Amen to all of that. I didn't know until I tried it with an unsupported host that it would turn around and provide a j.mp short URL, which is great, but non-obvious. Until then, I'm wondering why I'd want to use a link shortener with a seven-letter TLD.
It's a serious mistake to disallow shortening TLDs like "google.com", because that's what people are likely to type in to test the service. It's deadly to ever let a user see a hostile, wholly uninformative error message like "INVALID_URI".
Owing to unnecessary ajax and a submit button which doesn't reflect presses, if I fail with google.com and then try yahoo.com, there's no indication that the site has received or acted on my second attempt.
Labeling your only button with an image, in an oddball font, using text that's both passive voice and a negative statement, is not OK, as you implicitly acknowledge in the introductory blog post, when you have to tell your potential users which piece of screen real-estate to click.
I knew what rev=canonical and friends were, and even I didn't have any idea what the phrase "healthy link shortening" meant to you. I thought it might have been a point about not relying on the Libyan government. The introductory blog post isn't doing its job. The list of sites with their own shorteners should be on the front page, and should be longer — that's how you connect the site's value with potential users.
The service itself is useful, though I'd most like to see it incorporated into Twitter clients like Hibari, which use bit.ly for everything. Evangelizing an API is harder. Good luck.