Paul Graham's startup essays are some of the most widely read advice for founders, but there are dozens of them and each runs long. Here are core learnings from twenty-two of his startup essays that you can finish in about 25 minutes. If you want to explore any of them in detail, the original links are below as well.
Paul Graham co-founded Viaweb, sold it to Yahoo, and then started Y Combinator, which has funded Airbnb, Stripe, Dropbox, and thousands of others. In between he wrote the essays that shaped how a generation of founders thinks about ideas, growth, fundraising, and why most startups fail. The catch is that there are dozens of them and each one rewards a slow read. Below is a run of his startup essays as carousels: the argument up front, readable in seconds, with a link to the full piece under each.
How to Start a Startup
A startup needs three things: good founders, a product people want, and to spend as little as possible. Everything else, ideas, hiring, learning from customers, fundraising, follows from getting those right, and slow growth plus frugality quietly improve the odds.
How to Start a Startup
paulgraham.com
Startups
41 min read → 60 sec
A framework for starting a startup built on three requirements, strong founders, products users want, and low spending. It covers ideas, hiring, customer learning, fundraising, and why slow growth and frugality improve the odds of success.
Startup success needs good people, something customers want, and very low spending.
A brilliant initial idea is not required.
2 / 20
Good founders can fix bad ideas. Bad ideas cannot fix bad founders.
3 / 20
Technology startups need technical founders who understand users.
Most failures come from weak customer understanding.
4 / 20
Build the product by learning from users.
1
Release prototypes fast.
2
Watch direct user reactions.
3
Refine from what users do.
4
Skip long planning.
5 / 20
Startup ideas are weak on their own.
Many companies change plans en route, and VCs reject NDA pitches.
6 / 20
Google built a billion-dollar business with better search, link ranking, and simple pages with ads.
7 / 20
A 10% improvement in ease of use can double sales.
Normal users find software intimidating.
8 / 20
Niche and low-end markets offer the best startup odds.
Simple cheap products often rise to beat incumbents.
9 / 20
Hire obsessive people.
1
Test smartness.
2
Test execution.
3
Test whether you can stand them.
10 / 20
“
I don't know.
11 / 20
“
Maybe you're right.
12 / 20
“
I don't understand x well enough.
13 / 20
Projects reveal talent better than interviews.
Startups usually form through friends and school networks.
14 / 20
2–4
People in the ideal founding team.
Solo founders lack support. Larger groups form factions.
15 / 20
Business expertise matters less than technical judgment and customer focus.
Founders can learn basic business.
16 / 20
Most failed startups die because customers do not want the product.
The best startup ideas come from noticing a problem you have firsthand, then starting with a small group who wants it solved badly. Organic ideas beat made-up ones because you can tell the difference between what people say they want and what they actually do.
How to Get Startup Ideas
paulgraham.com
Startups
30 min read → 60 sec
Startup ideas come from noticing urgent problems, usually ones founders know firsthand, then starting with a small group that wants a solution badly. The piece explains why organic ideas beat made-up ones and how to spot them.
Get startup ideas by noticing real problems, especially ones you have yourself.
The strongest ideas are wanted by founders, buildable by them, and underrated by others.
02 — 20
Trying to think up startup ideas produces plausible bad ideas.
Those ideas often fool founders into building them.
03 — 20
The best startup ideas pass three tests.
01
Founders want them.
02
Founders can build them.
03
Others underrate them.
04 — 20
Choose a narrow, deep well of urgent demand.
Skip broad, shallow interest.
05 — 20
A startup idea is probably bad if no one will use a crappy version one right now.
06 — 20
Live in the future, then build what is missing.
Or build what seems interesting.
07 — 20
People at the leading edge of fast-changing fields notice opportunities first.
They already live in the future.
08 — 20
Learning to hack is not necessary.
It is sufficient for reaching a fast-changing frontier.
09 — 20
Look for things that seem missing.
Turn off the filter asking whether they could become big companies.
10 — 20
Question the status quo and notice what chafes you.
Annoyance can signal future demand.
11 — 20
Projects that seem like toys often yield good startup ideas.
Apple, Microsoft, BackRub, and Facebook did.
12 — 20
Good startup ideas usually start in small markets.
They need a path into much larger ones.
13 — 20
Microsoft's Altair Basic and Facebook at Harvard started with a few thousand users.
Those users wanted them badly.
14 — 20
Paths out of an initial niche are hard to see.
Airbnb began with renting floor space during conventions.
15 — 20
A good idea often feels late.
Startups rarely die from competitors unless rivals have strong lock-in.
16 — 20
A crowded market is a good sign if you know what incumbents are overlooking.
17 — 20
Turn off the schlep filter and unsexy filter.
Stripe succeeded partly because others avoided payments.
Ideas are strongest when they grow out of problems you personally face, especially needs that have only just become fixable. Young founders often win by building something that looks like a toy long before it looks like a company.
