This is about pixels being essentially point samples and not geometric shapes. In my opinion the geometric shape model is still a very useful mental image when

This is about pixels being essentially point samples and not geometric shapes. In my opinion the geometric shape model is still a very useful mental image when dealing with pixels and does not contradict the point sample model. When dealing with pixels as shapes the shape is not necessarily a square though. Pixels can be rectangles and this was common in the home-computer era. The article even acknowledges this (in a footnote): "In general, a little rectangle, but I will normalize to the little square here. The little rectangle model is the same mistake."