Airbnb CEO Rejects AI Chatbots for Shopping as Growth Hits 18%

营收增速达18%之际，Airbnb CEO明确反对在购物中使用AI聊天机器人 Airbnb首席执行官Brian Chesky认为，基于文本的聊天机器人不适用于旅行和电子商务，因为购物需要视觉对比和多用户协作，而非对话式的文本流。 Airbnb首席执行官Brian Chesky分享了重塑早期创业激情并采用“创始人模式”如何助力Airbnb将营收增速重新提升至18%。他对AI界面的行业共识提出质疑，认为与丰富、直观且基于智能体的用户体验相比，基于文本的聊天机器人在本质上并不适合旅行和电子商务。Chesky还指出消费级AI存在巨大且被忽视的市场机遇，并鼓励设计师直接运用AI工具，而不是将产品设计的未来拱手让给管理者和工程师。 Airbnb在最近几个季度通过采用“创始人模式”领导力并设立“Project Y”重塑早期创业激情，将其营收增速从10%重新提升至18%。 (0:00) Brian Chesky倡导亲力亲为的“创始人模式”领导力，认为真正的领导力需要躬身入局与深度参与，而非远远地向下放权。 (2:44) 基于文本的聊天机器人是用于旅行和电子商务的错误界面，因为旅行需要视觉探索、直接操作、并排对比以及社交协作。 (6:03) Airbnb的核心护城河依托于其线下现实世界的基础设施，包括本地化支付处理、全球支持体系、房东网络以及数万亿美元的房产保障。 (10:57) 科技行业对企业级软件的高度聚焦，为兼具视觉沉浸感与动态交互的智能体界面留下了巨大且尚未开发的消费级AI机遇。 (13:02)

Airbnb CEO Brian Chesky argues that artificial intelligence will transform consumer products through visual and multimodal experiences rather than conversational chatbots, while asserting that real-world operational scale remains an enduring defense against software disruption. Internal AI Adoption and Hands-On Leadership (0:00) Chesky attributes Airbnb's revenue reacceleration—from 10% annual growth to 18% in the latest quarter—to aggressive internal AI integration and direct operational restructuring. Sixty percent of Airbnb's software code is currently written by AI, doubling competitor benchmarks, while automated tools resolve 40% of guest support tickets and have lowered customer service ticket costs by 10%. To drive this efficiency across a marketplace processing roughly $100 billion in gross booking value annually, Chesky instituted "Project Y," a nimble internal group modeled on early startup intensity to optimize user conversion and eliminate corporate friction. Advocating for "founder mode," Chesky rejects conventional management consulting models in favor of continuous, hands-on CEO involvement across company tiers. Comparing executive leadership to a cavalry general on the battlefield, he argues that managers must stay immersed in technical details and concrete data. He envisions an era of "AI founder mode" where traditional people managers become obsolete as leaders directly manage autonomous agents. The Chatbot Limitation and Consumer AI Deficit (6:03) Chesky contends that conversational text chatbots are the wrong paradigm for travel and e-commerce. He identifies four core friction points: travel search is inherently visual rather than text-heavy; typing prompts lacks the speed of direct manipulation filters; comparing thousands of options side-by-side is unwieldy in a chat window; and text bots remain isolated, single-player tools for what is typically a social, collaborative decision. For a travel journey spanning destination discovery, flights, accommodations, activities, logistics, and spontaneous plans, he anticipates that future disruption will rely on rich video, dynamic maps, and agentic interfaces rather than conversational text boxes. Despite AI driving significant enterprise productivity, Chesky points out that creative consumer applications have lagged. He notes that AI research labs are heavily staffed by researchers rather than designers and product managers, while incubators channel capital into B2B feedback loops—highlighting a recent Y Combinator batch where only approximately 16 out of 175 companies targeted consumer software. Warning against a historical repeat of traditional designers delaying entry into the early web, Chesky urges creative professionals to adopt AI coding tools directly rather than ceding interface development to engineers. Physical Infrastructure as the Primary Marketplace Moat (10:57) Addressing technological displacement, Chesky stresses that the visible consumer-facing application accounts for only about 20% of Airbnb's overall business. Citing Airbnb CTO Ahmed, Chesky asserts that while an AI agent can easily generate frontend booking software, replicating the physical, real-world operational infrastructure is far more difficult. This offline moat encompasses complex global payment flows, customer service adjudication, and regulatory compliance across more than 100 countries. Airbnb manages a network of 5.5 million host accounts and accommodates 4 to 5 million guests nightly. Furthermore, the company underwrites listing safety through its $3 million AirCover guarantee, protecting approximately $3 trillion in aggregate real estate value across one million homes every night. Expansion Across Boutique Hotels, Services, and Marketing (29:42) Beyond its short-term rental core, Chesky forecasts that Airbnb's gross booking value could double from $100 billion to $200 billion as it captures demand from traditional hotels, which currently host eight to nine guests for every one staying in an Airbnb. Growth targets include distributing inventory for independent and boutique hotels burdened by high online travel agency fees, facilitating extended stays exceeding 30 days for remote workers, and consolidating up to 80 fragmented service verticals into a single marketplace. On customer acquisition, Chesky observes that standard marketing channels like search engine optimization and digital ad banners quickly hit saturation and consumer fatigue. To counter this, he advocates for unconventional, high-visibility brand stunts—citing Airbnb's Malibu Barbie DreamHouse activation—as delivering substantially higher return on investment than standard paid advertising buys. While Chesky outlines extensive growth targets across boutique lodging, unified services, and multimodal interfaces, the realization of this vision depends on whether designers successfully pioneer new non-text consumer software paradigms and whether independent hotels embrace Airbnb over traditional travel distribution incumbents.

Airbnb 首席执行官 Brian Chesky 指出，消费级 AI 面临严重的界面错配与人才失衡，基于纯文本的对话机器人无法胜任复杂交易与旅行体验。他认为，真正的商业护城河根植于线下运营与多模态交互界面，公司正通过精细化领导与业务多元化打破增长瓶颈。 创始人模式与内部运营提速 (0:00) Airbnb 在 2026 年迎来了自疫情以来的首次增长再加速，季度营收增速由上年的 10% 提升至 18%，年总预订额（Gross Booking Value）达到约 1000 亿美元。Chesky 将大体量平台的放缓比作难以抗拒的地心引力，为扭转这一态势，他摒弃了依赖外部管理咨询公司的传统模式，转而采用类似 Paul Graham 提出的“创始人模式”（Founder Mode）。他绕过管理层级组建内部先锋团队“Project Y”，通过逐个模块的精细打磨清除官僚主义，并推行高频直接审查的现场管理风格。技术效率方面，Airbnb 内部已有 60% 的软件代码由 AI 生成，高于同行基准；同时，AI 独立解决了 40% 的客户咨询，帮助客户服务支持成本降低了 10%。 文本聊天交互在电商与旅行中的局限 (6:03) 针对当前将对话式聊天机器人视为未来电商入口的观点，Chesky 持明确否定态度。他以 OpenAI 下架第三方 ChatGPT 插件为例，指出纯文本界面存在四大天然缺陷：其一，旅行决策重度依赖视觉，而聊天框将图像边缘化；其二，缺乏直观的操作与快速筛选控件，迫使用户输入每一步指令；其三，难以进行横向多维对比，模型要么输出过度简化的单一选项，要么输出难以浏览的长篇文本；其四，规划行程本质上是多用户协作的社交行为，而单人对话框割裂了互动流程。Chesky 强调，未来旅行搜索应当基于动态地图、高清图像与视频生成的多模态沉浸式界面，而不是单调的提示词窗口。 线下服务壁垒与消费级 AI 的设计缺失 (10:57) Chesky 认为，消费者可见的前端应用仅占 Airbnb 运营体系的 20% 左右，真正的技术与资产壁垒在于复杂的现实世界支撑体系。Airbnb 在全球覆盖超过 100 个国家，维护着 550 万房东网络，每晚接待 400 万至 500 万房客，并通过高达 300 万美元的 AirCover 保障计划为夜均价值约 3 万亿美元的房产提供安全防线。即便软件界面容易被 AI 代理复刻，底层的跨境支付、纠纷仲裁与实体服务能力依然难以被纯算法颠覆。此外，他指出当前硅谷过度偏向企业级 B2B 软件与代码助手——以近期 Y Combinator 入选企业为例，约 175 家初创公司中仅约 16 家聚焦消费端。这种失衡主要源于新兴 AI 实验室多由纯算法研究员主导、缺乏设计师深度参与，导致消费级应用迟迟未能迎来真正的突破时刻。 非对称品类扩张与高回报创意营销 (29:42) 在核心业务层面，Chesky 预测其平台年总预订额有潜力从 1000 亿美元翻倍至 2000 亿美元。当前全球酒店住客与 Airbnb 住客的比例约为八比一或九比一，公司计划切入占比达半数却常被大型在线旅游代理（OTA）高佣金困扰的独立精品酒店市场。同时，Airbnb 试图拓展超过 30 天的长期租住服务，并构建整合多达 80 种生活服务的统一平台。在获客与品牌推广方面，Chesky 强调成熟渠道已普遍面临转化衰减，传统买量模式容易遭遇“横幅盲区”（Banner Blindness）。相比常规广告，极具颠覆性的创意营销能产生远超投入的回报，例如配合电影发布的马里布真人版“芭比梦想豪宅”（Barbie Malibu DreamHouse）在线下引发病毒式传播，其投入产出比显著优于公司常规硬广投放。 Chesky 认为消费级 AI 的拐点将随着视觉生成技术的成熟在未来两年内到来，但这要求设计人才主动掌握 AI 编程工具，以弥合界面交互与技术底层之间的断层；然而，在企业级投资持续吸纳大多数资本的现实下，消费端多模态产品能否重塑复杂线下交易链条仍有待验证。

## Background 本期访谈由科技播客频道 TBPN 于 2026 年 5 月 22 日发布在 YouTube 平台，对话嘉宾为 Airbnb 联合创始人兼 CEO 布莱恩·切斯基（Brian Chesky）。根据视频简介，切斯基深入探讨了 Airbnb 如何通过重建初创公司烈度与践行“创始人模式”重振业绩增长、为什么当前 AI 聊天机器人并非旅行与电商的未来形态、Airbnb 内部 60% 代码由 AI 编写的技术重构，以及为何下一轮 AI 创新浪潮必将来自沉浸式消费级产品而非企业级软件。 ## TL;DR 在年预订额逼近千亿美元的体量下，Airbnb 实现了季度营收增速由 10% 跃升至 18% 的强劲再加速，切斯基将此归功于深入业务微观细节的“在场式领导”以及全公司 60% 代码由 AI 辅助编写的研发革命。切斯基旗帜鲜明地断言纯文本 Chatbot 绝非旅行与电商的终局交互形态，痛陈其缺乏直接操纵与比价能力，未来的智能体必须依托多模态驱动的沉浸式富视觉界面。针对行业盲目扎堆企业级软件和 AI 实验室缺乏产品审美的怪象，他呼吁设计师群体抛弃刻板线框图、主动下场写代码以抢占下一代交互定义权。此外，他指出 Airbnb 真正的护城河在于线下万亿级的履约保障与生态治理，并披露了向独立精品酒店、全品类服务平台及 30 天以上中长租期扩张的未来蓝图。 ## Atmosphere & style 整场对话节奏紧凑明快、信息密度极高，兼具严谨的商业推演与极客式的实操热忱。切斯基言辞犀利坦诚、充满反叛色彩与偏执感，对传统商学院的“放权假象”和硅谷跟风企业级的内卷生态展开了毫无保留的抨击。主持人与嘉宾之间的思想交锋张力十足，不仅在具体产品推演中碰撞出火花，更在即兴构想“观光飞艇空中民宿直播”的轻松笑声中收尾，让听众深刻感受到硅谷早期那种身先士卒、敢于冒险的纯粹创业精神。 ## Bulletin - [00:00](https://www.youtube.com/watch?v=TyNufBceyZo&t=0s) Airbnb 业绩迎来强劲再加速，最新季度营收增速由去年的 10% 跃升至 18%，年总预订额已达 1,000 亿美元大关。 - [02:11](https://www.youtube.com/watch?v=TyNufBceyZo&t=131s) 公司内部已有 60% 的代码由 AI 辅助编写，客服体系中 40% 的求助完全由 AI 独立解决，单笔客服工单处理成本下降 10%。 - [06:06](https://www.youtube.com/watch?v=TyNufBceyZo&t=366s) Brian Chesky 明确提出聊天机器人（Chatbot）绝非旅游与电商交互的终局，未来消费级 Agent 必须依赖高视觉密度的富交互界面。 - [17:34](https://www.youtube.com/watch?v=TyNufBceyZo&t=1054s) 针对当前 AI 创业团队过度扎堆企业端软件的现状，Chesky 预测多模态生成模型将在未来一到两年内引爆消费级产品的全新革命。 - [29:23](https://www.youtube.com/watch?v=TyNufBceyZo&t=1763s) 在短租核心基本盘之外，Airbnb 正在积极布局精品独立酒店、综合服务交易平台以及 30 天以上中长期旅居等三大全新增长曲线。 ## Key takeaways - [00:00](https://www.youtube.com/watch?v=TyNufBceyZo&t=0s) Airbnb 通过抽调精干特战小队 Project Y 专注房客端转化率，重拾创业初期的“初创公司烈度”，成功打破千亿美元体量双边平台的增长重力法则。 - [02:11](https://www.youtube.com/watch?v=TyNufBceyZo&t=131s) AI 已全方位深度嵌入 Airbnb 工程与运营体系：60% 业务代码由 AI 辅助完成，40% 的客服问题实现全自动闭环处理。 - [02:40](https://www.youtube.com/watch?v=TyNufBceyZo&t=160s) 推行“创始人模式”管理哲学，领导者应跳过中间层深入一线细节，提倡“先将一个房间装修到极致再逐步铺开”，反对传统商学院盲目放权的缺位管理。 - [02:40](https://www.youtube.com/watch?v=TyNufBceyZo&t=160s) AI 时代将消灭单纯向上汇报的纯管理岗，未来的领导者与管理者必须具备调度掌控 AI Agent 的亲力亲为（Hands-on）能力。 - [06:06](https://www.youtube.com/watch?v=TyNufBceyZo&t=366s) 聊天机器人无法胜任电商与旅游业务，原因在于其以纯文本为主缺乏视觉冲击力、缺少直接操纵控件、极难实现横向比价，且不具备多人协作能力。 - [08:47](https://www.youtube.com/watch?v=TyNufBceyZo&t=527s) 下一代消费级 Agent 绝非对话输入框，而是结合自然对话与动态渲染的高信息密度沉浸式富交互界面。 - [09:42](https://www.youtube.com/watch?v=TyNufBceyZo&t=582s) 前台客户端仅占 Airbnb 整体业务复杂度的 20%，真正的核心壁垒在于万亿规模的线下履约清算、纠纷仲裁、房东治理与 AirCover 保障体系。 - [14:48](https://www.youtube.com/watch?v=TyNufBceyZo&t=888s) 在旅行六步全流程中，AI 最容易在“目的地探索”与“机票交通”等前期阶段取得革命性突破，而后续履约重度依赖线下资产。 - [17:34](https://www.youtube.com/watch?v=TyNufBceyZo&t=1054s) AI 创业生态存在严重失衡：大量技术实验室缺乏产品与设计基因，导致企业级赛道同质化内卷，而消费级蓝海市场仍处在缺位状态。 - [17:34](https://www.youtube.com/watch?v=TyNufBceyZo&t=1054s) 随着图像与视频生成模型迎来质变拐点，未来一到两年内消费级软件将迎来类似初代 iPhone 般的视觉交互新范式。 - [20:56](https://www.youtube.com/watch?v=TyNufBceyZo&t=1256s) 创业必须遵循“在所有人向左拐时坚定右拐”的逆向思维，警惕过度拥挤的热门风口，坚信竞争往往是留给盲从者的陷阱。 - [23:00](https://www.youtube.com/watch?v=TyNufBceyZo&t=1380s) 呼吁创意人群与设计师不要在 AI 时代作壁上观，应主动学习使用 AI 编写代码，避免重蹈互联网早期被产品经理和工程师主导决策的覆辙。 - [26:37](https://www.youtube.com/watch?v=TyNufBceyZo&t=1597s) 市场营销打法失效周期正在急剧缩短，常规战术会迅速遭遇用户的“横幅盲区”，唯有极具颠覆性的产品和前所未有的新怪招才能突围。 - [29:23](https://www.youtube.com/watch?v=TyNufBceyZo&t=1763s) 除核心短租业务外，Airbnb 将目光锁定于全球独立精品酒店分发、80+ 类别的综合服务交易平台以及 30 天以上的中长期数字游民租住生态。 - [31:32](https://www.youtube.com/watch?v=TyNufBceyZo&t=1892s) 传统广告的投资回报率难以比拟具有超级文化属性的标志性事件营销，如芭比梦幻屋案例展现了破圈创意的巨大爆发力。 ## Detailed walkthrough ### [00:00](https://www.youtube.com/watch?v=TyNufBceyZo&t=0s) 业绩强劲反弹与重拾“初创公司烈度” 对谈伊始，主持人和 Brian Chesky 聊起 2026 年 Airbnb 的最新业务进展。Brian Chesky 兴奋地透露，公司自疫情爆发以来首次实现了业绩的显著再加速：去年 Airbnb 的营收增速为 10%，而最新公布的季度财报中营收增速已大幅跳升至 18%。他感慨道，双边交易平台（Marketplace）具备一种不可忽视的“重力法则”，当业务规模达到年交易额（Gross Bookings）约 1,000 亿美元的超大体量时，增长曲线一旦放缓，想要重新将其拉升是极其艰难的，而团队如今做到了这一点。 面对主持人关于“究竟如何做到的”提问，Chesky 坦言，几年前随着公司体量不断膨胀，他敏锐地察觉到团队正在失去早期的“初创公司烈度”（Startup Intensity）与充沛精力。为了找回这股劲头，他亲自抽调了一支精干的小团队，代号命名为“Project Y”。他们并没有铺大摊子，而是精准聚焦于一个极小的切入口——房客旅程中的“转化率”（Conversion Rate）。Chesky 带领这支团队重新找回当年与联合创始人 Joe Gebbia 在旧金山劳施街（Rausch Street）公寓创业时的感觉：彻底抛弃大公司的繁文缛节，像一家真正的初创团队一样日夜苦干、死磕数据，对用户体验展现出近乎偏执的狂热。 在小团队打出战果后，公司开始将这种模式复制到各个工作组（Pods）中，向全公司传授早期的高强度打法，剔除层层官僚审批，吸纳顶尖人才并为其扫清障碍。尽管大公司的财务表现往往是内部组织变革的“滞后指标”，需要数年时间才能体现，但 Chesky 强调现在的 Airbnb 比以往任何时候都更像一家初创企业，公司感觉更加年轻、敏捷。 - **财务再加速**：营收增速由去年的 10% 跃升至当季的 18%，年总预订额（Gross Bookings）规模已达 1,000 亿美元。 - **劳施街精神（Rausch Street）**：以初创时期的公寓为精神象征，通过小规模精锐团队“Project Y”主攻转化率优化。 --- ### [02:11](https://www.youtube.com/watch?v=TyNufBceyZo&t=131s) AI 深度重塑工程与客服架构 在重燃创业活力的同时，AI 的融入为 Airbnb 提供了强劲的推力。Brian Chesky 透露了一组颇为亮眼的数据：目前 Airbnb 全公司已有 60% 的代码由 AI 辅助编写，这一普及率是行业同行和竞争对手基准值的两倍之多。不同于许多在线旅游平台（OTA）对 AI 落地效果的迷茫，Airbnb 已经在端到端的业务旅程中全面铺开 AI 应用。 其中成效最为立竿见影的是客户支持（Customer Service）环节。接入 AI 系统后，Airbnb 单个客服工单的解决成本下降了 10%；更有 40% 的用户求助完全由 AI 独立解决，无需人工介入。技术重构从底层代码延伸到用户服务的每一个触点，极大推动了整个组织的运营效率。 ```mermaid flowchart LR accTitle: Airbnb AI-Driven Operational Architecture accDescr: Diagram showing AI code generation and customer service resolution impact at Airbnb subgraph Engineering["工程研发端"] A["60% 业务代码由 AI 编写"] --> B["研发节奏与产能翻倍"] end subgraph CustomerSupport["客服运营端"] C["用户求助入口"] --> D{"AI 自动化处理"} D -- "40% 自动化闭环" --> E["问题自主解决"] D -- "60% 复杂工单流转" --> F["人工客服支持"] F --> G["单次工单成本降低 10%"] end ``` --- ### [02:40](https://www.youtube.com/watch?v=TyNufBceyZo&t=160s) “创始人模式”：下沉细节与在场式领导 当主持人问及派驻特战小队进入既有业务线是否会引发大公司内部常见的部门政治冲突、乃至演变成类似“麦肯锡空降做 PPT 搞得人心惶惶”的尴尬局面时，Brian Chesky 分享了他在过去五年里摸索出的核心管理哲学，即保罗·格雷厄姆（Paul Graham）所定义的“创始人模式”（Founder Mode）。Chesky 指出，所谓的创始人模式，本质就是跳过中间管理层，亲自下沉到最微观的业务细节中去。 对于大公司领导者，Chesky 给出了一条极为实用的改造建议：“不要试图重修整栋房子，先挑选一间房间，把它做到极致完美，然后再一间一间推进。”在 Airbnb，他并未大动干戈换掉既有团队，而是直接在现有人员中挑人，通过每周甚至每天高频复盘工作产出，手把手向团队灌输追求卓越的烈度。这种做法起初让不少人感到极不适应，确实有人因此选择离职，但留下来的绝大多数人很快发现，有 CEO 亲自坐镇，反而能直接碾碎官僚阻碍，工作推进变得更快更顺畅。 Chesky 将这种深度介入比作高尔夫挥杆教练：当领导者长期与团队并肩磨炼，将高标准内化为团队的“肌肉记忆”后，管理者便可逐步退后，组织也能持续自发运转。他直言不讳地批判了传统商学院倡导的“充分信任下属并彻底放权”的说辞，认为真正的伟大领导力体现为“在场”（Presence）而非“缺位”（Absence）： > “真正的领导力是在场，而不是缺席。商学院和管理咨询教导 CEO 应该‘信任团队并闪到一边’，但我认为‘信任’绝不等于‘撒手不管’。作为骑兵将军，你必须跨在马背上、身处战场一线，而不是躲在海外大后方画蓝图。领导力永远来自身先士卒，而不是殿后观望。” 他进一步预言，在 AI 时代即将来临之际，企业将迎来“AI 创始人模式”。由于数据与细节变得触手可及，组织中将不再需要单纯依靠向上汇报、向下管人的“纯人员管理者”（Pure People Managers）——未来的管理者不仅要带人，更要具备调度和掌控 AI Agent 的硬核动手能力。 - **微观切入**：反对全盘盲目重组，提倡“先装修好一个房间，再逐间推进”。 - **在场式领导**：CEO 亲自进行日/周度细节复盘，建立团队肌肉记忆。 - **管理形态重塑**：AI 时代消灭单纯的管理岗，推行全员 Hands-on（亲力亲为）的作战方式。 --- ### [06:06](https://www.youtube.com/watch?v=TyNufBceyZo&t=366s) 偏执感与为何 Chatbot 并非旅行电商的终局 对谈随后转向竞争维度的推演。主持人 Jordi 追问，Chesky 是否会经常假想：如果自己穿越回十到十五年前，会如何从零开始颠覆掉现在的 Airbnb？Chesky 毫不避讳地承认这确实是一个让他感到后背发凉的思维实验。他不久前刚在内部会议上对核心团队放下狠话：“如果是 26 岁时的我、Joe 还有 Nate（Nathan Blecharczyk），我们绝对有能力把今天的 Airbnb 彻底干趴下。” 虽然 Airbnb 拥有全球知名品牌（品牌名称甚至变成了名词和动词）以及极其稳固的双边网络壁垒，但 Chesky 依然保持着强烈的偏执与危机感。他告诫团队，没有任何优势是理所当然的，若固步自封，随便一群充满活力的二三十岁年轻人都能通过全新的软件形态改写游戏规则。现有的 App 虽然精致美观，但必须进化为真正的“AI 原生”产品。 谈到这里，Brian Chesky 抛出了一个极为颠覆主流认知的判断：**聊天机器人（Chatbot）绝不是未来旅行和电子商务的正确交互形态**。他提到 ChatGPT 曾高调推出第三方插件/应用商店，但随后在去年三月将其下线，这本身就印证了对话流形态的局限。Chesky 总结了 Chatbot 无法胜任旅游电商的四大硬伤： 1. **以纯文本为主，视觉体验严重滞后**：旅行本质是由视觉与氛围驱动的体验，但在聊天流中，图片与视觉信息往往沦为次要的“附赠品”。 2. **缺乏直接操纵（Direct Manipulation）能力**：用户每一个动作都需要敲入完整的 Prompt，对于对话而言尚可，但在实际挑选过程中，用户无法直观地添加筛选标签（Filters）、任意点击交互，导致极高的“交互阻力”。 3. **极难进行横向对比（Comparison Shopping）**：电商与旅行的核心环节是货比三家。在成千上万个候选房源中，AI 很难精准命中唯一答案只推一两个结果；而一旦把大量选项堆在聊天框里，用户很快就会迷失。 4. **单人单向体验，缺乏协作属性**：现有的 Chatbot 绝大多数是单人交互工具，而旅行规划天然具有强烈的多人协作属性；加之 Airbnb 上 85% 的订单都需要在房客与房东之间建立账户并收发沟通，文本框无法支撑这种复杂的网络生态。 对此，主持人提出了即兴反驳与探讨：以自己妻子规划度假为例，如果目标很明确，就是在特定日期锁定三家备选酒店，理论上完全可以由一个 Chatbot Agent 代劳——输入人数与地点后，Agent 能迅速抓取房型、价格和图片，并输出一份权衡利弊的直观对比分析。两位围绕 AI 助手到底该如何重构消费级交易体验展开了激烈的交锋。 ### [08:47](https://www.youtube.com/watch?v=TyNufBceyZo&t=527s) 告别纯文本聊天：未来的 Agent 需要沉浸式的“富交互界面” 当主持人描述如果能让 AI 汇总全网信息生成攻略、并像协同聊天一样分享给同伴一起查阅时，布莱恩·切斯基（Brian Chesky）立刻打断并明确指出：“那是未来，但那绝不是我们今天所说的 Chatbot（聊天机器人）。”切斯基坚信，未来的核心不再是孤立的 App，而是各类智能体（Agents）。然而，这些 Agent 绝对不会是以纯文本对话为核心的。 切斯基认为，目前的聊天窗口虽然能够勉强处理这些需求，但绝非最优解。在电商与预订场景中，用户需要的是极具沉浸感、高度视觉化的交互界面，而不是单纯在打字框里一问一答。“想象一下，如果你在 iPhone 上只能用 iMessage 来处理所有事情，那该有多别扭？”手机之所以成功，正是因为苹果的应用商店（App Store）允许每一个应用根据自身业务定制独特的富图形界面。因此，下一代电商 Agent 尽管允许双向语音与对话，但展现形式必须以高信息密度的视觉组件为主，纯文本交互只适用于极其有限的窄门类场景。 ```mermaid flowchart TD accTitle: AI交互范式演进对比 accDescr: 对比传统Chatbot模式与未来沉浸式Agent模式在交互层与表现层的差异 subgraph Today["当前 Chatbot 局限"] A1["用户输入提示词"] --> B1["纯文本 / 基础卡片"] B1 --> C1["信息扁平、难以横向比对"] end subgraph Future["下一代消费级 Agent"] A2["意图理解与自然对话"] --> B2["动态富 UI 渲染"] B2 --> C2["高清图集 / 空间地图"] B2 --> C3["沉浸式视频 / 实时状态"] B2 --> C4["结构化参数与快捷操作"] end ``` ### [09:42](https://www.youtube.com/watch?v=TyNufBceyZo&t=582s) 软件的护城河：App 仅占 20%，真正的壁垒在线下万亿级运营 面对市场对“AI 时代软件公司护城河是否不复存在”、“人人都能通过 Vibe Coding 复制 Airbnb”的质疑，切斯基展现出了极强的底气。他回顾道，ChatGPT 刚推出第三方插件生态时，Airbnb 的股价曾单日下挫约 7%，资本市场一度恐慌旅行平台会被中介化。但切斯基强调，单纯复制前端界面在 AI 时代确实变得轻而易举，可 Airbnb 真正庞大的部分用户根本看不到。 曾领导 Meta 的 Llama 大模型团队、现任 Airbnb CTO 的艾哈迈德·阿尔-达勒（Ahmad Al-Dahle）刚加入公司时曾感叹：“Airbnb 远比一个 App 复杂得多。”切斯基透露，普通用户所接触的客户端只占整个 Airbnb 业务体量的 20%： - **全球履约规模**：每晚有四五百万人入住分布在全球 100 多个国家的 Airbnb 房源，背后涉及极端复杂的跨国清结算、纠纷仲裁与多语言客服网络。 - **万亿级风控与保险**：Airbnb 为房东提供高达 300 万美元的财产与防盗损坏保障（AirCover），乘以每晚在保的百万套房源，每天承保的资产规模高达数万亿美元。 - **双端生态治理**：平台背后还运行着面向 550 万名房东的独立管理系统与运营工具。 切斯基坦言，如果单看前台的住客端 App，AI 时代确实可能有人做出体验更好的版本，而 Airbnb 的目标就是在别人动手前率先用 Agent 将其彻底重构。但核心在于，Airbnb 本质上是一家运营驱动的线下实体体验公司，线下的重型基础设施与信任网络是任何代码生成工具都无法凭空变出来的。 ### [12:15](https://www.youtube.com/watch?v=TyNufBceyZo&t=735s) 消费级 AI 的缺位：从“文本加按钮”到全新视觉范式 切斯基指出了当下 AI 行业一个极其严重的失衡现象：几乎所有的优秀团队都在涌向企业级（Enterprise）市场和代码辅助工具。他提到，几个月前在 TBPN 节目上他就引用过 Y Combinator 的统计数据，最近这一趋势依然没有改变——YC 当期 175 家初创公司中，做消费级（Consumer）产品的仅仅只有 16 家左右。 切斯基认为，这给消费级产品留下了巨大的蓝海。目前的 AI 消费端体验基本停滞在“文本、图片与几个按钮”的粗糙阶段。然而，随着底层图像与视频生成模型的飞跃，产品完全能够实现前所未有的突破性视觉交互： - **颠覆静态对比**：传统的酒店或民宿比价往往受制于不同网站杂乱的网页排版，而未来的视觉 Agent 能够把不同房源的环境、实景图、空间布局在统一的视觉标准下并排重构。 - **沉浸式感知空间**：用户不仅能看到房间，还能通过动态视觉直接“走进”周边街区，将地图位置、周边氛围与自然对话无缝融合。 - **摆脱传统网页束缚**：用户不再需要跳进跳出各种 OTA 网站，Agent 应当提供比传统网页更加结构化、更富艺术感的高保真信息流。 ### [14:48](https://www.youtube.com/watch?v=TyNufBceyZo&t=888s) 解构旅行六步旅程：大模型如何颠覆目的地发现与行程规划 为了具象化说明 AI 与交互界面的落地场景，切斯基将一次完整的旅行拆解为六个连续阶段： ```mermaid flowchart LR accTitle: 完整旅行体验的六个阶段 accDescr: 展示从目的地探索到当地即兴体验的六阶段全流程 subgraph Discovery["AI 最易革命性突破的前期阶段"] S1["第一步：目的地探索<br>(去哪里？)"] S2["第二步：机票交通<br>(如何抵达？)"] end subgraph Logistics["依赖线下重资产与运营的履约阶段"] S3["第三步：住宿选择<br>(酒店 / 民宿)"] S4["第四步：游玩与餐饮<br>(做什么？)"] S5["第五步：地面后勤<br>(租车 / 落地服务)"] S6["第六步：在地即兴体验<br>(同城临时决策)"] end S1 --> S2 --> S3 --> S4 --> S5 --> S6 ``` 切斯基分析，大模型技术结合富交互界面，在“第一步：目的地探索”和“第二步：交通机票”上将带来真正的革命性突破： - **直觉式灵感推荐**：用户可以提出模糊且个性化的需求，例如“我想找一个类似巴黎、但物价更亲民、适合 8 月游玩且有歌剧演出的地方”，大模型能够给出极具洞察力的视觉化行程提案。 - **机票分发门槛降低**：全球航空分销基本被三大全球分销系统（GDS）垄断，任何团队接通 API 都能快速构建机票比价工具，技术壁垒极低。 - **现有 Chatbot 的转化真相**：目前的聊天机器人更像是“超级版 Google”，主要扮演着向外部网站输送推荐流量（Referral Traffic）的角色。切斯基透露，这类来自 AI 助手的引流转化率甚至高于传统 Google 搜索，因此当下对旅游平台反而是利好的补充；想要真正脱媒（Disintermediate）现有的旅游巨头，AI 必须深入后端的住宿、服务与在地全旅程，而这必须依赖更高级的交互载体。 ### [16:33](https://www.youtube.com/watch?v=TyNufBceyZo&t=993s) 从 Hipmunk 痛苦指数到行业困惑：代码革命爆发，为何 C 端产品尚未破局？ 主持人顺着切斯基的观点回忆起早期由 Y Combinator 孵化、Reddit 联合创始人史蒂夫·霍夫曼（Steve Huffman）参与创立的机票搜索工具 Hipmunk。Hipmunk 曾经推出了革命性的“痛苦指数（Pain Index）”排序算法：它不仅看票价高低，还会综合考量中转停留时间、是否需要清晨赶往机场等身体疲惫度因素，进行多维度加权展示。主持人感叹，放在今天，这种充满巧思的界面可能一个周末靠 Vibe Coding 就能写出来。 然而令人费解的现实是：打开当下的 App Store 排行榜，霸榜的 AI 工具依然千篇一律是 ChatGPT、Gemini、Claude 这些文本对话框。正如 Google DeepMind CEO 戴米斯·哈萨比斯（Demis Hassabis）在 YC 演讲时指出的那样，行业出现了一个极其诡异的割裂现象：一方面，企业端和工程端已经迎来了近乎“超级智能”的效率爆发——甚至在 Airbnb 内部，已经有 60% 的代码由 AI 协助编写；但另一方面，面向大众的消费级产品，却迟迟没有迎来属于自己的 iPhone 式破局时刻。 ### [17:34](https://www.youtube.com/watch?v=TyNufBceyZo&t=1054s) 为什么 AI 创业全扎堆企业端？消费级产品的破局时机与多模态突破 主持人抛出了一个尖锐的观察：现在的科技圈似乎缺少让人眼前一亮的全新 3A 级大作或风靡全球的爆款电子游戏，大家不禁困惑——这究竟是创意的缺失、冒险精神的匮乏，还是因为市场上的热钱实在太多了？对于擅长 AI 的创业者来说，进军企业级（Enterprise）市场几乎成了一条不用动脑的康庄大道：容易拿融资、容易签客户、变现快且风险低；反观消费级（Consumer）市场，风险大且极难捉摸。 Brian Chesky 认为这是一个切中要害的问题，并剖析了背后的多重深层原因： 1. **AI 实验室严重缺乏产品与设计基因**：最近有推文提到行业内涌现出了大约 60 家新成立的 AI 实验室（Neo labs）。Chesky 接触过其中不少研究员，发现他们几乎在跟风做一模一样的事情——不是钻研科学计算、代码生成，就是试图训练某种差异化底座模型。团队里清一色全都是纯 AI 技术背景的人，完全没有产品经理和设计师的身影。许多技术人员甚至抱有一种极其过分简化的想法，认为“只要通用人工智能（AGI）做出来了，消费级产品的问题自然就会被它解决”，完全忽视了做消费级应用必须具备对用户体验的独特洞见与品味。 2. **硅谷盲目追逐风口的“氛围感”（Vibe）与 YC 互买飞轮**：硅谷变得比 2007 年他刚来时更加依赖潮流风向。当大家看到做企业端赚钱，便一窝蜂涌入。在孵化器（如 Y Combinator）中，早期甚至形成了一种心照不宣的飞轮机制：大家互相把同期批次的其他初创公司当成自己的第一批付费客户。过去创业者还会努力拉彼此作为个人消费者来试用产品，而现在大家发现直接作为 B 端客户互相采购高效得多——就像五家做代码生成的初创公司聚在一起互相对彼此说：“用我的工具写代码吧！” 3. **企业级市场正在迅速变成红海**：每个人都认为做消费级太难，但 Chesky 直言：“在企业端与 10 家跟你的点子完全一样的公司惨烈内卷，同样极其艰难。” ```mermaid flowchart TD accTitle: AI 初创企业扎堆 Enterprise 与消费级破局路径 accDescr: 展示了当前 AI 创业团队过度集中于企业端的恶性内卷，以及多模态技术驱动下消费级创新的突破路径。 