There’s a conceivable reality where you have a spectrum of autistic traits, but whether someone is autistic is a strict binary. Imagine a lamp that can have any color, but that is either turned on or off. This would be quite funky, because there’d have to be some sort of mechanism that causes strict grouping - something you see in psychology maybe sometimes in sequence learning research and some types of reasoning research, but otherwise is quite rare.
However, this is obviously not reality.
There’s people with more pronounced autistic traits and those with less pronounced ones. There isn’t half an autist, but there can be someone whose autistic traits’ intensities are near the middle of those of a person who clearly is autistic and those of a person who clearly isn’t.
The categorical nature of diagnoses does not reflect the underlying phenomena, it reflects arguments about healthcare resource allocation. The actual phenomena are more nuanced
Narcissism is a somewhat multidimensional trait with gradients of intensity. Who told you that it’s a binary classification?
That’s a more realistic analogy, but my point here isn’t that autism is not a multidimensional and continuously distributed trait. My point is only that a spectrum and dichotomized group membership are technically conceivable, even if substantively absurd.