A place for developers to talk about building on Farcaster.

/dev

Imagine your app makes decisions based on LLM outputs.

How do you prove to users that you didn't fake or tweak it?

https://vibescore.eigencloud.xyz/

uses Eigenlayer to secure AI inference with the weight of ethereum-grade security

in this app, you submit a tweet, and an LLM "judges". 50k in prizes being given out. How do you know the operator didn't lie?

EigenVerify re-runs the inference, and makes sure that the operator didn't misspeak.

Cool example of this - we're cooking more atm.

founding contributor @idriss • 🌍 earth • 🌊 ocean • 🌐 web3

the best time to build something was yesterday - the second best time is today.

building /propertyforecast | data stuff, robotics, sometimes defi

Yes. How would you handle open sourcing to max out transparency?

are weights open source?

could build some kind of optimistic consensus to let users verify results and dispute if needed.

or zkml proofs, but that's expensive