subgraph Problem["当前行业困境：企业端内卷"] A["纯技术团队涌现<br/>（60+ 实验室缺乏设计师/产品）"] --> B["扎堆做代码生成/科学模型"] B --> C["互为客户刷 ARR<br/>（YC 式 B2B 循环）"] C --> D["过度拥挤的同质化竞争"] end subgraph Solution["破局转机：多模态驱动的消费级革命"] E["图像/视频模型突破<br/>（如 Seedance、高质图像生成）"] --> F["突破纯文本聊天限制<br/>（跨入实时生成界面）"] F --> G["标杆性消费级产品诞生<br/>（类似当年苹果开创范式）"] end D -.->|需逆向思维| G ``` Chesky 预测，眼下虽然是企业级软件的天下（甚至像 ChatGPT 这样现象级的消费级应用，其后续巨大营收很可能也是来自类似 Codex/Claude 这种代码生成与开发能力），但图像与视频生成模型正在迎来宛如“Claude 代码能力突破”般的关键质变节点。以字节跳动的 Seaweed（即即梦/Seedance）等视频与图像生成模型为例，AI 正在彻底冲破单调的纯文本对话框。在未来一到两年内，行业将具备实时生成视觉化动态交互界面的能力。未来两年必定会爆发一场消费级产品的壮丽革命，而这迫切需要几家领军企业站出来成为新的行业参照物（Reference points）——在过去，扮演这个角色的通常是史蒂夫·乔布斯（Steve Jobs）时代的苹果公司；而现在，包括 Airbnb 在内的探路者也必须尝试去踏出这条崭新的道路。 --- ### [20:56](https://www.youtube.com/watch?v=TyNufBceyZo&t=1256s) 拒绝千篇一律的同质化：竞争是给失败者准备的 当被问及是否担心全行业如今都在做同一件事——无论从云巨头（Hyperscalers）、顶级实验室到新生代实验室，大家几乎都在扎堆做“能够在电脑上替用户自动操作的 Agent（智能体）”时，Chesky 给出了清醒的建议。 他借用 Airbnb 最早期的投资人之一彼得·蒂尔（Peter Thiel）在《从 0 到 1》（*Zero to One*）中的经典名言警示创业者：“竞争是给失败者准备的（Competition's for losers）。”如果一个赛道里你只是第十家做同样事情的公司，除非你能展现出绝对压倒性的优势，否则在拥挤的赛道里极难生存。创业的秘诀永远是“在所有人向左拐时，坚定地向右拐（Zig when everyone else is zagging）”。 - **Airbnb 的逆向案例**：2007 年 Airbnb 创立之初，全硅谷都在疯狂追捧社交网络（Social Networking），团队当时也曾面临随大流做社交产品的诱惑。最终他们却在一个周末偶然充气了三张充气床垫做成了“AirBed & Breakfast”。当时所有人都不看好，甚至被嘲笑为“有史以来最烂的点子”，但它最终成为了“有史以来最成功的烂点子”，独辟蹊径开创了全新市场。 - **Uber 与 OpenAI 的历史相似性**：Uber 在创立时没有模仿任何人；OpenAI 和 Anthropic 创办初期，生成式 AI 根本不是所谓的热门风口，他们也并未去抄袭市面上的成熟公司。 - **巨大的未开垦蓝海**：Chesky 强调，一旦某个方向形成了全民狂热的趋势，它往往就已经太拥挤了。事实上，整个美国经济乃至全球商业中，还有大片传统行业和真实需求根本没有被 AI 真正触及。 --- ### [23:00](https://www.youtube.com/watch?v=TyNufBceyZo&t=1380s) 创意阶层勿作壁上观：设计师不该在 AI 时代沦为看客 针对目前在罗德岛设计学院（RISD）等顶尖艺术院校以及广泛的设计师社群中普遍弥漫的抵触 AI、担忧失业的情绪，作为 RISD 校友的 Chesky 表达了深切的忧虑与急迫的呼吁。他直言，最担心的就是一整代优秀的设计师、艺术家和创意人才选择在 AI 浪潮前“作壁上观”（Sit out），而这恰恰是他有生之年所见到的属于创意人群的最大历史机遇。 Chesky 指出，设计师在历史上并非第一次落后于时代技术浪潮： 1. **互联网与 Web 时代的惨痛前车之鉴**：30 到 35 年前互联网刚刚兴起时，主流权威的设计大师们根本不屑于涉足所谓的“网页设计”（Web design），甚至将其视作下等、低端的设计门类，全员死守传统纸媒印刷设计。这导致早期优秀的顶尖设计力量集体缺席互联网，造成了视觉设计与底层工程技术之间严重的鸿沟脱节。 2. **“产品经理”职能的诞生与反思**：正是因为当时的设计师自我设限、眼界狭窄，行业才不得不催生出“产品经理（Product Manager）”这一中介角色来填补沟通空白。Chesky 强调自己并不否定产品经理的价值，但在工业设计领域，工业设计师本身就是产品负责人；在建筑学中，建筑师就是掌控全局的最高指挥。如果当下的设计师依然拒绝拥抱 AI，历史就会重演，最终的结果只能是由工程师和产品经理来替设计师做决定。 > “如果设计师选择作壁上观，结果就是工程师和产品经理越俎代庖来替你们做设计。” Chesky 向创意群体发出了直言不讳的建议： - **消解技能边界**：现在的设计师完全可以身兼工程师与产品经理之职。他感慨道：“如果我今天 26 岁重新创立 Airbnb，我一定会直接用 Claude Code 现场实时写代码。”现在他在业余时间纯粹为了乐趣也正在这么做。 - **走出传统线框图，直接用代码构建**：未来的互联网绝不可能停留在冰冷乏味的纯文字对话框，它必将极其丰富且高度视觉化。先进的图像和视频模型正是赋予设计师打造下一代超丰富界面的绝佳画笔。设计师必须扔掉刻板的静态线框图（Wireframes），挽起袖子直接下场写代码，主动用 AI 武器来定义下一代消费级交互界面的模样。 ### [26:37](https://www.youtube.com/watch?v=TyNufBceyZo&t=1597s) 消费级获客渠道的变迁与“横幅盲区”困境 主持人抛出一个尖锐的增长问题：在当今的 App Store 中，获客结构与分发逻辑是否发生了根本性变化？早期 Airbnb 凭借极强的 SEO（搜索引擎优化）、Google 流量以及裂变邀请机制（Referral Program）取得了巨大成功，但这些传统打法在今天是否依然奏效，抑或已经彻底失灵？对于拥有优秀产品的创业者而言，将产品推向大众消费者的全新打法究竟是什么？ Brian Chesky 指出，行业确实需要全新的营销模式。他幽默地调侃道，在硅谷所有高管职位中，流动率最高的大概就是 CMO（首席营销官）——相比之下，CFO 和 CTO 的变动极少。这一现象背后的核心逻辑在于：**市场营销的有效法则每隔几年就会彻底改变，曾经奏效的策略会迅速过时淘汰**。营销是企业中最艰难的职能之一，因为任何打法一旦证明有效，所有人便会蜂拥而至争相模仿，导致其迅速贬值并失去新鲜感。例如网红营销（Influencer Marketing）最初效果极佳，但当所有人都在做时，受众便开始产生免疫，这在心理学和广告学上被称为“横幅盲区”（Banner Blindness）——即用户在反复看到同类形式的信息后，会在大脑中自动将其过滤。 因此，不仅产品界面设计需要从传统的输入框走向更具沉浸感、创意感的新范式，营销本身也必须推陈出新。回顾 Airbnb 的早期经历，Chesky 与联合创始人 Joe Gebbia 曾采取过许多在当时看来匪夷所思的“怪招”：手工制作并售卖极具政治收藏价值的奥巴马与麦凯恩麦片盒（Obama O's & Cap'n McCain's）、在社交媒体刚刚兴起时全力押注社媒传播、在纸媒时代主动围追堵截各路记者以争取曝光。如今常有创业者向他请教如何获取用户，他的回答非常直接：**所有 2008 年的旧战术在今天早已失效，必须根据当下的环境寻找属于自己的全新打法**。 与早期相比，消费级互联网的分发环境已经高度成熟。当年 Airbnb 创立时，App Store 才刚起步，互联网本身仍在以每年新增数亿用户的爆发式速度扩张，创业团队只需上线一个网站，就能顺应整个互联网的增长红利水涨船高。尽管如今成熟的分发渠道壁垒森严，但 App Store 榜单前列依然不断涌现出新应用（目前主要以各类 AI Chatbot 为主）。这一现象证明：**只要做出真正具有颠覆性、突破性的革命性产品，消费者依然会主动找到你**。 - **营销的核心两大原则**： 1. 做出足够颠覆性的产品，让用户主动去发现它； 2. 开拓属于自己的全新战术，因为任何标准化的常规打法往往已经失去吸引力。 --- ### [29:23](https://www.youtube.com/watch?v=TyNufBceyZo&t=1763s) 突破百亿核心基本盘：Airbnb 未来四大扩张曲线 主持人询问 Airbnb 的未来增长空间：在已经主导短租市场（STR, Short-Term Rental）的前提下，未来的核心机会是继续将现有品类打磨到极致并享受其自然增长，还是去寻找另一个具有类似体量的新市场？ Chesky 明确表示，**Airbnb 未来最大的增长机遇在于拓展核心业务之外的全新版图**。目前 Airbnb 的核心短租业务总预订价值（Gross Booking Value, GBV）已接近 1,000 亿美元。尽管随着渗透率提升，这块核心盘有望翻倍至 2,000 亿美元（目前平均每 8 到 9 个人选择入住传统酒店，才会有 1 个人选择入住 Airbnb，只要能再争取到 1 个人，总规模就能达到 2,000 亿美元），但更为广阔的增量空间存在于短租之外的四大方向。 ```mermaid flowchart TD accTitle: Airbnb 未来增长矩阵示意图 accDescr: 展示 Airbnb 核心短租业务与四大拓展品类之间的关系。 subgraph Core ["核心基本盘 (GBV ~1000亿美元)"] A["短租民宿业务 (STR)<br/>目标翻倍至 2000 亿美元"] end subgraph Expansion ["四大业务拓展方向"] B["精品与独立酒店<br/>(反抗 OTA 抽成与被迫加盟)"] C["本地与旅行服务平台<br/>(打造服务业的 Amazon)"] D["中长期旅居 (30天以上)<br/>(远程办公与数字游民生态)"] E["线下沉浸式体验与营销<br/>(如芭比梦幻屋等标志性 IP)"] end A --> B A --> C A --> D A --> E ``` 具体而言，Airbnb 正在积极布局的业务包括： 1. **精品与独立酒店市场**：全球酒店行业规模是 Airbnb 的 8 到 9 倍。全球约有一半的酒店属于独立运营的精品酒店，而非连锁集团。这类酒店在大型 OTA（在线旅行社平台）上面临高昂的抽成费率，且由于缺乏连锁品牌的会员忠诚度计划（Loyalty Programs）和议价筹码，正面临被万豪（Marriott）、希尔顿（Hilton）等巨头强制收编加盟的生存困境。Airbnb 计划成为这些独立精品酒店更友好的全球分发渠道，这是一个体量达数十亿美元的高价值市场。 2. **综合服务交易平台（Amazon for Services）**：当前消费者可以一键打车、一键叫外卖，但日常旅行和生活中还有 80 多种不同的服务需求缺乏统一的线上平台。虽然单个垂直细分市场的规模可能有限，但将这 80 多个垂直品类整合在一起，将汇聚成一个极其庞大的市场。 3. **30天以上的中长期旅居（Living / Extended Stays）**：随着笔记本办公和分布式远程工作的普及，数字游民与跨城旅居人群规模激增，超长租期住宿已成为平台增长最快的细分赛道之一。 --- ### [31:32](https://www.youtube.com/watch?v=TyNufBceyZo&t=1892s) 营销的本质是“与众不同”与打造超级文化符号 主持人谈及旧金山街头铺天盖地的科技广告牌，探讨传统广告渠道究竟是呈现周期性轮回（如网红营销降温后未来是否会重新复苏），还是会走向永久衰落。 Chesky 分析认为，旧渠道并不会真正消失，只是不再具备“单靠这一招就能带来爆发式增长”的神奇魔力；它们会变成企业必须维持的基础配置，但无法再充当破局的核心杠杆。谈及广告效果，Chesky 坦率承认：“**广告对我们每个人的影响，往往远比我们自己愿意承认的要大得多。**”他提到，Airbnb 每年在广告投放上的支出高达约 10 亿美元；很多用户主观上觉得自己不受广告影响，但心理学研究证实，当一个人把某条信息看到第 7 遍时，无论其内心有多抵触，潜意识都开始逐渐接受。 然而，常规广告投放的效率远不如全新创意所带来的投资回报率（ROI）。Chesky 列举了 Airbnb 历史上最成功的现象级营销案例：在电影《芭比》（Barbie）上映期间，Airbnb 将马里布的一栋真实豪宅改造成了全粉色的“芭比梦幻屋”（Barbie Malibu DreamHouse）。该事件引发全网狂欢和全球媒体自发报道，其 ROI 远超公司历史上投放过的任何传统付费广告。主持人随后也以红牛赞助鲍姆加特纳（Felix Baumgartner）完成太空跳伞为例：红牛仅花费不到 100 万美元，却在全球收获了超过 10 亿次播放，带来了近千倍的营销回报；但这种极具冲击力的大事件一旦有十家公司模仿，效果就会断崖式下跌，因此红牛后续转而挑战穿机库飞行等全新特技，绝不重复自我。 在对谈结尾，话题延伸到了荒诞而有趣的硬核营销点子。主持人提到 TBPN 长期以来一直梦想租用飞艇（Blimp）做全美巡回直播，甚至利用星链（Starlink）网络让嘉宾从空中私人飞机连线直播。Chesky 对此表现出极大兴趣，现场提议双方可以展开合作：打造一个可以过夜入住的 Airbnb 观光飞艇（空中巡航 Sky Cruise），沿着洛杉矶到旧金山航线飞行两到三天，并在飞艇上完成播客实况直播。双方在欢快的氛围中达成初步意向，为本期对谈画上圆满句号。 ### [35:13](https://www.youtube.com/watch?v=TyNufBceyZo&t=2113s) 节目收尾与致谢道别 在全场深度访谈的尾声，对话在轻松而热烈的氛围中画上句点。主持人在节目结束时向 Brian Chesky 致以真诚的赞赏与感谢，称赞道“你太棒了”（"You're the man"），随后向收看节目的观众道别致意（"Goodbye, guys"），正式结束了本期关于 Airbnb 创业心态重塑与 AI 时代消费级产品未来的精彩对话。 ## Notable quotes > [02:40](https://www.youtube.com/watch?v=TyNufBceyZo&t=160s) 不要试图重修整栋房子，先挑选一间房间，把它做到极致完美，然后再一间一间推进。 — Brian Chesky > [02:40](https://www.youtube.com/watch?v=TyNufBceyZo&t=160s) 真正的领导力是在场，而不是缺席。商学院和管理咨询教导 CEO 应该‘信任团队并闪到一边’，但我认为‘信任’绝不等于‘撒手不管’。作为骑兵将军，你必须跨在马背上、身处战场一线，而不是躲在海外大后方画蓝图。领导力永远来自身先士卒，而不是殿后观望。 — Brian Chesky > [06:06](https://www.youtube.com/watch?v=TyNufBceyZo&t=366s) 如果是 26 岁时的我、Joe 还有 Nate，我们绝对有能力把今天的 Airbnb 彻底干趴下。 — Brian Chesky > [08:47](https://www.youtube.com/watch?v=TyNufBceyZo&t=527s) 那是未来，但那绝不是我们今天所说的 Chatbot（聊天机器人）。 — Brian Chesky > [08:47](https://www.youtube.com/watch?v=TyNufBceyZo&t=527s) 想象一下，如果你在 iPhone 上只能用 iMessage 来处理所有事情，那该有多别扭？ — Brian Chesky > [09:42](https://www.youtube.com/watch?v=TyNufBceyZo&t=582s) Airbnb 远比一个 App 复杂得多。 — Ahmad Al-Dahle > [17:34](https://www.youtube.com/watch?v=TyNufBceyZo&t=1054s) 在企业端与 10 家跟你的点子完全一样的公司惨烈内卷，同样极其艰难。 — Brian Chesky > [20:56](https://www.youtube.com/watch?v=TyNufBceyZo&t=1256s) 竞争是给失败者准备的（Competition's for losers）。 — Peter Thiel > [20:56](https://www.youtube.com/watch?v=TyNufBceyZo&t=1256s) 在所有人向左拐时，坚定地向右拐（Zig when everyone else is zagging）。 — Brian Chesky > [23:00](https://www.youtube.com/watch?v=TyNufBceyZo&t=1380s) 如果设计师选择作壁上观，结果就是工程师和产品经理越俎代庖来替你们做设计。 — Brian Chesky > [23:00](https://www.youtube.com/watch?v=TyNufBceyZo&t=1380s) 如果我今天 26 岁重新创立 Airbnb，我一定会直接用 Claude Code 现场实时写代码。 — Brian Chesky > [26:37](https://www.youtube.com/watch?v=TyNufBceyZo&t=1597s) 市场营销的有效法则每隔几年就会彻底改变，曾经奏效的策略会迅速过时淘汰。 — Brian Chesky > [26:37](https://www.youtube.com/watch?v=TyNufBceyZo&t=1597s) 所有 2008 年的旧战术在今天早已失效，必须根据当下的环境寻找属于自己的全新打法。 — Brian Chesky > [31:32](https://www.youtube.com/watch?v=TyNufBceyZo&t=1892s) 广告对我们每个人的影响，往往远比我们自己愿意承认的要大得多。 — Brian Chesky ## Mentioned Brian Chesky, Joe Gebbia, Nathan Blecharczyk, Paul Graham, Jordi, Ahmad Al-Dahle, Demis Hassabis, Steve Huffman, Steve Jobs, Peter Thiel, Felix Baumgartner, Airbnb, TBPN, Project Y, AirCover, Meta, Llama, Y Combinator, Hipmunk, Reddit, ChatGPT, Gemini, Claude, Claude Code, Google DeepMind, ByteDance, Seaweed (Seedance/即梦), OpenAI, Anthropic, 罗德岛设计学院 (RISD), Uber, 万豪 (Marriott), 希尔顿 (Hilton), 亚马逊 (Amazon), 苹果 (Apple), App Store, 星链 (Starlink), 红牛 (Red Bull), 《芭比》（Barbie）, 《从 0 到 1》（Zero to One） ## Video description This is our full conversation with Brian Chesky, recorded live on TBPN. We discuss how Airbnb reaccelerated growth by rebuilding startup intensity and embracing “founder mode” leadership, why Chesky believes today’s AI chatbots are not the right interface for the future of travel and e-commerce, how Airbnb is using AI across the business with 60% of its code now written by AI, and why he believes the next major wave of AI innovation will come from immersive consumer products, visual interfaces, and agentic experiences rather than another wave of enterprise software. Sign up for TBPN’s daily newsletter at TBPN.com Follow TBPN: https://TBPN.com https://x.com/tbpn https://open.spotify.com/show/2L6WMqY3GUPCGBD0dX6p00?si=674252d53acf4231 https://podcasts.apple.com/us/podcast/technology-brothers/id1772360235 https://www.youtube.com/@TBPNLive [Source](https://www.youtube.com/watch?v=TyNufBceyZo)

[2:40] Brian Chesky: 不要试图重修整栋房子，先挑选一间房间，把它做到极致完美，然后再一间一间推进。 [2:40] Brian Chesky: 真正的领导力是在场，而不是缺席。商学院和管理咨询教导 CEO 应该‘信任团队并闪到一边’，但我认为‘信任’绝不等于‘撒手不管’。作为骑兵将军，你必须跨在马背上、身处战场一线，而不是躲在海外大后方画蓝图。领导力永远来自身先士卒，而不是殿后观望。 [6:06] Brian Chesky: 如果是 26 岁时的我、Joe 还有 Nate，我们绝对有能力把今天的 Airbnb 彻底干趴下。 [8:47] Brian Chesky: 那是未来，但那绝不是我们今天所说的 Chatbot（聊天机器人）。 [8:47] Brian Chesky: 想象一下，如果你在 iPhone 上只能用 iMessage 来处理所有事情，那该有多别扭？ [9:42] Ahmad Al-Dahle: Airbnb 远比一个 App 复杂得多。 [17:34] Brian Chesky: 在企业端与 10 家跟你的点子完全一样的公司惨烈内卷，同样极其艰难。 [20:56] Peter Thiel: 竞争是给失败者准备的（Competition's for losers）。 [20:56] Brian Chesky: 在所有人向左拐时，坚定地向右拐（Zig when everyone else is zagging）。 [23:00] Brian Chesky: 如果设计师选择作壁上观，结果就是工程师和产品经理越俎代庖来替你们做设计。 [23:00] Brian Chesky: 如果我今天 26 岁重新创立 Airbnb，我一定会直接用 Claude Code 现场实时写代码。 [26:37] Brian Chesky: 市场营销的有效法则每隔几年就会彻底改变，曾经奏效的策略会迅速过时淘汰。 [26:37] Brian Chesky: 所有 2008 年的旧战术在今天早已失效，必须根据当下的环境寻找属于自己的全新打法。 [31:32] Brian Chesky: 广告对我们每个人的影响，往往远比我们自己愿意承认的要大得多。

