../

Y Combinator & Paul Graham

What Paul Graham's essays and Y Combinator's partners actually say about ideas, launching, users, growth, money and founders, with the year and source for each idea and a pointer to where it applies in the idea-to-MVP playbook. The contrasting investor views are in a16z and Peter Thiel; operators' views are in other founders.

Who they are and why listen

WhoRoleWhy their advice carries weightMain written sources
Paul Graham (PG)co-founded Viaweb (1995, sold to Yahoo 1998) and Y Combinator (2005)has watched thousands of very early startups up close; writes from pattern-matching, not theoryessays at paulgraham.com (2005–now)
Jessica LivingstonYC co-founderinterviewed founders about their earliest days (Founders at Work)books, "How Not to Fail"
Sam AltmanYC president 2014–2019ran YC's growth era; wrote the most compact summary of YC adviceStartup Playbook (2015), blog
Michael Seibelco-founder of Justin.tv/Twitch and Socialcam; YC partner and later managing directorturned "launch now" into concrete MVP rulesStartup School talks, michaelseibel.com
Jared FriedmanYC partnersystematised how YC evaluates and generates ideasStartup School talk on startup ideas
Geoff RalstonYC president 2019–2022wrote YC's standard seed-fundraising guideA Guide to Seed Fundraising (2016)

YC's advice is unusually consistent because it is one motto applied to every question. PG says the phrase became the motto about a month after YC started, and that if choosing again it's still the one he'd pick (Be Good, 2008 (opens in a new tab)):

"Make something people want."

Almost every other rule on this page is a corollary: launch early (to find out what people want), talk to users (to find out what people want), do things that don't scale (to make a few people want it a lot), and measure growth (to check people want it).

The core rules

YC's own summary is its essential startup advice (opens in a new tab) page (YC Library, undated; it cites essays from 2008–2016). Its "pocket guide" boiled down:

PrincipleWhat it meansSourceApply it by…
Launch nowship a mediocre product early; it's the only way to learn what customers need, as long as it has a real kernel of valueYC essential advice; PG, 18 Mistakes (opens in a new tab) (2006)building the MVP to a deadline
Make something people wantthe one mistake that kills startups is not doing this; nearly all failure funnels through itPG, 18 Mistakes (2006)treating every other plan as secondary to validation
Do things that don't scalerecruit and delight the first users by handPG, Do Things That Don't Scale (opens in a new tab) (2013)first users
Find the 90/10 solutionPaul Buchheit's rule: get 90% of the value for 10% of the workYC essential advicescoping the MVP
10–100 customers who love youa small group who love you beats a large group who kind of like youYC essential advicepicking a narrow first segment in ideation
Write code, talk to usersthese two tasks should take almost all founder time early onYC essential advicethe weekly operating rhythm
Growth follows productYC says growth is the result of a great product, not the precursor; don't scale before product–market fitYC essential advice; Seibelthe PMF checks before spending on growth
Ignore competitorsYC's line is that startups "die of suicide not murder"YC essential advicecompetition analysis once, then back to users
It's not your moneyraised money carries a duty to spend it only on the company's prospectsYC essential advicebudgeting the six weeks in the playbook
Founder relationships mattermany companies fail early because founders fall outYC essential advice; PG 18 Mistakes #17agreeing roles and equity before building
Be nicePG finds almost no mean people among the most successful founders he knowsPG, Mean People Fail (opens in a new tab) (2014)hiring and co-founder choice

Startup ideas

How to get them (PG, 2012)

How to Get Startup Ideas (opens in a new tab) opens with its whole thesis:

"The way to get startup ideas is not to try to think of startup ideas. It's to look for problems, preferably problems you have yourself."

