I'm personally much more fond of the Unix 'tool approach', in which there are no task-specific modes or plugins, but tools can still be composed ad hoc, as in Plan 9's Acme.
The idea is that Light Table will be built from the ground up to support very dynamic environmental contexts in which you can place anything.
The comparison to emacs isn't meant as a slight to LightTable either, but please, let's not forgot the past when creating the future.
With that said, wouldn't his "It's hard for me to imagine being able to build anything I showed in say Eclipse in any reasonable amount of time" comment also apply to Emacs?
Light table looks to me (with all the respect due to the work of the authors) like a fad where people get all hyped up about a product. I mean there's nothing revolutionary and while looking at the video I was thinking "boy this is too confusing!". I also believe that when the project is going to have 500+ lines of code per file, the way it's done now, is going to be horrible to deal with it!
There are other stuff that bug me out: when I program I don't want rounded borders and everything. I don't want 40 different colors.
But if people are going to be it, great.
As for the rounded corners, I think it really is a matter of personal preference and they're definitely not going to affect my ability to program. Colors are just used to visual convey information seamlessly (think syntax highlighting). I think they can really be put to good use.
I do agree with you that huge "files" (the concept seems to be abstracted by the IDE) could get messy. However, I trust the Light Table guys to get it right :)
Isn't that the whole point? To me this looks pretty amazing.