Interface World Model

An Interface World Model is an AI approach where, whenever a user taps or drags on the screen, the response is freshly generated in that instant rather than pulling up a pre-set screen.

A conventional app is like flipping through pages of a comic book that were already drawn in advance. Press a button, and the next page—already fixed beforehand—appears. An Interface World Model, by contrast, is more like a cartoonist watching your hand gesture in real time and drawing the next scene on the spot. The screen isn't stored somewhere and called up; it's created only after the click signal arrives, right at that moment.

This approach is rooted in video-generation AI that predicts and continues the next frame. An Interface World Model swaps out the target of that prediction—from video frames to the computer screen itself. So a click or drag isn't treated as a command like "press this button," but as a hint that says "the next screen should change like this."

In articles, it appears in sentences like: "Runway is introducing this program as the world's first 'Interface World Model.'" Here, Interface World Model isn't the name of a specific product—it refers to the technical approach of predicting and generating screens in real time—while Solaris is the name of the program Runway built using that approach.