[2:44] Brian Chesky: Great leadership is presence, not absence. I think a generation of management consulting or management school taught us that CEOs should trust their people and get out of the way... If you're a cavalry general, you should be on a horse. You should be on the battlefield. [6:03] Brian Chesky: I do not think a chatbot is the right interface for travel or e-commerce. That might be a radical statement... I think the current chatbot paradigm—what you're describing, it can do it. I just don't think it's the best way to do what you just described. I think there's a much more immersive way to do that. [16:34] Brian Chesky: Enterprise makes a lot of sense until everyone does enterprise. And then suddenly, like everyone's saying consumer's hard. You know what's also hard? Competing with 10 other companies in enterprise doing your idea, too. [23:21] Brian Chesky: In industrial design, there's no product managers for the most part—they're industrial designers. In architecture, the architect is the product manager. [34:06] Brian Chesky: Okay, you heard it here first. Airbnb and TBPN are going to work together on a blimp partnership.

[0:00] Brian, how are you doing? [0:01] Good, how you guys doing? We're doing [0:02] fantastic. Welcome back to the show. [0:05] >> Yeah. [0:06] Why don't we kick it off with just [0:08] you know, an update on how 2026 is [0:10] going, how the business is going, what's [0:12] new in your world? [0:14] Yeah, I mean things are really good. Um, [0:15] we've accelerated growth for the first [0:17] time since the pandemic. We grew 10% um, [0:20] last year in revenue. And this quarter [0:23] we announced the revenue was 18%, so [0:25] from 10 to 18%, which is a pretty big [0:28] acceleration. Yeah, thank you. [0:29] Um, and and marketplaces are really [0:31] really hard, right? Like it's kind of [0:33] like gravity. Once a marketplace at our [0:35] size doing around 100 billion dollars in [0:36] gross booking a year start to come down, [0:39] it's really hard to tip that curve. So, [0:41] this has been a pretty big feat for the [0:42] team to be able to do it. [0:43] >> So, how did you do it? [0:45] Um, you know, it's kind of funny. A [0:47] number of years ago [0:48] we So, I noticed that, you know, [0:52] as we got bigger, we started to losing a [0:53] little bit of that startup intensity, [0:55] that startup energy. And I asked myself [0:57] like how can we get that energy back? [0:58] And we basically took a very small team. [1:01] We we named the team project Y. The name [1:03] doesn't really matter, but it basically [1:06] I took a very small team of people and [1:07] we said we're going to focus on a very [1:10] very small service area. And we decided [1:11] to focus on conversion rate, the guest [1:13] journey. And I basically tried to work [1:15] with the team like as if we were at [1:17] Rausch Street. Rausch Street was the [1:18] apartment that Joe and I started. And I [1:20] said we're going to act like a startup. [1:22] And the team just grinded really really [1:23] hard. We weren't working like a big [1:24] company. We were a very small team [1:26] grinding really hard, focusing on [1:28] obsessing over the customer experience, [1:30] really looking at the data. And we [1:31] really got a lot of points on the board. [1:33] Then we really started taking these pods [1:35] and we really started working with the [1:37] teams trying to coach them how we worked [1:38] in the early days. And I just think the [1:40] pace increased, the intensity increased. [1:42] We really like tried to bring in like [1:44] world class people onto the team. They [1:46] got very very focused. We tried to get [1:48] all the management and bureaucracy out [1:50] of their way. [1:51] And it's been a couple years. In fact, [1:53] we're doing this last year, but one of [1:54] the things is as you know, financial [1:56] results are lagging indicators, [1:57] especially when you're a big company. [1:59] So, it takes sometimes a while to get [2:01] the the the financial numbers to reflect [2:04] what's happening inside the company. [2:05] But, I feel like we're a startup again, [2:07] like more than ever. We're like We feel [2:09] much younger and smaller than earlier. [2:11] And I think with AI, that would be other [2:13] thing is 60% of our code is now written [2:16] by AI, which is twice our benchmark of [2:20] our competitors and peers. And it's [2:22] really, really helping us. AI, I think, [2:24] is a huge boon to us. I don't know if [2:25] it's helped the OTAs, but it's helped us [2:27] with customer service. The cost for [2:29] customer service tickets down 10%. 40% [2:32] people who contacted me, the AI solves [2:34] the problem for them. And we brought it [2:37] through the entire journey. So, [2:38] everything is really accelerating. On [2:40] the [2:41] on the the the the the special team that [2:44] went into optimize conversion rate, I [2:46] imagine you have folks internally whose [2:48] job it was basically the customer [2:50] journey already. Then, you bring in your [2:53] special team. And the is there some sort [2:55] of culture clash there? Like, how do you [2:57] set people up for success to, you know, [3:00] not get too political in that [3:02] environment? Actually see it as an [3:04] opportunity for a win, some fresh eyes, [3:06] some fresh ideas. Like, what what is [3:08] required to actually have success? [3:11] Because I think a lot of big companies [3:12] that bring in McKinsey and they put [3:14] together a big deck and everyone freaks [3:15] out and thinks they're getting fired. [3:16] And maybe some of the good ideas are [3:18] surfaced, but it never really goes [3:20] through. So, what like what what what [3:22] what do you have to communicate to the [3:24] team that is receiving information from [3:26] this new this new quiet team? [3:29] Yeah, so it's a great thing. I mean, [3:30] really what I'm talking about has played [3:32] out over like really 5 years. [3:34] >> Yeah. And the term founder mode, what it [3:36] really meant that Paul Graham wrote was [3:38] about me like skipping layers of [3:40] management and going into the details [3:42] with teams. And instead of trying to me, [3:44] here's my advice. If you're a CEO or a [3:46] leader of a company that's big, don't [3:47] try to change the whole company. Try to [3:49] change a corner of the company. It's [3:51] kind of like don't renovate the whole [3:52] house, pick like one room and make it [3:54] perfect and then go room to room. So, I [3:56] really told teams like, "Hey, like I [3:58] actually I actually didn't replace the [4:00] team. I took the team that was already [4:02] working. I like handpicked some people [4:04] on the team, but I really just taught [4:06] them the pace and I would review work [4:08] very regularly. So, I'd sometimes review [4:10] the work weekly or even daily to just [4:13] teach them a level of intensity, a level [4:15] of perfection. [4:17] And it was unfamiliar and uncomfortable [4:19] and I will tell you that not everyone [4:21] liked it and some people didn't stick [4:22] around the company cuz they didn't like [4:24] that way of working, but those who stuck [4:26] around and the vast majority did, they [4:28] realized, "Wow, actually when the CEO is [4:30] involved, it's actually easier. There's [4:32] less bureaucracy. I try to make the work [4:33] better and I try to clear all [4:35] obstacles." And so, I basically did that [4:37] group by group and the great thing is [4:39] like I was in massive number of details, [4:42] reviewing everything for years, but [4:44] eventually it's like muscle memory. It's [4:46] like [4:47] an instructor or a coach. You have to [4:50] teach them, but once they learn that and [4:51] they have the muscle memory like a golf [4:53] swing, you can step back and now it's [4:55] muscle memory. So, now I don't have to [4:57] be in all these details, all these [4:58] teams, but I would just say a couple [5:00] things to people listening like great [5:02] leadership is presence, not absence. I [5:04] think a generation of management [5:06] consulting or management school, [5:08] management teaching taught us that CEOs [5:11] should trust their people and get out of [5:12] the way and I don't think trust and get [5:14] out of the way are the same thing. [5:15] Again, I think leadership is presence [5:18] and if you if you you should actually be [5:19] partnering with your people. You should [5:21] be on the field with them. If you're a [5:23] If you're a cavalry general, you should [5:24] be on a horse. You should be on the [5:26] battlefield. You're not like you're not [5:28] overseas somewhere else just writing out [5:30] blueprints. You got to be on the field [5:32] with them and you got to be leading from [5:33] example. I think leadership is from the [5:35] front, not from the back. And so, these [5:36] are just some of the things we do and I [5:38] think [5:39] AI is going to create like the [5:41] equivalent of AI founder mode, which is [5:42] now you got to be even more hands-on and [5:44] I I think in the era of AI there should [5:46] be any pure people managers because [5:49] you're so close [5:51] to the details, to the data that [5:52] everyone has the opportunity to be [5:54] hands-on. And it's hard to imagine only [5:56] managing