Justin Kan & Brian Chesky: The Rejections That Built Twitch and Airbnb | E2344

Justin Kan’s 2007 Justin.tv stunt—broadcasting his life around the clock with a camera strapped on—eventually became a platform where anyone could stream. The company first tried to attract celebrities and other prominent creators, but they would not reliably produce content. By around 2010, Justin.tv had about 10 million monthly visitors and ranked roughly in Alexa’s top 300, yet growth had stalled after five years. Gaming accounted for only about 3% of content and a few hundred thousand monthly unique visitors, but as StarCraft 2 boosted interest, cofounder Emmett proposed concentrating on game streams and building a community around them. Kan backed the bet on a small audience, and the gaming section later spun off as Twitch.

The gaming focus came after Justin.tv had reached about 25 employees without sustained growth or a clear exit. Google explored buying it as a small feature alongside YouTube, interviewed the entire team, then rejected the deal, saying the team was not good enough. Kan said the rejection sent the founders back to the drawing board and warned other founders against letting a prospective acquirer interview the whole company. Twitch was eventually acquired by Amazon in 2014 for $970 million. Early game streaming also drew copyright strikes from publishers worried that people would see game content before buying; Kan said publishers later recognized streamers’ promotional value, while the discussion noted that issues involving Nintendo persisted.

Airbnb began with an idea Brian Chesky had in October 2007, around a San Francisco design conference, but early reactions reflected widespread doubt. In early 2008, Chesky and Joe sought $150,000 at a $1.5 million post-money valuation, offering investors 10% of the company. An introduction to 10–20 angel investors yielded few replies. The founding team had one software engineer and two designers, which investors viewed as a weakness; Chesky argued that design was central to the product because the problem was not purely technological. Investors also questioned whether people wanted to travel or would stay in strangers’ homes, and one told Chesky they liked everything except the founders and the idea.

At Y Combinator, Paul Graham asked whether anyone was using Airbnb and what was wrong with them, and suggested the founders start a payments company or online bank. They stuck with Airbnb. Near the end of the interview, Joe explained that selling novelty cereal had helped fund the startup: Obama O’s sold for $40 a box. Graham saw the stunt as evidence that the founders might persuade people to stay in one another’s homes, and Y Combinator admitted them. Chesky said Jessica and Graham later called the founders “cockroaches” because they would keep going even if fundraising dried up. Their persistence and resourcefulness helped turn rejection into evidence of resilience.

Kan’s advice was to concentrate on founders’ strengths, improve at weaker areas or delegate them, and avoid becoming a bottleneck by micromanaging capable hires. He also warned that success with one product can create survivorship bias: founders may assume their decisions are right because one venture worked. The host contrasted Kan’s emphasis on shared responsibility and people running their own areas with “founder mode.” Chesky said his roommate Phil worked at Justin.tv, which he described as a Y Combinator-funded precursor to Twitch. The host also noted that Airbnb launched Airbnb Rooms in the week of the May 2023 interview, returning to its original spare-room concept.

The discussion of AI emphasized that, so far, it has often made workers more productive and busier rather than simply replacing them. One workflow used Claude to turn human-curated stories into a show docket; Zach still built and refined the skill and reviewed its output, with automation cutting the work by an estimated 20%, from a previous estimate of three hours to about an hour or less a day. Speakers said current systems still rely on people to select inputs, judge results, and improve processes, though models could take on some expert work over the next three to five years. One forecast was gradual employment change and medium- to long-term job churn, not an 18-month shock.

Chesky predicted AI could enable millions of startups, while another speaker questioned whether it would produce that many new founders, noting that economies also need employees and consumers with income. Speakers described natural-language tools helping nontechnical workers build apps and creators or small businesses interpret analytics, suggesting AI may first reshape existing work. One newsletter writer kept the writing entirely human but used Grockbot for research, saying it enabled three researched Substack posts in a day. Other examples included surfacing AI demos and tutorials, summarizing trends, finding video timestamps, and locating Fei-Fei Li’s email; the transcript said her startup was sold to AMD in an $8.2 billion all-stock deal.