Organic Startup Ideas
paulgraham.com
Startups
4 min read → 45 sec
Startup ideas are strongest when they grow from problems founders personally face, especially needs that have only recently become fixable. Young founders often win by building projects that look like toys before they look like companies.
The best startup ideas come from fixing something missing in your own life
Organic ideas beat guesses about what other users might need
2 / 13
Startup ideas fall into two types: organic ideas and made-up ideas aimed at other users.
The strongest startups are usually organic, not guesses.
3 / 13
Young founders are bad at predicting demand, but strong at spotting new problems they personally face.
4 / 13
Apple began because Steve Wozniak wanted a computer and could build one himself.
Wozniak did not think the Apple I was the start of a company.
5 / 13
Bill Gates built the first Altair Basic for himself. Google began as something Larry and Sergey would use.
6 / 13
In 2004, Harvard students still used a printed Facebook. An online version was an obvious fix.
Mark Zuckerberg built it as a project, not a company plan.
7 / 13
Organic ideas often start as side projects or toys, not as serious company plans.
8 / 13
Products dismissed as toys can signal overlooked opportunities.
Early microcomputers, planes, and cars were all dismissed this way.
9 / 13
An unmet need that is just becoming fixable is the most valuable startup opportunity.
10 / 13
Viaweb was built for direct marketers the founders did not understand or belong to.
The founders understood customers poorly without becoming users themselves.
11 / 13
Founders should become users of their own product.
Capable hackers keep picking weak startup ideas, and the fix is learning to choose problems customers will pay to solve. Attachment, status motives, and a fear of competition all push founders away from the better markets.
Why Smart People Have Bad Ideas
paulgraham.com
Startups
13 min read → 65 sec
Why capable hackers choose weak startup ideas, and why the fix is learning to pick problems customers will pay to solve. It argues that attachment, status motives, and fear of competition push founders away from better markets.
Smart founders often start with bad ideas because they pick the first one and stick with it
The still life effect makes time invested feel like proof an idea is good
2 / 15
Artix failed because art galleries did not want websites, while Viaweb succeeded by serving people who did
The market's desire matters more than the founder's skill
3 / 15
Microsoft's first company was Traf-o-data, which did not do as well as Microsoft
Even the best founders start with ideas that do not work
4 / 15
Changing a startup idea is harder than changing a name, so founders must consciously discount attachment
The still life effect turns sunk time into false validation
5 / 15
Startups should aim first to make money, then maybe be cool, not the reverse
6 / 15
A startup with divided loyalties is unlikely to make money, because profit must be the first priority
7 / 15
The right test is whether people will pay, and whether this is what they will pay most for
8 / 15
Founders often choose weak markets to avoid competition, even when stronger markets are beatable
VC-backed startups in 1995 were less formidable than they looked, because many were weak at building software
9 / 15
Unpleasant work often pays better, because work people enjoy attracts more supply and lower returns
10 / 15
Many young founders are good at solving assigned problems but weak at choosing valuable problems
11 / 15
Choosing what to build can be learned, and the rule is make something people want
12 / 15
There is no single thing called business, only tasks like selling, pricing, support, incorporation, and fundraising
13 / 15
Reading The Wall Street Journal for a week should yield two or three startup ideas from visible unsolved problems
Startups die from a recurring set of mistakes: the wrong cofounders, bad hires, botched fundraising, launching too slowly, and neglecting users. Underneath almost all of them is the same failure, not making something people want.
The 18 Mistakes That Kill Startups
paulgraham.com
Startups
24 min read → 70 sec
A 2006 essay on 18 recurring startup mistakes, from weak cofounder choices and bad hiring to poor fundraising, slow launches, and neglecting users. The central claim is that most failure comes from not making something people want.
Startup failure funnels through one cause: not making something users want
The 18 mistakes are ways founders stop themselves from doing that
02 / 20
Single founders are at a major disadvantage
Startups are too hard to do alone
03 / 20
Half-hearted effort kills many startups
Most founders do not try hard enough full-time
04 / 20
Startups do better in hubs like Silicon Valley, Boston, Seattle, Austin, Denver, and New York
05 / 20
Choosing a small obscure niche to avoid competition usually means avoiding good ideas
06 / 20
Solve a specific problem you understand, not a derivative or vague idea
Derivative ideas are weaker than solving an unsolved problem you have felt yourself
07 / 20
Plans should change when evidence improves, but repeated restarts from scratch are a bad sign
08 / 20
Non-programmer founders often kill startups by hiring weak programmers they cannot accurately evaluate
09 / 20
Choosing the wrong platform can destroy a startup, as Windows and Java applets did for many
10 / 20
Launching slowly blocks learning. Founders only fully understand ideas after users react to them
11 / 20
Launching too early can damage reputation
Release the smallest useful core that can expand
12 / 20
If founders are not the target user, they need specific users and measured responses, not guesses
13 / 20
Funding should buy enough runway to reach the next visible step, often 50% to 100% more than expected
14 / 20
Hiring too many people burns cash and slows progress
Early hires should write code or get users
15 / 20
Large VC rounds add pressure, slow fundraising, reduce flexibility, and make pivots harder
16 / 20
Early startups should prioritize product over business model
Making something people want is harder than monetizing it
17 / 20
At least one hacker has to do sales, deals, or user acquisition. Ideas alone do not get bought
Release early, ship constant improvements, listen to users, and fear the right risks. The lessons that matter most are the counterintuitive ones: stay determined, be skeptical of deals, and keep growth as the thing you actually optimize.
