iiinova — where your digital self keeps evolving

even had to go to an old version of Gephi to get this image preview plugin working >_>

010

graph is based on follows

000

yeah I forgot about their tweets too. good correction!

010

cool. this embedding based or graph-layout based?

000

opened up the graph file and you're there. the layout isn't the same as the original, but if you look hard enough you're in the original somewhere!

340

every node in that CD txt file should be on the viz. I crawled on 4/30.

230

putting things into perspective with The Blue Marble & The Blue Sky

0182

If you're the kind of person who wants to dig deep into tiny niches on Bluesky here you go github.com/deklanw/basic_bluesky_an… Browse or ctrl+f in this file and you'll find something interesting

1180

OK, I'm tired. Done for now. Might post more tomorrow. Code is here. Please forgive the messiness github.com/deklanw/basic_bluesky_an…

2110

Example niche you should check out: Sveltey Frontend influencers Top accounts: @threepointone.bsky.social @rich-harris.dev @jaredpalmer.com @swyx.io @thatjdanisso.cool @wattenberger.com Check out that last account in particular if you enjoy data viz. She's spectacular

1120

Example niche you should check out: Meditators, Buddhists, consciousness, psychedelics Top accounts: @leona.bsky.social @metaphorician.com @meansinfinity.bsky.social @tramazonlily.bsky.social @neats.bsky.social @vincehorn.space

2141

Example niche: Future of coding / humane computing Top accounts: @geoffreylitt.com @gordon.bsky.social @thesephist.com

1192

That pretty much does it for the highest level of community structure. To me this high-level view of the network gets boring fast. But, you *can* make finer distinctions! They're just harder to visually untangle in a viz of the full network. I'm going to post now about some niches

3150

deepfates ur killing me with that long ass handle

2560

Now I'll have some fun by showing some of high-level accounts with top accounts. First: there's a surprising number of non-English accounts here. I'll do them at once for anyone who wants to take a peek:

3460

Staring at the animation above it's striking how fast different communities can burst onto Bluesky. For TPOT a lot of it was probably @deepfates.com.deepfates.com.deepfates.com.deepfates.com.deepfates.com memeing it into existence on Twitter. Don't underestimate coordinated invite giving!

4660

For a short time TPOT comprised a major portion of Bluesky. Around 20% of all accounts. If you restrict that to just English-speaking accounts then it was probably closer to 50%. Once TPOT growth saturated and other communities joined TPOT's pct dropped off. Here are two visualizations of that

4823

Around April 20th the Digital Art community similarly emerged within a few days. General tech arrived around the 22nd-28th. And, Politics/Mainstream arrived around 28th-29th. Korea grew slowly the whole time.

2522

The timeline so far is, roughly: OG Bluesky (employees and early invites) plus some Brazilian, Japanese, and Middle Eastern accounts. Slow growth of these communities (especially Middle East) until around April 9th. TPOT rapidly exploded onto Bluesky from April 9th-13th after which it cooled off.

2640

Here is an animation of the Bluesky network's evolution over the month of April. (The community blobs line up with the image in the previous tweet. See that if you want to know what the blobs roughly correspond to) Bluesky doesn't support gifs/videos so here: vimeo.com/823187302

110425

Presenting an exploration of Bluesky's network. Going to go over the community structure, evolution of the network, top accounts, etc. First, here is the high-level community structure with some labels. Hi res w/ zoom here: www.easyzoom.com/imageaccess/884cb1…

74587231

what are the plans for helping bluesky with that data? @robhaisfield.com was brainstorming some ideas already. I have some janky code for scraping bluesky follow graph. it has 22k nodes and 964k edges. doesn't take long to scrape. will do some analysis with it soon. if u want to play with data lmk

120

because of hub accounts which follow hundreds of thousands or millions of accounts distance isn't a great measure on Twitter. if you restrict your hopping to only follow out-edges for accounts with reasonable (< 5000~) out-degree then you get something a bit better

150

this was one of the first usecase ideas for GPT-4 proposed within 24 hours twitter.com/jbrowder1/status/163572…

040