If there has been one name that has been synoymous with “container management”, that name has been Portainer. I have used it for years now and have always thought of… The post Portainer 3.0 Changes Direction. What Does It Mean for Docker Home Labs? appeared first on Virtualization Howto.

Explore the capabilities of Retrieval-Augmented Generation (RAG) in AI and its operational foundation, combining retrieval systems and generative models to enhance NLP applications' relevance and accuracy.

So, in case you haven’t heard yet, TrueNAS has released their official plugin for Proxmox. It means they now have a sanctioned, official plugin that they develop and support for… The post I Installed the New TrueNAS Proxmox Plugin in My Home Lab. Here’s the Setup appeared first on Virtualization Howto.

Discover how Claude Code simplifies full-stack app development by managing your frontend and backend seamlessly. Learn to build, Dockerize, and deploy web applications efficiently.

If you haven’t heard about Pangolin, I wrote up a full walkthrough with the solution back in September of last year. You can read that post here: Why Pangolin is… The post Pangolin 1.23 Just Made High Availability Clustering Self-Service for Home Labs appeared first on Virtualization Howto.

AI slid into just about every layer of enterprise IT, largely unnoticed, from the productivity tools employees use day-to-day to the infrastructure that sits behind them. That embedding cuts both ways. The beef that security teams and researchers at CrowdStrike have with one another is that the technology those workers now use to identify threats […]

Learn how to enable and manage NVIDIA GPU workloads on Kubernetes, enhancing your cloud-native capabilities.

0.擦好防晒，岛上的紫外线都很强1.装备顺序，穿一条贴身泳裤，毕竟公用浮潜装备2.配重铅块，每块1kg，为4块在腰带左右两侧，1块绑在气瓶，为什么5kg未知3.检查气瓶，新的气瓶上面有胶套，小拧...

Proxmox has the ability to run a two node configuration for a cluster. And for many home labbers this is a great setup since you don’t have to use as… The post I Wouldn’t Run a 2-Node Proxmox Cluster Without a Third Vote appeared first on Virtualization Howto.

接上一篇《我做了个全球 AI 简报：不刷 X，也能跟上最新 AI 动态》，半个多月来xiaoz一直在对这个网站进行持续迭代，今天用这篇文章给各位老板汇报下最新的变化。网站名称和域名变更原网站名称 “AI Pulse” -> 变更为 “AI情报局”原网站域名https://ai.xphub.de

Claude can return the kind of answer that is worth keeping: a full architecture suggestion, an explanation of how data moves through a service, or a careful plan for changing several files. The useful part may be spread across twenty paragraphs, though, and reading all of them while jumping between an editor, a terminal, and […]

With all the recent announcements that we are seeing with Proxmox as of late, it has made me stop and think about what the future holds for the platform. The… The post TrueNAS Is Only the Beginning: 5 Proxmox Storage Predictions for 2027 appeared first on Virtualization Howto.

Well, as you all know, I love project based learning. I think it produces some of the best results of any type of learning you can do. When you have… The post 7 New Home Lab Projects to Try This Weekend: September 18, 2026 appeared first on Virtualization Howto.

昨天，努比亚发布了豆包手机「努比亚 NaviX Ultra」。这也是首款搭载豆包手机助手消费者版本的手机。 我一直对 AI Agent 手机很感兴趣，所以第一时间拿到了一台进行测试。 第五代骁龙 8 至尊版、16GB+512GB，在这两年手机存储、内存等硬件成本不断上涨的背景下，最终定价 6499 元，国补后 5999 元。 横向来看，目前同级配置的手机大多集中在 4000～6000 元区间。单纯...

Installing Proxmox VE Server on a new host or mini PC is the easy part in most cases. This process is usually a non event as you just have the… The post 7 Proxmox Checks I Make Before I Trust a New Host appeared first on Virtualization Howto.

Explore best practices for deploying OpenClaw AI agents to production including Docker setup, Kubernetes integration, and ensuring high availability and scalability.

这两周杜老师做技术选型时接连被两件事扎到：MinIO 在 5 月悄悄砍掉了社区版的 Web Console，11 万行代码删除；Mattermost v11 给免费版加了”消息历史最多 1 万条”的限制——旧消息还躺在数据库里，但 UI 看不见了。摆在一起看，“屠龙的少年长出了龙鳞” 这事好像不再是个比喻。写篇文章，把这两年”开源变收费”的剧本翻出来聊聊。

我们知道 ICANN 计划于 2026 年 10 月 11 日正式启用新的 DNS 根区密钥签名密钥 KSK2024，以完成对 KSK2017 的轮换。如果你在管理启用 DNSSEC 功能的递归服务器，你可能需要关注一下实际生效的根信任锚里是否已经存在 KSK2024（Key Tag 38696）。 DNSSEC 与 KSK 密钥 关于 DNSSEC，不得不先区分两个概念： KSK（Key Sig...