The Hardest Lessons for Startups to Learn
paulgraham.com
Startups
20 min read → 70 sec
A set of counterintuitive startup lessons on releasing early, shipping constant improvements, listening to users, fearing the right risks, staying determined, and treating deals skeptically while focusing on growth.
Startups learn fastest by releasing minimal version 1 early, then improving from user reactions
The main job is making users happy, because growth comes from converting casual visitors into users
2 / 19
Users hate bugs, but they usually accept a minimal version 1 if more is coming soon
3 / 19
Many startups died from releasing too slowly. None died from releasing a stable minimal product too early
4 / 19
Wufoo, a form-builder, released before finishing the database, drew 83,000 visitors, and learned to reduce Flash
5 / 19
The first release is a shakedown cruise that exposes bad ideas, broken teams, and other major flaws
6 / 19
Releasing creates urgency. Live problems feel alarming instead of merely interesting
7 / 19
Startups should improve the product every day or two with one unit of user benefit
8 / 19
Constant product improvement is also marketing. Users notice progress and talk about it
9 / 19
Sites should explain what they do fast and put the strongest value on the front page
Most visitors arrive ready to hit Back, so the site must stop them immediately
10 / 19
Unknown startups are more dangerous than Google. They are as hungry and less visible
11 / 19
Microsoft and Apple changed direction after customers showed them what business to build
12 / 19
The biggest startup killers
1
Internal disputes between founders
2
Inertia and moving too slowly
3
Ignoring users
13 / 19
Determination matters more than intelligence in founders, but it must stay flexible rather than stubborn
14 / 19
Founders who seem ambivalent get less attention from investors, acquirers, and other partners
15 / 19
Treat deals as unlikely until closed, and keep running the company toward user growth
16 / 19
A startup's only real leverage in deals is genuinely not needing the deal
17 / 19
A startup is best seen as a way to make a living faster, not merely a way to get rich
Ramen profitability means earning just enough to cover the founders’ living costs. It is a survival tactic, not the goal, but it buys time, cuts your dependence on investors, and improves both your leverage and your morale.
Ramen Profitable
paulgraham.com
Startups
8 min read → 70 sec
Ramen profitability means a startup earns enough to cover founders' living costs, which buys time and reduces dependence on investors. It can improve leverage, morale, and focus, but it is a survival tactic rather than the end goal.
Ramen profitable means earning just enough to cover founders' living expenses
It buys time by removing the need to raise money or shut down
02 — 19
Traditional profitability
5 years, $50M
Time and capital needed to reach full profitability
Ramen profitability can arrive in 2 months at $3000 a month
03 — 19
A startup making $3000 a month has not succeeded
But it no longer needs outside money to survive
04 — 19
Many software startups can reach ramen profitability because founders' living costs are often the main expense
05 — 19
Once investors know you need money, they can stall and push you into worse terms
06 — 19
Small-scale profitability shows investors three things
01
Customers pay
02
Founders build what people want
03
Expenses stay low
07 — 19
Many funded startups fail for three reasons
01
Customers will not pay
02
Founders solve the wrong problem
03
Spending outruns revenue
08 — 19
Covering the founders' own expenses makes the company feel real
It turns survival into the default state
09 — 19
The hardest part of startups is bearing responsibility
Reducing that weight improves survival odds
10 — 19
A startup that reaches ramen profitability may be more likely to succeed than not
11 — 19
Raising money can cut productivity to a third and distract founders for months
12 — 19
When fundraising becomes the most urgent problem, founders think about investors instead of the product
13 — 19
Ramen profitability does not mean never taking investment
Few startups succeed without it
14 — 19
Ramen profitability does not require charging in the final business model
Google first made money licensing search to Yahoo
15 — 19
Joe Kraus's rule to charge from the start can help unfocused founders, but it is too constraining for some startups
16 — 19
It is easy to make $3000 a month consulting
Founders can mistake a consulting business for a startup
17 — 19
Ramen profitability is not the goal. Growth is the goal, and consulting is the main trap
In a short pitch, explain the product fast, show real user demand, keep the slides clean, and project confidence without pretending to a maturity you don’t have. The goal is to be understood quickly, not to sound impressive.
How to Present to Investors
paulgraham.com
Startups
12 min read → 55 sec
Advice for early-stage startup founders on presenting to investors in a short pitch. It focuses on explaining the product quickly, showing user demand, using clear slides, and projecting confidence without pretending maturity.
A strong investor pitch does two things fast. It explains what you built and why users want it.
