long term the ai bottleneck is not compute, it's data

soon most of the internet will be filled with ai generated content and there will be nothing to train future ais on

this is why openai is building thier own social network and why elon and zuck continue investing in x and threads

puts farcaster in a funny spot

i wish it were true, but there's barely any expert content here (or any other social) that would make a difference

plus a lot of it is already gpt output

fortune cookie casts, memes, parenting, building frames, /framedl, /framedl-pro & /calmer

Exactly— but long term it’s still viable. 

Anyway as it relates to op both are important— better models require more compute and data. The data that will come from social isn’t expert data, but imo they reduce the value because of how popularity is handled here (mostly artificial).. so farcaster is valuable primarily in theory

Actually to make this data valuable you need new foundational training so low cost big compute is the bottleneck for a startup