A recent analysis from a Chinese developer blog has sparked an important conversation about how AI products structure their output. The core idea is that there is an emerging 'UI Output Protocol' with three main layers: Markdown for simple documents, HTML for rich web pages, and custom UI DSLs for interactive components like cards, filters, and charts. This framework helps developers and designers make conscious choices about which format to use for different types of AI responses. For example, using Markdown for quick answers, HTML for complex layouts, and a UI DSL for actionable widgets. The article highlights that this is not about replacing one format with another, but about understanding their complementary roles. For the global developer community, this provides a valuable mental model for designing more sophisticated and user-friendly AI interfaces. The trend is clear: AI products are moving beyond text, and a structured approach to output is becoming essential.
This article explores the architectural layers behind AI product outputs, from simple Markdown to rich HTML and interactive UI DSLs. It argues that understanding this protocol is key to building effective AI interfaces. The topic is highly relevant as AI products increasingly move beyond text-based responses.