PrincipleWhat it meansSourceApply it by…
Organic, not made-upideas that grow from your own experience beat "sitcom" ideas invented to sound like a startupHow to Get Startup Ideas (2012)listing problems you personally hit, in ideation
Dig a wellchoose something a small number of people want a large amount over something many people want a littlesamenaming the first user who wants it now, even as a crappy v1 from a two-person startup
"Live in the future, then build what's missing"work at the leading edge of a changing field; gaps become obvioussamethe "why now?" test
Turn off the filtersthe "unsexy" and "schlep" filters hide the best ideassame; Schlep Blindness (opens in a new tab) (2012)deliberately listing boring and tedious problems
Don't fear competitorsPG: it's exceptionally rare for startups to be killed by competitors; err toward markets that have themHow to Get Startup Ideas (2012)writing down what incumbents overlook, not whether they exist
Beachheadif no competitor does X and a subset of users urgently need X, you have a foothold; check who is in itsamesizing the first segment bottom-up

Schlep blindness (2012): the unconscious mind won't let you see ideas that involve painful, tedious work. PG's example is Stripe: thousands of programmers knew online payments were painful, but dealing with banks, fraud and regulation was so off-putting that they built recipe sites instead. His line: "A company is defined by the schleps it will undertake." The upside is less competition, because everyone else was scared off too.

Frighteningly ambitious ideas (2012 (opens in a new tab)): PG lists seven (a new search engine, replace email, replace universities, internet drama, the next Steve Jobs, bring back Moore's law, ongoing diagnosis) and argues the biggest ideas repel you because they seem to threaten your identity. His advice is to approach them obliquely: start with a small piece that could grow.

How YC evaluates them (Friedman)

Jared Friedman's Startup School talk How to Get Startup Ideas (opens in a new tab) (YC Library, transcript on the page) is the most systematic YC treatment.

PartWhat Friedman says (paraphrased)
4 mistakesbelieving you need an amazing idea; jumping on the first idea without a couple of weeks' thought; starting with a solution in search of a problem (YC calls it a SISP); believing ideas are hard to find
Idea scorerate four criteria (adapted from Dalton Caldwell): how big could it be; founder–market fit; how sure you are the problem is big; whether you have a new, important insight
Good signsyou want it yourself; it only recently became possible; successful companies do something similar elsewhere
4 bad filtersrejecting ideas that seem hard to start (schlep), boring, too ambitious, or crowded; he says no-competitor spaces often have no competitors because nobody wants the product
7 recipes(1) what your team is unusually good at, the best recipe, behind about half of YC's most successful companies; (2) what you wish someone would build for you; (3) what you'd work on for 10 years even if it failed; (4) what recent change makes possible; (5) new variants of recent successes, treat with scepticism; (6) crowdsource from people with problems; (7) industries that seem broken

Use this with the idea evaluation and founder–market fit sections of the ideation sheet.

Launch early, then iterate

PG on why speed matters, from The 18 Mistakes That Kill Startups (opens in a new tab) (2006):

MistakePG's pointWhat to do
#8 Slowness in launchingsoftware is always "85% done"; the delays are usually procrastination: fear of users, of judgment, too many things, perfectionismforce yourself to launch something fairly quickly
#9 Launching too earlyPG says launching too slowly has killed about a hundred times more startups, but a bad launch can ruin your reputation with early adoptersship a core that is useful on its own and can grow into the whole project
#10 No specific user in mindyou can't build things users like without understanding thempick one user type and watch them
#5 Obstinacystartups are like science, not an Olympic event: your original plan is probably wrong; but switching idea every week is equally fatalchange direction when each new idea reuses what you built and users get excited about it
#18 Half-hearted effortmost failed founders never quit their day jobsgo full-time once it's worth testing properly

In How to Start a Startup (opens in a new tab) (2005) PG says he nearly added a fourth essential, getting a version 1 out as soon as you can, but left it out because it's implied by making something customers want: the only way to find that is to put a prototype in front of them and refine it.

Michael Seibel on MVPs

Seibel's Startup School talk How to Build an MVP (opens in a new tab) (2023, transcript on the YC Library page) is YC's clearest MVP brief:

PointWhat Seibel says (paraphrased)Apply it by…
Launch, then learnyou only start learning about users when a product is in front of them; surveys and hundreds of interviews are not a substitutethe build–measure–learn loop
Early adopters forgivepeople who talk to startups are used to broken products; the ones who'd leave after one bug never would have tried yourecruiting early adopters, not the mainstream
"Fake Steve Jobs"the first iPhone had no App Store, no video and only 2G; even Jobs iteratedcutting v1 features without guilt
Three traits of good MVPsfast to build, very limited functionality, aimed at a small set of users (first Airbnb: no payments, no map, air beds only, conferences only; first Stripe: manual bank paperwork every night)the types of MVP
Hair on firebuild for customers so desperate they'll use a brick to put out the firetargeting the most acute segment first
Four tricksset a specific deadline (2–6 weeks); write the spec down; cut the spec to what a desperate customer needs; don't fall in love with the MVPscoping and appetite

He closes with a YC refrain: better a hundred people who love the product than a hundred thousand who kind of like it.

