One thing I was wrong about over the last few years: sign up costs would dramatically reduce spam. 

Turns out spam is a top 3 problem (aside from retention and infrastructure scaling) to solve for when building a permissionless decentralized social networking protocol.

Spammers are willing to pay for sign ups at prices that normal users aren't.

Spam is also relative: what's spammy for one person is not for another.

Corollary: when you talk to developers building on Farcaster, spam is a top of mind issue whereas users giving product feedback but not actively building in the ecosystem tend to think this isn't that big an issue / not that hard a problem to solve. 

Also a good proxy for the quality of first principles thinking when suggesting "why don't you just do this?" if you haven't considered how would spammers abuse this and what's the solution.

Working on Farcaster and Warpcast. Longer thoughts at https://dwr.email

It's the spammer + chatGPT combo that is really challenging here for me. The feeling that you might be interacting with a bot is so off putting. While most of these spammers seem to be below the fold in replies, it is frustrating to have them engaging here because it dilutes the value of the community for all of us. 

Wondering if there is something around *repetitive* AI usage in casts that could be used as a way to mark an account as spam? Similar to how schools use tools to mark students work as AI generated vs. an original thought?