Recent work on world models can already generate photorealistic streams in which a single agent wanders freely. But multi-agent social interaction is still clumsy. Scripts paper over the gap; agents don’t feel like they share a world.
My claim: the gap isn’t in the policy, it’s in the representation. If two agents don’t have a common substrate describing what is happening and what might happen next, they can’t coordinate. They can only imitate coordination.
The Reachy pilot study: what 38% means is one concrete example of this in miniature. Even with good perception, our VLM picked generic expressions because it lacked a persistent scene model of the human it was engaging with.
Related: Grounding signals beyond language, and a much smaller instance of the same failure in Shared state is a tiny world model.