He Built Uber Eats to $25B, Now He's Pushing AI Beyond Chat

Chau demonstrates Cove by planning a family trip to Tokyo, Kanazawa, and Osaka. Cove creates editable cards for city comparisons and an itinerary, then updates them with preferences such as a seven-day trip, food, nature, and kid-friendly activities. Chau describes the workspace as a shared canvas where people can work with AI in parallel and collaborate with others.

Chau shows how Cove can use material added to a workspace, including YouTube transcripts, PDFs, and documents, to build AI apps. Examples include a quiz generated from a lecture and a three-body-problem simulation. He also describes his 11-year-old daughter using Cove to make an aquarium water-cycle tracker and a snail identification app.

After leaving Uber, Chau and colleagues prototyped products including music generation and a travel planner, then questioned whether chat threads suit collaborative work. He says Cove combines models and services for different tasks, such as Llama for suggestions, Sonnet for coding, and Perplexity for web search. The team balances a long-term vision for a generative workspace with a near-term roadmap that must adapt as models improve.

Chau says Uber Eats began with a small team mixing experienced Uber operations and engineering staff with newer product and design perspectives. The team set an ambition to build a business as large as rides, and leadership treated new businesses as important to Uber's future rather than as a side experiment. Once the team found a promising direction, it scaled rapidly.

Chau expects AI tools may make generalists more valuable by letting individuals take on more kinds of work. He encourages PMs to treat AI as a thought partner and to build practical intuition by trying products, making mistakes, and experimenting. He calls that developing an “AI spidey sense” across different models.