When I used to work in games people would say to me, "hey I have this great idea for a game..." and then I'd have to gently explain that either the idea had already been tried or probably wouldn't be fun for xyz reasons. Or I would just smile and nod. Watching the GridOS videos they have on their website brought that feeling back. I cringed watching them. I imagined this dude telling some poor programmer, "hey I have this great idea for a user interface..."
"The desktop should never end! It should just go on and on..." Great idea, genius. So you want people to get lost finding their apps? Or worse, lose their apps?
"Oh well then we'll have this map in the top corner so you don't get lost!" How about you just don't let people get lost in the first place. Confine the apps to as small an area as possible.
"Grids are cool! Everything should be in a grid with massive blocks of dead space between groups." So, you want people to get tired of scrolling? You want to not use the available screen real estate to convey useful information?
"There should be this gadget for your left thumb that acts like a jog dial, and have scroll momentum and do this neat little shaking of the items attached to the wheel!" So you want to violate your grid design and rigid fade-in/fade-out visual semantics for this one different UI element. O.. K.. And you also want things to not actually be visible until they stop shaking / stabilize? And you want this under the user's thumb, so they can't actually see what they're controlling? Right.
"But the right thumb should be like a context menu, and have this grid pop-out with context sensitive options, (and all the options should be on the same background with single-color icons)." Is there any rhyme or reason to how these icons are arranged? "Sure! Arrange them based on the distance your thumb has to travel, so commonly used items are closed to the edge.. you know, like Dvorák and stuff." Alright, I guess people will have a hard time finding where things are the first time, but maybe they'll develop a muscle memory for where things are. Oh wait. This is a tablet. There is no physical feedback mechanism. They'll have to use visual feedback and get lost every time.
"We'll have the coolest UI for playing music!" Uhh, yea, isn't that kind-of a solved problem? Plus, you don't really need much of a UI for listening to music on a portable device. Unless you're developing a great UI for managing music--and sync'ing that management with wherever your music library is stored--then you might be onto something. But I suspect you're not.