 programs open. For the wider field, our rankings of the top 20 US accelerators and the top 20 European accelerators compare terms and outcomes.

You also do not need an accelerator to keep going. Our list of 40 free startup credit programs covers the infrastructure money many accelerators bundle, our directories of US VCs with an AI thesis map the investors who fund the same companies YC does, and for founders outside the US, our guide to EU AI grants and funding programs covers non-dilutive money. If what you need most is people, our ranking of founder communities worldwide is the place to look.

Why this matters: most failures are unforced, and a no is now a short detour rather than a verdict. How to apply it: run a hostile read and a logged-out link check before submitting, apply to one or two parallel programs whose terms you would genuinely accept, and if YC says no, use the next three months to produce the slope that makes the next application obvious.

Conclusion: A Decision Framework for the Next Thirteen Days

Every piece of advice in this guide reduces to one principle: the reader's attention is scarce and the cheap signals are worthless, so spend your words on evidence that is expensive to fake. That principle explains why YC asks for a one-minute unscripted video, why it now asks to see a real coding session, why it tells interviewees not to rehearse, and why applications that grew 60% longer did not get 60% better. It also tells you how to spend the time you have left.

Use these questions to decide what to do today. Should you apply at all? If you have a founding team, an idea you are excited about, and can be in San Francisco from January to March 2027, YC's own FAQ says waiting would be "a mistake," since 40% of each batch is just an idea. If you are already raising at strong terms, run the deal math in section 9 first. What should you write first? The founder section and the 50-character line, because they frame everything else and cannot be faked. What will move the decision most? Progress: real users, real usage, real numbers, made between now and your interview. What should you cut? Every word, adjective and framing device a reader could have predicted, especially the ones a model added.

The founders who get interviews this cycle will not be the ones with the most polished prose. They will be the ones whose applications read like a dated log of things that actually happened: who they are, what they built, who uses it, what they learned, and what they shipped in the two weeks after they hit submit. You have until November 2 at 8pm PT. Submit early, keep building, and let the interview be the update.

This guide reflects YC's published Winter 2027 process, deal terms and directory data as of October 20, 2026. YC changes its application, deadlines and terms between cycles, and the alternative programs' terms change often, so verify every date and term on the official pages before you apply.