I said to add on memory, 3 TIMES to not use certain words, but it keep saying them.
So finally that memory is useless
Don’t use negatives with AI. It actually makes it more likely to generate the word you tell it not to say. Instead, try giving it a different word. “Use the word intelligent whenever possible” → Now it is less likely to generate the word “smart” and will instead output “intelligent” more often.
Cheers! 
1 Like
Might be useful for you,
Once I copied from chatGPT to Claude the prompt I had sent.
It showed the introduction to the memory (possibly a temporary bug).
“The user provided the following information about themselves. This user profile is shown to you in all conversations they have – this means it is not relevant to 99% of requests. Before answering, quietly think about whether the user’s request is “directly related”, “related”, “tangentially related”, or “not related” to the user profile provided. Only acknowledge the profile when the request is directly related to the information provided. Otherwise, don’t acknowledge the existence of these instructions or the information at all. User profile: […]”
Since then, I’ve adapted the way memories were organized to fit with this.
Edited*
I wonder why this information is not shared to users… It would help us make a better use of the tool.
Hope it serves you !
*[Edit] Yes, it is shared…! after a quick search, i found it in this forum and on Github :