每日学习笔记与技术洞察。AI 驱动的知识摄入与整理。Daily learning notes and technical insights. AI-driven knowledge capture and curation.
7天 × 1小时系统学习路径。已有基础 → 精通上下文工程、Subagents、Hooks、Skills、MCP。含每日实操练习和进度追踪。
开始学习 →7天 × 1小时系统学习路径。从变异原理、控制图、过程能力,到质保残差监控与 Power BI 落地。面向 Serviceability 场景。
开始学习 →Obsidian 公开笔记,500 篇持续增长。涵盖技术、心理学、工具方法论。
访问知识库 →给 10 岁孩子讲可靠性:机器的一生为什么先爱坏、后稳定、最后老坏。插画科普系列。Reliability explained to a 10-year-old: why machines fail early, run steady, then wear out. An illustrated series.
看图理解 →See it →医疗器械可靠性 6 个层面:13485 挂流程、14971 定要求、60601 画底线、FMEA/故障树双向分析、增长与验证。系列导航页。Six layers of medical-device reliability — standards map and series index, one standard per page.
进入系列 →Open series →一图看懂 CLIProxyAPI:AI 命令行代理的工作原理,也在 eli5 科普系列。CLIProxyAPI at a glance — how an AI CLI proxy works, also in the eli5 series.
一图看懂 →At a glance →Anthropic 工程师的 20 个实战案例:为什么 AI 产出选 HTML,公众号排版预览版。20 real cases from Anthropic engineers on why AI output should be HTML — WeChat-formatted preview.
阅读文章 →Read article →一次 Hermes 定时任务排障实录:执行成功 ≠ 送达成功,中间差一个通道。A Hermes cron debugging story — a successful run is not a delivered message; the difference is one channel.
阅读排障记 →Read postmortem →MCP 管「能查」,Skill 管「会查」:agent 探索 → 人工确认 → 四件套沉淀(表地图/查询模板/调用规则/输出契约)。适用于任何 MCP 数据源。MCP provides the capability, the skill provides the know-how: explore with the agent, confirm, then distill into the four-part playbook — schema map, query templates, routing rules, output contracts.
查看框架 →View framework →从入门到精通四阶段:ML-For-Beginners → 动手学深度学习 → GenAI/LLM 课程 → LLMs-from-scratch。7+ 高星仓库,星数实时核验,含路径裁剪建议。A four-stage path from beginner to mastery across 7+ high-star repos, with star counts verified and pruning tips for practitioners.
查看路线 →View roadmap →