The Anti "Anti-AI" Labeler, @antiantiai.bsky.social, is great for the non-"Anti-AI" community, but apparently setting up a labeler that tracks users who subscribe to certain lists is challenging.
There is probably an easier way to do this, especially if you intend to serve a smaller community.
Limit the people the labeler needs to work for to those who "like" the labeler.
For each person, review all people who have interacted with their posts in the time since the last update, and check if they meet the criteria for the labeler.
For any person on the list, check if they no longer meet the criteria to be on the list.
For labelers that serve < 1–5k users, which should be a pretty maintainable model that can run on a laptop for no additional costs, given Bluesky's current API structure.
Here are some quick calculations without any caching or backup optimizations.
Assume 5k labeler users that post 10x/day and receive ~3 interactions/post, and assume it takes 1 API call to check the labeler criteria
5k API calls to fetch their feeds
50k API calls to fetch replies to their posts
50k API calls to fetch quotes to their posts
<150k API calls to check criteria for all users that turn up for review
Total is ~250k API calls per day
Is this manageable? I don't actually know because I haven't tried to set it up. I would be curious to hear from people who have operated labelers.
Links to relevant material below:
