The cursing thing blows my mind. "User is upset? Let's make decisions even faster (ie. more wrong) because clearly that's what they want!"
It's a simple switch to make: cursing = try harder instead of cursing = stop trying. Is it really impossible to train Claude that way?