A nice new paper from Wyman, Hoetzlein, and Lefohn out of NVIDIA research digs deeper into the idea of a shadow map sampled in screen space. The paper and video are online. This basic idea has been kicking around for over a decade and most people thought it was very promising. This paper seems to have finally really made the idea live up to its promise (watch the video!).
I think this will be a big paper: my biggest surprise talking to video game programmers has been how much effort they put into shadows and yet that is still the source of their most annoying dynamic artifacts. I look forward to seeing good shadows in games!