发布于2026年6月17日Multi-Agent Research System:什么时候值得拆多个 Agentagentagent-learninganthropicmulti-agentresearchreading这篇笔记整理了 Anthropic Multi-Agent Research System 的核心理解:multi-agent 不是省 token 的方案,而是用更多上下文、工具调用和工程约束换取更高质量的宽搜索。
发布于2026年5月24日Building Effective Agents:从最小智能单元到 Agent 工程边界agentagent-learninganthropicreadingworkflowaci这篇笔记整理了 Building Effective Agents 的核心理解:从 augmented LLM 最小单元,到 workflow、agent、tool、memory 和 ACI 的工程边界。
发布于2026年5月24日Effective Context Engineering for AI Agentsagentagent-learninganthropiccontext-engineeringreading这篇笔记整理了 context engineering 的核心概念:它不是单纯写 prompt,而是为 agent 设计一套信息供应链。