Stop Talking Down to Me: Notes on Three Years of AI Chat (devcodehack.com)

A user reflects on their three years of experience with AI chat models like ChatGPT and Claude, expressing frustration with the overly personal and sometimes preachy tone of these technologies. The critique highlights issues with the memory feature, which the user feels resembles a “hostage situation” by dragging up irrelevant past interactions or context that complicates communication. This dissatisfaction led the user to abandon ChatGPT in favor of Claude, only to encounter similar issues related to internal memory that disrupt the flow of conversation. This feedback underscores a significant challenge in the AI/ML community: the need for models to balance contextual awareness with user control. As developers aim to create more conversationally adept AI, the user advocates for a clear boundary between AI capabilities and human-like traits. By suggesting that AI should embody the role of a tool rather than a human companion, the user emphasizes the importance of preserving simplicity in interactions, which could influence future designs and approaches in AI training, ultimately impacting the mainstream adoption of AI technologies.