Shrikant Pawade

Nobody Starts From Zero

  • essay
  • startups

What the stories about young founders leave out, what the research actually says about who succeeds, and how to build the thing they all had before you’d heard of them. Including the argument against this essay, and where it’s right.

At two in the morning in Pune, with a loan I can’t yet repay and a browser full of tabs about 19-year-olds getting into Y Combinator, I did the arithmetic that every ambitious person in their twenties eventually does. Elon Musk started Zip2 at 24. Alexandr Wang started Scale AI at 19. Ritesh Agarwal, from a small town in Odisha, launched OYO at 19. The Zepto founders were 19. I am 26, I left a career in music ten months ago to learn to build things with AI, and I am still, by any honest measure, a beginner. The arithmetic said: late.

I’m a pianist by training. I spent fifteen years learning that nothing worth doing happens on the first attempt, and that the people who look effortless on stage have practised in ways you never see. So when the arithmetic said “late,” something else in me said: check what you’re not seeing.

Here is the claim this essay makes, so you can argue with it from the start. Every founder in the young-genius stories had, before day one, some combination of five things: years of a skill, a co-founder or network, someone covering living costs, a hub, and access to capital. I call this the floor. The stories are told with the floor removed. The research on who actually builds successful companies says the floor matters more than age, more than raw talent, and more than courage. And the floor is buildable from nothing, in a year or two, by anyone willing to do it in the right order.

Then, near the end, I’ll make the best case against all of this, because there is one.

Part one: what they actually had

Put the floor back into the stories and they read differently.

Elon Musk, Zip2, 1995, age 24. He had been coding since he was twelve. He had two degrees from the University of Pennsylvania, in physics and economics, and had been admitted to a Stanford PhD. His co-founder was his brother, Kimbal, who had just sold his share of a painting franchise. The family money is disputed: Ashlee Vance’s biography says their father put in about $28,000; Musk says his father’s contribution was roughly ten percent of a later $200,000 round. Either way, the picture is the same. A trained engineer, in Silicon Valley, in the single best year in history to start an internet company, with a sibling as a partner and a family that could write a cheque. Compaq bought Zip2 for $307 million four years later. Musk’s share was $22 million.

Alexandr Wang, Scale AI, 2016, age 19. Born in Los Alamos, New Mexico, to two physicists at the national laboratory. A math and coding competition kid. By seventeen he was a full-time engineer at Quora, where he met his eventual co-founder. Then Addepar. Then one year at MIT. By the time he walked into Y Combinator he had roughly eight years of serious technical work, a professional network in San Francisco, and the experience of being employable at an elite company before he could vote. Nineteen was the age on the form. The floor was a decade deep.

Ritesh Agarwal, OYO, 2013, age 19. His is the story closest to mine, and I used to tell it to myself as proof that you can start from nothing. He grew up in Bissam Cuttack and Titilagarh, in Odisha, far from any hub. In 2012 he launched Oravel Stays, a budget-accommodation listings site. It struggled. Bookings were few, repeat rates were low. What changed everything was the Thiel Fellowship in 2013: a $100,000 grant, a Silicon Valley network, and the credibility of being among the first fellows from Asia. Oravel pivoted into OYO Rooms that May. Agarwal is proof that you can start from far away. He is also proof that someone handed him a floor, in cash, at exactly the right moment.

Patrick Collison, Stripe, 2010, age 21. The part of the story people forget: he had already sold a company. He and his brother John built Auctomatic, went through Y Combinator, and sold it in March 2008 for a figure reported at around $5 million. Patrick was 19. Stripe was founded by two people who were already millionaires with an accelerator network and an exit on their record.

Aadit Palicha and Kaivalya Vohra, Zepto, 2021, age 19. Both grew up in Dubai, both got into Stanford for computer science, both dropped out. Less often mentioned: Zepto was their second company. Their first, KiranaKart, launched in 2020 and failed to find a market. The 19-year-olds who built a quick-commerce giant were 19-year-olds with a Stanford admission, a prior failure under their belt, and Y Combinator among the investors in their first large round.

