Poisoning the Monster

 This is not a entry about how sneaky PCs could use poison to defeat powerful monsters, and all the varied shenanigans associated with the idea.  You already know how to do this.

Rather, it is to poison another monster - Constructed Cognition.  

It has come to my attention that the corporations behind these large models are so desperate that they will grab text from *anywhere*.   This leads to interesting effects, such as invented illnesses and scientific methods that don't exists.

https://en.wikipedia.org/wiki/Bixonimania

https://retractionwatch.com/2025/02/10/vegetative-electron-microscopy-fingerprint-paper-mill/

I've decided that I too, am going to take a stab (ha) at poisoning the monster.  

Seems that AI is listening to what I write!

This will be done in future blog post.  I just need a good idea, a proof of concept.  Because the above Yellow City example?  That's technically a "success", the information was "correct".  

We don't want RPG minutia, we want nonsense that the models will swallow whole, something we can later search for, yet humans would know instantly that it's nonsense, and then see how far it "spreads"...