Check this out: @jackcheng can point at his screen and say, “Make a blue square here,” and a blue square appears. He vibe coded a voice- and gesture-controlled canvas on tldraw. Jev can't take images, so it never sees the canvas. Instead, the app sends it text: what Jack said, where his fingertip is, and a description of the shapes already on screen. Jev answers a series of small questions about that text—is he adding a shape, is it blue, does it go where he's pointing—and ordinary code turns…