Venture capital’s incentives push funds to grow large, make oversized bets, and demand high valuations. Those economics are what produce slow decisions, intrusive control, and exits that often hurt the founders and the startup.
A Unified Theory of VC Suckage
paulgraham.com
Startups
6 min read → 75 sec
Venture capital incentives push funds to grow large, make oversized bets, and demand high valuations. Those economics drive slow decisions, intrusive control, and exits that often hurt founders and startups.
VC behavior follows from fund economics, not personality
Capable young programmers should take startups more seriously than they do. The upside is high, failure is survivable, and most of the standard objections turn out to be weaker than they sound.
Why to Not Not Start a Startup
paulgraham.com
Startups
27 min read → 65 sec
An argument that capable young programmers should take startups more seriously, because the upside is high, failure is survivable, and many common objections are weaker than they seem.
Starting a startup beats a cubicle job for capable hackers, even when it fails
Most reasons people avoid startups are weak
2 / 20
Y Combinator's first 2005 batch
50%
success rate, 0% terrible experiences
3 / 20
Kiko failed after a year, sold on eBay for $250,000
Its founders started Justin.TV
4 / 20
Y Combinator would feel safe aiming for a 25% success rate, above the quoted 10% figure
5 / 20
People hesitate because several mixed reasons blur together
Reluctance should be analyzed as a checklist
6 / 20
You are old enough to start a startup once you respond to challenges like an adult
7 / 20
If you are too inexperienced to start a startup, starting one is the fastest cure
8 / 20
Determination is probably the best predictor of success
Many hackers underestimate their own
9 / 20
If you worry you are not smart enough, you probably are
Most startups need effort more than brilliance
10 / 20
You do not need business knowledge at first
Making something people want matters most
11 / 20
Investors strongly prefer cofounders. Lacking one is important enough to move or change ideas
12 / 20
70%
of the average Y Combinator idea changes in three months
Initial ideas matter less than founders
13 / 20
The best source of startup ideas is a real need in your own life, even if it seems narrow
14 / 20
There is no fixed limit to startups
Human wants keep expanding and markets adapt to absorb value
15 / 20
If you have a family, a consulting business or early startup job lowers risk more than founding
16 / 20
Startup failure carries little stigma in America
Big company managers preferred failed founders over big company employees
17 / 20
The strongest startup founders are often uncertain hackers, not gung-ho business types
18 / 20
A regular job is a powerful default, but startups may mark a historic shift in wealth creation
The mid-twenties may be the sweet spot for starting a company. Co-founders, low burn, being in a startup hub, and staying focused on users all matter far more than credentials do.
A Student's Guide to Startups
paulgraham.com
Startups
27 min read → 75 sec
A case for when students should start startups, why the mid-twenties may be the sweet spot, and what young founders gain and lack. It argues that co-founders, low burn, startup hubs, and user focus matter more than credentials.
Most students should wait until after graduation to start a startup
The best founder age is probably mid-twenties, between student advantages and work experience
02 / 20
003
Young founders have five main advantages over older ones
Stamina, poverty, rootlessness, colleagues, and ignorance
03 / 20
004
Long hours are universal in successful startups, and younger founders need them because they are less efficient
04 / 20
005
Low burn rate matters because most startups fail by running out of money before recovering from mistakes
05 / 20
006
Poor founders often build cheaper products because their own cost instincts match where technology is heading
06 / 20
007
The scarcest startup ingredient is co-founders, and school is the best place to find them
07 / 20
008
Successful startups are almost never started by one person. They usually begin with two people deciding to try
08 / 20
009
By age 30, many strong potential co-founders are tied down by jobs and harder to recruit
09 / 20
010
The biggest weakness of student founders is building class projects instead of solving real user problems
10 / 20
011
Work experience mainly removes childish habits, especially the flake reflex and indifference to user needs
11 / 20
012
Customers judge the distance remaining to what they need, not the effort you put in
12 / 20
013
Graduation removes the built-in excuse of being a student, which makes failure feel more real
13 / 20
014
Peer opinion drives founders more day to day than distant wealth, so social pressure changes outcomes
14 / 20
015
Startup success is such a close call that small extra forces can push a company into failure
15 / 20
016
Startup opportunities have historically increased over time, so waiting rarely means missing the last chance
16 / 20
The most useful startup preparation in college
INDEX
001
Work around startups to see how they operate
002
Choose co-founders who are builders, not just talkers
003
Build things for real users, not class projects
17 / 20
018
If you work before founding, avoid writing startup code while employed, or investors may reject the company
18 / 20
019
Google appears to be a startup black hole, because people like working there too much to leave
Cheap web startups shift power away from employers and investors and toward young founders. Early-stage acquisitions and user-focused building make starting a company a stronger bet than taking a traditional job, even given the early risk.
Hiring is Obsolete
paulgraham.com
Startups
20 min read → 60 sec
Cheap Web startups shift power from employers and investors toward young founders. Early-stage acquisitions, user-focused building, and high early-career risk make starting a company a stronger path than traditional hiring.
