Anthropic tries to hide Claude's AI actions. Devs hate it (www.theregister.com)

Anthropic has rolled out an update to Claude Code, its AI coding tool, that minimizes the visibility of file actions by only displaying a condensed output. Developers have voiced significant frustration with this change, stating that they need access to file names and details to maintain oversight, ensure security, and avoid potential token wastage from ineffective AI prompts. The new version, 2.1.20, collapses outputs to basic summaries like "Read 3 files," which many view as a detrimental simplification compromising their workflow. In response to developer backlash, Boris Cherny, head of Claude Code, asserted the update aims to declutter the user interface and highlight essential information such as diffs and outputs. Although a verbose mode is now available that provides more detail, many developers feel it falls short of their needs for transparency. The ongoing debate underscores a critical issue in the AI/ML community: balancing user experience with functionality. As AI tools evolve, maintaining clear communication of their internal processes will be crucial in preventing errors in complex coding scenarios.