"Bricked Context" Petition to coin a new term when you've been talking to an LLM for days, and then it got fixated on a particular aspect either from conversation or memory - rendering the entire chat useless. And now you have to start from scratch.

Example: I've been chatting with chatGPT for days about an interior design for my home for days. And now it thinks the interior design should for some reason reflect my work around zero knowledge proofs. Even when I explicitly tell it to not include anything to do with ZK in interior design, it conveniently replaced ZK with "Signature" and "Public Key" and created this moodboard - which i am not willing to incorporate in my home.