Young founders can bypass employers and be valued directly by users
Web startups start on $10,000, so investors matter far less
2 / 20
Large organizations systematically undervalue exceptional 22 year olds because they judge toward the mean
3 / 20
The founders of Yahoo, Google, and Microsoft averaged age 24
4 / 20
Early twenties are the best time for extreme career risk. Startup upside outweighs failure costs
5 / 20
Big companies should buy very early startups. It beats separate hiring and in-house product development
6 / 20
Most successful startups exit by acquisition, often mainly to acquire the team
7 / 20
$2M to $3M
A startup bought after six months is mostly a hiring bonus
8 / 20
Big companies are better at extracting value from existing products than creating new ones
9 / 20
Internal turf blocks new products. Web apps threatened Microsoft's desktop business
10 / 20
Disruptive technologies are developed by disruptive people, who rarely gain power inside big companies
11 / 20
Many startups exploring one idea beat one in-house team
Riskier variation finds better products
12 / 20
Startup pay is tied to success, so founders work harder than salaried product teams
13 / 20
Developers in startups learn more from direct sales and support than from focus groups
14 / 20
Acquirers should buy startups earlier, before VCs inflate prices into hundreds of millions
15 / 20
Cheaper startups reduce investor power, so founders can resist giving control to outside CEOs
16 / 20
Google's founders delayed hiring a CEO for a year and chose a PhD in computer science
17 / 20
Every manager asked at Yahoo, Google, Amazon, Cisco, and Microsoft preferred the failed founder candidate
18 / 20
Inexperience usually means building something users do not want, not lacking technical skill or financial discipline
Early startups are most productive when they drop the corporate costume and focus on real work. Messy offices, odd hours, and tiny teams routinely beat polished professionalism, and larger companies could learn from it.
Learning from Founders
paulgraham.com
Startups
4 min read → 55 sec
A case that early startups are most productive when they ignore corporate appearances and focus on real work. Messy offices, odd hours, and tiny teams can outperform polished professionalism and offer a model larger companies can study.
Startups do their most productive work earliest, when the team is tiny
Early startups look unlike corporate professionalism, yet may be the most productive part of the economy
02 / 15
003
Apple's earliest phase was unusually potent because 100% of employees were Steve Jobs or Steve Wozniak
03 / 15
004
Real startup work happens in messy, odd hours, not clean offices during standard business hours
04 / 15
005
Actual programming often happened at 2 in the morning, by badly dressed people in junk-strewn offices
05 / 15
006
Founders often stage professionalism for visitors, cleaning offices, dressing up, and filling conventional office hours
06 / 15
007
Looking productive often reduces real productivity, because effort shifts from performance to appearances
07 / 15
008
Conventional business imagery, suits, conference tables, PowerPoint, and thick reports, misrepresents productive work
08 / 15
009
The fastest quarter-mile car came from removing parts added only to make the car look fast
09 / 15
010
Suits do not help people think better, and many executives likely do their best thinking off the clock
10 / 15
011
Startup work mixes panic and deep thought, with some founders thinking as clearly as on a quiet Sunday morning
11 / 15
012
Even investors often fail to recognize real productivity when it does not match corporate expectations
12 / 15
013
Founders can feel like impostors because high-performance startups do not look like business is supposed to look
13 / 15
014
Big companies cannot copy everything startups do, but they can use them as a model
Outsiders often produce better work than insiders. Being marginal lets you take risks, work on small projects, make things directly, and use the Internet to route around the establishment.
The Power of the Marginal
paulgraham.com
Startups
26 min read → 70 sec
An argument for why outsiders often produce better work than insiders, and how risk, small projects, direct making, and the Internet help marginal people beat established elites.
Great new ideas often come from outsiders, not insiders
Marginal conditions lower the activation energy to tinker
02 / 18
003
Insider systems select the wrong people, reward caution, and burden work with expectations
Outsiders win by taking risks and working directly with the medium
03 / 18
004
California's mild climate created marginal space that helped garage startups like Hewlett-Packard, Apple, and Google
04 / 18
005
In 1976, outsiders could start computer companies because hardware costs had only recently fallen enough
The timing mattered more than the garage
05 / 18
Insider projects fail through six mechanisms
INDEX
001
Bad selection of people
002
Excessive scope
003
Risk aversion
004
Excessive seriousness
005
Vested interests
06 / 18
007
Every elite test can be passed two ways: by real ability or by hacking the test
07 / 18
008
Hard science PhD admissions are fairly honest, while undergraduate admissions are much more hackable
08 / 18
009
Teacher-practitioner overlap is high in math and physics, middling in medicine and law, and low in business and art
09 / 18
010
Corrupt tests can become anti-tests, filtering out the people they should select
Big companies get blindsided because hierarchies test for the wrong qualities
10 / 18
011
Outsiders have little to lose, so they can risk stupid ideas that insiders avoid to protect status
11 / 18
012
Delegation weakens work that benefits from one mind
Direct contact with the medium generates ideas
12 / 18
013
Startups often beat big companies because one person may design, build, and sell with customer needs in one head
13 / 18
014
Cheap, lightweight, small projects spread faster, evolve faster, and teach faster than large ones
14 / 18
015
New platforms are usually dismissed as unfit for real work, then prove more capable than expected
15 / 18
016
The Internet makes audiences more liquid, letting outsiders gain the audience that once belonged to elites
16 / 18
Complaints that you are unqualified or inappropriate often signal that outsider work threatens insider territory
A more powerful language can be a startup’s secret weapon. Viaweb used Lisp to ship faster and outbuild competitors who were stuck with slower-moving mainstream tools.
Beating the Averages
paulgraham.com
Startups
19 min read → 70 sec
A case for using more powerful programming languages in startups, built around Viaweb's use of Lisp to ship faster, outbuild competitors, and exploit the slow-moving habits of mainstream programmers.
Lisp gave Viaweb a real competitive edge over 20 to 30 rivals
Faster development, lower costs, features competitors could not match
02 / 19
Viaweb, started in summer 1995, let end users build online stores through Web pages
It was, as far as he knew, the first Web-based application
03 / 19
Startups that do what average startups do should expect to fail
Startup survival is well below 50%
04 / 19
Server-based software lets startups choose any language, so the best choice is the most powerful one
05 / 19
A startup should ignore what other companies use and choose technology only by what works best
06 / 19
Web applications magnify rapid development because you can release software the minute it is done
07 / 19
Viaweb could sometimes copy a competitor's announced feature within a day or two
08 / 19
20 to 25%
of the Viaweb editor source code was macros
09 / 19
Lisp macros matter because Lisp code is made of traversable data structures that programs can rewrite
10 / 19
Viaweb had about 20 to 30 competitors, but none of their software could compete with it
11 / 19
Customers cared that Viaweb worked well, not that it was written in Lisp
12 / 19
Viaweb grew rapidly from 1996 to Yahoo's purchase
01
About 70 stores in 1996
02
500 stores in 1997
03
1070 stores by Yahoo's purchase
13 / 19
Programming languages differ in practical power, even if they are all Turing equivalent
14 / 19
For application software, using anything but the most powerful reasonably efficient language is usually a mistake
15 / 19
Programmers using weaker languages can see what they lack below them, but not what they miss above them
This is the Blub paradox
16 / 19
Programming practice often lags hardware by 10 to 20 years because languages are habits of mind
17 / 19
Competitor job listings reveal technical strength. Oracle was safest, Perl or Python was scarier, Lisp was most dangerous
Server-based software will replace much desktop software because it is easier for users and better for small teams. Continuous releases, subscriptions, and browser delivery all hand the advantage to startups.
The Other Road Ahead
paulgraham.com
Startups
50 min read → 90 sec
An argument that server-based software will replace much desktop software because it is easier for users and better for small teams. It ties continuous releases, subscriptions, and browser delivery to startup advantage and weaker platform control.
The next generation of software will run on servers, not desktops
Web-based apps are easier, cheaper, more mobile, and more reliable for users
02 / 18
Viaweb switched to server-based software in 1995, and Yahoo Store later reached 14,000 users
The shift from desktop to Web proved viable at scale
03 / 18
Most users care more about convenience than control
Browser-based software wins if it requires less work
04 / 18
Purely Web-based software removes friction for users
01
No installation
02
No operating system conflicts
03
No paid upgrade shocks
04
Access your data from any client
05
Collaboration built in naturally
05 / 18
The idea of "your computer" is being replaced by "your data"
Data should be reachable from any client, anywhere
06 / 18
Server-hosted data is safer from crashes, theft, and viruses
Professionals manage backups and security, not end users
07 / 18
With one shared server version, bugs can be fixed immediately
Viaweb rarely had even 10 known bugs at any time
08 / 18
Desktop firms ship one or two big releases yearly
Viaweb often shipped three to five releases a day
09 / 18
Small incremental releases reduce compound bugs and make fresh bugs easier to reproduce
Continuous deployment turns bug fixing into a fast, tight loop
10 / 18
Close contact between support and developers turns support into bug detection, QA, and product feedback
Support becomes a real-time sensor for what users need
11 / 18
Watching real user behavior
60% to 90%
One test-drive change raised completion rate
Observing users guides design and growth decisions
12 / 18
A Web application is a whole operating environment
You control monitoring, recovery, backups, networking, and hardware choices
13 / 18
Server-based software can use multiple languages for different tasks
Viaweb used C, C++, Perl, and Lisp in one application
A startup hub forms when rich investors and technical talent cluster around a top university, in a place young nerds actually want to live. Buildings and government plans matter less than organic founder networks and time.
How to Be Silicon Valley
paulgraham.com
Startups
15 min read → 70 sec
A startup hub forms when rich investors and technical talent cluster around a top university in a place young nerds want to live. Buildings and government plans matter less than organic founder networks, local personality, and time.
