The second one where they just prompt it to agree with the user about everything is somewhat defensible. But yeah, this is just a polarity inversion of openai/

The second one where they just prompt it to agree with the user about everything is somewhat defensible. But yeah, this is just a polarity inversion of openai/google/etc, not exactly an improvement.