This problem showed up in the MIT integration bee last year: $$\int_0^1 \frac{dx}{\lfloor{}1-\log_2(x)\rfloor}$$
Basically, after doing a lot of tedious work I graphed out part of the function and noticed that the function was just a bunch of rectangles. The width of rectangle $n$ was $2^{-n}$ and the height was $\frac1n$ so the total area was: $$\sum_{n=1}^{\infty}({2^{-n}})(\frac1{n})$$
After some more work I recognized this to be the taylor series of $-\ln(1-x)$ evaluated at $x=\frac12$ and since $-\ln(1-\frac12)=\ln(2)$, I concluded that the integral was equal to $\ln(2)$, which some numerical integration verified for me.
My question is, is there any easier way to do this that maybe isn't so tedious?
Let's see, if you do $u=1-\log_2 x$ then
$$-2^{1-u}(\log 2)\, du=dx$$
Then you get
$$(\log 2)\cdot \int_{1}^\infty{2^{1-u}\over \lfloor u\rfloor}\,du$$
This is clearly
$$2(\log 2)\sum_{n=1}^\infty \int_n^{n+1}{2^{-u}\over \lfloor u\rfloor}\,du$$
But on the interval $[n,n+1)$ by definition the floor function is equal to $n$, so the integrand is just ${2^{-u}\over n}$ on that interval, which allows us to rewrite this as
$$2(\log 2)\sum_{n=1}^\infty {1\over n}\int_n^{n+1}2^{-u}\,du$$
the inner integral is easily seen to be
$$\left({1\over2\log 2}\right)2^{-n}$$
and of course this gives
$$\sum_{n=1}^\infty {1\over n2^n}=\log 2$$
Of course this is still the same sum, but you don't have to do any guessing, it's all definitely the right thing, without experimenting--and you don't need any numerical tests to verify it as a guess.