I'm lucky enough to get to use Haskell in my AI class. I liked my implementation of the DPLL algorithm so well that I got permission to post it! I tried a new (to me, in Haskell) methodology of writing a bunch of code at a fairly high level using intentionally abstract types/functions that didn't exist, and then defining them later on based on how I was using them. It worked pretty well and the result was cool.
The writeup that I turned in is in dpll_writeup.lhs, and a PDF/Latex render is in dpll_writeup.pdf.
The intended audience is my AI professor, who doesn't know Haskell, so some concepts might be blatantly obvious (like what maybe does) while some might be obtuse (what is the DPLL algorithm for anyway?).
I'm writing blog post aimed at Haskellers to accompany the code.