In this article, the examples are artificially constructed to support their claims in a dishonest manner. For example, they talk about how “floating label” text is smaller than the “IFTA label” text even though there is no reason they couldn’t be the same size (and, of course, the text is larger and more visible in the floating label’s initial state). The same thing is true of the “crammed spacing” claim, there is nothing inherent to the floating label that reduces spacing.
Finally, the article suggests that “IFTA label” is better by using colour to differentiate between complete and incomplete fields. This manages to ignore well-established research showing users with achromatopsia and low vision won’t be able to differentiate between those states at all since the contrast difference is only 1.25:1! At least unfilled fields with floating labels have no label at all above the field—which is not a great UI clue either, but is at least one that is legible for everybody. (In any case, unfilled fields should always be discoverable by form validation at submission time pointing out which fields are invalid and why.)