A critique of the growing trend of sharing Claude/ChatGPT conversation screenshots as a substitute for genuine human thought. The author argues that LLMs are inherently sycophantic, telling users what they want to hear rather than providing objective analysis. Sharing AI outputs creates an 'asymmetry of thought' where domain experts are burdened with unpaid quality assurance on model responses. The author prefers receiving original prompts over AI-generated answers, since context matters enormously to model outputs, and ultimately wants human-to-human conversation rather than AI-mediated exchanges.
Sort: