some thoughts on paul graham's "do things that don't scale"

i recently reread paul graham's essay do things that don't scale, and it reminded me of a lot of things i've experienced working on growth for startups.

i've worked on different products, from AI tools to crypto platforms, and one thing i've noticed is that getting your first users is usually a lot messier than people think.

everyone sees the final numbers. the views, signups, users, and revenue. but nobody really sees how much manual work goes into getting there.

when i first started doing growth, i thought the hardest part was coming up with a great idea for a campaign. something creative enough to go viral.

but after spending hundreds of hours reaching out to creators, writing posts, testing different angles, and distributing content, i started looking at it differently.

sometimes a post gets 100 views. another gets 100,000. you can use the same account, talk about the same product, and still get completely different results.

and the only way i've found to understand why is to keep testing.

the part nobody sees

one of my favorite examples from paul graham's essay is how stripe got its first users.

when someone agreed to try stripe, the founders didn't just send them a link. they asked for their laptop and helped them set everything up right there.

airbnb did something similar. the founders went door to door in new york, meeting hosts and helping them improve their listings.

looking at those companies today, it's easy to forget how they started.

and i think the same thing happens with growth.

people see a startup getting hundreds of thousands of views and assume there must be some clever growth hack behind it.

sometimes there is a good idea behind the campaign. but there's also someone spending hours finding the right creators, messaging people, rewriting copy, and trying different ways to distribute it.

i've done plenty of that myself.

some of my best-performing campaigns involved work that would have been difficult to automate. i had to talk to people, understand their audiences, and figure out what would actually make them interested in sharing something.

it's not particularly exciting work, but it teaches you a lot.

distribution is something you learn by doing

i think there's a difference between knowing growth strategies and actually knowing how to get attention.

you can read hundreds of posts about distribution, study successful launches, and learn every marketing framework.

but until you've tried getting people to care about something yourself, you're missing a big part of it.

for example, i've learned that a small change in how you describe a product can completely change how people react to it.

sometimes the product isn't the problem. you're just talking about it in a way nobody finds interesting.

other times, the content is good, but you're showing it to the wrong audience.

you don't really learn these things from a dashboard. you learn them by publishing, talking to people, seeing what happens, and trying again.

that's probably the part of growth i enjoy most.

every new product forces me to understand a different market, different people, and different reasons why someone might care.

being small is actually useful

another thing i liked about paul graham's essay is his point about how much attention small startups can give their first users.

big companies can't personally talk to every customer. small startups can.

and i think founders sometimes forget that's an advantage.

when you're early, you can message your users directly, ask what they think, help them personally, and even build something specifically for one customer.

you don't need a complicated process for everything.

the same applies to distribution. you can experiment with ideas that a bigger company would probably never approve. you can change your messaging in an afternoon, test something the next morning, and learn from it immediately.

you have very little to lose by trying things.

what i took away from it

the biggest thing i've learned from working on growth is that there isn't one playbook that works for every startup.

something that worked for an AI product might completely fail for a crypto platform. a campaign that gets millions of views might bring very few actual users.

that's why i still spend so much time doing things manually, even when there are tools that could make parts of the process faster.

i want to understand why something works before i try to scale it.

and that's what i like about do things that don't scale.

paul graham wrote it back in 2013, but it still feels relevant today, especially when everyone is trying to automate everything with AI.

i'm still learning, still testing, and still getting things wrong.

but the more startups i work with, the more i realize how valuable it is to actually do the work yourself before trying to build a system around it.