Runway announced Solaris, its first Interface World Model — "a new kind of operating system that generates interactive interfaces frame by frame, in real time, with no code." Instead of an LLM writing code that then executes, Solaris renders the interface layer directly: click a button and the next frame is generated in response, the way a video model generates the next shot. It's built on Gen-4.5, adapted to understand interaction and respond in real time, and continues the world-model line Runway started with GWM-1. The engineering is a checklist of what real-time generation demands: frames generated autoregressively, each depending only on what came before; the many-step denoising process distilled to a few steps; and the fast model trained on its own outputs so quality holds over long sessions. A language model still decides how the interface should evolve — Solaris just is the screen. Runway's benchmarks make the case in two ways: frontier multimodal LLMs (Claude Fable 5, GPT-4o, Gemini 2.5 Pro) all lose information when asked to reconstruct interfaces as complexity rises, and in a 250-person preference study Solaris beat Claude Opus 5-coded interfaces 61%–24% on following instructions and 71%–21% on natural behavior. Runway is candid about the hard part: stable, legible text — the thing interfaces need most and video models do worst. Early access is open by request. For this timeline the significance is the direction of travel: announced the same day fal.live turned a video model into a live broadcast, Solaris turns one into an operating system. The technology built to generate films is becoming a general real-time visual medium — and the money and research pushing it forward no longer needs a screenplay.