There was a paper about that actually, results get better when asked politely

That’s because high-quality sources usually use formal and/or polite language, so priming the model with this sort of input enhances output

Also, it’s just nice