One of the more interesting editors I've seen being worked on is edlib (https://github.com/neilbrown/edlib). The theory is that rather doing a buffer-based approach as in Emacs, the text is actually just a representation of a file. The consequence being that you can have a presentation open as one representation and also be modifying the presentation as a different representation. It's very young (and is a toy project as well), but it's definitely a cool idea IMO.
Emacs buffers are copies of the representation, not actual representations. The author had a talk at a recent linux.conf.au (can't find the link right now), which explains it much better than I could.