Runway imagines a computer where windows become video simulations

Runway Labs is experimenting with Project Continuum, a computing environment built around real-time generated video interfaces. Portals, Visual Thinking, Responsive Video Interfaces, and Interactive Worlds explore four ways Interface World Models could move beyond conventional graphical interfaces.

An interface that generates what it needs to show

An explanation can take the form of an animated simulation directly overlaid on a page. A video can change its composition when its window is resized. A code agent can make its actions visible while it works.

Project Continuum brings these experiments together in a research application developed by Runway Labs. The project is built around Interface World Models, including Solaris, which are used here to generate and modify video interfaces in real time.

Four concepts make up this first demonstration: Portals, Visual Thinking, Responsive Video Interfaces, and Interactive Worlds. Portals generated during browsing

Portals appear as interactive video zones generated from the content being viewed. They can represent a concept, a place, or a simulation without leaving the current page.

Runway intentionally gives them soft edges to visually distinguish the generated content from the rest of the interface. Once moved within the page, a Portal remains displayed alongside the existing content.

The video interface does not function as a fixed sequence. When resized, the model recomposes the scene to account for the newly available space. It is also possible to click directly inside to interact with the simulation in real time. Watching a code agent work

Visual Thinking applies the same principle to programming agents.

Instead of limiting their activity to a succession of messages, the video model visually represents the different steps of the work and the tool calls as they are executed.

Here, Runway aims to maintain a visible representation of a complex autonomous workflow as it progresses. However, the demonstration remains a research exploration and does not yet document compatible code models, performance, or access methods for this interface. A video can become an interactive environment

With Interactive Worlds, an existing media asset serves as the starting point for a real-time simulation. An image or video can be expanded into a multimodal environment in which a person or an agent interacts directly.

Meanwhile, Responsive Video Interfaces explore generative content capable of modifying its composition based on the size and layout of the interface, rather than simply scaling a pre-existing video.

These two concepts bring the graphical interface closer to an environment generated on the fly: its visual content can be recalculated based on the available space or user interactions. A research application, not yet a new OS

Despite being presented as research around a new operating system, Project Continuum remains an experimental application from Runway Labs at this stage.

Runway has not shared a release schedule, system requirements, or pricing. For now, the demonstration serves to showcase four interaction pathways built around real-time video generation and Interface World Models.