Turns out if you repeat your prompt, the model gives you a better answer. Not a smarter model. Not a...