Farcaster should prioritize "Client-Algorithm Separation" (CAS). 

I want goated AI teams building alternative feed algos that I can plug into my favorite client. Tightly coupling these two pieces slows innovation. Let the AI nerds cook!

I think there is lots of room for interesting innovation here. There's quite a distance between data on Hubs and data in clients. Right now most clients build on top of the replicator (raw cast data in Postgres), but there could be an middle layer.

I'm eth newbie..just learning
(working on @farcaster)

terminally.online

Exactly. In my head, I see it working a bit like RSS feeds. I go to a service, sign in, and pay for them to generate a killer feed for me. Then I subscribe to it in Warpcast. Makes it possible to have “headless” algos that can be profitable businesses