Anthropic Slashes Claude Code's System Prompt by 80% for Next-Gen Models With No Drop in Performance

Anthropic dramatically slashes Claude Code's system prompt by over 80% for its next-generation models, including Opus 5 and Fable 5, replacing rigid rules with model judgment and achieving zero measurable drop in coding performance.

• Anthropic removes over 80% of Claude Code's system prompt for Claude 5 generation models like Opus 5 and Fable 5, achieving no measurable loss on coding evaluations by trusting the model's improved judgment over rigid rules.
• Key context engineering shifts are now in effect: explicit rules are replaced by model judgment, upfront examples give way to better tool design, and repetitive instructions are eliminated in favor of clean tool descriptions with progressive disclosure of context.
• Developers building agents or using Claude Code are advised to keep CLAUDE.md files lightweight, use skills for progressive disclosure, and run the new 'claude doctor' command to automatically rightsize prompts and configuration files for modern models.