people and not agents. So, I [5:59] think this hands-on approach is for [6:01] everyone inside of a company. How do uh [6:03] Jordi, please. [6:06] Do you think a lot about what you would [6:08] do if you were you from 10 or 15 years [6:11] ago and you were trying to disrupt [6:13] Airbnb? Like is that is that a helpful [6:15] exercise and that [6:17] >> it's slightly and it's slightly scary [6:19] because I think that I've told I had a [6:21] meeting with the team recently. I said [6:22] that 26-year-old me and Joe and Nate [6:25] could could f us up [6:28] um if we wanted to. And so I told them [6:30] this is what I would do. [6:31] >> because but that's because you [6:33] understand this market better than [6:35] anyone else. You know you know the key [6:37] drivers. Like I don't think any [6:39] 26-year-old off the street Hey, I'm [6:41] paranoid. I think that if Airbnb sits [6:43] still, I think a different group of 20 [6:46] or 30-year-olds could also disrupt us. [6:48] And I think that's true of every one of [6:50] us. And maybe the by the way, maybe [6:51] that's not true. Maybe that's not [6:52] reassuring cuz we do have a brand that's [6:54] a noun and a verb. We've got a global [6:56] network. [6:57] >> Yeah. I I think it's actually hard to [6:59] build a network. That may not be [7:00] possible. But the software and the app I [7:03] told our team like we can't sit still. [7:04] Like our app is beautiful. It's really [7:06] nice. Yeah. But we got to be in a world [7:08] of AI native. And here's a key point I [7:10] would make. I do not think a chatbot is [7:13] the right interface for travel or [7:14] e-commerce. That might be a radical [7:16] statement. ChatGPT launched third-party [7:19] apps last year. In March, they shut them [7:20] down. I don't think a chatbot's the [7:22] right interface. It's got four or five [7:24] problems. The first problem is it's [7:26] text-based. Um you know, photos are an [7:28] afterthought. The second is there's no [7:29] direct manipulation. You have to type [7:31] every single prompt, which is fine for a [7:33] conversation, but you can't like add [7:35] filters. You can't click around. Um the [7:37] third problem is um it's hard to [7:39] compare. A lot of e-commerce and travel [7:42] is comparison shopping. If you have [7:44] thousands of options, the AI has to know [7:47] exactly what you want to be able to show [7:48] you one or two things, but you usually [7:50] want to see more choices and you get [7:52] lost. And most AI is single player. It's [7:54] not collaborative. Let alone the fact [7:57] that Airbnb it requires people to have [8:00] an account. That 85% of people send a [8:02] message. So, what I'm trying to do is [8:04] imagine [8:05] >> sorry sorry to push back there, but I'm [8:07] I'm uh Yeah, go for it. [8:08] >> my wife's planning a vacation and [8:12] there's like [8:14] three-ish hotels that we're looking at [8:17] and we're looking for specific dates and [8:21] she theoretically, I'm not saying the [8:23] products are there yet, but [8:24] theoretically you could ask a chatbot [8:26] I'm interested in staying in this [8:28] location. I'm interested in these [8:29] hotels. I have XYZ number of people and [8:33] it could go and the agent could pull go [8:35] and pull together [8:37] like relevant sort of like listings or [8:40] or room types, etc. pricing, show me [8:43] pictures and then actually do an [8:45] analysis of the tradeoffs [8:47] based on all the information available [8:49] as well as [8:50] >> Exactly. [8:50] >> information in other parts of the [8:51] internet and pull it together and then [8:53] she could share that chat with me and we [8:55] we could both review it together. [8:57] >> Yes, I agree. That's the future and [8:59] that's not a chatbot we described. [9:01] That's not a chatbot. It's going to be a [9:02] completely different interface. It's [9:04] going to be um well, I guess you'll have [9:06] to wait and see, but [9:07] >> Yeah. I I I think the future are not [9:10] apps. The future agents, but I don't [9:12] think they're going to be text forward. [9:14] I think they're going to be really rich [9:15] user interface. Okay. [9:16] >> Um and so I think the current chatbot [9:18] paradigm, what you're describing, it can [9:20] do it. I just don't think it's the best [9:22] way to do what you just described. I [9:23] think there's a much more immersive way [9:25] to do that. Yeah. Uh as you think back, [9:28] I mean it feels like uh at at various [9:30] points, clearly there's been like oh, is [9:33] is all of this going to move to chatbots [9:36] or is all of is are there going to be a [9:38] million competitors that are all vibe [9:39] coding exactly what you have and so is [9:41] software remote? [9:42] The stock has not been beaten down [9:45] during the SAS apocalypse, but have you [9:47] had to process those with investors? [9:49] Have you had to walk people through [9:52] Airbnb's strengths again? [9:54] >> I'm sure you did after the like the 2028 [9:56] intelligence crisis. Thankfully, they [9:58] picked DoorDash instead of you guys. [10:01] But it was still I really appreciate [10:03] them picking up someone else for once. [10:05] But yeah, no, absolutely. I think I [10:07] there were entire like actually when [10:08] ChatGPT launched the third-party apps, [10:10] their stock probably went down like 7%. [10:12] And by the way, I thought it was a [10:14] really good idea for them to do [10:15] third-party apps. I think it could have [10:16] been successful, but they would have [10:18] needed a richer SDK for it to work. And [10:21] it would have like like the App Store, [10:23] Apple's App Store was good for Airbnb [10:24] and was good for every company because [10:26] they had a really rich user interface. [10:27] Maybe this is the point I'm making that [10:29] imagine using iMessage on your iPhone to [10:31] do everything when in fact like every [10:34] other app has a unique interface. So [10:36] what I think is e-commerce, you want a [10:38] very rich user interface. It would be [10:40] agentic. You would be able to have a [10:42] conversation with it. You can talk to [10:44] it. It could talk back to it, but I [10:45] think the point is it has to be more [10:47] visual. I think a text-based interface [10:49] is for some solutions and a chatbot [10:51] that's visual yeah, that would work. I'm [10:53] just saying today's chatbots aren't the [10:55] right solution. But to answer your [10:57] question, [10:58] I had a couple new employees. Like one [11:00] of our one of our new team members is a [11:02] name a guy named Ahmed who is the CTO of [11:04] Airbnb now. He led the llama models at [11:06] Meta, and one of the comments he made to [11:08] me was he said, "Wow, yeah, Airbnb is so [11:12] much more than the app." And in fact, [11:14] the app that you see is like 20% of [11:16] Airbnb. I mean, we have typically four [11:18] or five million people staying in Airbnb [11:20] every night in more than 100 countries [11:21] around the world. So there's a lot more [11:23] to do around payments, around customer [11:25] service, adjudicating everything. We [11:27] have to you know, we have a $3 million [11:29] guarantee against theft or property [11:31] damage for a million homes a night. [11:32] That's $3 trillion. [11:34] You know, there's just so many types of [11:36] things around managing 5 and 1/2 million [11:37] hosts. We have a host app. So, there's a [11:40] lot of things that are beyond the guest [11:42] app. I actually think the guest app [11:43] would be pretty easy to copy and I think [11:45] in the age of AI, you can make a better [11:47] app than ours and we want to make that [11:48] app before anyone else does. We want to [11:51] be authentic, but I think the key is [11:53] most of Airbnb is not the app that you [11:54] use. It's mostly the off-line [11:56] experience. It's the operation. Yeah, [11:58] that makes sense. Yeah, [11:59] the iMessage example is good. It's even [12:01] maybe even going deeper into like old [12:03] SMS because iMessage has hydrated so [12:06] many things with the reactions and you [12:08] paste a link and it hydrates it and it's [12:10] becoming more of a visual tool, but it [12:13] is a long road. Question for you. [12:15] >> things Yeah. Yeah, well, before you said [12:16] one of the things is the last time I was [12:17] on, I did make a comment and I'll I'll [12:19] make it again. [12:20] Almost every AI company is an enterprise [12:22] company and the last stat I think I said [12:25] on TBPN was like 3 months ago and stats [12:27] haven't changed. I think it was 175 [12:29] companies in the YC batch and I think 16 [12:31] were consumer. [12:33] That trend hasn't changed. Almost [12:34] everyone is going into coding. Almost [12:37] everyone's going to enterprise. I think [12:38] that's great. My whole point though is [12:41] that the consumer experience hasn't [12:44] fundamentally changed that much beyond a [12:45] chatbot and I think the consumer is a [12:47] massive opportunity for AI and I think [12:50] it's going to need to be a richer user [12:52] interface. I think people that want to [12:54] do more things, they wanted to be I [12:56] mean, the modern AI today is text, [12:58] photos, and buttons. Yep. Mostly and [13:00] some videos and I think there's a more [13:03] breakthrough visual paradigm that could [13:05] be much more immersive and with the new [13:06] image and video generation models, you [13:08] could do something so much more [13:09] immersive. So, the exact example you [13:11] gave of trying to book a hotel, yeah, [13:13] you can do it on a chatbot, but there's [13:14] probably some more breakthrough way to [13:16] imagine that, to visualize it, to see [13:18] the neighborhood, to see the map, to be [13:20] able to talk to it, to understand where [13:22] it is, to be able to compare photos of [13:23] different hotels. There's something I [13:25] think richer on the horizon. Yeah, that [13:27] was I was I was surprised at the push [13:30] into browsers last year from some some [13:34] different uh AI companies. Specifically [13:36] because I always I already felt like an [13:38] LLM, like the chat the chatbots were uh [13:41] helping me browse around the internet. [13:43] It was like just double down on that and [13:45] helping me find information all over the [13:47] internet. But I like that it's brought [13:49] together in a standardized way, right? [13:51] If I like the example um you know, of of [13:54] comparing like different hotels, right? [13:56] It's nice to have it just like formatted [13:58] the same way. So I'm looking at it, I'm [13:59] not being like influ- I don't want to be [14:01] influenced by the images and like the [14:03] the property, but not necessarily the [14:05] design of the website, right? Because [14:07] that they they can be disconnected. Um I [14:10] wanted to ask if you think LLMs will [14:12] impact travel in the world in the way [14:15] that social media did uh in travel [14:17] trends specifically because Instagram [14:20] nowadays, like a place like Marfa in [14:23] like uh Texas, right? It's like it's [14:25] random town or you know, artistic type [14:28] town in Texas, and then it just becomes [14:29] a cool place because of Instagram, and [14:32] then it and then and then the entire [14:34] kind of [14:35] town evolves because of that. I can [14:37] imagine like people researching like [14:40] places to go with LLMs could ultimately [14:42] drive some of that and then ultimately [14:45] um reinforce each other. But but what do [14:47] what do you think? [14:48] Yeah, maybe the simple framework is [14:51] think about a travel journey. Um step [14:53] one is destination discovery, where [14:55] should I travel? Step two is flights, [14:57] how do I get there? Step three is where [14:59] do I stay, Airbnb or hotel? Step four is [15:01] like what do I do when I get there, [15:03] restaurants, activities? And step five [15:05] is typically like logistics, car rental, [15:08] services. And then step six is you're in [15:10] the city, and then you might want to do [15:12] things that time spontaneously. I think [15:15] LLMs or that let's call the LLM [15:17] technology applied to I'm arguing a [15:20] slightly richer user interface than [15:21] typical chatbot will be revolutionary [15:24] for step one, step two, for destination [15:26] discovery and flights. Also, because [15:28] flights are not very hard to build, [15:30] there's just three global distribution [15:32] systems, anyone can pipe in an API and [15:35] have a flight booking app pretty pretty [15:36] quickly. And so, I think the LLMs are [15:38] really really good at destination [15:40] discovery. If you want to say like, [15:41] "Hey, I want to go somewhere that's like [15:42] Paris, but it's a little more [15:44] affordable, it's good to go in August, I [15:46] like opera." Like, it's going to give [15:48] you like a a very rich like suggestion [15:50] of places to go. Now, I think the [15:53] current chatbot doesn't have really rich [15:55] maps. I think it could do a lot more [15:57] visually. I think eventually chatbots or [16:00] this new interface could be very [16:01] video-based. So, imagine a chatbot that [16:03] was actually video-based or very very [16:05] photo-based, rather than a little less [16:07] text-based. I think that would probably [16:09] be what you'd want for travel. I think [16:10] what I'm describing would disrupt travel [16:13] um more than the current chatbots, which [16:15] are more acting like Google sending [16:16] referral traffic. And they actually the [16:19] referral traffic from these chatbots are [16:21] converting higher than Google. And so, [16:23] actually the chatbots are actually [16:25] additive to travel companies. To [16:27] disintermediate travel company, you [16:28] would really want to disintermediate the [16:31] travel journey. It would have to be much [16:32] richer. Yeah, I I feel like Do Do you [16:34] remember the company Hipmunk? I think it [16:36] was founded by some YC founders, right? [16:39] Uh yeah, actually the CEO of Reddit. Um [16:41] That's right. Steve Huffman, right? Uh [16:43] yeah, Steve You're You're talking about [16:44] Hipcamp? No. Uh Hipmunk. [16:47] >> Hipmunk was a Hipmunk was a really cool [16:49] flight booking It was a really cool [16:50] flight booking uh dashboard and it would [16:51] it would uh rank the flights by pain, [16:54] like a pain index. So, it would say, [16:57] "Well, you're not going to face [16:58] financial pain because it's a really [17:00] cheap flight, but you do have a [17:01] stopover. Or, you do have to get to the [17:04] the the airport super early." And so, it [17:06] would blend all these things together [17:07] and it's something that it feels like [17:09] could be vibe-coded over a weekend now. [17:13] Uh and yet, it is weird that when I open [17:15] up the App Store charts, I see ChatGPT, [17:18] Gemini, Claude, and then it's just AI [17:20] chat apps. And we haven't had that [17:22] breakout moment. Demis was on stage at [17:24] YC actually talking about the fact that [17:26] it feels weird that we have this super [17:28] intelligence in the enterprise and vibe [17:30] coding. And you're 60% of Airbnb's code [17:32] is written by AI now, and yet we don't [17:34] have like a new AAA game or a new video [17:37] game that everyone's playing. And I'm [17:38] wondering like is that a lack of [17:40] creativity? Is it a lack of of [17:43] risk-taking that there's just too much [17:45] money? It's too It's too It's too [17:47] obvious to be able if you're if you're [17:49] good in AI, just go into enterprise [17:51] because you'll just raise a bunch of [17:52] money, get a bunch of customers. It's [17:54] really easy to make money. And it's more [17:55] risk-on in the consumer. Like what do [17:57] you think needs to happen? Or is it just [17:59] the intelligence isn't quite there? And [18:00] maybe the next model's the one that [18:02] unlocks it. Man, well, this is such a [18:04] good question. Um I think there's like [18:06] three or four factors going on. [18:08] Um the first factor is I saw a tweet [18:10] recently that I think there's like 60 [18:12] new Neo labs being formed. [18:14] >> Yeah. And um I've met a number of these [18:16] researchers, and a lot of them are [18:18] interested in doing the same things. And [18:20] I think what I'm noticing is most these [18:22] teams are purely AI people. They don't [18:25] have product people or designers on [18:26] their team. And they're picking things [18:28] like science. They're picking things [18:30] like coding. They're picking things like [18:32] we want to create a different type of [18:33] model. This is great, but I don't know [18:34] if all 60 companies should be doing the [18:36] same problem. And so what you're seeing [18:38] is you don't see a big focus on people [18:40] wanting to do consumer. I think some [18:41] people think that the AGI can just [18:43] figure out consumer. I think that's um [18:47] maybe simple overly simplistic thinking. [18:49] I do think you have to have a a point of [18:51] view about consumer. So I think that's [18:53] the first point. I think the second [18:55] point is, you know, Silicon Valley's [18:56] become more vibe and trend-based than [18:58] when I came to Silicon Valley in 2007, [19:00] although it was back then. And one of [19:02] the things I think when people see [19:04] enterprise companies doing well, they go [19:06] into it. And I think there's this like [19:08] natural flywheel. Um the other thing is [19:10] I know in Y Combinator, for example, [19:12] we teach the Y Combinator companies use [19:14] the other companies as their customers. [19:16] Now, back in the day, we would get them [19:18] to use our product as consumers. But, I [19:20] think everyone's figured out actually [19:22] it's way more efficient to get them [19:24] to to become customers. [19:25] >> It's like five code five code gen [19:28] startups all being like, "You code with [19:30] me. [19:30] Yes, exactly. Yeah. And so, I think [19:33] enterprise makes a lot of sense until [19:35] everyone does enterprise. And then [19:37] suddenly, like everyone's saying [19:39] consumer's hard. You know what's also [19:40] hard? Like competing with 10 other [19:42] companies in enterprise doing your idea, [19:44] too. That's also hard. I think [19:46] additionally, the image and [19:48] in video generation models are having a [19:51] I think image two is almost like a [19:53] Claude code breakthrough. It is so good [19:56] now in the C dance video models that I [19:58] think we're about to enter this new era [20:01] where suddenly going beyond text is [20:03] possible. I think you're going to even [20:05] be able to see real-time AI generated [20:07] interface in the next year or two. So, [20:08] my prediction is this has been the era [20:10] of of of enterprise. Even ChatGPT's [20:13] breakout success, I would predict that [20:15] most of our revenue is going to go be [20:16] going to Codex. That's my pre- I might [20:19] be wrong, but just given how much [20:20] money's going to Claude. [20:22] And I think in the next two years, [20:24] you're going to see a massive revolution [20:25] in consumer. And I think you're going to [20:27] need a couple companies to lead the way. [20:30] I think we need reference points. I [20:31] mean, even me saying the chatbots not [20:33] the interface, people are probably [20:34] asking, "Well, what is?" And someone [20:36] needs to be the one to do that. I mean, [20:38] we're