Runway demonstrated Continuum, a system that streams an interactive screen generated by a video model. The demonstration included a coding agent using Continuum to reason through a problem involving a 3D bike model.
Runway showed how next-frame prediction can power interactive interfaces rather than just video clips. In one demonstration, users could turn on a car’s lights, rotate it or open its bonnet from a single image, without Unity, Unreal or Blender underneath.
Cristóbal Valenzuela argued that larger video models learn properties such as gravity, motion and object permanence without hand-coded rules. He said the company now evaluates physics and geometry, and that its models have passed its test of whether people can distinguish generated footage from real video.
The Continuum demonstration also turned highlighted text about subjects including batteries and photosynthesis into live, resizable 3D visualisations. For builders, this points towards generated interfaces and visual tools for AI agents, but the demonstrations do not establish their reliability for everyday business use.
Generating interactive scenes without a conventional 3D engine could change how companies build visual interfaces.
A coding agent using a generated visual environment suggests a possible tool for reasoning about spatial problems.
Reliable physics and practical business use remain unproven by these demonstrations, despite Runway’s realism claims.
Written automatically by the Made in 10 Newsroom from the source above.
Traversal has raised $120 million in a Series B funding round for AI agents that investigate software failures. The deal values the New York startup at $1 billion.
Tencent Holdings has released Hy4, an open-weight AI model with visual understanding. Tencent says it can handle desktop tasks across command-line and graphical interfaces, expanding its scope for AI agents.
OpenAI says ChatGPT now handles search queries equal to 20 percent of Google’s volume. The company did not disclose the source of the Google figure used for the comparison.