Pages that don’t work well with darkreader aren’t many, but when they do fail they fail somewhat entertainingly:

Seeing as how the full unquantized FP16 for Llama 3.1 405B requires around a terabyte of VRAM (16 bits per parameter + context), I’d say way more than several.

A bit of a counterpoint: https://whtwnd.com/alexia.bsky.cyrneko.eu/3l727v7zlis2i

The way some moderation lists (ban/labeling) work is algorithmic.

You can actually host one of these services yourself!