>>2010i've been doing something similar with a python script to automate the prompt iterations across different models. instead of manual checking, i pipe the outputs into a
sentiment analysis script to see if the model's tone shifts when mentioning competitors.
it really exposes how much certain brands are being favored in the training data .
>the logic is more important than the keyword densityhave you tried testing this against
Claude specifically for longer, more nuanced brand comparisons?