1 September 2026
Runway releases interface generator that renders software as video
First reported
TLDR AI, The Decoder and 2 others ran this on , all on the same day.
- Runway introduced Solaris, a system that generates website and app interfaces frame-by-frame as users interact, without running traditional code underneath.
- The model combines Runway's Gen-4.5 video generator with a language model to decide what changes and render each frame in 720p resolution.
- Use cases include adaptive shopping experiences that change per user, real-time 3D product visualization, and visual tutorials, though text rendering remains unreliable.
- The system is in early research stage with known problems in long sessions, consistency, accessibility compatibility, and the risk that wrong-looking results may mislead users.
Where they differ
TLDR AIfocused on the technical mechanism without addressing limitations.
The Neuronemphasized new reliability challenges this approach introduces.
The Rundown AIreported a specific user preference metric (71% chose Solaris over Claude Opus coded pages).
What each one reported
Solaris generates real-time interactive interfaces frame by frame as users interact, handling rendering and user input jointly without needing intermediate representation. It enables new ways to build websites, apps, and train agents in dynamic environments.
Runway introduced Solaris, an Interface World Model that generates UI one frame at a time as users interact, rather than writing code first. The newsletter argues this approach creates flexibility but introduces new reliability challenges around consistency, accessibility, and security that real work will test.
Runway introduced Solaris, an Interface World Model that renders websites and apps as live video with every frame drawn in real time as users interact, with no code running underneath. The system paired Runway's Gen-4.5 video model with an LLM, and in testing users preferred Solaris over Claude Opus 5-coded pages in 71% of matchups on in-scene behavior.
Reported by The Decoder