Thread - Nostr Hypermedia

Patrick _@patrickulrich.com 11 months ago

🤯 by nostrdb

Very interesting 🤔

jb55

we have a really cool local algo feed cooking at damus. in nostrdb we're building a feature called "local analytics" that stays on-device. local algos can tap into local analytics data to build dynamic feeds based on people and threads you've interacted with. cool thing about this is that its completely private and doesn't leak anything to the outside world. algos done the right way.

View quoted note →

chrizzz 11 months ago

Ok that’s dope

average_bitcoiner average@primal.net 11 months ago

Hm. Would something like this be able to localize a DVM? Or is that what this is doing already?

1 replies ↓

caleb caleb@zap.stream 11 months ago

Are parts of it up to date in the nostrdb repo? I understand it’s a WIP Just a data scientist here that would like to check out the logic + some pretty interesting machine learning could be done and still run locally once a user has the app long enough. It’s a great idea though

1 replies ↓

jb55 _@jb55.com 11 months ago

DVMs are the output of some algorithm. local algos are just code that dynamically generate feeds. These algos can eventually be described by a single nostrscript that you can share with people. This would enable offline, personalized feed generation without spamming or leaking metadata to the network.

jb55 _@jb55.com 11 months ago

it's not in there yet, it's something we've been actively discussing and planning.

GitHub

local analytics · Issue #610 · damus-io/notedeck

tracking clicks on certain profiles, threads, replies. all this data only stays on device and can be used by local nostrscript algos

there are some things like interactions which you can just query the database for which would work today though.

TheWildHustle TheWildHustle@primal.net 11 months ago

Good Afternoon Nostr

jb55

we have a really cool local algo feed cooking at damus. in nostrdb we're building a feature called "local analytics" that stays on-device. local algos can tap into local analytics data to build dynamic feeds based on people and threads you've interacted with. cool thing about this is that its completely private and doesn't leak anything to the outside world. algos done the right way.

View quoted note →

royster⚡️ royster@primal.net 11 months ago

Hell yea 🤙🏻

npub1x836...tpjd 11 months ago

If only there was a good way to pepper in high information content that’s surprising to your local model. A way to periodically burst your bubble.

Replies (10)