Trying this out now. Interesting idea! It sure does load slow on 75,000+ reblogs. Also, I wasn't quite sure how it worked when I clicked (didn't realize it showed up below the post) and clicked it twice. It's loading the graph twice. Maybe make a second attempt on the same post not load another copy of the graph?
I wish there was metadata about individual nodes or the structure. I would also like links to the nodes to find tumblr blogs that get a lot of reblogs.
Ouch! I wouldn't try to do it on 75,000+ reblogs. Since Tumblr doesn't expose reblog data via their API, I have to manually go and load 50 notes at a time via the urls from the 'more notes' link. Each request takes about 700ms and is ~75kb of markup. For 75,000 reblogs that would take about three hours. Plus, rendering them all - keep in mind this is O(V^3). Definitely not optimized. That's why I suggest a typical maximum of around 500 notes to visualize...
Adding metadata for each node on-hover is definitely, definitely on my to-do list. That will make it useful and not just a toy, I think.
Some other metadata besides just which tumblr each node represents that I want to get on there eventually is its centrality.
It's raw canvas. I didn't want to use any libraries to help me out in the graph-building - that was the main focus of the weekend project. My next steps, of course, are to learn more about graph theory so I can implement less naive physics calculations.
Given it's a hacker audience, I'd say there's an implicite meta aspect to all such submissions. So yes, I appreciate knowing this was a weekend project... and a very cool one at that!
I wish there was metadata about individual nodes or the structure. I would also like links to the nodes to find tumblr blogs that get a lot of reblogs.