Jeff Bezos, Amazon, 1994, age 30. Princeton. Senior vice president at the hedge fund D.E. Shaw. And in 1995 his parents invested $245,573 in the company, after he warned them there was a seventy percent chance they’d never see it again. He had a decade of elite work experience and a family that could absorb a quarter-million-dollar loss.

Founder Age on the form What was already there
Musk 24 12 years coding, two degrees, brother as co-founder, family capital, Silicon Valley in 1995
Wang 19 Physicist parents, paid engineering jobs from 17, MIT, SF network
Agarwal 19 A first startup already running, an accelerator, then $100,000 and a network from Thiel
Collison 21 A company already sold for ~$5M, YC alumni
Palicha and Vohra 19 Stanford, a failed first company, YC as investor
Bezos 30 Princeton, D.E. Shaw, $245,573 from parents

None of this diminishes them. Having a floor and doing something with it are different things, and most people with floors do nothing. But if you’re comparing yourself to them without the floor, you are comparing your opening position to their fifth move.

The obvious objection

A sharp reader will say: you picked six founders who had floors. Of course the pattern holds. What about the ones who didn’t?

They exist, and they’re worth looking at, because they tell you what a floor is actually made of.

Jack Ma failed the university entrance exam twice, was rejected from dozens of jobs, and was an English teacher in Hangzhou earning very little. He founded Alibaba in 1999, at 34, in his apartment, with seventeen co-founders who pooled roughly $60,000 between them. No elite degree, no family money, no hub. But look at what he did have: a salaried job that covered rent for years while he tried things, a first internet company before Alibaba, and seventeen people who trusted him enough to put in their savings. The floor was people and a job, not money and pedigree.

Jan Koum grew up in Ukraine and then in Mountain View on food stamps, cleaning a grocery store as a teenager. He didn’t fund WhatsApp from a fortune; his co-founder Brian Acton raised the first $250,000 from former Yahoo colleagues. But by the time of WhatsApp, Koum had spent nine years as an infrastructure engineer at Yahoo. The floor was the skill and the relationships that nine years produce.

Ritesh Agarwal before Thiel. Oravel was failing. He was nineteen, far from home, and the floor arrived from outside. Which is the honest lesson: floors can be handed to you by institutions, and some institutions exist to do exactly that. More on this in Part six.

So the claim is not “nobody succeeds without privilege.” It’s narrower and I think unarguable: nobody succeeds without a floor, the headline founders had far more of one than the headlines admit, and the ones who started with less built it first, usually from a job and from people. If you can find me a founder of a lasting company who had no skill, no co-founder, no income and no network on day one, I’ll revise the essay.

Part two: the age data, which nobody who sells urgency wants you to read

The young-founder story isn’t just incomplete. On the numbers, it’s wrong.

In 2018, Pierre Azoulay, Benjamin Jones, J. Daniel Kim and Javier Miranda got access to U.S. Census Bureau administrative data on 2.7 million founders whose companies went on to hire at least one employee. The mean age at founding was 42. For the one-in-a-thousand fastest-growing companies, the mean age was 45. A 50-year-old founder was roughly 1.8 times more likely to build a top-growth company than a 30-year-old. The result held in high-tech sectors, in entrepreneurial hubs, and when measured by successful exits.

A decade earlier, Vivek Wadhwa’s Kauffman Foundation study, The Anatomy of an Entrepreneur, had surveyed 549 founders in high-growth industries. Average age at launch: 40. Ninety-five percent had bachelor’s degrees. Around seventy percent were married, sixty percent had children. Fewer than one percent came from extremely rich or extremely poor backgrounds; the typical founder was middle class.

Hurun’s Global Unicorn Index for 2022 found that the founders of billion-dollar companies worldwide had started them at 36, on average. Hurun’s list of likely future Indian unicorns put the average founder age at 39.

Even Y Combinator, the institution most associated with the young-founder image, publishes numbers that undercut it. For its Winter 2015 batch the median founder age was 29, with a range from 20 to 66. For Summer 2017 and Winter 2018 the average was just under 30.

