I made a couple of tech demos so you can see what I'm talking about more easily:
http://www.gayheap.com
http://blog.gayheap.com
http://tube.gayheap.com
I whipped each of these sites up in less than 30 minutes, including design, using the API.
The first example pulls the most recent scene from 3 different sites, then displays a hosted video (if one exists). If not, it shows the "large" picture. It then pulls the scene title, the models in the scene, and the scene teaser text. It pulls the URL for the tour, and a couple of other images as well, and then slaps them up on a page. Easy-peasy, and it updates when the sites update. Surfers can scroll to older updates with the navigation at the bottom of the page. I didn't add any search because my goal was to finish each site in less than 30 minutes, as a proof of concept.
The second example was a (CSS) re-skin of the original site, without the thumbnail row, and without the flash videos. More of a "blog" feel to it. It has the "full" descriptions for the sets, rather than the shorter ones used on the previous site.
The third example is just a simple attempt at making a MGP masquerading as a simple tube. It just refers the surfer to the tour pages, but it serves as a proof-of-concept for handling video-type sites.
Does this help to make more sense about APIs? The "components" (a picture, a description, a model's name) were all available to me, and I chose how to lay them out on the page. Compare and contrast that to what you get with RSS, or another format.
If you want an example of a killer API, check out Pink Visual's implementation:
http://api.pinkvisual.com/