4 ms·
Simply search user prompts for curse words and then measure hostility sentiment. User hostility rises as agents fail to meet expectations.
by Dowwie 8mo ago
Simply search user prompts for curse words and then measure hostility sentiment. User hostility rises as agents fail to meet expectations.
- Trufa 8mo agoI'm glad I'm not the only one.
- mrbananagrabber 8mo agoI uh might be skewing that as I generally just use a lot of curse words with Claude by default
- ctxc 8mo agoI 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 8mo agothere are many times where I just do it myself and it thinks it did well.
- preuceian 8mo agoMaybe im overlooking something obvious but how do you 'simply' scan the content of Claude users their prompts?
- gordonhart 8mo agoGP 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.
- mhl47 8mo agoOr there are global events that stress people out .. or their expectations change over time. Not that simple ;)
- nateberkopec 8mo agoGood thing expectations are perfectly constant!
- mbm 8mo agoThis might be strangely effective.
- F7F7F7 8mo agoThere’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.