--- title: Your first users are a search problem description: Most people have never been among a product's first ten users, so your first ones won't come from convincing the crowd. They come from finding the few who love being early or have a burning problem. Notes on Y Combinator's How To Get Your First Users, where Ankit Gupta swaps the minimum viable product for a minimum evolvable one. date: 2026-10-19 tags: startups, first customers, product launch, Y Combinator --- Your first users won't come from convincing people. They come from finding the few who already want to be first. Ask yourself how many products you use today that you were among the first ten users of. For most people the answer is zero, which is Ankit Gupta's point: finding your first users is more of a search problem than a persuasion problem. So your first version only has to survive contact with a tiny group of people who might actually try it, and then change fast. The video is [How To Get Your First Users](https://www.youtube.com/watch?v=0kARDVL2nZg), five and a half minutes on Y Combinator's channel, an episode of their Main Function series published in January. Ankit is a general partner at YC. His update to the classic playbook: you don't just need a minimum viable product. You need a minimum evolvable product. I'm writing this one for myself first. I ship side projects all the time, and [mulewatch](/projects/mulewatch.html) and [vibetidy](/projects/vibetidy.html) are on npm. Both are free and open source, so nobody pays for them. If you're also on the near side of that line, with something built and no first client yet, this note is for you too. ## Almost nobody wants to be first Most people aren't early adopters. Almost no one wants to be a startup's first paying customer, and yet every great product still finds a few people willing to take that leap. Ankit names two kinds. Some people just love being early: his colleague Gustav spent years at Airbnb trying products from startups and bringing them into the company. Others have a problem so pressing that they'll give any new product a shot if it looks like it could make their life easier.

The lineFinding your first users is more of a search problem than a persuasion problem.

The few you're looking for, in a crowd that will never be first. The proportions are schematic: the video gives no share.
## Ankit was somebody's first customer Ankit has been the second kind himself. When his team needed to ship their first inference API, they wanted it out fast and didn't want to deal with billing or a public endpoint. Within three days he found a startup whose product solved that, and paid it. He was their first customer. Their size or reputation didn't matter. His team's problem did. So they took a chance on the startup, and it delivered. That's the whole lesson in one story. You're looking for the Gustavs and the Ankits of the world: the ones who try new things, or who have a burning need you can solve.

  1. NeedShip the team's first inference API

    And ship it fast

  2. SkipBilling and a public endpoint

    Nobody wanted to deal with those

  3. Find and payA startup that solved it

    Within three days

  4. FirstAnkit is their first customer

    They took the chance. It delivered.

Didn't matterThe startup's sizeIts reputation

MatteredThe team's problem

Found and paid within three days, as Ankit tells it. A burning problem made him an early adopter.
## Charge real money from the start The search framing has a few counterintuitive consequences, and the first is about price. Early adopters and people with a burning problem are rarely price sensitive. The goal isn't revenue, it's feedback, and paying customers give sharper feedback than free users ever will. His version: you're more likely to get feedback from an angry customer paying a lot of money than from a nobody who isn't willing to pay. It's the rule my own projects break. mulewatch and vibetidy are both free.

Free user

Won't pay

"A nobody who isn't willing to pay"

Less likely to tell you anything

vs

Paying customer

Pays a lot, and is angry

More likely to tell you what's wrong

Sharper feedback than free users ever give

The goalNot revenue: feedback. Early adopters are rarely price sensitive.

The same product, two kinds of early user. The bars are schematic: Ankit compares them, he doesn't measure them.
## Knock on doors, then launch early Second: targeted, personal outreach. The way you find these people probably doesn't look like the way you reach everyone else. A billboard is way less likely to reach them than a targeted cold email or a knock on their door. Third: launch early, something YC has preached from the beginning. In the early days you don't know much about who these users are, so you want to engineer a wide surface area for them to find you. So you work both ends. One by one toward the people you can guess, and a launch for the ones you can't.

Reach out

  • BillboardWay less likely to reach them
  • Targeted cold emailTo someone with the problem
  • A knock on the doorPersonal, one at a time

Launch early

Your product, with a wide surface for people you can't name yet

One by one where you can guess, wide open where you can't. The dashed people are the ones you haven't met.
## Watch them like an anthropologist Fourth: study your early users closely. Ankit says you should be like an anthropologist who has discovered a hidden civilization. How do they make decisions? Why would they make the strange choice to trust you? You want to understand how they think and what they want. Fifth: experiment fast and don't fear churn. Run constant experiments on pricing, landing pages, onboarding, features, everything. Talk to your early users and try to make them love the product, but don't stress if you lose one. If you annoy someone, you can usually fix it, because the relationship is personal. And if they churn, that's fine too: plenty of others haven't even heard of you yet. That's an advantage a startup has over a big company. When you run a bad experiment, no one writes about it. You're fighting relevance, not headlines.

Field notes

Your early users

  1. How do they make decisions?
  2. Why would they make the strange choice to trust you?
  3. How do they think?
  4. What do they want?

and

Experiments

Constant, on everything

  • Pricing
  • Landing pages
  • Onboarding
  • Features

ChurnAnnoyed someone? It's personal, so you can usually fix it. Lost them? Plenty haven't heard of you yet.

