Readers discussing a new suggested message feature in Claude Code note that it often frustratingly suggests reverting unwanted changes the AI made autonomously. While users debate the jarring all-lowercase formatting of these automated prompts, many suspect the feature is actually designed to harvest frictionless user feedback for future model training. Ultimately, the discussion highlights the irony of an AI tool where the underlying model appears to be the primary beneficiary of the user interaction.