Do Things that Don't Scale

Claude

Summary. Paul Graham argues that early-stage startups almost always need to do things that don't scale: recruiting users one by one, providing lavishly personal customer service, and manually chasing growth before any process can be automated. He cites Airbnb's founders going door-to-door to photograph listings and Stripe's "Collison installation" (installing the product in person for early users) as proof that manual, unscalable effort is often the engine that later makes scale possible. He also debunks the myth that good startups simply take off on their own, and suggests starting in a narrow market you can dominate before expanding.

On the note. The essay's real bite is in the psychological claim, not the tactical one: founders resist unscalable work not because it's inefficient but because it feels like an admission that the idea isn't good enough to spread itself, and Graham is saying that feeling is just wrong — virality is usually manufactured, not discovered. It's worth noticing this is the same move as The mosquito bucket of doom works — LessWrong: a trivial, hand-tended, non-scalable intervention outperforming the elegant systemic fix people assume they need.

Related

• manually recruiting early users one at a time ↔ trivial-cost DIY fix with outsized payoff (The mosquito bucket of doom works — LessWrong) · Both cases show a small, manual, non-scalable action outperforming the wait-for-the-elegant-solution instinct: a bucket of larvicide and a founder cold-emailing users both work because they bypass the assumption that impact requires scale from the start.