doing things that don’t scale

The talk explains why early startups should focus on their biggest immediate problem, usually finding customers and learning what they need, instead of building for scale too soon. Founders can do work manually, provide hands-on service, or use improvised tools to test demand and improve the product before investing in automation. Examples include Airbnb’s founders taking listing photos, Fleek’s founders selling clothes by hand, Algolia implementing search for Product Hunt, Instacart photographing grocery items, and DoorDash using simple tools to handle orders and delivery. Once demand and a workable service are proven, founders can invest in scaling while avoiding the trap of relying indefinitely on consulting or other labor-intensive work.

Paul Graham explains that doing things that don't scale means serving early customers through hands-on, manual work that would be impractical at a much larger size. Startups should do this anyway because they have little to lose at the beginning, and the work helps them learn what customers need. He also notes that many startup lessons he encountered firsthand later became common advice taught at Y Combinator.

Paul Graham and Geoff Ralston discuss how Viaweb became an early web app because its founders wanted to avoid writing software for Windows. Graham argues that startup ideas often seem implausible at first, that founders learn by doing manual work for early customers, and that persistence and a strong founding team matter more than raw intelligence. He recommends talking to users, shipping as soon as a product has a small amount of real utility, and not letting fear of competitors or imperfect launches slow progress. The conversation also covers choosing co-founders, YC’s counterintuitive startup advice, fundraising, and the costs of starting a company too early in life.

Apoorva Mehta recounts how he left Amazon, tried many startup ideas, and eventually founded Instacart to solve a grocery shopping problem he personally experienced. He explains how early failures, direct customer feedback, and doing work manually helped him find product demand and improve the service. His story includes winning a late entry to Y Combinator, handling early operational problems, and persuading investors despite the history of failed grocery delivery companies. The talk teaches founders to solve problems they care about, stay resilient through failure, and scale only after finding demand.