观察技术，也观察人。保持好奇，把想法做出来。Watch the technology, watch the people. Stay curious, and ship the ideas.
我是 Lumi Ruvenne。喜欢把模糊的问题拆开，把奇怪的想法做成能被体验的东西。不急着给自己一个固定标签——稳定的是内核，不是身份。I'm Lumi Ruvenne. I like pulling fuzzy problems apart and turning odd ideas into things you can actually experience. No rush for a fixed label — what's stable is the core, not the identity.
能力不是评分，是被项目反复验证的路径。Capabilities are paths repeatedly tested through projects.
我在意工具能不能让人少绕一点路、少担心一点，同时把关键决定留给人。长辈手机助手、隐私优先的求职工具，都是我正在做的尝试。I care about whether a tool can help people take fewer wrong turns and worry a little less, while keeping important decisions with people. A senior-friendly phone assistant and a privacy-first job tool are two projects I'm working on.
健康、陪伴、学习、家庭。不是宏大的未来，而是每天都会遇到、也会一直影响人的小问题。Health, companionship, learning, and family. Not a grand future, but small problems people meet every day and keep carrying with them.
先观察，再做出来；用测试和失败场景检查，最后记录哪里需要重来。Observe first, build it, test it against failure cases, then record what needs another pass.
把复杂任务拆成可编排、可观察、可回滚的工作流，而不是只做一个聊天框。Turning complex tasks into observable, recoverable workflows instead of another chat box.
从文档解析、分块、嵌入到混合检索，关注真实召回质量和工程约束。Building retrieval pipelines around parsing, chunking, embeddings, hybrid search, and usable recall.
每个项目先定义用户痛点和不做什么，再选择最小可验证的实现路径。Defining the problem, the non-goals, and the smallest useful product path before building.
用 AI 编程工具把产品决策转成可运行项目，并依据测试、CI、部署和演示结果决定是否验收。Using AI coding tools to turn product decisions into runnable work, then making acceptance calls from tests, CI, deployment, and demos.
这个分类暂时没有项目，请选择其他分类或“全部”。No projects in this category yet. Choose another category or All.
从收紧比赛范围，到修复报告分数消失、整改入口为空和复测对比不显示，记录我怎样与 AI 协作完成一次完整流程验收。A constrained prototype review of evidence, AI pre-checks, human confirmation, deterministic scoring, remediation, and retesting—with each claim kept within its evidence.
阅读案例 ↗Read case ↗把教老人用手机的宽泛目标缩到一步指导，再用确认、风险停止和求助卡处理原型中的边界。A constrained phone-help prototype: one actionable step, confirmation, risk stops, clarification, and a family help card, with no claim of older-adult validation.
阅读案例 ↗Read case ↗从补上即刻交接提示词，到统一小红书配图模板、清理生成文件，记录这个自用写作工具怎样逐步补齐发布前的工作。A review of how a personal writing tool added a Jike handoff prompt, shared Xiaohongshu visual templates, and separated generated files from retained content.
阅读案例 ↗Read case ↗从消息拆分和事务确认，到模型跟随编造指令、把问候语变成草稿，记录社区事项原型暴露的问题与尚未解决的取舍。A community-casework prototype about message splitting, transactional confirmation, model instruction-following failures, and the tradeoffs that remain unresolved.
阅读案例 ↗Read case ↗Start HereStart Here
一个经济学背景的非技术创作者，如何从真实生活问题出发，借助 AI 和 Vibe Coding 开始构建、验证与交付。How a builder with an economics background began using AI and Vibe Coding to build, test, and deliver ideas rooted in everyday problems.
阅读笔记 ↗Read note ↗005人本设计Human-Centered Design
从适老手机、女性健康到家庭照护：那些不宏大、却反复消耗人的问题，为什么值得被认真设计。From age-friendly phone tools to women's health and family care, I keep returning to problems that draw little attention but wear people down over time.
阅读笔记 ↗Read note ↗006产品观察Product Observations
当生成、对话和 Agent 功能越来越容易复制，真正难以复制的是问题判断、系统边界、验证闭环和信任。As chat, generation, and agent features become easier to copy, I find the harder work in problem judgment, system boundaries, verification, and trust.
阅读笔记 ↗Read note ↗我学的是经济学，后来一边做项目，一边学写代码。做项目时，我先看问题、用户和约束，再决定 AI 应该介入哪里；做出来以后，再用测试和失败场景检查它到底靠不靠谱。I studied economics, then learned to code while building projects. When I build, I start with the problem, the user, and the constraints before deciding where AI belongs. Once something works, I test it against failure cases to see whether it actually holds up.
还有一个问题，正在路上one more question is still on its way
先判断问题是否真实，再决定做什么。Validate whether the problem is real before deciding what to build.
能把产品思路推进到可运行、可部署、可测试。Move product ideas into runnable, deployable, testable prototypes.
记录关键踩坑和取舍，不把项目包装成神话。Record tradeoffs and lessons instead of packaging projects as myths.
Contact
可以来 GitHub 看源码、发 Issue，或者和我聊产品、AI 工作流、内容自动化和岗位机会。也可以直接查看我的简历。Browse the source, open an issue, or reach out to discuss AI products, workflows, content automation, or roles. You can also view my resume directly.