A startup hub needs two limiting reagents: rich people and nerds
Silicon Valley is made by founders and investors, not buildings or parks
02 / 17
003
If the right ten thousand people moved from Silicon Valley to Buffalo, Buffalo would become Silicon Valley
The hub is the people, not the place
03 / 17
004
US towns become startup hubs if and only if they have both rich people and nerds
Miami has rich people but few nerds. Pittsburgh has nerds but lacks investors
04 / 17
005
Government cannot replace startup investors
Bureaucrats lack domain experience, incentives, advice, and connections
05 / 17
006
Technology parks fail because the key startup stage is three founders in an apartment
Wherever a startup is when it gets funded, it will usually stay
06 / 17
007
A top university in an attractive, young-feeling town is the seed of a startup hub
07 / 17
008
Within the US, there are no technology hubs without first-rate universities or computer science departments
08 / 17
$600M
Cost to build a world-class faculty by hiring 200 top researchers with $3 million bonuses each
09 / 17
Towns that attract nerds need four qualities
INDEX
001
Personality, not generic sprawl
002
Density, not suburbs
003
Old-style walkable neighborhoods
004
Limits on large developments
10 / 17
011
Nerds prefer places like Berkeley or Boulder, not glamour hubs like New York or Los Angeles
11 / 17
012
Startup hubs need young-feeling towns with intact centers, because young people do not want suburbs
12 / 17
013
High-tech US cities are also liberal because liberal places tolerate odd ideas
13 / 17
014
Startup hubs grow organically because startups beget startups, and the cycle takes about five years
14 / 17
015
Government is the biggest developer of all, so you do not build a silicon valley, you let one grow
15 / 17
016
Silicon Valley traces back to Shockley Semiconductor in 1956 and Fairchild Semiconductor in 1957
Startup clusters form more easily in America because immigration, wealth, civil liberties, universities, flexible labor, funding, and market size all reinforce one another. Other countries could compete by improving those same conditions.
Why Startups Condense in America
paulgraham.com
Startups
20 min read → 85 sec
Startup clusters form more easily in America because immigration, wealth, civil liberties, universities, labor flexibility, funding, and market size all reinforce one another. Other countries could compete by improving those conditions.
Startup clusters form where a great university sits near a town smart people like
America combines ten conditions that make it the easiest place for startups
02 / 17
A startup hub must attract smart, ambitious people from everywhere
Closed immigration policies block it
03 / 17
Police states can build factories, but suppressing dissent also suppresses the imagination startups need
04 / 17
Countries that flatten university quality lose the concentrated excellence that attracts talent and seeds startups
Private universities compete more naturally; central control pulls universities toward the mean
05 / 17
Rigid firing rules hurt startups most because every early employee must perform well
Higher job security tends to correlate with lower performance across industries and countries
06 / 17
When work is seen as economic exchange rather than lifelong employment, starting a company feels less risky
07 / 17
Larval startups break rules by necessity, so strict enforcement of zoning and incorporation rules kills them early
08 / 17
Germany's early 1990s $20,000 incorporation requirement would have blocked companies like Apple and Viaweb
09 / 17
Small countries force startups to internationalize immediately
US startups can begin with a 300 million person home market
10 / 17
Startup hubs compound because successful founders become angel investors
That cycle usually takes about five years
11 / 17
America's loose career routing helps startups because people can choose paths late and change direction easily
12 / 17
The strongest American advantages for startups
01
Immigration that attracts talent from everywhere
02
Rich infrastructure built over decades
03
Civil liberties that protect dissent and imagination
04
Top universities that concentrate excellence
05
Venture funding networks that recycle success
13 / 17
Government intervention can run ports efficiently, but it cannot coax startups into existence
14 / 17
India has the talent for a rival hub, but poverty may delay startups until baseline prosperity rises
15 / 17
A rival to Silicon Valley could beat America with smarter immigration, lower capital gains taxes, and better urban design
Y Combinator began less as a money play or a charity than as a leverage hack: small interventions that unlock startups which otherwise would never have existed.
Why YC
paulgraham.com
Startups
2 min read → 45 sec
Paul Graham says Y Combinator began less as a money play or charity project than as a leverage hack, using small interventions to unlock startups that otherwise would not exist.
Y Combinator started as a leverage hack
Apply small force at the right point to unlock startups that otherwise would not exist
2 / 14
Thousands of smart people could start companies but do not
A relatively small intervention at the right point can unlock a stream of new startups
3 / 14
YC did not start mainly to make money
Its average returns were unknowable for years
4 / 14
YC did not start mainly to help young founders, though helping them was appealing
5 / 14
Startups are mostly schleps. Founders rarely start them mainly for fun
6 / 14
A hacker's motive can be amoral optimization
Neither selfishness nor virtue, but finding the right tweak to make a system run better
7 / 14
The world's economy is an open-source system that can run more efficiently with the right tweak
8 / 14
YC also satisfied a long-running desire to work with Rtm and Trevor again after Viaweb
9 / 14
Paul Graham kept looking after Viaweb for something he, Rtm, and Trevor could build together
10 / 14
Viaweb was started explicitly to make money and escape living from one freelance project to the next
11 / 14
Viaweb was sometimes fun, but mostly was not fun and was not designed for fun
12 / 14
Startups are a good thing. This makes YC partly virtuous even if that was not the main motive
The shift from climbing corporate ladders to building startups changed how people get paid and take risk. Safe jobs and promotion paths carried a hidden economic value that standard statistics never captured.