Do things that don't scale

The most-cited YC essay, Do Things That Don't Scale (opens in a new tab) (2013). PG says the most common unscalable thing founders have to do at the start is recruit users manually, and that almost all startups are fragile at first.

TacticWhat it meansPG's exampleApply it by…
Recruit users manuallygo out and get them one by one; you can't wait for users to comeAirbnb going door to door in New Yorka named list of 50–100 prospects in launch
Collison installationYC's name for Stripe's habit: when someone agreed to try it, the Collisons said "Right then, give me your laptop" and set them up on the spotStripeonboarding each early user in person or on a call
Delight insanelythe first users should feel signing up was one of their best choices; make up for a buggy product with attentivenessWufoo's hand-written thank-you notespersonal onboarding, founder-run support
Contained firestart in a narrow market so it gets hot before you add logsFacebook at Harvard firstlaunching to one community, one campus, one city
Pull a Merakihardware founders assemble the first units themselvesMeraki routers, Pebble watchesbuilding the first batch by hand
Consult for one userfor B2B, build as if for one customer until it fits perfectly; don't charge by the hourPG's advice to B2B startupsa design partner, not a consulting contract
Be the softwaredo by hand what you'll automate laterStripe's "instant" merchant accounts were set up manually behind the scenesa concierge or Wizard-of-Oz MVP
Skip the Big Launcha launch only needs to produce an initial core of users; PG calls believing otherwise solipsism and laziness–treating launch as a process (launch sheet)
Distrust partnershipsbig-company partnerships rarely give startups much–not waiting on a partner to start growth

PG's summary: startup ideas are vectors, not scalars. Plan the product and the unscalable thing you'll do to start it, which is usually: recruit users manually and give them an overwhelmingly good experience.

Growth as the compass

Startup = Growth (opens in a new tab) (2012) defines a startup as "a company designed to grow fast"; technology, venture funding and an exit are optional. Growth is the compass for almost every decision.

ClaimDetail
three phasesslow or no growth while figuring things out → rapid growth → slowing as a big company (an S-curve)
the one numberthe growth rate; "If you don't know that number, you don't even know if you're doing well or badly."
what to measurerevenue first; active users if not charging yet
YC benchmarks (weekly)"A good growth rate during YC is 5-7% a week. If you can hit 10% a week you're doing exceptionally well. If you can only manage 1%, it's a sign you haven't yet figured out what you're doing."
methodpick a rate you think you can hit and try to hit it every week: it turns the startup into an optimization problem

Weekly rates compound fast: annual multiple =(1+g)52= (1+g)^{52}.

Weekly growth gg×/yearDoubling time
1%1.7×~70 weeks
2%2.8×~35 weeks
5%12.6×~14 weeks
7%33.7×~10 weeks
10%142×~7 weeks

These rates are for the early ascent from a small base; no company keeps 5% a week for years. Use the numbers as a signal of whether you have found something, as PG does, not as a forecast.

Default alive or default dead? (2015)

Default Alive or Default Dead? (opens in a new tab) asks one question of any startup more than 8–9 months old: if expenses stay flat and revenue keeps growing at the recent rate, does it reach profitability on the money in the bank? PG says half the founders he asks don't know.

PointWhat PG says
ask earlyworrying too early is cheap; realizing too late leads to the fatal pinch: default dead + slow growth + not enough time to fix it
don't count on investorsinvestor interest is a function of growth, and fickle; treat fundraising as plan A and write down plan B
spending ≠ growthfast growth usually comes from a product that hits a nerve, not from spending
main cause of default deathhiring too fast, often encouraged by VCs, because kill-or-cure strategies suit a portfolio but not a founder

Run it on your own numbers every month. The companion idea is ramen profitable (2009 (opens in a new tab)): revenue that just covers the founders' living costs. Its main value, PG says, is that it buys time and changes your position with investors.

