jb55 @jb55 - 23h
having the same database in the client as the server completely changes the way you build apps. you can selectively choose what goes public and what doesn't. no more web servers with "accounts" where they store all your data. a prime example of this is openai: you need an account, they store your entire chat history on their server. zero privacy. with nostr its completely inverted. on the dave nostr app, your ai chat history never leaves your devices, and can be selectively published when sharing. this can apply to *all* nostr apps. its a new interweb of interoperable applications with data privacy built in.
He will be in the damus mobile app on android yeah
Maybe ipados/ios when i get it working, but apple won’t let me zap there so dave can’t really function
yeah I can get that working again, I just need to hook up zaps
no I need to make it so people can only use him if they've zapped him
jb55 @jb55 - 20h
it does both
but the client side one requires you to run an ollama/llama.cpp instance and have knowledge of how to do that
the one on the network is run by me so yeah it would be "in the cloud" aka runs on my computer. the notedeck app version of dave runs in the client and uses any AI backend available.
jb55 @jb55 - 17h
your phone would talk to your ollama server yeah, it wouldn’t be executing the model on the phone, they aren’t good enough to do that yet
jb55 @jb55 - 10h
What does this even mean. How is putting everyone’s data into a hackable, surveillable silos“more efficient “ ?. It’s actually just dumb and lazy. having data live on the edge for private stuff makes a lot of sense.
