The article argues for a standardized approach to agent-to-UI communication, highlighting the current fragmentation and its cost. Developers face a challenge because each AI assistant renders user interfaces differently, forcing those who maintain button components to adapt. While connecting agents is becoming easier, the rendering process across various clients remains inconsistent, with no single source of truth for component ownership or appearance. Google's A2UI aims to solve this by using declarative JSON, where agents name components that clients already possess, and clients handle rendering.This declarative JSON approach, similar to early HTML, is presented as a neutral proposal without vendor ownership. However, widespread adoption is hindered by various methods clients use to display UI, including sandboxed HTML and vendor-specific formats like Slack's BlockKit and Microsoft's Adaptive Cards. The core issue is the ownership of styling and component vocabulary, which is currently handed over to the client, complicating design system consistency. The author draws a parallel to the browser wars of the late 1990s, where competing standards led to duplicated effort and stagnation.The article emphasizes that the "canvas" or more complex UI areas beyond simple chat are where the real cost lies, with each vendor maintaining its own set of components. While declarative UI like A2UI offers a path to a shared "parts list," the current reality requires significant effort to make design systems work across numerous agent surfaces and platforms. The solution proposed is to choose a target standard, with A2UI being the most promising, to unify the rendering process and reduce the "surface tax" imposed by managing multiple implementations. Ultimately, the success of any standard depends on its neutrality and the willingness of platform vendors to absorb the complexity of rendering.