Has anyone thought about using Maildrop with LLM/AIs for spam filtering?

People

I originally started using Maildrop because it conveniently sorted mail 
from my many lists and for my many projects and interests into their 
appropriate folders.  I now have 2,983 Maildir folders.  Some years ago 
I stopped using Spamdyke (to prevent mail being accepted into my system) 
because of it not doing well enough and started using Maildrop to filter 
out spam after the spam had actually been accepted.  However, the amount 
of spam has been getting out of control ie:

2015    771
2016    2410
2017    7178
2018    3181
2019    2711
2020    6527
2021    9341
2022    9740
2023    5530
2024    16674
2025    10610 - so far . .

Milan recently helped me improve .mailfilter so that spam was more 
efficiently stopped but now I am thinking that an LLM/AI solution might 
be better at recognising spam that my regexes . . and I have a lot of 
saved spam to train it on!

Has anyone done anything with Maildrop using LLM/AIs? - I have no idea 
about how this integration might work but it seems like a good idea to 
me . .

Regards,

Phil.
-- 
Philip Rhoades

PO Box 896
Cowra  NSW  2794
Australia
E-mail:  [email protected]