One of the essential benefits of this protocol above and beyond content stored on IPFS is: you get human-readable names for your content and your peers, and you can use those names to form links that refer to other data. Memos give us a way to deliver free-form headers up-front that are also decoupled from the raw data of the content, which is very nice to have when resolving links to content you don't know much about.
Other protocols in this domain start with the premise: in order to organize thought, we should crystalize the data structures to represent thought.
In our approach we seek to develop a substrate that such data structures may emerge from, but without a rigid presumption of how they ought to form (or even that there should be only one representation).