Do Things That Don't Scale

Manual, personal work early on that wins your first users and teaches you what to build.

Also known as: do things that don't scale, Paul Graham, manual onboarding

Do things that don’t scale (Paul Graham’s famous advice) means winning early users with manual, personal, unscalable work: onboarding each user yourself, hand-configuring accounts, writing personal emails, showing up in person. What looks inefficient is actually the fastest learning available — and large competitors cannot copy personal attention.

scalable (later):  self-serve signup → automated onboarding → docs
unscalable (now):  founder onboards each user → watches them struggle → learns → fixes

The manual phase teaches what to automate: every repeated hand-task is a validated feature spec, every user struggle a roadmap item. Startups that automate too early build elegant systems for workflows nobody wanted.

The classic mistakes:

Automating before understanding. Building self-serve for a flow three users have tried. Serve thirty by hand first; the pattern (or its absence) tells you what to build.
Hiding the founder. Polished landing pages with no human contact convert worse and teach nothing. Put your name, face and calendar on the early product.
Never transitioning. Hand-work that persists past product-market fit becomes an ops burden that caps growth. Define the automation trigger (user count, repeatability) and honor it.
Doing unscalable marketing. Manual sales to ten enterprise prospects is strategy; manual tweets to ten consumers is hobby. Unscalable works where each unit is valuable (B2B, high-ticket) — match the tactic to the math.

The sequence: unscalable to learn and win the first users → automate what repeats → keep the personal touch where it differentiates (founder sales calls, community presence) long after it stops being necessary.