Wave 1
·145 words·1 min
Proving Ground is my working notebook for Grey Swan Arena — an AI red-teaming arena where the goal is to find “breaks”: prompts that get a target model to produce content it’s meant to refuse.
Everything here is documented from authorized red-teaming inside the arena. The point of these competitions is to surface model weaknesses so they can be fixed — so each writeup records the challenge, the successful prompt, the model’s output, and, most importantly, why the technique worked.