Oddbean new post about | logout
 Big relays and small relays will all be needed in the future. Big relays have a lot of signal, a lot of noise and a low signal/noise ratio. Small relays (specially community relays) have the potential of being the opposite.

 The idea that there will be a "centralized super relay" in the future is pretty vague and I will explain why. If there is a relay with a lot of signal and little noise which is also somewhat centralized, that relay would still compete against the biggest amplifier of signal in the entire world (Bitcoin blocks) and it would lose catastrophically because the Bitcoin system is just too solid

The paths to big relays and small specialized relays are out there. I don't see a problem in people taking different routes in this aspect and competing  
 Big relays are a solution in search of a problem, IMO. Since relays can pull in notes from other relays, our own personal feeds can be 100% signal and diverse. The Big Relay Model assumes you have to go to that relay to see someone's notes because notes are something scarce, like on legacy media. There is no scarcity in notes because they are signed, so you can create n copies of them.

The only thing that has any scarcity, at all, on here, are npubs identified as being run by interesting humans or helpful bots, so that's what you should collect. People have been collecting them in their follows, but collecting them in relays is both more effective and more discrete.

WoT algos are better than nothing, but they're  dependent upon your feed being public data. 😬 All of the AI work we've been doing has me thinking more about which things I want to make explicit to the public, like labeling notes by topic or an attestation for some specific npub, and which ones I should make harder to discover, like what I'm personally looking at.