going to try to do our part. I'm [20:39] not sure Airbnb will be the company to [20:42] blaze the trail, but I think someone has [20:43] to. I think in another era, it would [20:45] have been Apple. Like the Steve Jobs era [20:47] Apple would have done that. Maybe Apple [20:50] will do that, but they would have been [20:52] typically the company that would have [20:53] done that. Yeah. Are you worried that [20:56] everyone is working on the exact same [20:58] thing? [20:59] Um [21:01] And when I say exact same thing, I mean [21:03] agents that can do things on on computer [21:06] and or well, you know, look at the [21:07] iPhone app store chart or look at uh [21:10] look at the early stage startup market, [21:12] right? A lot of companies, you know, [21:15] what the full spectrum of companies from [21:17] hyperscalers to labs to [21:20] neo labs to [21:21] >> not I'm not too worried. I might be a [21:24] little worried for the entrepreneurs [21:28] if they're like the 10th company doing [21:30] something. I mean, like Peter Thiel was [21:31] one of our earliest investors and he [21:33] used to he wrote this book Zero to One [21:35] that a lot of people listening probably [21:36] read. He has a saying, "Competition's [21:38] for losers." And you want to kind of try [21:40] to do something that no one else is [21:41] doing. Back in the day, that was [21:43] enterprise cuz everyone was making an [21:44] app. And now, I just I have to be [21:46] careful about being the 10th company um [21:48] doing something unless you're going to [21:50] be a lot better decisively. It's just [21:52] really, really hard in a crowded market. [21:54] So, you want to kind of zig when [21:55] everyone else is zagging. Um so, I'm not [21:57] too worried, but maybe a little bit and [21:59] I would just encourage entrepreneurs to [22:01] try to claim a space for their own. And [22:04] by the way, Airbnb was that. It was kind [22:06] of accidental, but like like in 2007, [22:09] everyone wanted to do like a social [22:10] networking something. And we were like [22:13] it was tempting for us to try to be a [22:15] social networking something something as [22:16] well. It was accidental that we inflated [22:20] through airbeds one weekend and created [22:22] airbed and breakfast and people thought [22:24] it was the worst idea ever. I guess it [22:26] became like the worst idea that ever [22:27] worked, but we carved our own space as [22:30] did Uber [22:31] um and we weren't trying to be anything [22:32] else. And so, I kind of And by the way, [22:34] OpenAI [22:36] wasn't trying to be anyone else either. [22:37] Maybe they were a little bit like [22:38] DeepMind, but AI was not the thing when [22:41] OpenAI and Anthropic got started. So, I [22:43] do think there is something about not [22:46] chasing trends. I think once the trend, [22:49] it can be a little crowded and to try to [22:51] claim out your own space. And I would [22:52] say there's so much of the US economy [22:55] that AI hasn't yet touched. Yeah. [22:58] Um [22:59] on that note, [23:00] how are you talking to or how how would [23:03] you talk to designers or RISD students [23:08] about applying non-technical skill sets [23:13] or or non-AI disciplines, non-computer [23:16] science backgrounds to consumer I [23:21] The reason I ask you this is I feel like [23:22] there's a lot of there's a lot of [23:23] pushback against AI in design [23:26] communities uh in colleges right now [23:30] broadly AI is not particularly popular [23:33] and there's a lot of fear of job loss, [23:35] but at the same time, you know, I'm sure [23:38] like the tools have changed throughout [23:40] the history and there's an and I feel [23:43] like you believe that there's incredible [23:44] opportunity, but I'm worried that some [23:47] people are not jumping on that [23:48] opportunity because they have [23:51] hesitations about [23:53] various oh well, does this displace this [23:56] tool or what you know, how is this made [23:57] or you know, is this uh it it it is this [24:00] a is this the right tool for the job? [24:02] Should I even be using this? [24:04] Yeah, I'm really worried that um an [24:06] entire generation of designers, artists, [24:08] and creative people are going to decide [24:10] to kind of sit out AI and I think it's [24:12] the biggest opportunity for creative [24:14] people in my lifetime. By the way, this [24:17] is not the first time designers were [24:18] late. Um we we creatives and designers [24:21] were really late to the internet and [24:22] web. So, going back 30 35 years um most [24:26] of the prestigious designers did not get [24:28] into so-called web design and web design [24:30] was considered a lower tier, lower [24:32] status design and they stayed with [24:34] print, all the established people. And [24:36] so, they did not really chase design. [24:38] What ended up happening was because you [24:40] had a lot of these really excellent [24:42] designers that did not go into web [24:43] design um I think what happened was a [24:46] lot of the there were a lot of good [24:47] people that became web designers, but I [24:49] think there became a gap between web [24:51] design and engineering. You had people [24:52] that, you know, maybe didn't have the [24:54] full full design skill set and what [24:56] happened was this function emerged [24:59] called product management. Product [25:00] management, I'm not arguing against, we [25:02] have product management, it's very [25:02] important. But in industrial design, [25:04] there's no product managers for the most [25:06] part, they're industrial designers. In [25:08] architecture, the architect is the [25:10] product manager. And so I think what [25:12] ended up happening in web design, [25:13] designers were very very narrow, there [25:15] was a void, product managers filled that [25:16] void. I'm not arguing against that [25:18] model, but I think if designers sit out, [25:20] what you're going to see is engineers [25:22] and product managers designing for them. [25:24] And I think the counterpoint is [25:26] designers can be engineers and product [25:27] people for all intents and purposes. [25:29] Said differently, if I were starting [25:31] Airbnb today, I'd be live coding in [25:32] cloud code. Um Yeah. And I'm trying to [25:35] do it now. I'm I'm doing it for fun. I'm [25:37] not really going to be doing anything [25:38] productive at Airbnb. But if I was 26, I [25:40] would have thought of myself as a [25:42] technical person. I think all designers [25:44] should think of themselves as product [25:46] people and as, you know, front end [25:48] design engineers, if if anything. And so [25:51] I think this is a boon. I don't think [25:53] that the future of of the internet has [25:56] to be all text-based. I think it can be [25:57] very visual. I think the photo and video [26:00] generation models allow you to design [26:02] incredibly rich interfaces. And this is [26:05] I think the best time in the world for [26:06] designers and creative people to get [26:08] involved. I think they're just a little [26:09] bit afraid. And I whatever we can do to [26:11] enlist them to just get their hands [26:13] dirty. Many designers are. I just think [26:15] more should get involved and not be in [26:18] be getting getting out of typical [26:19] wireframes. Just start coding. Do you [26:21] think so assuming we we we solve this [26:25] problem of not enough, you know, [26:28] creative people working in consumer, we [26:30] we build, you know, we as a society [26:33] build cool new consumer products and [26:35] experiences and apps. There's something [26:37] that deserves to be at the top of the [26:38] App Store. Has distribution and the [26:42] acquisition of customers and and and [26:44] users changed structurally. I I you [26:48] know, you were very good at SEO and [26:51] Google and there was a referral program. [26:54] Do those patterns still work? Are they [26:56] broken? Is there a new playbook that [26:58] needs to be [27:00] that needs to be rolled out for anyone [27:01] who has something that's great, but they [27:03] need to get it in the hands of consumers [27:05] broadly? [27:06] I think I think we need new patterns. [27:08] This is a funny saying. I think the [27:10] highest turnover job of any executive in [27:13] Silicon Valley might be the CMO. I don't [27:15] know for sure, but like you know, you [27:16] don't see a lot of turnover CFOs, you [27:18] don't see a lot of turnover CTOs. But [27:20] like I Amazon was famously and I I don't [27:23] know if [27:24] part of my theory is that like what [27:26] works in marketing changes every few [27:28] years and you have to be adaptable and [27:30] your old playbook gets outdated. So [27:32] marketing is unfortunately one of the [27:33] hardest functions. I have a huge amount [27:35] of respect for people in marketing cuz [27:37] once something works it almost becomes [27:39] stale because then everyone does it. And [27:41] so like influencer marketing was really [27:43] successful until everyone did it and [27:44] then people kind of tune it out. It's [27:46] this thing called banner blindness. [27:48] Banner blindness is after you see [27:50] something over and over you tend to be [27:52] blind to it and so you need a new [27:53] tactic. So I think that a couple [27:56] thoughts. Number one, [27:57] I think that just like we need new [28:00] interface design, new creative, [28:01] experiential approaches, we also need [28:04] new types of marketing. I don't know off [28:05] the top of my head what that is. It's [28:07] probably something we haven't done [28:08] before. I mean Joe and I, we did like [28:10] bizarre things like we sold collectible [28:12] cereal boxes and we like when social [28:16] media was