Tried @miranetwork's Sterling today and it's actually really useful tbh. Everyone's building AI agents rn, but almost nobody checks if the model behind them is the one you're paying for. Even Anthropic is looking at this now. Nobody can really prove a provider is running the model it says it is. Their fix costs 70-95x more compute though. Sterling gives you public scores and only routes to providers that pass. Pay in USDC or MIRA. Makes a lot of sense IMO.