So yesterday I talked about some different relationships that could possibly be explored in Project Greenshift. These relationships are localized meaning that they are derived from the current piece of content in some manner. Many of these relationships are arbitrary, but some of them could be useful in uncovering new content.
But as mentioned last time, we need some place to start. Even if we have these relationships to generate new interesting pieces of content in the next round of exploration, we need initial content to start the entire process off. We need to start with some content not derived from these relationships.
The easiest thing I could think of is to take the top five pieces of content from trending, but then I was like, no. That's not happening. Trending is too stagnant and those people have been rewarded enough by themselves that they don't need additional exposure or attention.
We could instead use the hot and new sections since they are less stagnant, but they suffer from their own issues. The hot section is often dominated by whales and sellouts and the new section contains mostly spam. Neither really makes a legitimately good spot to begin consuming content.
I think we can all agree that the way content is organized ranges from absolute dumpster fire to somewhat mediocre and we certainly don't want users to start on the wrong foot. But thankfully there is another way.
Most rational people follow people they generally like and most people that follow other people also follow people that in some cases have overlapping interests. So, we simply pull our initial content from the feed of a target user. Hell, you could even specify someone you hate and see where that goes. Just in the rare case you want to go on a "flagventure".
So, given a seed user, we will randomly select five pieces of content from their personal feed and use that as a reference point. Since all feeds are curated to some extent this allows us to hopefully bypass a lot of the promoted and spammy content found in the default feeds. And for new users, they can simply type a different author's name and begin exploring that way.
Now with a starting point and five pieces of content, users can now select one, and we can generate five new pieces of content using our relationships. We simply rinse and repeat until the user gets tired and quits. Hopefully traversal works similarly to going down rabbit holes that happen on other media sites and makes them interesting.
Steem is so underdeveloped in terms of content discovery that on the surface level it may appear boring even though a treasure trove of content may be hidden on the blockchain. Let's dig for that buried treasure!
