The speaker argues that a startup’s first version should be a **minimum evolvable product**—simple enough to survive contact with a tiny group of early users who are willing to try something new or have a burning problem. Finding those users is a search problem, not a persuasion problem: look for early‑adopter types (people who enjoy testing new tools) or those with a pressing need you can solve. Charge them real money early to get sharp feedback, use personalized outreach, launch quickly to broaden your discovery surface, and treat early users like anthropologists study a hidden culture—observe their decisions, experiment relentlessly, and don’t fear churn. Their feedback steers the product’s evolution, much like natural selection shapes a phylogenetic tree; early adopters’ preferences (e.g., Tesla Roadster buyers valuing speed and tech over comfort) determine the trajectory of later, mass‑market versions. Ultimately, the product’s final form depends on where you start and who you start with.
1. Most people are not early adopters.
2. Most people have used zero products where they were among the first 10 users.
3. Almost no one wants to be a startup's first paying customer.
4. Every great product still manages to find a few people willing to take that leap.
5. The earliest version of a product only needs to survive contact with a tiny group of people who might try it.
6. Gustav, a colleague, worked at Airbnb for years and enjoyed trying out products from startups and bringing them into the company.
7. Some people have a burning issue that makes them willing to try any new product that could make their life easier.
8. When the speaker's team needed to ship their first inference API, they wanted to ship it fast and avoid dealing with billing or a public endpoint.
9. Within three days, the speaker found and paid a startup whose product solved that issue.
10. The speaker was that startup's first customer.
11. The startup's size or reputation did not matter; the speaker's problem did.
12. Early adopters and people with a burning problem are rarely price sensitive.
13. Paying customers give sharper feedback than free users.
14. An angry customer paying a lot of money is more likely to give feedback than a non‑paying nobody.
15. Targeted personal outreach (e.g., cold email, door knock) is more effective than billboards for reaching early users.
16. Launching early creates a wide surface area for early users to find the product.
17. Founders should study early users closely, like anthropologists studying a hidden civilization.
18. Founders should run constant experiments on pricing, landing pages, onboarding, features, etc.
19. Losing an early user is not catastrophic because many others have not yet heard of the product.
20. When a startup runs a bad experiment, no one writes about it; they fight relevance, not headlines.
21. The average personal software spend is tiny (e.g., about $150 per month).
22. Corporate cards often have multiple tools each costing more than that personal monthly spend.
23. In the AI era, consumer apps struggle because ads often don't cover AI costs and subscriptions must fit into a small personal budget.
24. Many AI founders start by selling to prosumers, businesses, or users with high advertising value (e.g., doctors).
25. Early users give feedback that steers how the product evolves over time.
26. Tesla's early product, the Roadster, was a high‑margin vehicle used to fund capex for the Model S and later models.
27. Tesla searched for early adopters willing to buy an impractical, expensive car with limited range and charging infrastructure.
28. Early adopters cared more about tech and acceleration than comfort, influencing the Model Y's performance characteristics.
29. The Model Y has a faster zero‑to‑60 than a Lamborghini and better tech than a BMW, but worse suspension and comfort than a Toyota.
30. If early adopters had preferred a slow, plush vehicle, Tesla's cars would look different today.
31. A product's first version should be a minimum evolvable product—simple enough to respond to market pressures and evolve.
32. The product will change a lot, so it does not need to be perfect from the start.
33. The final product will depend on the starting point and the early users involved.