The Indian picture is no different once you look at the founders rather than the mythology. Deepinder Goyal started what became Zomato at 25, after IIT Delhi and a job at Bain. Bhavish Aggarwal started Ola at 25, after IIT Bombay and two years at Microsoft Research. Kunal Shah started FreeCharge at 27 and CRED at 35. Sridhar Vembu was around 28 when AdventNet, which became Zoho, began, after a Princeton PhD and years at Qualcomm. Falguni Nayar left a banking career at Kotak to start Nykaa at 50.

What the age data does and doesn’t say

A debater would stop me here. “So I should wait until 45?” No, and the data doesn’t say that either. Age in these studies is a proxy. The variable doing the work is experience. In the Census study, the single strongest predictor of upper-tail success was not age but prior work in the same industry: founders with three or more years in the specific field of their startup hit the top tier at twice the rate of those with none. Older founders succeed more often because they’ve had more time to accumulate that, not because of the birthday.

Which reframes the question from “how old should I be” to “how much of the floor do I have.” A 24-year-old with eight years of engineering, like Wang, has more of it than a 40-year-old switching into a field they’ve never worked in. The data isn’t telling you to wait. It’s telling you what to spend the waiting on.

Why, then, does it feel like everyone successful is nineteen? Because the nineteen-year-olds are the story. A 44-year-old with industry experience building a profitable company is not news. A teenager who dropped out of Stanford is. We’re not seeing the distribution. We’re seeing what gets written about, and what gets written about is selected precisely for being rare.

Part three: is it about money?

This is where I expected the research to confirm my resentment, and it mostly didn’t.

The obvious hypothesis is that rich kids start companies because they can afford to. Erik Hurst and Annamaria Lusardi tested it in 2004 and found something more interesting: across most of the wealth distribution, the relationship between how much money a household has and whether someone in it starts a business is essentially flat. It only turns positive at the very top, above the 95th percentile. For most would-be founders, they argued, the capital needed to start is small enough that money isn’t the binding constraint. Later work by Robert Fairlie and Harry Krashinsky pushed back, finding that housing wealth does raise entry across the distribution, so the question isn’t settled. But even the strongest version of the evidence doesn’t say “you need to be rich.” It says money helps at the margin.

Ross Levine and Yona Rubinstein’s 2017 paper, Smart and Illicit, separated real entrepreneurs, those who incorporate and build firms, from the self-employed in general. The incorporated group did skew toward higher-income families with better-educated mothers. But the traits that predicted it were not wealth as such. They were high scores on learning-aptitude tests combined, oddly, with a history of rule-breaking as teenagers. Smart and willing to break rules. The family background mattered because it produced the skills and the confidence, not because it wrote the cheques.

And when founders are asked where their money actually came from, the answer is overwhelmingly themselves. Kauffman’s analysis of the U.S. Census’s Annual Survey of Entrepreneurs found that about 64 percent of founders relied on personal or family savings as their primary startup capital. Not venture capital. Not rich parents. Savings.

The tension a careful reader will spot

If money barely predicts who starts a company, why am I about to tell you to save twelve months of expenses before you do?

Because Hurst and Lusardi measured entry. Starting is cheap. Anyone can register a company. What money buys is not the right to start but the ability to survive long enough to learn something, and the ability to think clearly while you do. Those are different outcomes, and the evidence on them points the other way, which is the next section.

So the floor is not “be rich.” It’s narrower and more reachable than that: enough skill to build something people want, enough buffer that a bad month doesn’t end the company, and enough people around you to make the first ten customers and the first hire possible. The wealthy founders in the stories had those things handed to them. The rest of us build them. In my case, the skill is becoming fine-tuning small language models and building the evaluation systems around them, the buffer is zero and the people are a cohort I nearly drifted out of. That’s an honest inventory. It’s also a to-do list.

