

This sounds horrific to me! The question in my mind is not how many ghosts to allow over here, but how to generate authentic interactions between real humans. We don’t need every single niche covered, we just need people to enjoy their time spent here.
I do not want to have to scroll down reading through tens to hundreds to thousands of comments, none of which represent real humans who would respond if I replied to them - that is the precise opposite of what I want.
I do not think that the overall lack of content is the problem here - it’s the overall lack of content that is worthwhile to read, …
I didn’t think I conveyed the idea well. It would not be an automated or bot process (Reddit comments wouldn’t be added as Threadiverse comments). The good/useful comments from Reddit would be utilized by a poster to add to their own threadiverse post to help add more relevant content (Example that got 57 comments).
A further 90% improvement can be attained by blocking Lemmy.ml …
While that may be the fastest solution, the triad haven’t been doing great in at least the last few months, even without major intervention (Post with Data):
Lemmy.ml:
- Difference from 3 Months ago: -103 MAU (-5.5%)
- Difference from 1 week ago: -23 MAU (-1.3%)
- Posts and Comments (per week): 727 (+106 vs historic)
- Difference from 3 Months ago: -78 MAU (-5.7%)
- Difference from 1 week ago: -18 MAU (-1.4%)
- Posts and Comments (per week): 1659 (-3545 vs historic)
- Difference from 3 Months ago: 22 MAU (+3.9%)
- Difference from 1 week ago: 0 MAU (0%)
- Posts and Comments (per week): 311 (-1 vs historic)
Average of the other 16 Instances (of Top 20, excludes lemmy.world who’s similarly crashing):
- Difference from 3 Months ago: -16 MAU (-1.4%)
- Difference from 1 week ago: +3 MAU (+0.3%)
- Posts and Comments (per week): 387 (+22 vs historic)
It looks like hexbear.net is suffocating by itself. Maybe some users move from hexbear.net to lemmy.ml or the declining hexbear.net activity is forcing lemmy.ml users to be more active (I have no evidence of either, but for argument’s sake say they do), even then lemmy.ml is shrinking too.
In this very conversation thread there was a particular onerous troll who kept following our conversation and downvoting everything I said …
I did wonder what happened to that person.
New joiners, it seems obvious from the comments they leave (usually on other platforms after already leaving), don’t bother to put in the work to curate their experiences and instead simply flee …
Seems like some efforts toward better on-boarding could return good results (could include default blocks, etc.). I have some other analytics I need to get working again first, but after, I think I’ll take a stab at looking at decoupling and identifying key instance on-boarding differences that effect retention.
Edit: moved comma.
Edit 2: Looks like Pylova likes the subscribable block list idea: https://codeberg.org/quokkau/PyLova/issues/30




My understanding is federated “payment” systems are difficult. I think there is active discussion on the topic, but I’m not sure where code/development is at, or if it’s even started.
I think flohmarkt (“federated decentral classified ad software”) is the closest we have right now. It has low (~10-30) Monthly Active Users (MAU), but is in active development.