I have considered this, but this option is intended as a quick filter that the user can reset/update on a daily basis, while the kind:10000 mute list seems more suitable for permanent muting (it is a replaceable event).

fevela.me allows that:



however, without supporting any NIP, it is only a local option.

I think https://vertexlab.io is doing that.

Yes, since the second order is feasible, I could explore this possibility. Thanks!

The duplication of reposts is actually another problem.
By the way, and UI change doesn't help with filtering out annoying reposters, for this reason, I would like to understand who implements kind:30007.

For what? Search or network?
Unless you have a niche target audience and are able to tailor your campaign very well, it's a bloodbath. Then in the last years they forced an annoying autopilot-mode, removing the possibility to precisely control the keywords matching, so you are basically feeding and blind trusting the algorithm. An the algorithm wants a solid budget just to start running your ads.

How does this solve the problem? I want to filter kind:6 for specific people that use them too much, without giving (to me) any real value.

This is something that is very hard to achieve purely client side, since it is data and cpu intensive.
About the visualization: graphs are cool, but they should really offer a useful perspective, and I don't see much possible innovations here; a graph of names/pictures is quite useless without any segmentation about interests and activity patterns.
A WoT score like Coracle.social could probably be sufficient to highlight contacts that have a good trust level.

@hodlbod can you share your experience about implementing the WoT?

Lightning?!

In fact I noticed that NIP-85 doesn't have a last_created_at result, I was going to open a PR.