Well, there actually have been quite a few interesting developments lately. In how you describe the heuristic functions that determine as how desirable a certain position in the game tree is seen by the AI, as well as the check that determine how far to look down in a certain position (quiesence search, for instance), there is a lot of room for creativity and cleverness.
A whole different approach would of course be to make a Monte Carlo AI, which would work a lot different from a Negamax one.
In any case, I was thinking a bit about creating my own algorithm for Negamax-esue games a while back. One of the things I was wondering about at that time, was how to represent game boards efficiently in a functional programming language.
In the end I got sidetracked, and instead of building a Negamax player, I am now building the Tensor library, which allows you to create and manipulate Vectors, Matrices and other Tensors. It was made with, amongst other things, two-dimensional game boards like Chess’ in mind.