Google has published an alpha GenUI SDK for Flutter — a package, now live on pub.dev, that lets an AI agent generate and assemble parts of an app’s interface at runtime rather than a developer hand-coding every screen. The mechanism is an orchestration layer: the AI agent decides what content to show and which components to use, drawing from the app’s own established widget catalog and brand guidelines, and returns a JSON description of the layout that Flutter renders natively. As the user interacts with what’s on screen, those state changes feed back to the agent, creating a live loop between interface and AI rather than a one-shot generation. It was spotlighted at Google I/O 2026 under the session title “Flutter + A2UI = GenUI” and has been widely described since as the most architecturally significant change to the framework in years — though Google is upfront that it’s experimental and the API will keep changing.
The distinction worth sitting with is between this and the AI-assisted coding most founders already associate with faster app builds. Tools like Claude Code and Cursor speed up how a human writes an app’s code; GenUI is about the app itself deciding, live, what interface best serves a given piece of AI-generated content — a conversational agent that renders a form instead of a wall of text, or a dashboard that reshapes itself around what the user just asked for. It’s the mobile equivalent of the shift chat interfaces have already made from “text in, text out” to rich, structured responses.
So what
This is early — alpha, Flutter-specific, and not yet something to build a production app’s core UX around. But it’s a real signal of where AI-native mobile development is heading: apps that don’t just call an AI model for answers, but let it shape parts of the interface itself. Worth tracking if you’re specifying a Flutter build with an AI assistant or agent at its core, so the architecture isn’t locked into a shape that fights this pattern later. See our AI-assisted development work, or get in touch to talk through what an AI-native mobile build looks like today.