a16z's Justine Moore on the potential of multimodal AI in 2026, where anything-in, anything-out models give users the freedom to iterate rapidly across many kinds of media: @venturetwins (a16z)