Bad testAt a startup no one writes about it. You're fighting relevance, not headlines.

What to learn about them, and what to keep changing. The runs are schematic: the video names the areas, not a pace.
## Who pays shapes who you build for All of this can shape who you're building for in the first place. Most people don't pay for a lot of consumer apps. Ankit's own personal software spend is about $150 a month, total. Meanwhile his corporate card carries several tools that each cost more than that alone. In the AI era that gap matters. Ads often don't cover AI costs, and a subscription has to squeeze into an already small personal budget. Consumer companies will still be made, but it's why many AI founders start by selling to prosumers or businesses, or target users like doctors who have high advertising value.

His personal software, all of itabout $150 a month

One tool on the corporate cardmore, alone

Another onemore, alone

And anothermore, alone

So many AI founders start with

ProsumersBusinessesHigh ad value, like doctors

WhyAds often don't cover AI costs, and a subscription has to fit a small personal budget.

Only the $150 is Ankit's number. The corporate bars are schematic: past his whole personal budget, by an amount he doesn't give.
## Start as an amoeba Here's the bigger point: your early users don't just give you feedback. They end up steering how your product evolves. Ankit's analogy, and he asks you to bear with him: think of a startup as a phylogenetic tree. The root is an amoeba, and the leaves are complex multicellular organisms like humans or dogs. Almost every product you buy has run that evolutionary process, from amoeba to the maturity of a human or a dog: millions of users, a refined sales pitch, clear value. Early startups are amoebas. They have just the basic functions needed to get exposed to outside pressure. From there, the founders run an evolutionary search through the tree of possible directions.
The tree from Ankit's analogy. Which leaf you reach is schematic: that's the point, nobody knows it at the root.
## The Roadster was a search Tesla is his case study, and its amoeba was the Roadster. The usual story is that Tesla needed a high-margin product to fund the capital investment for the Model S, and later the Model 3 and Y. Ankit says that's probably true, but offers a second reading: Tesla was searching for early adopters. It wanted the people crazy enough to buy an impractical $150,000 car that didn't go very far, didn't fit much, couldn't charge publicly anywhere, and looked strange. And the product then evolved along what those people wanted. Why does the Model Y, a mass-market car, have a faster 0 to 60 than a Lamborghini and better tech than a BMW, but worse suspension and comfort than a Toyota? Turns out early adopters cared much more about tech and acceleration than comfort. Would a mass-market car designed in a vacuum do 0 to 60 in under three seconds? Probably not. It's an outcome of the search Tesla ran. If early adopters had been willing to pay $150,000 for a slow, plush car, Ankit bets Tesla's cars would look very different today.

The amoeba

Roadster, $150,000

Didn't go very far

Didn't fit much

Couldn't charge publicly anywhere

Looked strange

Its buyers cared about

Tech

Acceleration

Comfort

Mass market, years later

Model Y

0 to 60 faster than a Lamborghini

Better tech than a BMW

Worse suspension and comfort than a Toyota

Ankit's betHad they paid $150,000 for a slow, plush car, Tesla's cars would look very different today.

Path-dependent: the car followed what its first buyers wanted. The trait bars are schematic, the comparisons are his.
## Ship something that can evolve This is the algorithm YC helps founders run, and it's why Ankit says your first version shouldn't just be a minimum viable product. It should be a minimum evolvable product: something simple that can respond to market pressure, survive contact with early users, and adapt fast to whatever they push it toward. He calls that freeing. The product will change a lot, so it doesn't have to be perfect from the start. What it becomes depends on where you begin and who you begin it with.

Where you beginSomething simpleA minimum evolvable product

Who you begin it withYour first usersThe ones you searched for

What it becomesWhatever they push it toward

Minimum viableNot enough on its own

Minimum evolvableSurvives contact with early users, adapts fast, doesn't have to be perfect

His last sentence, laid out. Both inputs are choices you make before the first user shows up.
## Try it The whole video is five rules and one picture. Here they are as a checklist, in the order Ankit gives them, for anyone still before a first client, me included.

  1. 01Search

    The burning problem, or the people who love being early

  2. 02Reach out

    A targeted cold email, a knock on the door

  3. 03Charge

    Real money, from the start, for the feedback

  4. 04Launch early

    A wide surface for the ones you can't guess

  5. 05Study

    How they decide, and why they trusted you

  6. 06Evolve

    Experiment on everything, don't fear churn

Six steps, the first one most of us skip. The order is the video's.
If you want more: - [How To Get Your First Users](https://www.youtube.com/watch?v=0kARDVL2nZg), the video, 5 minutes 41 seconds - [Do Things That Don't Scale](https://paulgraham.com/ds.html), Paul Graham's essay on recruiting users by hand, the older half of the same advice - [mulewatch](/projects/mulewatch.html) and [vibetidy](/projects/vibetidy.html), my two free tools on npm, still waiting for a first client If you're before your first client too, skip the billboard this week. Name the burning problem, find a few people who have it, and write to them one by one. Then tell me who answered.