The article said the technology is very similar to Google Goggles. From what I could deduce, Goggles uses some kind of local invariant descriptors (like SIFT or SURF), and TinEye uses some kind of global descriptor (maybe something like this: http://www.hackerfactor.com/blog/index.php?/archives/432-Loo... ).

If that is right, Google will be able to retrieve different photos in which the same object appears, whereas TinEye only retrieves the same image, with or without some changes. So, they're quite different beasts.

Can't wait to see if I am right...