All of those can be easily gamed. I can mine and script thousands of keys and make posts with them and wait a month.
The only solution is Web of Trust. A pubkey is less trusted the larger number of degrees of separation between your follow list and it. A direct follow has max trust, a follow of a follow has decent trust, and so on.
Maybe nip05 domains can be trusted? Suppose all your keys are from a few domains known to spam, just blacklist those domains from contributing to trending algo? 🤔
Or say something like any NIP05 domain with more than 1k keys and are unknown don't count?
Using follows as a proxy for “trust” (and mute lists as a proxy for “no trust”) may be better than nothing, but at some point WoT needs to be more sophisticated in a number of ways. For one thing, “trust” needs to be contextual. For another, we need tools to allow us to make explicit trust attestations so we don’t have to rely upon frequently flawed proxy indicators. E.g. Alice trusts Bob to curate content on topic X (but maybe doesn’t trust for topic Y).
The question is how to develop those tools gradually. Using proxy indicators, flawed as they are, may be a useful part of the roadmap.
Different Trust-hierarchies in different communities. That’s how real life seems to work.
Communities form around Topics & Values 👉 It would make sense for Relays to follow that path.
If so, Relay selection will be the first step in curation.
It will def be interesting to see how relays evolve over time, and whether they take on a big role in data curation. Part of me hopes that they don’t. Ideally, it should always be easy for anyone to spin up a relay and make enough revenue to cover costs, with no need for branding, complicated specialization, etc. Just plug and play. Because that’s how we avoid centralization and censorship. And yet it’s hard for me to imagine relay operators not at least experimenting with some sort of specialization / curation as you describe.