new, we were all over social [28:17] media and when when there were [28:19] newspapers, we would hunt down reporters [28:21] and like try to get them to write about [28:23] us. But like people ask me how do I get [28:25] users? I'm like all the tactics from [28:27] 2008 aren't relevant anymore. So you got [28:29] to be relevant and you got to find new [28:31] tactics. [28:32] The second thing I'd say though is [28:34] you're right, the distribution is [28:35] mature. When we started Airbnb, the App [28:38] Store was young. In fact, the internet [28:40] was still young. I mean, we could just [28:42] you could launch a website and just ride [28:44] the growth of the internet. And it's [28:46] hard. The internet [28:47] is not growing like it did before. In [28:49] other words, people were coming on the [28:51] internet in hundreds of millions a year, [28:53] new users. So, I do think distribution [28:55] for consumers mature. At the same time, [28:58] the top apps in the App Store are new [28:59] apps. They're They're They're mostly [29:01] chatbots. So, what that tells you, if [29:03] you do something truly breakthrough and [29:04] revolutionary, consumers will still [29:06] probably find it. And so, that's I think [29:09] I think the two principles are do [29:11] something so revolutionary, [29:13] consumers will find you. And number two, [29:15] you're going to have to find your own [29:16] new tactics. And you know, anything that [29:19] is standard is probably stale. [29:22] Well said. [29:23] I have one more, but Yeah, looking out [29:25] into the future, [29:26] you've ridden the growth of the [29:28] short-term rental market. You dominate [29:30] it. Now, is [29:33] is the biggest opportunity to just [29:34] continue to make the best product in the [29:36] category, ride the continued growth of [29:39] it, or do you think there's another STR [29:42] size market for Airbnb? [29:46] I think the biggest opportunity for [29:47] Airbnb is to go beyond our core [29:48] business. [29:49] I do think you know, our core business [29:51] does close to a hundred billion dollars [29:52] in gross sales, if you net out all the [29:54] other businesses, gross gross booking [29:56] value, the total amount of reservations [29:58] going through the site. I think that [30:00] could probably double one day. I don't [30:02] know how long that one day is, but for [30:04] every person who stays in Airbnb, eight [30:06] or nine stay in a hotel. I think we can [30:07] get one extra person in a Airbnb, [30:10] eventually. And I think that can get you [30:11] to two hundred billion. I think there's [30:13] a market much larger than Airbnb, which [30:15] is hotels. Again, hotels are about eight [30:17] or nine times the size of Airbnb. I [30:20] don't think we'll ever be a hotel [30:21] dominant site, but we are going more [30:23] aggressive into hotels. A fun thing is [30:25] that about half the hotels in the world [30:27] are independents and boutiques. They're [30:28] not chain hotels, and they're not really [30:30] happy listing on the OTAs, because they [30:33] pay a higher commission than chains. A [30:35] lot of these independent hotels are [30:36] being forced, or they feel like their [30:37] hand is being forced to franchise, to [30:39] Marriott, to Hilton, or other brands, [30:41] because they have loyalty programs, they [30:43] can you know, they can negotiate lower [30:45] commissions. So we think we can be a [30:47] distribution channel for these boutiques [30:48] independents. That is a multi-billion [30:51] dollar market. I think another one would [30:52] be services. There's no Amazon for [30:54] services and you know, think about like [30:57] you can hit a button and a car can pick [30:59] you up. You can hit a button and food [31:01] can be delivered to you. But what about [31:02] hitting a button having 80 other things [31:04] possible? You know, I don't know if any [31:06] one market is large, but if you add up [31:09] the 80 different service verticals, that [31:11] to me is another pretty big market. And [31:13] then maybe the last one is like living, [31:15] stays of longer than 30 days. Um more [31:18] and more people have a job via laptop. [31:20] More people are nomadic, more people are [31:21] moving around. That's another really big [31:23] market. So I think for Airbnb, we're [31:25] looking at really expanding to a lot of [31:28] different categories and I think that's [31:29] where most of the growth is going to be [31:30] in the future. [31:32] Do [31:33] billboards in San Francisco work on you? [31:35] Going back to marketing, it's one of [31:37] these old things where like yes, I I [31:39] completely agree with you. Like the the [31:41] the CMO who did the first billboard [31:43] campaign back in the '50s, probably [31:45] printed legend, you know, books written [31:47] about them. Uh Well, the billboard [31:49] campaign that worked fell off. Now it's [31:51] a couple years was just buy so many [31:54] billboards that people are like, wait, [31:55] how did this company buy so many [31:57] billboards, right? Yeah. And so I'm [31:58] wondering is your is your thesis on [32:00] marketing that the the these things go [32:02] in cycles and yes, we might be out of [32:04] the influencer marketing meta right now, [32:06] but it's going to come back in 5, 10 [32:08] years, who knows. That's possible. Or or [32:10] is it rise and fall and then never [32:12] again? [32:12] >> these channels though they they don't go [32:15] away. They just stop being like if you [32:17] do this thing, you're going to grow like [32:19] crazy. You still need to do that thing, [32:20] but it's no longer an article. [32:22] >> Okay. Yeah, what do you think? [32:23] >> are Okay, so so one thing I have to [32:24] acknowledge is I think all of us [32:26] advertising works better on us than we [32:29] admit. [32:29] >> Yep. I I honestly I I we we spend a lot [32:32] on like all these ads and I kind of [32:34] think to myself, does do these really [32:35] influence people? Like and if they do, [32:38] and if they didn't we wouldn't be [32:39] spending like a billion dollars a year [32:40] in advertising. So I want to admit that [32:42] advertising does influence us more than [32:44] we realize. There's I think a lot of [32:45] studies that show like oh that doesn't [32:46] change my mind but you see something [32:48] seven times and it probably even if you [32:50] don't want to believe it you might start [32:51] believing it. So clearly it works. I do [32:54] think though the ROI of new ideas works [32:57] a lot better. Just for example we spend [32:59] a lot in advertising but the most [33:01] popular marketing we've ever done is [33:03] like when the Barbie movie came out we [33:05] took a house in Malibu we turned the [33:07] Barbie Malibu Dream House and it was [33:09] like new and the whole internet talked [33:10] about it and the ROI of that was better [33:12] than any ad we've ever done right? We've [33:14] done things like that. So I do think [33:16] doing kind of crazy slightly unhinged [33:18] things that people notice like I mean [33:21] you guys are successful cuz you're [33:23] different right? I think. I think being [33:25] different is the key to marketing and so [33:28] if you have a billboard people are [33:30] probably influenced but if you if it [33:31] looks different than another billboard [33:33] it's going to work. So I think the key [33:34] to marketing is to be different because [33:36] you got to stand out. [33:39] Yeah the the the Red Bull [33:42] Baumgartner the the skydive from space [33:45] is I think like a thousand X ROI. I [33:48] think they spent less than a million [33:49] dollars on that and it would it I think [33:50] it has a billion views or something like [33:52] that. Great example of that. But if like [33:54] ten people did that and the ten people [33:56] did that it wouldn't work as well. And [33:57] that's probably why they didn't do a [33:59] sequel and then another one and another [34:00] one. They went and found something else. [34:01] Let's put a plane through a tunnel or [34:03] fly a you know [34:05] terrible and upside down or something. [34:06] Get a blimp get a blimp on Airbnb that I [34:08] can go on a sky cruise. Sleeping in a [34:11] blimp yeah. So that would be that would [34:13] be a that would be a very good idea. [34:14] Maybe we can talk about it. I'd love [34:15] that. We'll podcast. Yeah I don't know [34:18] how long it takes to get to Yeah LA to [34:20] San Francisco. If it takes me two days [34:23] three days but I have a beautiful view. [34:24] >> Would you guys do that? Like do we talk [34:26] about that? I would 100% do that. That'd [34:28] be good advertising for us and you. So [34:30] so so so so so we say things like this [34:32] all the time where we're like, "Oh, [34:33] yeah, like we'll definitely like [34:35] do something." And often times it's hard [34:36] to schedule, but we have been talking [34:38] about blimps for like over a year. [34:41] >> can broadcast. We've been having [34:42] >> to happen with a blimp. So bad. [34:44] >> call in from their jets with Starlink, [34:47] and the connection is perfect. So much [34:49] so that people in our chat sometimes [34:52] don't know that the person's on a plane. [34:55] Okay, you heard it here first. Airbnb [34:57] and TBPN are going to work together on a [34:59] blimp partnership. Yes. [35:01] >> are going to broadcast across the [35:02] country on a blimp. I love it. [35:04] >> going to be a part of that. I want to [35:05] make this work. Let's let's Let's let's [35:07] figure it out. We'll talk. [35:09] Uh fantastic. Well, thank you so much [35:10] for coming on the show. Have a great [35:11] weekend, and we'll talk to you soon. [35:13] >> You're the man. Goodbye, guys.