Thanks, man!
These are pretty valid concerns and are difficult problems. So far, what I've observed in the relay is a variety of npubs that I've never seen before, which is a great surprise. This is because the algorithm allows everyone to publish, even totally unknown pubkeys, but only kind 1 at day notes that don't contain links in the content. The other key here is that Wotrlay itself works by plugging in an external and replaceable Relatr instance. Relatr is where the rank computations happen, and it is very flexible in what makes up that rank. As you run your instance, you could decide not to take transitive mutes into the computation, maybe average mutes over time, or whatever you prefer. With Relatr in place, computing ranks following your own model, it is plugged into Wotrlay using ContextVM. So when the relay receives a new event, it first checks the configured Relatr instance, gets the rank for the author of the event, checks the bucket for that pubkey, and determines if the event can be published or not, taking these two factors into consideration.
We've designed all of this to improve for these concerns. Cold starts are solved by always allowing anyone to publish a kind 1 note a day, just to say "good morning" every day or introduce themselves. Then, as the user participates in that relay or other relays, the rank can evolve and the bucket grows, so it has more publishing capacity. I think, this is a good way to model the onboarding experience for someone in an organic way, letting social dynamics play out. Of course, it can also be bootstrapped, and new users can get full rights if the operators of the relay allow it in some way, for example, using a badge or similar, which can be part of the Relatr computations. So, relay operators can give you a badge, and that is already the right to get access.
Echo chambers are also mitigated by the same fact that relay operator are able to determine how the ranks are going to be calculated. You can come up with an algorithm that benefits diversity and creates ranks based on that. Chilling speech and fragmentation are also related to this. And capturing abuse and gaming is an inevitable risk from my point of view because we are in a permissionless land, which also means that anyone can create their own rules even if they are not fair. But it is always up to you to opt out. There is no perfect solution for something so relative and dynamic like trust and ranks, but the approach we are following with Relatr and Wotrlay, I think, is a step forward in the ability of the users to determine how these things work and tailor them to their needs.
There are also some related comments and more details in the ContextVM blog posts:
https://www.contextvm.org/blog Also available as long-form content in Nostr.
