Hacker News Reader

Don't think so unless all moderate actions are public.

I don't know about that, political slop is sometimes manually unflagged here.

People would find exploits in the automod. For example if a trivial keyword filter flags comments mentioning Palestine but not ones that mention Paluhstein, or "the country of Isn'treal"

This is called Algospeak and has infected a lot of people's everyday language. People say "unalive" instead of "kill" or "murder", because the internet has drilled into them that you have to use that word.

An LLM would easily understand that, no?

If trained enough, but the commenters will just start calling it watermelon instead. So you train your LLM to block comments about watermelons, so they say thing. So you train your LLM to block comments about things. Congrats you blocked all comments.