It's fractal a la the Apollonian Gasket, to wit place a large circle within a space, then repeatedly fill all available outside-a-circle spaces with the largest possible circle.
Interestingly in this case, the limiter is the number of gear teeth - when down to two, stop. It's a fractal if you start your first gear with an infinite number of teeth. That being problematic, when down to 2 teeth scale back up to resume filling space with the largest gear possible ... making the real trick determining whether a gear should be placed such that it does not immobilize the chain.
"Fractal" has always meant "self-similar" not "self-identical". There are subtle and some not-so-subtle differences within all of the repeating shapes in a Julia or Mandelbrot set.
In the most classical meaning, "fractal" stands for "an object with a fractional fractal dimension". Since there is a clear cut on the level of details, the fractal dimension of this image is integer (i.e., precisely 2).
I don't think there's much agreement about the exact, formal definition of fractal, but "Hausdorff dimension greater than the topological dimension" is probably a better one than "Hausdorff dimension is not an integer". Otherwise, not even the boundary of the Mandelbrot set is a fractal! (It has Hausdorff dimension 2).
Gears by themselves are just gears and those gears barely form a fractal structure, though once there was something a little bit resembling third/fourth iteration of a Sierpinski triangle[0].
Quasi-fractal makes sense to me, I'm assuming that the gears are created from generic fractal-generating code (pass a basic geometric pattern, draw a portion of a self-similar fractal) that's been modified to create finite gears instead. Or I could easily be talking nonsense.
The page itself is really cool, and I love the effect. I wonder if I would be able to use screenshots from it for personal projects? I assume not.
As the most minor point that could possibly be made – never seen · before, seems like it could be quite useful.
The MBTI and Big 5 are simply rotations of one another. They explain most of the same variance. This was demonstrated by McCrae & Costa, who created the Big 5, in the 80s.
Particularly the abstract, in which the magic words "However, correlational analyses showed that the four MBTI indices did measure aspects of four of the five major dimensions of normal personality." appear.
The thing is that of the correlations found, the strongest is between the 5-factor introversion-extraversion and the MBTI introversion-extraversion (.74 for men, .69 for women). This is hardly ground-breaking stuff. The next best is intuition and openness (.72 for men, .69 for women).
So on two axes, MBTI stumbles into something. Jung was a pretty smart guy who interview a lot of folk, so it figures he'd have noticed the surface features that anyone can notice (some people are introverted! some people are more conscientious!)
But all of this is by the by. It doesn't matter that MBTI is sometimes right. It matters that it is more wrong than right. If it is demonstrably wrong, most of the time, on most of its types, then it is dangerous and foolish to use it.
You might as well argue that "some people get better from eating bread with mould". Sure, that's true, but I'd rather take antibiotics.
Personality could be strongly typed, we don't know if it is yet. The continuous distributions we see in traits could be due to people behaving in a socially desirable way, while their mental algorithms are distinctly categorical.
To link something like "type theory" to personality theory, I would look to computational complexity theory. The link is conceptually simple - different types of people might be optimized for solving different classes of problems.
There might be a more direct link with mathematical types - I'll have to put it in the burner and see.
The difference between strongly typed and duck typed is not exactly equivalent to the difference between categorical and continuous variables, but it's close enough for now.
To speak plainly, the problem is that with duck-typing in computer science, it's trivial to determine whether or not an object has a given method. With regards to people, it's very difficult to reliably assess whether or not someone has a given personality trait, and it's even more difficult to know whether or not associating various personality traits together is beneficial. Different measures yield different results, both with regards to personality type and with regards to interactions between types.
