So, the one thing I notice is that in every trio of original image, GPT-4.1 image, and Krea image where the author says GPT-4.1 exhibits the AI look and Krea avoids it (except the first with the cat), comparing the original inage to the Krea image shows Krea retains all the described hallmarks of the AI look that are present in the GPT image, but just toned down a little bit (in the first, it lacks the obvious bokeh because it avoids showing anything at a much different distance than the main subject, which is for that aesthetic issue what avoiding showing hands is for dealing with the correctness issue of bad hands.)