its a good solution but its not elegant - sCrypt is a GPU killer, and i assume (without reading anything on it) its still going to have the memory hard function and thus the higher voltages - ?
Actually Neoscrypt handle memory little different than scrypt.
However code is not 100% optimized for GPUs yet so it is hard to tell how much the performance will be better.
Sounds good -
but i have a question why try to do something radically new if a hard fork is required - why not do something that the market knows ( like i previously mentioned ) for example taking well known algos from the "X" bunch or the "Quark" type and mixing them into a new Random driven function?
then rename this your own the point is :
Will the market see this as more of the "CPU only" reset that gives the designers of the function a distinct advantage over the rest of the market at inception as they will have an optimized GPU and the rest of the market will have whatever they can develop.
wouldn't there be risk that the market will see his as manipulation?
We are working on GPU miner, it should go live with miner ready so GPUs can be used from start.
+ Neoscrypt will be completely opensourced in few days so anybody can develop/modify his gpu mining software before live launch in few weeks

Completely transparent, no ninja launches...
Meanwhile take a look at awesome rPI project of tmuir12 who just uploaded his progress on youtube:
http://www.youtube.com/watch?v=DJdsMX-eeFM&feature=youtu.be