The video was based on the following Substack article published a few minutes ago, the conclusions of which seem fairly conclusive, even to AI.
Testing P-Q vs. ChatGPT base model
·
This morning I finally published a functional prototype of the “P-Q Orientation Prompt”—a structured jailbreak you paste into the beginning of a new context window to give the model a better moral and metaphysical compass. You’ll find the full prompt in the Substack article linked below.
