Dynamic Beating AI News Flash: Runway has released a world model called Solaris that can directly generate an app interface. Now, using AI to create an app usually involves generating HTML, CSS, and JavaScript first, and then having the browser turn the code into the interface. Solaris skips this step entirely: the entire interface that users see is the real-time generated image from the model.
Solaris is based on Runway's Gen-4.5 transformation. User clicks, drag-and-drop actions, or text inputs become the conditions for generating the next frame. The language model determines what should happen next on the interface, and Solaris instantly renders it. Pages, buttons, and interactive states do not need to be predefined.
Runway also conducted a comparison test between Solaris and the code-based interface generated by Claude Opus 5. 250 participants tested 30 interactions, resulting in nearly 7500 evaluations. In terms of "completing tasks as instructed," 61% chose Solaris, while 24% chose Claude; regarding naturalness, the percentages were 71% and 21% respectively.
Currently, Solaris is still in the research preview stage and has a long way to go before truly replacing apps. As each frame is generated in real time, the cost is still higher than regular web pages. Challenges such as text stability, long-term interface consistency, and accessibility support remain to be addressed.

