Oddbean new post about login | logout As a dev, I dont want client devs creating standardized WoT algos. Thats a centralizing technique. Instead, we should be building tools to empower users to define, share, and refine their own algorithms/filter rules. By far the best clients today that do this are... 1) Amethyst - for letting users define word/phrases to filter out 2) Gossip - for list management 3) Nostrudel - for list management Clients going in the wrong direction imho are those tying to create or apply dev controlled built in algorithms.. Notable offenders include 1) Primal - for the trending lists that just focus on most quotes and replies as a popularity contest amongst influencers 2) Coracle - for incessant WoT scoring algo based on follow lists and mute list 3) Damus ? is it now joining the follow/mute algo bandwagon? Reliance on DVMs as a WoT filter is also perilous. Though using it for answers and discovery has merit. #devstr #nostr #grownostr
Damus had wot filtering before anyone. We also have word muting. I just never made it apparent or obvious because i think follow wot filtering is the nuclear option. WoT is user empowering. It says “i trust these people and people they follow… etc) we can also make that more customizable over time with non-follow wot like interactivity, zaps received, etc
Follows ≠ Trust and we need to stop acting as though that is the case. Trust is context driven, hence why I favor lists over follows. Similarly, Mute is a tool, not an indicator of distrust. For most, its implying they dont want to see content from the referenced pubkeys, but there are users that leverage Mute as a temporary negation filter to follows because their client of choice does not adequately accomodate lists.
Taking the blacklist approach against spam is not that interesting, users can't be bothered to keep updating mute/whatever list, so you just become like an anti-virus company sending out definition updates every month just to stop spam.
I disagree, @Fabian, but I may be blind also. The way I see, people are used to muting (and reporting) spam from their messaging threads and email inboxes. This feature is well used across many platforms. The nature of WoT filtering means that each dataset is pulled from a given user’s entire network (of follows follows or whatever) to provide an even richer dataset than the user alone has created. Even if one user is a bit “lazy” in muting or reporting, blacklist spam filtering can still be effective. Am I missing something?