I've come full circle building @tangle_network's sandboxing and cloud infrastructure. The private payment protocol at @webbprotocol I designed -- a cross-chain zk payment system that can live on any EVM chain -- never saw the light of day, but it might soon. With AI becoming increasingly more relevant, it seems that there is in fact great use cases for private payments and anonymity -- for LLM inference, image generations, cloud services. @VitalikButerin wrote a post a while back about using protocols like Tornado Cash for anonymizing LLM inference through, iirc, tickets purchased from anonymized TC payments. The crux is that inference providers are learning about our identities every day through the things we prompt them. They can build profiles of us, a la Cambridge Analytica, and sell this data to advertisers or anyone willing to pay to manipulate us. We are increasingly spending more time prompting LLMs and increasingly giving up more of our data. This can be different. Instead, we can treat LLM inference providers as untrusted, while still using them. We can anonymize our usage through private payment protocols like Tornado cash and similarly built systems. Anonymize your holdings and use stablecoin payments to pay different LLM providers for queries. Use mixnets to distribute these requests and build a marketplace that incentivizes operators who put up more and more infrastructure so as to help grow provider diversity. I've started combining things again here because this seems super relevant and important now.
