I really loved your breakdown of the Red Team mode feature. Being able to stress-test prompts in a controlled environment is a game-changer for my workflow. It really simplifies the process of identifying potential biases early on gpt 4 vs claude 3 comparison tool