Brian Chesky: the chatbot is the MS-DOS of AI — Airbnb is designing for the next paradigm

Summary

Brian Chesky is framing today's AI chatbot interfaces as the MS-DOS era of computing — functional but primitive — and positioning Airbnb as a company designing for the paradigm that follows. His central argument: the chatbot is a pre-AI interface bolted onto a powerful new engine. The real interface evolution, he contends, will be more visual, more contextual, and ultimately agentic. Agents, not apps, are likely the next platform.

New CTO Hire

Chesky has appointed Ahmed Aldal as CTO. Aldal led the generative AI team at Meta that created the LLaMA models, which have been downloaded over 1 billion times and spawned more than 60,000 derivative models. Before Meta, Aldal joined Apple as a college hire in 2005, worked on the original iPhone's multi-touch technology, and later led Apple's autonomous vehicle program. Chesky cites the combination of frontier AI expertise and Apple-grade design craft as the decisive fit.

Consumer AI and the YC Data Point

87% of Y Combinator companies per batch are now enterprise-focused, a figure Chesky finds telling. He argues enterprise dominance in AI makes short-term sense — models are expensive to run, and enterprises offer immediate, high-value workflow automation. But the larger prize, in his view, is consumer, because that is where AI will reach billions of people in daily life. He identifies three viable consumer business models: advertising, subscriptions, and transactions. Airbnb has the third, with high-dollar transaction volume and 50 billion searches per year on its platform.

Airbnb's AI Product Strategy

Chesky is deliberately not rushing LLM integrations. When ChatGPT launched commerce features, Airbnb declined while Expedia and Booking moved first. His rationale is that 90% of Airbnb bookings involve a pre-stay message to a host and require verified ID — constraints that demand a high-quality integration, not a fast one. Notably, traffic arriving from ChatGPT already converts to bookings at a higher rate than traffic from Google Search, a signal he takes seriously.

On the product roadmap, Chesky points to several structural advantages Airbnb can exploit. The platform's conversational, multi-turn booking process is a natural fit for chat-adjacent interfaces that build richer user profiles over time. Most Airbnb trips involve multiple travelers with different preferences, a collaborative planning problem that single-player chatbots do not solve well. Collaboration tools and persistent memory — which current LLMs handle poorly due to stateless architectures — are areas where Chesky sees technical differentiation opportunities.

Authenticity and Generative Imagery

Generative AI image manipulation poses a real risk to listing integrity. Chesky's current answer is Airbnb's proprietary network of more than 5,000 professional photographers worldwide, a service built before Uber launched that dispatches photographers on demand. Photos taken by Airbnb-contracted photographers carry an implicit authentication layer. He views real-world verification as increasingly essential as synthetic media makes on-screen trust unreliable.

The Broader Platform Argument

Chesky does not see the rise of foundation model platforms as existential. His view is that some degree of application-layer specialization will persist — users will not rely solely on one or two generalist agents any more than they rely on one person for every need in life. He expects to exist on platforms like ChatGPT, Gemini, and Claude, but as a specialized travel and living agent rather than a commodity listing aggregator.

On Apple specifically, Chesky argues the company should redesign its devices from the ground up as AGI-native hardware, not continue running AI features on architectures conceived before the current wave. His benchmark for whether the AI revolution has actually arrived is simple: daily life for average people has not materially changed in the past three years, and it will not until devices, operating systems, and consumer applications are rebuilt around the new paradigm.