Gregory's ServerAnd here's a fun little hint at some of the annoying behaviour in the base model that they've tried to knock out of it with some system prompt instructions
Seriously, stop saying "certainly"!
|
5 comments
 @judell you can get a bit of that with Claude "Projects" - I have one which thinks it's a manatee who's an expert at vegetarian recipes and salads I don't use it for ChatGPT because I don't like the idea that my interactions with the models might be biased in unexpected ways by my custom instructions - I want to gain as much experience as possible with the base model that everyone else is using @judell here's my manatee reverse engineering a photo of some avocado toast for me https://gist.github.com/simonw/7ae554005d788ca233440f18d067e4c6  @simon I haven’t had much success using negative prompting. I’m surprised they get this to work. @com yeah same here, telling models not to do things isn’t that effective in my experience I’m pretty sure I’ve still seen Claude say “certainly”! |
@simon Yeah, no kidding. Any sign of a customizable base user prompt a la ChatGPT?
And, follow-up question, assuming you're using that feature of ChatGPT, how helpful do you think it has been?