Runway introduces AI model that renders software interfaces as video

Runway introduced a new AI system named Solaris that creates software interfaces on the fly. Instead of executing underlying code to render button clicks or layout changes, the technology predicts and draws the next visual frame directly on screen in real time as a person interacts with it.

In traditional software development, designers create visuals that programmers must convert into strict computer instructions. Runway claims this middle step limits how applications respond to users. By bypassing code entirely, Solaris generates a responsive visual environment where actions like trying on virtual clothes, altering room layouts, or testing physical simulations happen fluidly. The company also suggests this visual approach could help automated software agents learn to navigate unpredictable website layouts more effectively.

However, relying entirely on visual prediction raises significant practical questions about precision and security. Because traditional apps depend on deterministic code to process sensitive actions like payments or form submissions, it remains unclear how a system generating raw pixels can reliably handle strict data logic without making visual errors or introducing security risks.