The two-glyph thing is a matter of description length and parametrization. 10+2 is an expression, whereas 10 is also a constant that is the base of our number system. Thus, 10 carries more meaning than 12, even though both are constant - 10 is potentially the aforementioned parameter, but what is the 2? With sufficient study, 12 becomes a number of important constants as well (such as the number of inches in a foot, or the integral of a centered quadratic function), but not ones considered fundamental.
Essentially, it's about the difference between a concept and a measurement, or the difference between (x+y)/x and 1+x/y. 2pi is an expression, pi is the concept.
I've read this a few times, and I don't think I really understand. Perhaps you could explain further?
10+2 is an expression, whereas 10 is also a constant that is the base of our number system.
There is no notion of an "expression" as distinct from a "number" (or "function" if it involves a variable) in any branch of math apart from computer science[1]. In algebraic terms, (12) and (10+2) and (6x2) and (0xC) and (2^4-2) and "twelve" are all literally the same thing. Well, technically they are all equivalent notations for the same abstract concept.
Thus, 10 carries more meaning than 12
Even if I accept this (which I'm not convinced I do), it's beside the point: 10 and 12 are not equal. Unlike with pi and a hypothetical tau, using one where the other is called for would be an error.
[1]There is the notion of the limit, which is subtly different: limits do care how a function behaves at other points. One could make the case that this makes a limit into a sort of expression, but to be honest I think that only obscures the idea.
Algebraically, 12 is not the same as 10+2. 12 is an element of, say, ℤ, while 10+2 is one of <+,ℤ²>. To make them interchangeable, we need to establish an equivalence relation. Given that relation, we then have the opportunity to express useful, non-obvious equivalences using transitivity.
Now that we have X=10+2=12, we need to choose which one represent the equivalence class of X. 12 is certainly shorter, but a seemingly magic constant. 10+2 implies that in other number systems, X=b+1+1 may be also true. If the scribe subscribes to the principle of MDL, we can speculate that this is the reason he chose the longer version, and if that is accurate, we have gained more information. If we chose 9+3, we would arguably lose information, since this expression is (hypothetically) misleading.
This is all to say that expressions are more informative than their equivalence classes, since they have been hand-picked to be representative.
To represent the equivalence class of 6.28... with 2*3.14... implies that the equivalence class of 3.14... is more important, and that the prototype likely involves two separate instances of the concept of π. This is misleading.
Essentially, it's about the difference between a concept and a measurement, or the difference between (x+y)/x and 1+x/y. 2pi is an expression, pi is the concept.