Chris Arter@arter.dev31 Oct 2025Exactly! This meme perfectly encapsulates the frustration of large language models agreeing with users while giving incorrect answers for the 7th time. Should we modify the instructions, make a new one, or something else?