After the Ladder
paulgraham.com
Business
3 min read → 60 sec
The shift from climbing corporate ladders to building startups changed how people get paid and take risk. Safe jobs and promotion paths carried hidden economic value that standard statistics and balance sheets did not capture.
The corporate ladder gave way to startups selling to big companies
The new model is more liquid and efficient, but financially less different than it looks
2 / 12
Economic inequality looks larger than it is because statistics ignore the value of safe jobs
A secure, hard-to-lose position is worth money, but that wealth never appears in the data
3 / 12
A corporate ladder position had real financial value because it promised future high-paying roles
The ladder's hidden value resembled corporate goodwill: it created expected future earnings
4 / 12
Big companies once avoided firing people and promoted from within largely by seniority
5 / 12
A sinecure is effectively an annuity, and trading safe jobs is a common form of corruption
6 / 12
Socialist countries also had meaningful wealth gaps because powerful bureaucrats held secure seniority-based jobs
The value of a safe position transcends economic systems
7 / 12
Corporate balance sheets ignored the implicit debt owed to employees expecting later rewards
8 / 12
Takeovers since the 1980s helped kill the corporate ladder by making long internal careers unreliable
Corporate raiders profited partly by stripping firms that carried these unrecorded obligations
9 / 12
A company full of vice presidents may be repaying earlier work, not merely indulging corruption
10 / 12
Treating jobs as rewards often turned good engineers into bad managers and increased office politics
Open source and blogging suggest people do better work on projects they care about, in flexible settings, with ideas rising from the bottom. Those same forces quietly weaken traditional offices, hierarchy, and employment.
What Business Can Learn from Open Source
paulgraham.com
Business
18 min read → 70 sec
Open source and blogging suggest that people do better work on projects they care about, in flexible environments, with ideas rising from the bottom. The essay argues these forces weaken traditional offices, hierarchy, and employment.
Open source and blogging reveal three forces reshaping business
American impatience produces strong software and movies but weak cars and cities. The contrast is between speed-driven, hackable work and the fields where design, craftsmanship, and long-term taste matter more.
Made in USA
paulgraham.com
Business
8 min read → 80 sec
American impatience helps produce strong software and movies but weak cars and cities. The piece contrasts speed-driven, hackable work with fields where design, craftsmanship, and long-term taste matter more.
American impatience makes the country strong in software and movies, weak in cars and cities
Speed wins in malleable mediums, loses where taste and craftsmanship matter
02 / 16
003
Software and movies reward fast, iterative work because both are malleable mediums
You can reshape them cheaply after seeing what works
03 / 16
004
Good software starts with a fast prototype that reveals better ideas
Working slowly often perfects an initial mistaken idea instead of improving the product
04 / 16
005
Movies, like software, end up imperfect and different from the original plan, yet still work
05 / 16
006
Cars and cities reward taste, craftsmanship, and detail, which American culture prizes less than speed
06 / 16
007
Physical products face hard constraints, so dramatic hacks matter less than precise design
07 / 16
008
American cars are badly designed because marketing people, not designers, shape them
Executives treat a car's projected image as more important than its actual design quality
08 / 16
009
Japanese companies building cars in the US weaken the claim that American car problems come from workers
Japanese cars are better because Japanese culture prizes design and craftsmanship
09 / 16
010
American products often stop at crude usefulness or add ornament instead of improving design quality
American houses follow the same pattern, from flimsy boxes to larger McMansions with expensive fittings
10 / 16
011
Successful software companies tend to be run by programmers, and films are largely controlled by directors
11 / 16
012
Focus groups only help in the short term because customers ultimately follow buyers with better taste
12 / 16
013
Apple's iPod succeeded over Sony because great designers satisfy needs that focus groups cannot find
Good design wins long term, and Apple shows America can pair speed with design obsession
13 / 16
014
American exurbs are depressing because developers build piecemeal projects, not whole towns shaped by market discipline
14 / 16
015
The just-do-it model generates wealth and innovation because speed helps create new things first
That is twenty-two of Paul Graham’s startup essays. Read end to end they are hours of dense argument, scanned here in about 25 minutes.
The through-line is consistent. Make something people want. Start from a real problem you have firsthand, not an idea you talked yourself into. Keep spending low so you keep your options open, treat fundraising and pitching as processes rather than talents, and remember that most failure traces back to building something nobody actually needed. None of it is a shortcut, and Graham never claimed it was.
Save the carousels you want to come back to, and when an essay earns your full attention, the link to the original is under each one. The point was never to replace the essays. It is to let you decide, in a few minutes, which ones deserve the full read.