It really feels like Proxmox is on fire lately with announcements and momentum behind the solution. We just came off the heels of Proxmox announcing they were bringing on 24×7… The post TrueNAS Just Fixed One of Proxmox’s Biggest Storage Headaches appeared first on Virtualization Howto.

xiaoz在开发新项目的时候，经常需要给网站生成简单logo和网站图标，网上找的图标参差不齐且不一定符合项目情况，于是自己做了SVG图标生成器，在结合AI大模型的情况下，你不需要懂任何设计，只需要一句自然语言描述即可生成适配的SVG图标，如果你也有类似需求，可以试试我这个SVG工具箱：SVG Bla

There are so many cool projects now that we have access to in the the realm of home lab, especially when it comes to automation and AI Ops for our… The post I Gave AI Control of My Docker Home Lab With Komodo MCP appeared first on Virtualization Howto.

Explore the open-source LLM landscape in 2026 with a deep dive into Llama 3, Mistral, and Gemma, highlighting their features and deployment strategies.

A handful of people I follow and know about have announced their recent departure. Not from a job (although plenty of that going around), but from social media, AI usage, or in more extreme cases the Internet entirely. In Goodbye, Forever, Probably, Salma announced her retirement from the internet. She makes a sound case for leaving DevRel, a demanding job with mushy metrics that struggles to find a home between product and marketing, because community building is neither of those. I could see why Salma wanted out of the metric chasing and the daily requirement to be a public face. Salma foreshadowed this move for awhile, there was a palpable discontent with the industry in her online voice. And she did it. She made the jump to the Offline and she’s been dark on the internet since her post…

One of the coolest things that you can do is centralize your authentication in the home lab. So, instead of having all kinds of separate usernames and passwords that are… The post I Added Break-Glass Accounts Before My Home Lab Locked Me Out appeared first on Virtualization Howto.

Discover how to deploy scalable large language model inferences on Kubernetes, ensuring efficient resource management and performance optimization.

When it comes to putting a new service into the home lab that is something I will feel like I can rely on and trust, it has to be able… The post I Don’t Trust a Home Lab Service Until It Passes These 7 Tests appeared first on Virtualization Howto.

There are SO many tools out there that allow you to keep your containers updated and I have used just about all of them, including Watchtower, Dockcheck, and others. Running… The post I Stopped Manually Updating My Docker Container Stacks. I Use This Instead appeared first on Virtualization Howto.

Explore the intricacies of Agentic AI and autonomous agents in this comprehensive guide. Understand how these AI systems operate independently, their architecture, and the technologies powering their autonomy.

Data lake implementation is the process of designing, building, and governing a centralized environment where an organization can store raw and curated data for analytics, machine learning, reporting, and other business needs. A well-designed data lake can improve access to information while giving teams the flexibility to work with structured and unstructured data. The key […]

Let’s face it, our time in the home lab among other things we have going on in our lives and families may be hard to come by and there are… The post If I Had 2 Hours for My Home Lab This Weekend, I’d Do These 7 Things appeared first on Virtualization Howto.

合适的才是最好的，基于AIOpslab项目，考察模型在解决物理世界问题的能力，直接看效果

Explore the differences between Retrieval-Augmented Generation and fine-tuning for AI applications. Learn which method suits your project best.

Discover how Claude Code can transform your DevOps processes through intelligent automation directly from your terminal. Learn installation, features, and practical applications.

A demanding sprint can put even a close-knit tech team under pressure. Tight deadlines, stubborn bugs, shifting requirements, and a growing backlog can quickly drain energy, especially when several difficult sprints arrive back-to-back. Boosting morale during these periods requires more than ordering lunch or telling everyone to push through. Managers can protect their teams by […]

每天有大量重复工作占用我们的时间:批量重命名文件、爬取数据、发送邮件、操作 Excel、自动部署……学会用 Python 把这些重复任务自动化,每天节省 2-3 小时不是梦。本文通过 12 个真实案例,带你入门 Python 自动化。

每天有大量重复工作占用我们的时间:批量重命名文件、爬取数据、发送邮件、操作 Excel、自动部署……学会用 Python 把这些重复任务自动化,每天节省 2-3 小时不是梦。本文通过 12 个真实案例,带你入门 Python 自动化。

Proxmox is really good with all types of storage and I have used it as a dedicated NAS appliance many times. I have used dedicated NAS devices. I have also… The post I Turned My Proxmox Server Into a NAS Without Installing TrueNAS appeared first on Virtualization Howto.

