I think that is a great model. Make relays front and center, and users travel around browsing them.

I also think some users want to know "What is Edward Snowden saying?" and they don't really care where he is saying it, they just want the "it" being said.

So we need multiple models working in parallel.