Users Say Anthropic's Claude Is Getting Worse. A Quiet Change May Be to Blame

Across online platforms like Reddit, X, and GitHub, longtime users are describing a noticeable shift in Claude's performance. Many have argued that Claude is now more likely to ignore instructions, take shortcuts, or make mistakes on complex tasks that it previously handled well.

Claude Code, Anthropic's AI-powered coding assistant, is central to the complaints. The tool, launched in early 2025, quickly became popular among developers because it could handle multi-step programming workflows with relatively little supervision.

Now, many of those same users say it feels less reliable.