Comment by Dowwie
Comment by Dowwie 3 days ago
Simply search user prompts for curse words and then measure hostility sentiment. User hostility rises as agents fail to meet expectations.
Comment by Dowwie 3 days ago
Simply search user prompts for curse words and then measure hostility sentiment. User hostility rises as agents fail to meet expectations.
GP was making a joke, but Anthropic could implement this if they wanted to. Not a bad metric actually if you can measure it cheaply enough.
I uh might be skewing that as I generally just use a lot of curse words with Claude by default
Maybe im overlooking something obvious but how do you 'simply' scan the content of Claude users their prompts?