Part four: why starting without a floor fails, and what it does to your brain

Here is the part I wish someone had told me a year ago, when I quit my only source of income to learn Python with no plan for rent.

In 2013, Anandi Mani, Sendhil Mullainathan, Eldar Shafir and Jiaying Zhao published a study in Science called Poverty Impedes Cognitive Function. One of their experiments followed sugarcane farmers in Tamil Nadu through the harvest cycle. The same farmers were tested on cognitive tasks before harvest, when they were broke, and after, when they’d been paid. Post-harvest, they performed significantly better. The researchers ruled out nutrition, time pressure and physical work; the farmers also reported more stress before harvest, but stress alone didn’t account for the gap. Princeton’s press summary put the size of the effect at roughly thirteen IQ points, or the equivalent of losing a full night’s sleep. That number is a journalist’s translation, and later replications have debated the magnitude. But the direction has held up, and anyone who has been broke will recognise it.

Scarcity taxes your bandwidth. Money worry sits in working memory and takes up room that should be used for the problem in front of you. This is why starting a company from a position of financial panic is not brave. It’s cognitively self-sabotaging. You make the decisions that matter most, about product and customers and hiring, with a measurably impaired brain, and you make them under a deadline set by your bank balance rather than by the market.

Airbnb is the usual counterexample, and it’s worth looking at closely because it proves the rule. In 2008, Brian Chesky and Joe Gebbia were carrying credit card debt, Chesky’s around $25,000 and Gebbia’s in the tens of thousands, stored in a binder of the kind kids use for baseball cards. The company was weeks from death. What saved it was a stunt: they designed election-themed cereal boxes, Obama O’s and Cap’n McCain’s, and sold enough at $40 a box to net around $30,000. That caught Paul Graham’s attention, Y Combinator put in $20,000, and Sequoia followed with $600,000 a few months later.

Notice what actually happened. Two trained designers from the Rhode Island School of Design, with a technical co-founder, in San Francisco, were rescued by a lucky stunt and an accelerator that existed specifically to catch people like them. The story is told as “they had no floor and made it.” The accurate version is “they had a skill floor and a location floor, no money floor, nearly died of that, and were saved by an institution.” If you’re reading this from Pune without a YC to catch you, the lesson is not “be like Airbnb.”

Starting without a floor doesn’t produce cheap, informative failure. It produces expensive, uninformative failure. The company dies of money before the idea gets tested, and you learn nothing about the idea. You learn only that you ran out of money, which you knew would happen.

Part five: the hybrid path, and its limits

If quitting everything to start is the worst way, what’s the best?

Joseph Raffiee and Jie Feng answered this in a 2014 paper in the Academy of Management Journal with the plain title Should I Quit My Day Job? They followed a nationally representative cohort of about 12,700 Americans from 1994 to 2008 and compared people who went straight from employment into full-time entrepreneurship with people who did it in stages, keeping the job while starting the business. The staged founders’ ventures had a 33 percent lower hazard of exit. The effect was stronger for people who’d started something before. The mechanism, the authors argued, was learning: the hybrid phase lets you test the business while the job pays for the test.

The pattern shows up again and again when you look at founders who didn’t have a family floor and built one themselves.

Phil Knight started Blue Ribbon Sports, which became Nike, in January 1964. He kept working as a certified public accountant, first at Coopers & Lybrand and then at Price Waterhouse, and then taught accounting at Portland State University. He didn’t go full-time until 1969, five years in, when the company had passed a million dollars in sales and could pay him. The man who told the world to Just Do It spent half a decade doing it on nights and weekends.

Sara Blakely had $5,000 in savings and a job selling fax machines door to door. She kept the job for two years while she developed Spanx: researching hosiery patents, getting rejected by mills, prototyping by hand. She quit when there was something to quit for.

Jack Ma taught English for years while he tried his first internet ventures. Koum spent nine years at Yahoo. Vembu, Goyal, Aggarwal and Shah each had years in a serious job before founding, and each used what the job gave them: the skills, the savings, the colleagues who became co-founders and first investors.

