Post

@jeffbach.bsky.social Hi. It looks like we are working on similar automated block lists, but you were smart enough to put yours on a different account. One thing I've been looking into lately is high churn for following/unfollowing.

0 reposts0 likes

That would seem like a really good metric. A problem I’ve run into is that my (terrible) Ruby code stores the current profile for users it sees on the firehose in a (concurrent wrapped) hash in memory that gets too big.

So every few days I’ll be working from a new cache because I didn’t safely close the app in time and it got oom killed. To keep track of follow(er/ing) churn, I’d need to be much better with data handling.

I have terrible Python code. :D I'm using Jetstream instead of the full firehose to watch for follows and unfollows. I cache counts in memory per did and then write it to a SQLite as cumulative count by (did, day, hour). I can then query for dids with high activity (either total, daily, or by hour).