Is it the idea or the person?

24 hours ago @balajis.eth suggested using mods and NFTs to enhance Farcaster. A mere mention and now there’s a live working group motivated to tackle it. If a rando shared it, would it have inspired so much action? Probably not. The true effect of “influencers”. We are influenced.

Crypto payments for small businesses. Contributing to the mainstream use of crypto. @gogh
ongogh.com /shop-la

The only problem is externality pricing.

Building https://negationgame.com

Thanks for asking! Two things:
1) Dictatorship based moderation (a la Reddit) is best when channel creation is permissionless
2) NFTs don't change the dynamics in any useful way (other than inciting speculation)

Here's a not-so-kosher take on nfts:

Thanks for asking! Two things:
1) Dictatorship based moderation (a la Reddit) is best when channel creation is permissionless
2) NFTs don't change the dynamics in any useful way (other than inciting speculation)

Here's a not-so-kosher take on nfts:
https://warpcast.com/nor/0xcfa4b1

1) I think channel creation should be permissionless, so that‘s easy.

2) Disagree. Our social actions online are important but locked to servers. You can’t easily export your Uber star ratings for example.

But NFTs change things. They allow lightweight reputation portability. Uber stars can be imported by Lyft.

1) Agreed

2) The thing to be careful to remember is that data is interesting only as a function of its method of generation. In other words, stars from Uber can’t merely be aggregated with Lyft stars, and it’s not easily to govern inference on the generating method. 

Is that clear or should I give an example?

Yeah I think this is exactly where we disagree.

Data is not enough to perform inference, you also need models of how the data was generated. The great lie of statistics is that statistics is enough for knowledge.

Of course, I’m some cases it is, but in the vast majority of valuable use cases it’s not sufficient

There’s usually a way to do sensor fusion across data sets of similar type.

Simplest approach: build a dataset of paired examples of drivers on Uber and Lyft, form a scatterplot of their star ratings, and look for deviation.

If it’s mostly linear (which one would expect) then a simple 1:1 import may just work.