I've been beta testing this for a while now, and I have to say that it's a great concept, but has a long way to go.
It is web based, and works quite well once you get past the Java problem of running the blocks editor.
It has a lot of short-comings, and I'm hoping that it's because it's an early beta.
I'm still programming the Android version of my game directly in Java because the App Inventor is way off from being able to do it.
App Inventor has great potential should Google choose to put the work into it.
My fear is that it will not become as advanced as is needed for serious Indie development, and is aimed at anybody and everybody to create simple apps that will sustain the deployment of google ads via free games.
I've asked Google for a roadmap for App Inventor, but I am yet to hear back.