Money: fundraising and spending

PrincipleWhat it meansSourceApply it by…
Spend as little as possibleone of PG's three essentials (with good people and making something customers want); aim for "cool and cheap, not expensive and impressive"How to Start a Startup (2005)a budget measured in months of runway
Raise when you have tractionRalston: raise once you know the market and customer and have a product being adopted fast; 10% a week for several weeks is impressiveA Guide to Seed Fundraising (opens in a new tab) (2016)waiting for launch metrics before pitching
Raise for a planideally enough to reach profitability; otherwise enough for 12–18 months to the next milestonesamea plan per amount raised
Dilution10% in a seed round is great; most take up to 20%; try not to exceed 25%samemodeling rounds before talking to investors
Size the roundhis 2016 rule of thumb: about $15k per engineer per month all-in, so 5 engineers × 18 months ≈ $1.35Msameredoing the sum with today's salaries
Use standard documentsmost Silicon Valley seed rounds use SAFEs or convertible notes; the valuation at this stage seldom decides successsameYC's post-money SAFE
Raise fast, then workgrowth curves flatten while founders fundraiseYC essential advicea tight fundraising window
Rounds are not milestonesvaluation is not success or even its probabilityYC essential advice (Seibel)celebrating users, not term sheets

Founders and operating

PrincipleWhat it meansSourceApply it by…
Relentlessly resourcefulPG's two-word description of a good founder: not just determined but constantly trying new ways round obstacles; the opposite is "hapless"Relentlessly Resourceful (opens in a new tab) (2009)judging co-founders (and yourself) by how they handle a blocked path
"Animals"hire people who take their work a little too seriouslyHow to Start a Startup (2005)early hiring
Maker's schedulemakers work in half-day blocks; one meeting can blow an afternoonMaker's Schedule, Manager's Schedule (opens in a new tab) (2009)clustering meetings at the end of the day
Don't diestartups die of demoralisation more than money; "Startups rarely die in mid keystroke. So keep typing!"How Not to Die (opens in a new tab) (2007)expecting lows; finding the few users who love you
Make something someone lovesPG cites Paul Buchheit: a few ecstatic users means you're on trackHow Not to Die (2007)growing from a core of fanatics
Don't obsess over the business model earlymaking money is easier than making something great; nearly all startups that make something popular find a way to make moneyBe Good (opens in a new tab) (2008); Why to Not Not Start a Startup (opens in a new tab) (2007)charging early anyway when it's a validation signal (see critiques)
Most reasons not to start are bogusPG lists 16 reasons for reluctance and says which are real (e.g. family to support) and which are not (e.g. knowing nothing about business)Why to Not Not Start a Startup (2007)writing down your own reasons and grading them
Founder modethe advice to "hire good people and give them room" damaged many founders' companies; founders should stay involved in detail, with skip-level contactFounder Mode (opens in a new tab) (2024)(for later) staying close to product as the team grows

Founder Mode (2024) came out of a Brian Chesky talk at a YC event. PG admits nobody yet knows exactly what founder mode consists of; he predicts it will break the rule that a CEO engages only through direct reports. See Chesky and the leadership angle in management.

Sam Altman's playbook

Altman's Startup Playbook (opens in a new tab) (2015) opens: a successful startup needs a great idea (including a great market), a great team, a great product and great execution.

PartKey advice (paraphrased unless quoted)Apply it by…
Ideaexplain it clearly; know who desperately needs it (best case: you); test consumer ideas by launching and enterprise ideas by trying to sell (a letter of intent before code); "The best ideas sound bad but are in fact good"; YC asks how the company will one day be a monopoly, using Thiel's termvalidation
Teammediocre teams don't build great companies; the best founders are unusually responsiveco-founder choice
Product"Here is the secret to success: have a great product." Build a product-improvement engine: talk to users, watch them, fix the worst part, repeat; 5% better a week compoundsthe weekly iteration loop
Diagnostic questionsdo users come back? are they fanatical? would they be truly bummed if you disappeared? do they recommend you unprompted? B2B: at least 10 paying customers?the PMF section
Executiongrowth and momentum; "Never lose momentum"; one growth metric; don't fool yourself with vanity metrics: retention matters as much as acquisitionmetrics
Focus and intensityhis two-word version of operating advice; say no a lotcutting scope

His later essay How to Be Successful (opens in a new tab) (2019) is career-level: 13 points including compound yourself, have almost too much self-belief, think independently, get good at sales, make it easy to take risks, focus, work hard, be bold, be wilful, be hard to compete with, build a network, get rich by owning things, and be internally driven.

