Seems plausible. I am imagining a voice that reads all the submitted link titles and then lets you choose one and go to the link or listen to the comments. Having comments be actual recordings of the commenter's speech would actually be quite interesting and more personal than reading them in the fairly generic text form.
Text is much easier to skim than audio. Visual bandwidth is higher than auditory (though less so for communicating emotions), and the receiver has much more control over the cursor.
Intelligent audio communication is often more concise than its equal sum of text. Spoken poetry like rap is a vector for expression that can convey cultural plights and experiences in a few words or sentences. Ideas are compressed through memetics and words can share dimensions across the scope of the song as well as the (+/-) zeitgeist.
