Good Ideas Don't Die

Brian Chesky is the co-founder and CEO of Airbnb, which he started after moving to San Francisco at 26 and which is now worth about $95 billion. He trained as a designer at the Rhode Island School of Design and could not code when he arrived in Silicon Valley 19 years ago. He is now rebuilding Airbnb as an AI-native app and expanding it from homes into services such as chefs, photographers, ski rentals and grocery delivery. He first met the host 7 years ago when she approached him at LAX as he boarded a commercial flight. These signals are from Brian Chesky's interview with Marina Mogilko on Silicon Valley Girl.

1. Agents Over Chatbots. Brian Chesky says people conflate agents and chatbots because most agents today happen to be chatbots. To him an agent is the next version of an application, one that knows you and acts on your goals, while a chatbot is only a user interface. Airbnb's app will feel more like an agent over time, coaching hosts on goals and pricing. Chesky expects agents to win because they can talk to each other, and says chatbots are the wrong interface for travel and shopping.
2. Travel as Comparison Shopping. The average Airbnb booking has 3 travelers, so search has to be shareable, visual and full of filters, which a text chat handles badly. Airbnb's answer is a toggle for natural-language search that still pre-sets the filters, plus AI-written highlights telling each guest why a listing suits them. Guests can ask the listing page a question in place of messaging the host, and compare homes side by side on pool, price, ratings and location. New neighborhood maps start in Paris and will roll out to 1,000 cities.
3. A Travel Graph. When Chesky arrived in Silicon Valley, Facebook connected you to your friends. Now none of his friends are on it, and he calls Instagram, X, YouTube and TikTok broadcast platforms. Airbnb now lets you connect with friends and family and see where they've stayed, so before a trip to Chicago you can find who's been and ask what to do. He says people increasingly trust recommendations from people they know over celebrities or writers, and he wants Airbnb to help people meet strangers offline, the way his parents met people at malls and theaters.
4. Weeks Instead of a Year. Airbnb is shipping 80% more features than a year ago, mostly because of AI. A new service used to take a year to launch and now takes weeks. The current list covers gaps between homes and hotels: baby gear, laundry, ski equipment sized and waiting for you (being tested in Italy), and Instacart groceries the host can stock in the fridge before you arrive. Airbnb tests each service in 1 city and expands only if it works.
5. The Long Tail of Services. Chesky's example is a masseuse in Tulum who today gets work only from locals or a resort, and could now sell directly to every traveler arriving in town. A photographer at the Eiffel Tower can offer families 30-minute professional shoots so both parents end up in the picture. Ride sharing and food delivery are big enough to support apps like Uber and DoorDash, while most services never reach that scale. He wants people to "Airbnb anything," and says the company takes every lead from anyone who applies to become a services host.
6. Product Velocity Over Tokens. Early this year CEOs ran token leaderboards and shamed employees who used too little AI. Chesky says that worked for a couple of months to force adoption and failed as a strategy, and now companies are asking why their AI bills are so high. Customers don't care whether a feature was built with AI. Airbnb measures how many features it ships, how fast, and whether they're high quality and successful.
7. AI as Amplifier. Every company has access to the same AI coding tools, the way every designer once had the same Photoshop and Illustrator. Chesky says the tools make good people much better and help weaker people far less, so the performance gap between top and average widens. AI gets an idea out of your head faster without making the idea better. Since anyone can now build, the question becomes "what do you build and should you build it."
8. Idea Volume. Chesky quotes Linus Pauling: "The best way to have a good idea is to have a lot of ideas." He cites IDEO's rule of putting every idea on a Post-it with no critique at first, since critiquing makes people edit their thinking out of fear, and a bad idea is sometimes one step from a good one. He writes about 1,000 words a day in a notebook, roughly 100 one-line ideas, which comes to about 30,000 a year. He says 99% of them are bad, and that doesn't matter because nobody sees them.
9. The Bolding Habit. At the end of each day Chesky bolds the 5 to 10 lines out of 100 that jump out at him. Each month, AI now reads the bolded lines, finds patterns, and points out ideas he wrote off as stupid, which become his short list. He finds that good ideas resurface on their own: he'll write the same idea a month later without realizing it. Ideas that keep coming back are the ones he takes seriously.
10. The Hidden Volume. Leonardo da Vinci painted only 12 to 15 pictures, including the Mona Lisa, yet left notebooks that Chesky says are 80% writing and 20% drawing. Vincent van Gogh did most of his work in his last 3 years and painted more than one painting a day, on top of sketches. Hemingway wrote from 6 a.m. until noon, when he had nothing left. People see the few finished works and miss the volume behind them, so Chesky's advice is to edit hard and publish only the best.
11. Blinders On. Chesky admits his vice is X, and says founders spend too much time watching competitors like kids chasing a soccer ball. "Van Gogh couldn't doom scroll." He recommends checking news and competitors once a day for 20 to 30 minutes and spending the rest of the day uninterrupted with the phone off. He argues attention span tracks screen size: you'll sit through an IMAX film, but on a phone you last 15 seconds before scrolling, so serious work belongs on a computer.
12. 100 More Anthropics. Before ChatGPT, Chesky felt lucky to have arrived at 26 just as the iPhone and cloud computing made startups easy; now he wishes he were 26 again. If technology doubles every year for 10 years, that's 1,000x, so he expects 50 to 100 more companies the size of Anthropic, plus millions of 1 to 3 person businesses. He would look at education, health and an "AI Disney," and warns that forcing AI into a messaging interface is like putting a jet engine on a bicycle. He was intimidated to try Claude Code until he had an employee tutor him, and tells anyone hesitant to find someone who knows more and jump in.

Watch the full video at https://www.youtube.com/watch?v=iJlyNupMOzA.