Am I being naive if I imagine:
- nostr:npub17304velluajf6lylvjynpj2f3ndg396w063gj2gef5qk0nwtcyjqfj9yky implementing this
- it becoming an interface where you switch between privately or publicly chatting with various DVM bots, specialised in various domains, that orchestrate job chains etc for you
- you and nostr:npub1az9xj85cmxv8e9j9y80lvqp97crsqdu2fpu3srwthd99qfu9qsgstam8y8 then decide on a nostr:npub1qdcakl75gd7wv0nqmmwrz09ddm5tzl7xj8lq2gclng2qzd8up5yqjpzclt episode to call it Vendata
No, with confidential compute, even the service provider doesn't know the compute the user is doing.
Because with a temporary nsec you separate the compute from:
- the input data you provide
- the model a DVM provides
Right?
We have an API there, nostr:npub1l2vyh47mk2p0qlsku7hg0vn29faehy9hy34ygaclpn66ukqp3afqutajft can start making a proof of concept with a DVM-in-the-middle-coordinator-as-a-service. Curious to see how this could play out. I still need to wrap my head around how we build this glue on our end.
Aha I see. So it's more likely that Unleashed becomes a confidential DVM (because that's similar to the API you've got set up) than that it would be an interface for discovering and using DVM's?
