Google provided all the facts. Everything else is opinion, preference, and hearsay. Here’s mine:
The thing I’m most likely to use App Engine for in the short term is to whip together those tiny site ideas I’ve had floating around; stuff that’s probably not going to make me any money and therefore doesn’t make sense to spend money on hosting for. EC2 can’t compete here, as even a tiny site means at least $70/month, and even that doesn’t automatically scale. A cheap VPS also can’t offer me easy scaling if a small project blows up. This is the first visible deploy-and-forget offering out there.
I like that the way to give your App Engine app a custom domain is via Google Apps (nee For Your Domain). It’s too bad Google Apps accounts accounts are still second-class citizens in Google-land.
You actually can’t get an App Engine account with a Google Apps account. Or use Google Groups with an Apps account. Or sign up for AdSense with an Apps account. Etc. etc. etc. ad nauseam.
Python. Is anybody surprised? Python is easy to learn, clean, clear, and fast enough. Google employs Python’s author, so cooking up a sandboxed version of the language probably wasn’t so bad. Python web frameworks still lag a year behind Rails in features, but it doesn’t matter. Python is a better fit, and people will learn it just to play in the big leagues.
Clever developers will build service-oriented architectures around App Engine. Many applications won’t fit Google’s constraints, and it’s bad business to be completely dependent on an inaccessible vendor. But people will want some Google magic, and they’ll build REST services as App Engine apps to backend sites hosted elsewhere.