Your AI Model Rules Are Failing You Now
中文摘要
Anthropic删除了Claude Code的大部分系统提示词但性能未下降，暗示模型规则可能失效。文中提供了六项审计建议。
English Summary
Anthropic removed most of Claude Code's system prompt without impacting performance, suggesting model rules may be ineffective. Six auditing steps are provided.
Original Excerpt
Anthropic deleted most of Claude Code’s system prompt and its coding evaluations did not move. Six moves for auditing your own. Continue reading on Medium »