项目、判断和成长过程，像叶脉一样展开。Projects, decisions, and growth unfold as a living archive.
这里不写空泛介绍，只保留能解释我怎么做判断、怎么推进项目、怎么从一次次反馈里修正方向的记录。These notes keep the evidence behind decisions, delivery, feedback, and changes in direction.
YUEXUAI｜月序智能创始人在做企业 AI 创业探索YUEXUAI: exploring enterprise AI as a founder
以真实业务流程为起点，把企业 AI 从概念、工具和 Demo 推进到可验证、可交付、可持续迭代的工作系统。Building a founder-led enterprise AI practice around real workflows, evidence, permissions, and delivery.
AI 眼睛：我怎样把视频与文章变成可追溯知识卡How AI Eye turns video and articles into traceable knowledge cards
从用户主动选择的页面出发，先保留原始字幕或正文，再把碎片上下文整理成可编辑、可检索、可追溯的文字知识卡。A build log about preserving source evidence before structuring reusable knowledge.
小叶认知系统：从碎片信息到可复用认知资产From fragmented information to reusable cognitive assets
把不可变来源、候选认知、人工确认和复合框架分层，让有价值的判断沉淀成可追溯、可检索、可复用的数据。How source, candidate, confirmation, and recall become a controlled cognition pipeline.
为什么这个站最后留下了一片叶子Why a leaf became the lasting mark of this site
叶子不是临时找来的图形，而是从“烨”的同音和少年时期的称呼里长出来的个人标识。The origin and meaning of Lay's leaf identity.
把模糊 AI 想法做成可判断原型Turning an ambiguous AI idea into a testable prototype
记录我处理 AI 产品想法的方式：先拆目标和限制，再做界面、流程和前端原型。A note on product structure, evidence, and implementation boundaries.
小叶是怎么成为我的长期 Agent 伙伴How Xiaoye became my long-term agent partner
记录小叶从一次次项目协作里形成的角色：帮我实现想法，记住偏好，也提醒我回到真实问题。A record of the role, memory, preferences, and verification loop behind Xiaoye.
为什么要持续写项目日志Why I keep a continuous project log
把项目、工具、学习和判断过程持续记录下来，形成一条能回看的成长线。Writing down decisions makes change visible and future work more reusable.
我刚开始做 Web Coding 时，为什么总是做出玩具 DemoWhy my early web coding projects became toy demos
能快速做出 Demo，不代表 Demo 有真实用户、真实问题和可持续价值。Fast output is not the same as a real user, problem, or sustainable value.
Vibe Copilot：在 AI 写代码前，先判断这个点子值不值得做Vibe Copilot: judge the idea before AI writes the code
把模糊点子拆成用户场景、问题、MVP 和验收标准。Breaking a vague idea into users, constraints, and an evaluable MVP.
Codex Harness / OS Brain：我想解决 AI Coding 记不住我的问题Codex OS Brain: addressing memory loss in AI coding
记录记忆、上下文压缩和 Token 成本问题。A research note on context, memory, evidence, and personalized collaboration.
上下文压缩会不会让 AI 忘掉重要信息？Can context compression make AI forget what matters?
压缩会节省 Token，但也可能丢掉关键约束。A note on preserving decisions, constraints, and evidence while reducing context.
我为什么想在哈尔滨做黑客松Why I want to organize a hackathon in Harbin
哈尔滨也需要一个更开放的 AI 创新场域。Building a local environment for collaboration, practice, and visible outcomes.
面向高中政治老师的 AI 备课系统，我是怎么开始的Starting an AI lesson-preparation system for high-school teachers
备课不只是生成教案，而是资料、教材、案例和课堂活动的长期工作流。Beginning with real teaching material, cases, classroom activity, and workflow needs.
从设计学背景到 AI 产品，我这条路是怎么走出来的My path from design studies to AI products
只会做视觉不够，AI 产品需要理解系统、用户和技术边界。How visual training, computer science, and product practice began to connect.
申研准备中，我如何整理自己的项目和研究兴趣Organizing projects and research interests for graduate applications
项目很多，但需要形成清晰主线。Turning scattered experience into a clear evidence map and research direction.
我正在使用哪些 AI 工具，以及它们分别解决什么问题The AI tools I use and the problems they solve
如果只列工具名，就看不出它们在工作流里的真实位置。A practical division of tools by task, boundary, and verification method.