I've noticed that $a \bmod b$ (with 'mod' in the operational sense) can also be represented using various tricks of formulation. For instance, that
$$a \bmod b = a - b\left\lfloor\frac{a}{b}\right\rfloor .$$
There are several ways to do it by (ab)using various functions like ceiling, max, abs, and the like. However, I realized yesterday that
$$a \bmod 2 = \frac{1-(-1)^a}{2}.$$
I find this interesting as I consider basic exponentiation to be a purer operation in some sense than something like floor, perhaps in that it doesn't have a built-in sense of conditionality or knowledge of fractional parts. Furthermore, you can use substitution to reach arbitrarily high powers of $2$, as in
$$a \bmod 4 = \frac{1-(-1)^a}{2}+1-(-1)^{\frac{a-\frac{1-(-1)^a}{2}}{2}},$$
which amounts to right-shifting $a$ by one spot and repeating the process to get the second bit you need for the $\bmod 4$. I realize that's not pretty, but I'm interested that it's possible. The motivation here is identifying situations where formulae may implicitly support a richness of computational complexity one wouldn't expect, which parity detection and manipulation goes a long way towards. Which leads me to my question:
Is there some way using exponentiation or other basic operations to find a comparable expression for $a \bmod 3$, or ideally, $a \bmod b$?