Comment by Dowwie

Comment by Dowwie 3 days ago

12 replies

Simply search user prompts for curse words and then measure hostility sentiment. User hostility rises as agents fail to meet expectations.

preuceian 3 days ago

Maybe im overlooking something obvious but how do you 'simply' scan the content of Claude users their prompts?

  • gordonhart 3 days ago

    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.

mrbananagrabber 3 days ago

I uh might be skewing that as I generally just use a lot of curse words with Claude by default

Trufa 3 days ago

I'm glad I'm not the only one.

  • sejje 3 days ago

    One time I cussed Claude out so hard that it actually quit his doom-loop and fixed the thing.

    It's the only time cussing worked, though.

    • bn-l 3 days ago

      I don’t know. My gut feeling is it seems to help.

ctxc 3 days ago

I feel bad about it but sometimes it's so daft, I can't even xD

It's not my fault, they set high standards!

smotched 3 days ago

there are many times where I just do it myself and it thinks it did well.

F7F7F7 3 days ago

There’s a correlation between getting the “How’s Claude Doing This Session?” (Or whatever) and four letter words.

It’s not always then, but it often follows it.

mhl47 3 days ago

Or there are global events that stress people out .. or their expectations change over time. Not that simple ;)

mbm 3 days ago

This might be strangely effective.