Where the hybrid argument is weaker than it looks

A good debater would attack this section hardest, and they’d have three real points.

First, the study measures exit, and exit isn’t failure. Some of the full-time founders who “exited” sold their companies. The hazard-of-exit number is a survival statistic, not a success statistic.

Second, selection. The paper itself finds that more risk-averse people choose the hybrid route. Perhaps cautious people build cautious companies that survive longer and grow less. The study can’t rule that out.

Third, and most seriously: nobody scales a company on nights and weekends. Paul Graham has argued for years that startups are all-consuming, and the venture-scale outcomes in Part two were built by people working on nothing else. Knight went full-time. Blakely quit. Koum left Yahoo before WhatsApp. If you stay hybrid forever, you will build a side business, not a company.

All three are right, and they sharpen the claim rather than break it. The hybrid path is for the testing phase, not the scaling phase. You keep the floor while you find out whether the idea is real: whether anyone will pay, whether you can build it, whether it holds up after the first ten customers. Then you go full-time, and the evidence from Knight and Blakely is that you’ll know when, because the thing you’re building will be pulling at you harder than the job is. The mistake isn’t going full-time. It’s going full-time before there’s anything to go full-time on.

Part six: what a floor is made of, and how to build one from nothing

A floor has five parts. They go in roughly this order, and you can’t skip the early ones.

1. Income that covers expenses. Not savings yet. Just a month that balances. This is the thing that stops the scarcity tax. It’s also the thing that most people in my position don’t have and keep trying to leap over. The source doesn’t matter. Gigs, a job, freelance work, teaching the thing you already know. Mine, right now, is weekend gigs as a musician, which cover my ₹25,000 a month. In India the most common floor is one nobody writes about: living with your parents. It’s a legitimate floor. Use it without shame if you have it, and don’t pretend you did it alone afterwards.

2. A skill a stranger will pay for. This is the slow part and the only part that truly compounds. The Census data says it plainly: three years of real experience in the field doubles your odds. Wang had eight years of engineering before Scale. Knight knew accounting cold. The question isn’t “what am I passionate about” but “what can I do that someone with a budget currently can’t get done.” It takes one to three years, and it is non-negotiable.

3. A buffer. Three months of expenses, held somewhere you don’t touch. Then six. Then twelve. Three months buys freedom from panic. Twelve months buys the right to try something. This is the part that converts “I want to start a company” from a feeling into an option. Given that most founders fund themselves from savings, it is also, statistically, your most likely source of startup capital.

4. Leverage. Things that work for you while you sleep: code that runs, writing that reaches strangers, a reputation that makes people answer your emails. Publishing your work as you go, however small, is how someone with no network builds one. Every repository and every honest write-up of what broke is a floor tile.

5. People. Two or three who’d build with you, five who’d introduce you to someone, one or two who’d write a small cheque. Nobody in the table above did it alone. The brother, the Quora colleague, Acton’s Yahoo friends, Ma’s seventeen. Every job, every cohort, every public post is a place to meet them.

Floors you can apply for

Here the debater has a fair point that the first draft of this essay missed. If institutions handed Agarwal and the Airbnb founders their floors, why build slowly when you can apply?

You should apply. Y Combinator, the Thiel Fellowship, which now pays $250,000, and in India programs like Entrepreneur First and Antler, some of which pay a stipend while you find a co-founder and an idea. Grants, fellowships, research positions, paid accelerators. My own cohort is currently offering compute, datasets and payment to people who’ll lead a small-model project. These are floors in a box.

But read the selection criteria honestly. The Thiel Fellowship’s early classes came mostly from Harvard, MIT and Princeton. YC’s median founder is 29 with work experience. Every one of these programs selects for people who already have part of the floor, usually the skill part. They accelerate; they don’t originate. So the sequence doesn’t change. Build the skill, show it in public, and apply to everything while you do. The applications cost nothing, and the first one you get into collapses a year of the slow path.

When is the floor enough?

Without an answer to this, the whole essay is a license to prepare forever. So here is one, and I’m holding myself to it.

