<?xml version="1.0" encoding="UTF-8"?><rss version="2.0"><channel><title>weishh 的思考笔记</title><link>https://blog.weishh32.com</link><description>AI Agent、系统工程与学习思考</description><language>zh-CN</language><item><title>流畅性错觉与连接门：用 AI 学习的陷阱与对策</title><link>https://blog.weishh32.com/blog/fluency-illusion-and-connection-gate</link><guid>https://blog.weishh32.com/blog/fluency-illusion-and-connection-gate</guid><description>和 LLM 讨论让一切都感觉通了，但重建知识的认知重活是 LLM 做的，不是你。讨论结束时的理解是工作记忆，不是长期记忆。</description><pubDate>Sat, 04 Apr 2026 00:00:00 GMT</pubDate></item><item><title>LLM 的功能性情绪：Prompt 工程和 Harness 设计的隐性维度</title><link>https://blog.weishh32.com/blog/functional-emotions-in-llms</link><guid>https://blog.weishh32.com/blog/functional-emotions-in-llms</guid><description>读完 Anthropic 的情绪论文后，我意识到 prompt 的情绪氛围不是装饰，而是一个有因果效力的隐性参数。</description><pubDate>Thu, 02 Apr 2026 00:00:00 GMT</pubDate></item><item><title>如果我来设计 Memory 子系统</title><link>https://blog.weishh32.com/blog/designing-agent-memory-systems</link><guid>https://blog.weishh32.com/blog/designing-agent-memory-systems</guid><description>基于对原版 ~5,200 行 TypeScript 的深度阅读和 Go 重实现的完整循环，我总结出的 Agent Memory 设计洞察。</description><pubDate>Wed, 01 Apr 2026 00:00:00 GMT</pubDate></item><item><title>生产级 Agent 系统的架构模式</title><link>https://blog.weishh32.com/blog/production-agent-architecture-patterns</link><guid>https://blog.weishh32.com/blog/production-agent-architecture-patterns</guid><description>读完一个 512K 行的生产级 CLI Agent 系统后，我发现之前学的六层架构模型终于有了具体的实现参照。</description><pubDate>Wed, 01 Apr 2026 00:00:00 GMT</pubDate></item><item><title>Vibe Coding 方法论：从 Prompt 到可运行代码的系统化流程</title><link>https://blog.weishh32.com/blog/vibe-coding-methodology</link><guid>https://blog.weishh32.com/blog/vibe-coding-methodology</guid><description>我发现&apos;让 AI 写代码&apos;和&apos;用 AI 系统地做工程&apos;是完全不同的事——前者靠运气，后者靠流程。</description><pubDate>Tue, 31 Mar 2026 00:00:00 GMT</pubDate></item><item><title>Harness Engineering：从职业焦虑到能力迁移路径</title><link>https://blog.weishh32.com/blog/harness-engineering-career-path</link><guid>https://blog.weishh32.com/blog/harness-engineering-career-path</guid><description>我发现焦虑已久的&apos;抽象层能力&apos;有了一个精确的名字：Harness Engineering，而游戏后台经验与之高度同构。</description><pubDate>Thu, 26 Mar 2026 00:00:00 GMT</pubDate></item><item><title>大问题拆解法：从模糊目标到可执行层次</title><link>https://blog.weishh32.com/blog/breaking-down-big-questions</link><guid>https://blog.weishh32.com/blog/breaking-down-big-questions</guid><description>面对&apos;赚够钱再体验生活&apos;这种模糊目标，我用逐词审查、假设翻转、依赖排序三步把它拆成了可以回答的问题。</description><pubDate>Wed, 25 Mar 2026 00:00:00 GMT</pubDate></item><item><title>显示性偏好 &gt; 陈述性欲望：从行为推理真实需求</title><link>https://blog.weishh32.com/blog/revealed-preferences-over-stated-desires</link><guid>https://blog.weishh32.com/blog/revealed-preferences-over-stated-desires</guid><description>与其问自己&apos;想要什么&apos;，不如观察自己过去实际选择了什么——25年的行为数据比任何自我分析都可靠。</description><pubDate>Wed, 25 Mar 2026 00:00:00 GMT</pubDate></item><item><title>I Want to Become a Claude Architect (Full Course)</title><link>https://blog.weishh32.com/blog/claude-architect-full-course</link><guid>https://blog.weishh32.com/blog/claude-architect-full-course</guid><description>A comprehensive breakdown of the Claude Certified Architect exam — covering Agentic Architecture, Tool Design &amp; MCP, Claude Code Workflows, Prompt Engineering, and Context Management.</description><pubDate>Sun, 15 Mar 2026 00:00:00 GMT</pubDate></item></channel></rss>