We've been continuing to make progress, but several teams have been overtaking us. In particular, there seems to be a huge number of points to be had for the "ray tracing" problem in the two-dimensional language. Last I heard from Peng he was making good progress on a robot to solve the "adventure game" puzzle. We're still in the top ten, which regardless of where we place in the end, is pretty respectable out of 238 teams (so far).
I finally cobbled together solutions for the O'Cult puzzles. After finishing the first one, which involved addition and multiplication, I thought I had figured out a way to generalize the solution to such that I could write a small compiler for the "document optimization" problem. However, I hadn't quite found the right pattern and wound up mostly constructing it by hand with a few small bits of code to generate the appropriate patterns.
Overall, this contest seems to have been much more about quickly writing throw-away code, rather than producing beautiful, extensible, and maintainable code. It has still been fun, at least when we aren't cursing their sadistic puzzles, but it is unclear that this contest says anything specific about the benefits of functional programming. Our Subversion repository is a melting pot of languages. So far I believe we have used C++, shell scripts, Haskell, OCaml, and Perl. I am hoping that the organizers release the tools that they used to construct the contest, because I'm very curious as to how they did a few things.
Anyway, it is getting late and I should probably get some sleep to prepare for a final push tomorrow.