Where each piece applies

Pipeline phaseYC advice that matters mostWhere
Ideationorganic ideas, narrow-and-deep wells, schlep blindness, Friedman's score and recipesideation
Validationtalk to users; make something people want; LOI before code for enterprise (Altman)validation
Product design90/10 solution; a core useful on its own; cut the specproduct design
Buildinglaunch now; deadline; don't fall in love with the MVPbuilding the MVP
Launch & iteratedo things that don't scale; weekly growth target; default alivelaunch & iterate
Whole journeywrite code and talk to users; don't dieplaybook

Critiques and limits

  • Survivorship bias. The examples (Airbnb, Stripe, Dropbox, Twitch) are the winners. PG's evidence is what he sees in YC companies, and in How Not to Die (opens in a new tab) (2007) he said YC expected about a third of funded startups to succeed and hoped for half, which is far above base rates for startups in general. Plenty of failed companies also launched early and did things that didn't scale.
  • Incentives. PG himself says kill-or-cure strategies are optimal for VCs because of the portfolio effect, while founders "want above all to survive" (Default Alive, 2015). "Startup = growth" also means: if your business can't grow 5–7% a week, YC's model isn't for you. A profitable small business is a fine outcome; it's just not a YC outcome.
  • Selection effects. YC picks founders it thinks are unusually capable. Advice that works for them (drop everything, move fast, raise quickly) may not transfer.
  • Software bias. Most examples are software and marketplaces. "Launch in two weeks" doesn't apply as-is to hardware, biotech or regulated products (Altman's playbook does note that for hard tech you test by talking to customers and building the smallest subset of the technology).
  • Internal tensions.
    • "Don't worry about the business model" (PG, 2007–2008) vs "it never makes sense to take 80 cents from a customer and then hand them a dollar back" (Buchheit, quoted on the same YC essential-advice page) and Altman's unit economics. Reconcile it by charging early when payment is your validation signal.
    • "Ignore competitors" vs Altman's question about how the company becomes a monopoly.
  • Conflict with Thiel. PG says to err toward markets with competitors and that startups are rarely killed by them; Thiel argues competition is for losers and you should target a market you can monopolise. They agree more than it sounds: both want a small niche you can dominate first.
  • Conflict with Jobs. "Launch embarrassed" sits badly with Jobs's obsession with polish; PG's reconciliation is that the experience of being an early user, not the product, should be insanely great (Do Things That Don't Scale (opens in a new tab)). See Steve Jobs.
  • Founder mode is a hypothesis. PG's 2024 essay is explicit that nobody yet knows what it consists of; it can also become a license to micromanage. Weigh it against extreme ownership and mission command.
  • Dated numbers. The seed-round sums (2016) and growth-rate benchmarks come from a specific era and funding climate.

Takeaways checklist

IDEA
[ ] Problem I (or my team) personally have or know deeply
[ ] Named first user who wants it now, even as a crappy v1
[ ] Checked for schlep/unsexy filters: am I avoiding a
    better but tedious idea?
[ ] Scored: size, founder-market fit, problem certainty,
    new insight
 
MVP
[ ] Deadline set (2-6 weeks); spec written, then cut
[ ] Core is useful on its own and can grow into the whole
[ ] 90/10 solution found for each feature
 
FIRST USERS
[ ] 50-100 prospects listed; recruiting them by hand
[ ] Onboarding each early user personally
[ ] Doing by hand what I'll automate later
[ ] Not waiting on a big launch or a partnership
 
GROWTH AND MONEY
[ ] One growth metric (revenue or active users), weekly
[ ] Weekly target set; missing it treated as alarming
[ ] Default alive or dead? Checked monthly
[ ] Plan B written down in case fundraising fails
[ ] Not hiring ahead of growth
 
FOUNDER
[ ] Most time on code and users
[ ] Meetings clustered; maker time protected
[ ] Expecting the lows; not quitting in them

References