Comment by gpm

I suspect the non-tech-demo version would be multimodal, and include support for textual tool calls...

Converging on what AI companies are already doing... just from the other direction.