It really does seem that the AI is only really trained in drawing anime women with big balloon tits, and even then, unless the anime character is facing straight forward, it struggles with figures.
That's unless you give it a really good prompt, it's as only as good as the information you give it.
It seems good at landscape, but rather because there's a lot going on in those pictures and you need to sit there and gawk at it for a while to pick apart where it goes wrong, the fine details it struggles with, but overal it's not too bad, bit samey... If you refine the images a lot, it becomes visually quite nice to look at, though some times it doesn't even know what it's trying to portray, it still looks cool.
Give it a prompt that is just nonsensical, and it'll come out with some truly horrific shit, give it a prompt where it is asked to create something that isn't common anime women or human characters, like mech warriors or knights in armour, or maybe even animals among different animals, it becomes uncanny, like it gets confused.