http://uguu-archive.appspot.com/misaka/replay.html
http://ioccc.org/2013/misaka/spoiler.html (mirror, in case if I ran out of quota)
I would be curious in some implementation details for this project, particularly how the intermediate deltas are stored.