Founders tend to picture launch day as the moment a switch flips and users arrive. It rarely works that way. In his essay "Do Things that Don't Scale," Paul Graham argues that the most common unscalable thing founders do at the start is recruit users manually, and that this is not a temporary embarrassment but the actual job.

Recruit, don't broadcast

The essay's best-known example is Stripe: in its early days, when someone agreed to try it, the founders said "Right then, give me your laptop" and set them up on the spot, a tactic dubbed the Collison installation. Waiting for people to find you feels safer, but it confuses passivity with marketing. A direct ask - a DM, an email, a demo at a meetup - gets you a yes or a no you can learn from today.

Manual recruitment also protects you from a quiet failure mode: shipping something "for everyone" that no one in particular wants. When you have to name the next ten people you will ask, you are forced to decide exactly who this is for.

Treat early users like consultants

Graham's second point is that early users deserve an unreasonable level of care. The experience you can give ten people is qualitatively different from what you can give ten thousand: you can watch them use the product, fix their problem the same day, and sometimes write the feature they ask for by hand before automating it. That attention buys two things money can't: people who genuinely love the product, and a stream of specific, grounded information about what it should become.

How to know it's working

The essay suggests judging an early startup by weekly growth rather than absolute numbers - a small base growing 5-10% a week compounds fast. But the softer test matters as much: are the users you recruited coming back, and would they be genuinely disappointed if you shut down? If yes, the scalable channels will have something real to amplify later. If no, more traffic only tells more people a weak story.

The phase ends on its own. You stop doing things that don't scale when demand starts breaking them, and not before. Until then, recruiting by hand and over-serving the people who say yes is not a detour around the real work - it is the real work.

Further reading: Do Things that Don't Scale - Paul Graham