62 posts. Most of them are corrections to something I used to believe.
Windows 报磁盘告警，1.4T 的 C 盘只剩 20MB，而 WSL 里实际只用了 272GB。中间踩到两个坑：稀疏 VHD 已被微软禁用，以及 wsl --shutdown 并不会释放 vhdx 的文件句柄。
Every agent product demo shows the same thing: it did the work by itself. That is precisely the part the buyer is afraid of, and almost nobody is selling the other half.
In eighteen months my paid Astro template sold once, for $29. The free MIT one collected $110 in unsolicited payments through a button with no sentence next to it. The shape of that money was more useful than the size.
有时候觉得不用把自己太当人去看待，去考虑太多情绪，未来的困难，只是把自己的身体当做工具，让大脑下达指令，在抵触和习惯的思考来临前先做出行动，就能让自己行动力更强。
Most AI products ship two labels: AI-generated and human-reviewed. That pair is a lie the moment an agent does the reviewing. What I changed after it broke.
最近把工作环境挪到了 WSL，用 Codex 的时候三天两头超时，界面上一直转圈 Reconnecting... / Request timed out。第一反应当然是，WSL 里代理没配好嘛…
A founder-focused take on why GEO is better approached as a vertical, productized service before building another AI visibility dashboard.
A code-led walkthrough of Zota as the successor to YouChannel: session memory design, context compaction, and model-generated frontend surfaces rendered by stable Next.js components.
A code-led postmortem of YouChannel, an open-source YouTube language-learning product built around bounded agent loops, YouTube Data API reads, Gemini video analysis, Gemini Live practice, Supabase, Fastify, and pg-boss.
A practical measurement framework for AI answer visibility, with prompt panels, repeated sampling, mention scores, citation tracking, and honest reporting.
Why AI visibility reaches beyond your own site, and how to think about review sites, community threads, docs, GitHub, benchmarks, and category pages.
A practical guide to writing GEO content with facts, comparisons, statistics, citations, and useful structure instead of stuffing keywords or pretending schema alone will save you.
A hands-on process for checking whether a brand appears in high intent AI search answers across ChatGPT, Perplexity, Gemini, Claude, Copilot, and Google AI Overviews.
A practical explanation of query fan-out, citation drift, source bias, and why ranking in Google is no longer the same thing as being quoted by AI answers.
A grounded introduction to generative engine optimization, AI search visibility, and why the market is real without being nearly as certain as the sales decks make it sound.
记录炼云小说工厂 v0.1 的开源过程，以及我对 AI 长篇写作、agent 工作流和可观察生产系统的一些想法。
When one agent becomes many agents, the bottleneck moves from model capability to task, context, permission, state, and acceptance management.
Context rot is not one disease. Long-context distraction and poisoned-context inertia need different product and engineering responses.
My working model for context engineering: the industry moved from adding more context, to pricing every token, to letting agents decide what should enter the working set.
A product note on why useful AI often removes the friction that makes thinking happen, and what a slower, less obedient AI interaction might look like.
When you cannot describe the website you want, that is not always a prompt failure. Some design judgment is tacit and only becomes visible through iteration.
A practical note on AI writing: voice is not copied from a sample, it is found in the delta between the model's default draft and the version you actually publish.
Adding an agent to a product is not the same as making an agent-first product. The hard parts are permissions, tool boundaries, context, task contracts, and treating AI as a real user.
A product note on consumer AI agents: the better the agent works, the more compute it burns, while model providers keep absorbing the visible feature surface.
A practical framework for agent memory: decide who the memory is for, what should be remembered, how long it should last, and how it will be retrieved before picking a tool.
A reflection on the lean SaaS stack: VPS, Go, SQLite, and hybrid local/API AI are not just cheap choices, they buy runway and reduce operational drag.
A skeptical note on Hermes-style agent evolution: useful self-improvement is less about magic autonomy and more about captured feedback, skills, and offline evaluation loops.
A product-builder reading of model interpretability work: if internal concepts can be found and steered, the UI should stop pretending behavior is only prompt text.
A reflection on the viral 'colleague skill' idea: if a few documents can recreate your workplace behavior, that behavior was probably not the scarce part of your work.
Most AI agents do not remember you. They reconstruct enough history inside a context window, then continue as if that were memory.
A short note on why agent skills are useful: they keep capabilities out of the main prompt until the agent actually needs them.
My practical take after reading prompt engineering research: prompt templates are useful, but treating them as universal recipes is the wrong level of abstraction.
A short builder-facing note on model training cost: the headline number is not only money, it is energy, hardware, operations, and risk.
一次实时语音语言学习 agent 的产品复盘，记录 YouTube 素材、Gemini Live API、hints、replan、记忆和 agentic 对话体验的取舍。
Say you're opening a neighborhood bar. You've got the perfect location, the vintage industrial a…
这篇文章介绍了一种非常实用的内外网隔离摸鱼方案！
A practical take on vibe coding after shipping with Cursor, Claude Code, Copilot, Bolt.new, and Lovable: what AI tools accelerate and where product judgment still matters.
