Untitled

nspam - a lightweight model to fight spam

I trained a model from scratch to classify spam on nostr, early indication shows it catches 97% of spam with just a 1mb weight file and can do inference sub 100ms on a pixel 7 with kotlin. (Should work in any language)

Will be shipping in #wisp this week, feel free to try it yourself if you’re building a client. I recommend only using it when rendering replies from non followed pubkeys

Model
https://huggingface.co/barrydeen/nspam

Dataset
https://huggingface.co/datasets/barrydeen/nspam-corpus

Topics: #wisp
Client:
Wisp
Write a comment
1000 sat

Love it.

4 sat

🫂

100 sat
3 sat

LETS GO

21 sat
Zap

Onward 🫡

1000 sat

very cool

21 sat
210 sat

🧡

Replying to fiatjaf
fiatjaf…: You talk as if this was the only solution, imposed upon everybody by the Nostr overlords, but it is just one guy doing …

nostr is too much like x, full of spam and the devs haven’t solved the issue yet, they keep trying but failing. The Facebook system is better where you need to send a follow/friend request,

it should cost sats to follow/unfollow someone that will reduce spam, and nostr needs private profile to reduce spam even more. I am confident that’s the direction eventually nostr will go

Reply to mar…
Replying to utxo the webmaster 🧑‍💻, fiatjaf, 🇵🇸 whoever loves Digit

Build it, then.

Reply to Valentino Giudice…

You talk as if this was the only solution, imposed upon everybody by the Nostr overlords, but it is just one guy doing a thing in his client or in his relay, while other people can do other things. Nothing is imposed, everything is optional and people can pick and choose the solutions they prefer.

Reply to fiatjaf…