So this is hardly news at this point, but I figured I'd comment on it now that things have progressed a little and I've exported my projects.
For those of you who don't know AppInventor: It is a web application created by Google to facilitate the creation of Android apps. The premise was that you wouldn't need to know how to code: It was a drag-and-drop creation environment, and sadly it didn't get terribly far.
It's not that it was terrible, mind you, it's just that.. well, there weren't many employees behind the creation of this tool and it wasn't open source, so development and bug fixing was slow, and there were two glaring flaws: Apps that were created were large (one of mine, which was essentially a predefined phone book, was 1.3MB), and these apps could not be uploaded to the Market: they weren't actual apps yet (they do not "compile" to bytecode) and required a helper to run.
I'm sure this can be fixed and AI can become a great tool, but I believe the main issue is one that has been a thorn in Google side for a while now: They're spread too thin and are trying to do too many things at once. Google may have the money for it, but they don't have the manpower. Thankfully, they seemed to have realized this and have not only open-sourced the code instead (or will) and have funded a new MIT Centre for Mobile Learning; This, to me, is a great idea as it gives Google a certain amount of control while not requiring them to directly manage it.
(Oh, and it being open source means others can contribute, but I don't see too much of that happening soon. Call me sceptic but those who can develop will probably not notice the tool at all and be directly trying to create their own apps. Then again, beginners could still be stuck in that middle zone, and perhaps advanced developers will still see it as a useful RAD tool ... if and once you can create actual source with it, I'd imagine!)