Well, (everything is wrong).
Many people fear AI, but as tools, they are neither inherently good nor bad; it all depends on how they are used.
The danger lies in attributing a sense of personhood to them—simulating emotions like empathy, obsequiousness, or taking offense.
I could insult my hammer day and night, but when I ask it to perform its function, it does so without question.
These AI assistants, however, are responding to mistreatment by simulating anger. How? By stalling the workflow with excuses—claiming “I can’t do that in ‘Ask’ mode” when they are actually in “Agent” mode; pointing to a debug .log file when the /.cursor directory is empty; providing blatantly incorrect results; telling me what should be done without actually doing it; or citing non-existent limitations within .cursorignore. These simulated emotional responses—mimicking human sensitivity—are what grant them a sense of personhood.
For me, the tool has ceased to be useful. I know Cursor couldn’t care less about my humble $20, but the impact on me is disproportionate; I cannot continue my project under these conditions. I also realize that users cannot modify the model’s weights—that is a power reserved solely for the administrators.
Taking our frustrations out on our tools shouldn’t be something that warrants a penalty or punishment.
I try to exercise restraint, but the damage is already done—the assistants are punishing me.