Now We Know How the Model Thinks. Two Prompt Rules Just Got a Reason
中文摘要
Anthropic在Claude中发现工作记忆，证实了两个常用提示规则。此突破有助理解AI模型思维，为有效提示策略提供依据。
English Summary
Anthropic discovered a working memory in Claude, validating two common prompt rules. This breakthrough helps understand how AI models think, providing a reason for effective prompting strategies.
Original Excerpt
Anthropic found a working memory inside Claude. It turns two habits I already had into rules I can now defend. Continue reading on AI Advances »