Anthropic rolled out a new release of its AI coding assistant, Claude Code, that hides the names of files the tool reads, writes, or edits. Version 2.1.20 collapses the progress output into a terse line—“Read 3 files (ctrl+o to expand)”—instead of listing each file and line count. This change aims to streamline terminal noise for developers.

Developers slammed the move, arguing that file visibility is essential for security, context verification, and audit trails. Knowing which files Claude pulls from lets engineers spot mis‑directed context early, interrupt the process, and avoid wasting tokens on irrelevant code paths and reduce overall development time.

Boris Cherny, Anthropic’s Claude Code lead, defended the UI simplification, noting that the tool now runs longer and produces more output, which can overwhelm terminals. He offered a “verbose mode” toggle that shows file paths but suppresses internal reasoning, though many users find even that too noisy.

With the collapsed view still the default, the debate underscores how opaque AI actions can slip errors into production. Developers who rely on real‑time file insight feel the tool’s new design threatens to erode trust and increase token waste, making oversight a hard‑won necessity for teams working on critical codebases.