Absolutely. I agree completely. And I am stoked that LLMs are about text, but accessible (because it's natural language adapted to the situation, and you can ask questions to it).
And, as you say, the canvas needs to be right there too, to see what you're doing with all that text. No context switching. This is reminiscent of Netscape being the bomb when it launched because it could show images right there on the page (no context switching!).