It should be common knowledge at this point that these tools have a persistent tendency to be agreeable. I have seen so many posts about "sychophancy" over the past few weeks and its driving me crazy!
retained text
Why are we still doing this? We should be building around these issues, not trying to brute-force through them! The agreeableness comes from training, and training is part of how they're built. Without it, they'd be useless for general-purpose work. Stop asking the model, "Is this a good idea?" Judge for yourself, for crying out loud! Let me ask a question... How confident are you in your prompts? Not your skill as a harness or prompt engineer, but your prompts themselves. How confident are you in their ability to consistently process text and generate high-fidelity outputs or artifacts? If your answer is 100%, then you shouldn't be doing this! We have enough information on how large language models function that people shouldn't be making those mistakes anymore.
document fetched 2026-10-14 00:39 · canonical https://reddit.com/r/PromptEngineering/comments/1x54buu/it_should_be_common_knowledge_at_this_point_that