已经很久没有写过blog，最近的日子像是被生活推着走，很多个时刻都有突然冒出的想法想记录。但又不断推迟，以至于我现在打字的时候，一点东西都想不起。所以，想到哪里写到哪里。 工作内容和之前大差不…
想象一下我们写字时用的横格本，那条让大部分字母“坐”在上面的横线，就是基线。它是字体垂直度量的核心参考线，几乎所有字母的底部都与它对齐。 作用： 确保一行文字看起来整齐划一。 例子： 字母 "…
关于“80分万岁”的生活反思，记录在做事态度、专注和自我安慰之间重新校准的一次个人复盘。
文字逐字显示，通过滚动进行触发，支持中断、回退、继续
Looking for the perfect Astro portfolio template? Discover Wholesome Dev Portfolio - the most fun, interactive, and SEO-optimized developer portfolio template built with Astro and Tailwind CSS
A minimal, responsive, and SEO-friendly blog template built with Astro, delivering an exceptional writing and reading experience
Astro Sassify is a free Astro, Tailwind CSS, and Alpine.js starter for SaaS and content websites, with a live demo, repository, and setup guide.
记录用 bolt.new、Copilot、Cursor 快速搭建 Astro 模板、截图工具和个人产品的过程，以及 AI 辅助开发对前端工作方式的影响。
An open-source Apple-inspired portfolio website built with Astro, Tailwind CSS, GSAP, and Three.js, with a live demo, design notes, and setup steps.
24年5月，我搬到公司附近的一个高层公寓，此前我从未住过超过6楼的房子里，也鲜有机会进入过高层建筑。我当前居住在23楼，整栋公寓一共29层，三部电梯，一层没记错的话有12户左右。按照我后面了解…
开源了当前博客的模板 看了James Blunt的演唱会 年终述职 深入使用了Midscene.js 给龟缸换了个过滤器，养了几条斑马鱼 去看了Feday 2024 bisole 加绒厨师鞋（…
How I used Bolt.new, GPT-4 prompts, GitHub Copilot Edits, and Astro to ship a minimal static blog in two hours.
FEDAY 2024 参会笔记，记录 Ling 流式 JSON、Web AI、midscenejs、OpenSumi、Qwik 等前端和 AI 产品技术启发。
去广州的飞机上 二十几岁的年纪，奢望房子，车子这些超出自己能力范围的外物，不过是让自己更加焦虑罢了。人生并不是要在二十多就准备好、完成一切，我还有大好岁月，坚持锻炼的话，我的壮年期应该还有二十…
上周末去了趟合肥，为了看安溥的演唱会，也趁这个机会逛了逛合肥。 这篇博客记录我在合肥路上的一些所见所闻和内心的想法，可能没什么逻辑，没什么实际上的内容，只是一些零碎的感受。 一些碎碎念 自己现…
非常反感那些用‘咱家’、‘兔子’、‘咱妈’等词汇称呼中国的人，仿佛他们的言论代表了所有人，站在绝对正确的一方，不是所有人都少个爹。并且使用这类称呼的人所发表的言论通常给人一种站在道德制高点的感…
可能比较乱，想到什么写什么 即使当前形势很差，也不要躺平，或者说放弃自己，对自己的主体性保持觉知；保持思考，关注自己的精神、情绪、身体的变化； 写周报是工作至今我认为很重要的事情，不仅是对工作…
历时一个半月，终于达成了黑神话全成就。 一个玩了83个小时，之前玩过类似的游戏，比如只狼，但是感觉难度太大，也没有坚持玩下去。 这次玩黑神话，也是看到发售后的动画，感觉不错，决定跟一下风。 刚…
大概是100天前，我报名了2024年9月28日的雅思机考，在厦门理工考场，果然如小红书上所说，服务很好。对我而言最好的还是不用打印准考证（是谁考前一天晚上才想起来这回事啊），美中不足的是，离我…
这两天在尝试将我的mac通过ssh连接到我的另一台windows主机时，遇到了一些问题。 一个就是windows上明明已经安装了go，但是mac连接ssh后，却报错没有找到go。 不过node…
11.22 广州见 <audio controls> <source src="https://media-bucket.larryxue.dev/Hush!%2B-%2B%E7%AC%AC%…
目前的生活状态，七分麻木，三分不知所措😕。 连续的雨天让我怀念晴天的日子，多想在阳光下沐浴 去攀岩吧！🧗 脑子里突然有了这个想法，虽然从未尝试过，虽然过往尝试的其他事物都放弃了 低头看了下…
用 Cloudflare r2 + PicGo 搭建了图床
拖延了这么久的博客终于是搭建好了，虽然还有很多不完善的地方，但是至少可以开始记录了。 记起来第一次搭建博客还是在 2019 年，当时我还是大二吧，正值covid期间，每天热衷于折腾一些奇怪的东…
Nothing tagged that yet.
No schedule. It arrives when something turns out to be wrong.