The sentence nobody quotes

Do things that don't scale is the most repeated advice in startups and the most misread: Stripe's laptop trick worked because Y Combinator had already handed them a building full of buyers, and Paul Graham says so in the same essay

Paul Graham's Do Things that Don't Scaleis thirteen years old and still the most quoted piece of startup advice in existence. You have read it, or you have read forty tweets that were made out of it. The headline is simple and correct: "The most common unscalable thing founders have to do at the start is to recruit users manually. Nearly all startups have to."

So you did it. You wrote the individual messages. You offered to get on a call and set the thing up yourself. Nothing happened, and the reasonable conclusion was that you are bad at this.

I want to point at a sentence in that essay that almost never survives the quoting.

The famous part

The most repeated passage is about Stripe. Graham writes that "at YC we use the term 'Collison installation' for the technique they invented", and describes it: when anyone agreed to try Stripe, the founders would say "Right then, give me your laptop" and set them up on the spot.

It is a great story. It is concrete, it is a little bit rude, and it makes an excellent tweet. It is also the bit everybody remembers, which is why founders come away believing the lesson is about aggression and hustle.

The part that gets dropped

Two paragraphs of the same essay explain where those laptops were. "Startups building things for other startups have a big pool of potential users in the other companies we've funded, and none took better advantage of it than Stripe." And then, plainer still: it was "one of many unforeseen advantages of the YC model (and specifically of making YC big) that B2B startups now have an instant market of hundreds of other startups ready at hand."

Graham is not hiding this. He states it directly, in the same essay, about the same company. It just never makes it into the quote, because "give me your laptop" is a story and "they were standing in a building full of pre-qualified buyers who had a reason to take their call" is a footnote.

Read together, the advice is not do the unscalable thing. It is: get access to a pool of buyers, then do the unscalable thing to each of them individually. Two steps, and the famous one is the second.

This is why the tactic keeps failing when it is lifted out. A founder with no users copies step two and does it into an empty room. The manual effort is identical. The Collison installation without the room is just a stranger asking for your laptop.

Dropbox ran the same play twice, and the video is not the lesson

The other tactic founders copy from this era is the Dropbox demo video, usually remembered as "make a good video and it will spread". The actual sequence is more instructive.

Drew Houston posted the first video to Hacker News. By his own account the strategy was to post the demo there and hope Paul Graham and Jessica Livingston saw it; it hit the top of the site and Dropbox was in the next YC batch. He was not broadcasting. He was posting to a forum he read, aimed at two specific people.

The second video went to Digg and Reddit, and he salted it with in-jokes for those audiences: references to XKCD, to the Obama campaign, to whatever those communities were laughing at that month. Houston has said the reasoning was simply about where they hang out. By his telling the waiting list went from 5,000 to 75,000.

Notice what the transferable skill actually is. It is not video production. It is knowing a community well enough to make its in-jokes correctly, which is not something you can do in an afternoon and is not something you can buy. He had the membership before he had the product. The video was step two again.

What the essay quietly assumes

Graham says of Airbnb that the unscalable work "consisted of going door to door in New York, recruiting new users and helping existing ones improve their listings". Existing ones. There were already listings in New York to go and improve, and the essay does not spend a word on how they got there, because that is not what the essay is about.

That is the reading error, and it is not Graham's fault. He is writing for founders who already have a trickle and need to be told to go and touch it by hand. It gets read by founders who have nothing at all, and to that reader it sounds like the trickle is what the hustle produces. It is the other way round.

The question that is actually transferable

Not "what unscalable thing should I do?" but "where are my buyers already assembled, and do I have standing to speak there?" Three tests, in order:

Is there a room at all?Somewhere your buyer already gathers and talks: a subreddit, a Discord, a conference floor, a specific hashtag, a newsletter's reply-all. If your buyer does not congregate anywhere you can name, that is a real finding and it rules out a lot of advice in one move.

Do you have standing? Being able to post is not standing. Standing is being someone whose message does not read as an intrusion, which usually means having been there first. This is the expensive part and the part that cannot be skipped, which is why finding the threads where you have something to add beats finding the threads with the most traffic.

Can you do it at the scale of one? If you can only make it work by sending the same thing to two hundred people, you are not doing the unscalable thing, you are doing the scalable thing badly. The whole mechanism of the Collison installation is that the person on the other end could tell it was for them.

The uncomfortable version

Sometimes the honest answer to the first test is no. Your buyers do not gather anywhere reachable, they are not on the forums, and there is no room to be a member of. That is not a personal failing and it is not fixed by more hustle. It is a signal that manual recruiting is the wrong first channel for this product, and that a different one, or a different buyer, is where the effort should go.

Almost nobody will tell you this, because the advice industry is built on tactics and a tactic has to be presented as universally applicable to be worth selling. Do things that don't scale is excellent advice. It is excellent advice conditional on having somewhere to do it, and the condition is printed right there in the source text.

So before you rewrite the outreach message again, go and answer the first question. The order to work through when nothing is landing is written up in what to fix first when nobody is using your app, and if you are not sure your buyer congregates anywhere at all, the questions that rule channels out will get you to an answer faster than trying them. Our own tool is free for that first pass and needs no account: give it your URL and see which rooms it says you are already standing near.