Patrick Collison

The pattern they actually represent

When somebody agreed to try Stripe, Patrick Collison did not send them a link. He asked for their laptop.

Paul Graham named the practice in his 2013 essay Do Things that Don't Scale, and recorded that the brothers invented it: inside Y Combinator, the term was the Collison installation. Graham sets the two versions of the same moment side by side. The more diffident founder asks "Will you try our beta?" and, hearing yes, says "Great, we'll send you a link." The Collison brothers "weren't going to wait." Theirs was "Right then, give me your laptop", and they set the person up on the spot.

His own account of that period is less ceremonious. The early setup work was done entirely by hand, he said in 2012, for a user base of two people who were both good friends of his.

Notice what is missing from that scene. No pitch. No follow-up sequence. No decision for anybody to render, because there is nothing left to decide — the work is already being done. Both versions of the moment begin with the identical yes. Only one of them ends with a working account, and the difference between them is not persuasion. It is who performs the next ten minutes of labour.

They did not go hunting for strangers either. The first users came out of the room the brothers were already standing in: the first ten or twenty, he said in that same interview, were all companies funded by the same programme that funded Stripe. People who would take the call anyway.

The pattern, stated plainly: when interest appears, do not convert it into a request. Convert it into work you perform in front of the person, on their machine, while the interest is still in the room.

The blockage it speaks to

In January 2011 Stripe had about five customers. It was already backed by Peter Thiel, Elon Musk and Sequoia Capital. Collison knows the number precisely because all five of them emailed him on the same morning, after the product changed its name overnight and nobody remembered to tell them.

Read that again before you decide your own numbers are embarrassing.

The reader this is for has something that works, or close enough. What they do not have is one person who has been asked to use it. The avoidance rarely arrives as fear. It arrives as queue management: the empty state, the error copy, the onboarding, the second integration. The list of things that must be true first regenerates faster than you can clear it.

Which is why perfectionism and premature optimisation are not siblings of the avoidance. They are its socially acceptable surface. An hour on the empty state is an hour in which nobody can say no, and that is what the hour is really buying.

The correction is not bravery. It substitutes the act. A request for a decision hands somebody a verdict to render on you, later, in private, with no obligation to tell you the result. An install is service: you are doing work for them, in the room, now. Nobody is asked to endorse you. Somebody is asked to lend you a keyboard for ten minutes. Different act, different fear attached.

Three moves you can steal

Delete "I'll send you a link" from your vocabulary and rehearse the replacement. The substitution has to fire in real time, under mild social pressure, so decide the sentence before the conversation starts: great, let's do it now, give me your laptop. The link is where an enthusiastic yes turns into a task on somebody else's list, ranked below everything else on it.

Install for the room you are already standing in. Write down twenty named people who would take your call today — your last team, your cohort, the group chat — and install for five of them this week. That removes prospecting, the part of selling that actually feels like selling. What is left is showing something you built to people who already know you.

Do the install, hand the keyboard back, then go quiet for five minutes. Ask for one real task and say nothing while they attempt it. Write down every pause, every re-read, every wrong click. Not their opinion — their behaviour. You have been optimising a guess, and five silent minutes replace the guess with a ranked list of defects.

Where the pattern breaks

The pattern has an expiry date, and Collison sets it himself. In 2018 he bounded the advice: Stripe has salespeople and marketing people who are immensely important to the organisation now, and he was "specifically talking about strategies to get started in the early days."

The destination is on the record — $1.4 trillion of payment volume in 2024, half the Fortune 100 as customers. Nobody hand-installs at that size, and the product was never built to need it: the 2011 launch coverage already noted it "can be set up in minutes." The install was never the distribution model. It was the instrument that told them what the self-serve flow had to remove.

What to do this week

Pick one person off that list of twenty and do one install before Friday. Not a demo. Not a call. Not a deck. Their machine, your hands, their account live at the end of it.

Then take the five silent minutes and fill a page with what they did, in their words rather than yours. If nobody on the list is reachable before Friday, the list is too short. Write ten more names.

The bar for the product is not "ready". The bar is "works once, for this person, in this room". If something breaks in front of you, that is not the failure case. It is the most valuable thing you will collect all week, and it arrived in ninety uncomfortable seconds rather than six months of silence.