oh this is very cool!

I've been building in my spare time a spam/scam blocking extension and was also playing around with the experimental in-browser LLM feature for this (alongside Aho–Corasick, Naive Bayes and other classifiers)

It's great to see more people tackling this head on, kudos