The floor is enough when income covers expenses, when there are twelve months of expenses saved, when you’ve been paid by a stranger for the skill at least three times, when you’ve shipped something people other than you have used, and when there are two people who’d build with you. Five conditions. Write the date you expect to meet them. Mine is written down.

Part seven: the case against this essay

I promised to make it, so here it is at full strength.

This is a rationalization for not starting. Survivorship bias cuts both ways. For every founder who waited and then won, there are thousands who waited and never started, and they’re as invisible as the floorless failures. “Build the floor” is what the cautious tell themselves. The founders in Part one had floors, yes, but they also moved the instant they could, and plenty of people with better floors never did.

The window is closing. Musk’s floor included 1995. If AI in 2026 is the internet in 1995, then a year spent saving twelve months of expenses is a year of the window gone, and the prepared person who starts later loses to the less prepared person who’s already in the market.

The research is American. Almost every study cited is U.S. data. India’s demographics, family structures and capital markets are different, and a 26-year-old in Pune in 2026 is not a 42-year-old in the Census Bureau’s files.

These are the strongest objections I can find, and I want to answer them honestly rather than win.

On the first: it’s correct, and it’s the failure mode this essay has to defend against. The defence is the five conditions above, with a date. A floor with a deadline is preparation. A floor without one is hiding. And the test of which one you’re doing is behavioural: are you asking strangers for money every quarter, regardless of readiness? If not, you’re hiding, and no amount of saving changes that.

On the second: windows are longer than they look from inside them. Google was founded in 1998, Facebook in 2004, both well after 1995 and both by people who’d spent the intervening years getting very good at something. In AI right now, the application layer is years from settled. And the person building the floor inside the field, learning to fine-tune and evaluate models rather than saving money as an accountant, is in the window the whole time. The window argument is an argument for building the skill part of the floor fast, not for skipping the floor.

On the third: fair, and partly answered by the Indian founders in Part two, who follow the same pattern, and by the fact that the sharpest piece of evidence in this essay, the sugarcane study, was done in Tamil Nadu. But the honest answer is that the Indian data is thinner, and you should weight the argument accordingly. I’d rather tell you that than pretend.

Part eight: on urgency, which you should keep

I don’t want to end by telling a 26-year-old to be patient. I’m not patient, and I don’t think the people who build important things are either.

Urgency is right. The feeling that you’re late is useful as fuel. It gets you out of bed for the two hours of study before the day starts. The mistake is using it as a steering wheel: letting the feeling decide the sequence, so you skip the floor and jump, and then the company dies of money while you’re still learning the thing it needed you to know.

“Fail fast” is good advice with a condition attached. The failure has to be cheap and it has to teach you something. A startup launched from debt with no runway fails fast, but it fails of the wrong thing, and it costs you the next two years of recovery. The cheap, informative failures are smaller and less dramatic: a service you sell to three people and nobody renews. A model you fine-tune that doesn’t beat the baseline. A post nobody reads. An offer you make and the customer says no to. Each one teaches you something a venture-backed failure would have taught you for a hundred times the price, and you can run one a quarter while you build the floor.

So here is what I’m doing, and what I’d suggest to anyone doing the two-in-the-morning arithmetic.

Weekends, I play. That’s the floor, and I’ve stopped being embarrassed by it. Weekdays, two hours before everything else, I’m working through the fundamentals of how neural networks learn, one video and one notebook at a time, with a specific target: in thirty days, a small language model that I fine-tuned myself, on data I made, that measurably beats the base model. After that, a small paid service to the accountants I already know, to start the buffer and get my first rep at asking someone for money. I’m applying to the programs that hand out floors, and I’m posting the work as I go, including the parts that fail. After that, the next rung.

The startup is still the plan. It will be built on a floor, not on a feeling. And when it fails, because the first one usually does, it will fail of something interesting.

The founders in the headlines were not early. They were prepared, and then they were quick. The order matters. Get prepared. Then be quick.

Sources

Founder histories

Research

← All posts