The current movement toward AI interface standards echoes the past fight for web standards led by Jeffrey Zeldman. In the 1990s, competing browsers created incompatible experiences, forcing developers to rebuild sites multiple times. Zeldman advocated for shared conventions, not by creating specifications, but by convincing the industry of their value. His efforts led to the Web Standards Project and A List Apart, grassroots initiatives that pushed for browser consistency. This philosophy of practitioners establishing standards from the ground up is now relevant for AI interfaces. Today's AI assistants like ChatGPT and Copilot exhibit similar fragmentation, each with its own conventions for displaying information and reasoning. This lack of interoperability hinders effective evaluation and comparison of AI outputs. Just as Zeldman emphasized separating structure, presentation, and behavior, AI interfaces need similar distinct layers for their reasoning, output, and actions. The success of Zeldman’s movement was also due to presenting a business case for standards, emphasizing cost and reach over initial moral arguments for accessibility. Finally, the coalition approach, where practitioners collectively shaped norms rather than receiving top-down mandates, proved effective. This historical lesson offers a roadmap for establishing AI interface conventions.