hi hi today's HF dataset: All user relationships (follows) ever, you can use this to:

1. Grab following information in bulk instead of using an API
2. Investigate growth over time of the average user vs power users
3. Draw (non dick shaped) visuals

https://huggingface.co/datasets/jc4p/farcaster-links

first attempt at clustering this data, surprise surprise the biggest cluster is airdrop and tech enthusiasts

Full report: https://gist.github.com/jc4p/4818b8be89147b0b43dd50d7b3b450e1
Code for reproducing (you'll need a hefty CPU): https://github.com/jc4p/farcaster-links-analyzer

and yes i did write this code in Rust so i wouldn't get a reputation as a .NET dev

prev founded rocketr.net, swish.id now exploring aiXcrypto identity ai. expired sprint car driver & energy data engineer.

both wonderful ideas, i think i also need to implement neynar user trust scores + try to get clusters that have higher density, rn it's very unimodal against cluster ID 11

wondering if channel labels might effect this meaningfully 

or maybe bucketed by major personas/profilings if that data exists through your past frames

Now we need a frame to show which cluster we are in!

"Crypto & Affectionate Companions"

I will refrain from drawing any conclusions there