请问HN：如何学习构建自主智能系统（如Claude Code）？
大家好，
我正在尝试学习如何构建类似于Claude Code的自主智能系统，并最终将这些知识应用于特定领域的用例（例如，“面向医疗、金融、教育等领域的Claude Code”）。
对于那些研究或构建过这类系统的朋友们，我非常希望听到你们的建议，具体包括：
• 基础学习：哪些书籍、课程或论文能够为理解基于大型语言模型（LLM）系统及其决策过程提供最佳基础？
• 架构模式：哪些设计模式值得研究，以应对上下文管理、记忆、推理和编排等问题？
• 构建与部署：你如何看待构建内部系统与将其打包/分发为API、SDK或产品之间的区别？
• 开源项目：哪些开源项目在内部机制（决策、评估、上下文工程、工具使用等）方面最值得研究？
• 评估与可观察性：有哪些工具或产品可以帮助评估质量、测量系统行为以及观察在实际使用中的性能？
• 模型：哪些模型更适合“思考”（推理、规划、分解问题）与“执行”（执行、编码、检索）？
• 学习路径：你会如何从理论→原型→生产级系统进行推进？
我的目标是发现高质量的资源，让人们能够真正花时间学习和构建——通过迭代和实践——同时分享我所学的知识，以便同样在这条道路上的其他人也能受益。
提前感谢大家分享你们的经验和指导！
查看原文
Hello folks,<p>I’m trying to learn how one can build agentic AI systems similar to Claude Code, and eventually adapt that knowledge toward domain-specific use cases (e.g., “Claude Code for healthcare, finance, education, etc.”).<p>For those of you who’ve studied or built these kinds of systems, I’d love to hear your recommendations on:<p>• Foundational learning: What books, courses, or papers provide the best grounding for understanding LLM-based systems and their decision-making?<p>• Architectural patterns: What design patterns are worth studying for things like context management, memory, reasoning, and orchestration?<p>• Build vs. deploy: How do you think about building internal systems vs. packaging/distributing them as APIs, SDKs, or products?<p>• Open source projects: Which ones are most valuable to study for internals (decision making, evals, context engineering, tool use, etc.)?<p>• Evals and observability: What tools or products help evaluate quality, measure system behavior, and observe performance in real-world use?<p>• Models: Which models are best suited for “thinking” (reasoning, planning, decomposing problems) vs. “doing” (execution, coding, retrieval)?<p>• Learning path: How would you approach going from theory → prototype → production-quality system?<p>My goal is to discover high-quality resources that one can truly spend time learning from and building with—through iteration and practice—while also sharing what I learn so others on the same path can benefit.<p>Thanks in advance for sharing your experiences and guidance!