The Exhaustion of Talking to a Tool
The article discusses how interacting with LLM-based tools can feel mentally exhausting due to the need for constant, precise prompting and the cognitive overhead of treating them as conversation partners rather than obedient utilities.
Background
- The essay critiques a growing dynamic where AI assistants (like ChatGPT, Claude, Gemini) are treated as tools to be directed, but users must exhaustively prompt, explain context, and correct them — turning the human into a manager rather than a collaborator.
- It explores the psychological and cognitive cost of "tool-talking": the feeling of having to constantly supervise, re-prompt, and debug the AI's output, which can be more draining than doing the work oneself.
- The author, Ohad Raviv (an Israeli software engineer and writer), ties this to broader questions about human-computer interaction, labor division, and whether current AI design optimizes for usefulness or for shifting burden onto the user.
- This resonates with ongoing debates in tech circles about AI "alignment," prompt engineering as a new job skill, and the sustainability of current chat-based AI interfaces for complex, ongoing work.