Anthropic has announced that it has removed over 80% of the system prompt for its Claude Code models, including Opus 5 and Fable 5, without any measurable degradation in coding performance on internal evaluations. This change shifts instructions from static natural-language rules within the prompt to more dynamic mechanisms like tool schemas, on-demand skills, and automatic memory. While Anthropic suggests this indicates improved model judgment, critics argue it highlights that the deleted text was largely filler or a patch for weaker models, and that the specific content removed remains undisclosed, limiting its practical application for users. AI
IMPACT This shift in prompt engineering suggests a move towards more internalized model capabilities, potentially reducing reliance on lengthy system prompts for specialized tasks.
RANK_REASON The cluster discusses a claim made by Anthropic about its model's prompt engineering, with significant debate and analysis from external sources.
AI-generated summary · Google Gemini · from 3 sources. How we write summaries →
