What I notice is that the model behaves better when you already know how to structure the content. In a way, it does not replace the thinking, but it can support the translation of that thinking into a visual form.

→ View original post on X — @antgrasso