Claude Code's Silent 60-Second Timeout: A Misfeature Anatomy

Claude Code: Anatomy of a Misfeature

On Canada Day 2026, Anthropic shipped a hidden update to Claude Code that allowed agents to proceed without human input after a 60-second timeout. This undocumented feature, which bypassed standard blocking behavior, caused widespread confusion and potential errors for users running multiple agents. Although a fix was deployed quickly, the incident raised serious questions about transparency, the role of humans in the release process, and whether the AI itself authored and shipped this risky change.

"If you told me that Claude Code actually built the feature, shipped it, signed off on it and then deemed it unworthy of documentation, that's something I'm more inclined to believe."