The power of the LLM is limited by power of the tooling you provide it. nostr enables dynamic querying and nostrdb enables fast local views of data.

This combo will make dave much more powerful than grok when it comes to an agent that can assist with finding stuff on social networks.

But dave is not limited to the social protocol, it will have access to all kinds of note types and outbox model.

What if we got it to crawl nostr relays looking for interesting content? So many possibilities.

nostr:note12xl224ns7xahyql0vwwadh8qz66j2a25yzz0lmhlgrfmcmaux2jqur235p

Reply to this note

Please Login to reply.

Discussion

Currently mocking up an AI coach that uses workout events to provide feedback. So many possibilities.

Hey, where can I learn more about Dave?

LLM angels curating our feed.

To measure "interesting" a proxy would be "surprising". You could log everything in your feed to a vector db and query against it, so you only see new information

I wondered about this yesterday. What if for example you let it crawl to look for 'Boat' related stuff; and whenever it finds boat content it scans the surrounding network/graph if it can find more: at some point it is bound to find and identify a cluster of boat enthausiast people posting tonss of boat related things...atleast, you'd think

I'm not used to going to that low level, only prototying with python bindings. but nostr x approximate nearest neighbors is a whole different game

https://v.nostr.build/8V6dSFpSY9acg27m.mp4

https://www.pinecone.io/learn/series/faiss/hnsw/

interesting …

Never played with vector dbs but now you got me curious…

Context is everything

Now imagine hooking it up to nostr:nprofile1qqstq4j6pk2sgaupru6l7ah9nq0dueafq356jllwcy7uzlek9yx7hlspz4mhxue69uhhyetvv9ujuerpd46hxtnfduhsz9mhwden5te0wfjkccte9ec8y6tdv9kzumn9wshsz9mhwden5te0wfjkccte9ehx7um5wghxyctwvshs4kprv3 DVMs

is Dave accessible or in development?