Intelligent UI needs a new HTML (www.openui.com)

OpenAI recently introduced a new Interface Inference Language (IIL) named DIL alongside its Intelligent UI, allowing models to describe interfaces dynamically without needing to rewrite components from scratch. This innovation highlights a shift in how software can be generated and interacted with, treating interfaces as transient rather than permanent artifacts. However, the emergence of multiple IILs like A2UI, OpenUI Lang, and json-render raises concerns about compatibility and fragmentation in the developer community, reminiscent of the early web where proprietary tags led to a disjointed ecosystem. The significance of a shared IIL framework lies in its potential to unify disparate models and tools, enabling smoother transitions across different platforms and models without requiring extensive reimplementation. Key technical features of IILs include using constrained components, streaming token-by-token rendering, data references instead of direct data inclusion, and forgiving error handling. Establishing a shared standard could enhance model evaluation processes, facilitate the development of common debugging tools, and reduce implementation costs, much like how HTML created a cohesive web environment. As development becomes increasingly machine-driven, the need for a foundational IIL that fosters an interoperable ecosystem is more crucial than ever.