Learn how to transform LLM outputs into structured JSON using OpenAI, Claude, and other models for seamless integration in applications.

Many home labbers are starting to experiment with AI in the home lab and are using things like GPU passthrough on their Proxmox hosts. But passthrough is not just something… The post I Changed How I Do GPU Passthrough in Proxmox After Finding This Feature appeared first on Virtualization Howto.

本文是“Phorge 现代化改造实战”系列第十七篇，记录把旧版简体中文翻译迁入当前 fork 的过程：新增 zh_CN 语言定义和约 2.7 万行翻译表，并接通设置页与请求运行时的加载逻辑。同时迁入 scripts/i18n 维护工具，用扫描缺失键、批量补键、在线模型翻译和占位符校验的流程持续维护翻译文件。

本文是“Phorge 现代化改造实战”系列第十六篇，记录用 gorge-db-api 接管节点健康、复制状态、结构差异和升级状态查询的过程：接口返回 200，PHP 却可能因为字段名不一致生成错误对象。r3 用 /api/db/meta 运行时握手、两仓共享标准 JSON 和跨仓测试建立契约，跨仓契约不能靠手工同步。

Discover Claude Code, the revolutionary AI-powered terminal coding assistant by Anthropic, and learn how to install and maximize its potential in your workflow.

If you have spent a lot of time over the years building redundancy into your home lab, you probably have looked at many of the same layers that I have.… The post I Was Running Redundant VMs on the Same Proxmox Host Without Realizing It appeared first on Virtualization Howto.

本文是“Phorge 现代化改造实战”系列第十五篇，记录用 gorge-taskqueue 和 gorge-worker 两个 Go 服务接管 Phorge 有状态的后台工作队列：继承队列的线协议，处理 ID 分配、活跃表、租约抢占、yield 和 Conduit 请求格式，并在停掉旧 taskmaster 后证明任务只由预期的消费者处理、新旧消费者不会互相抢任务。

本文是“Phorge 现代化改造实战”系列第十四篇，记录给 Go 服务建立一条统一、受控的 Phorge API 入口：gorge-conduit 接收其它 Go 服务的 ANY /api/:method，完成服务令牌认证和按 IP 限流，再转发给 Phorge 自带的 Conduit API。网关可以换框架，但对外的 Conduit 协议格式不能变。

本文是“Phorge 现代化改造实战”系列第十三篇，记录把 Phorge、六个 Gorge 服务及其数据库、搜索和邮件后端整体联通的过程：容器 Up、/healthz 200、/readyz 正常都不代表业务已经切换，按“运行、可达、就绪、被选中、业务生效”五个阶段逐层确认才算真的接上。

本文是“Phorge 现代化改造实战”系列第十二篇，记录一次只有接上真实 Elasticsearch 和 Meilisearch 才暴露的搜索故障：ES 7.17 建索引报 mapper_parsing_exception、Meilisearch 因属性不可筛选返回 400。核心结论是版本配置决定整个索引结构，单元测试证明程序发出了自认为正确的请求，真实后端仍可能拒绝。

本文是“Phorge 现代化改造实战”系列第十一篇，记录把六个 Gorge 服务共享的 HTTP 底座从 Echo v4 迁到 Fiber v3、WebSocket 从 gorilla 换到 gofiber 的过程：接口、状态码、错误码、探针和协议都不能变，用“测试只升不降”守住那些反直觉的兼容语义。

Explore how to connect OpenClaw agents to external APIs and tools, enhancing their functionality within AI-driven ecosystems.

Just when we thought that Broadcom wouldn’t be able to slow down customers moving off of vSphere to another platform. They may just now have done something that will have… The post Leaving VMware Just Got Harder After Broadcom Pulled VDDK Downloads appeared first on Virtualization Howto.

本文是“Phorge 现代化改造实战”系列第十篇，记录把 Webhook 投递迁进 `gorge-webhook` 的过程。这次调用方向反了过来：Go 直接轮询 Phorge 的队列表，所以要先用乐观抢占、接管开关和只读诊断接口，把“只投递一次”变成可以证明的结果。

本文是“Phorge 现代化改造实战”系列第九篇，记录把 `gorge-file-storage` 迁进单仓库、接上 MySQL、本地磁盘和 S3 三个后端的过程：Phorge 管元数据、Gorge 只管字节，以及为什么保留 8MB 上限反而能稳定处理更大的文件。

本文是“Phorge 现代化改造实战”系列第八篇，记录把 `gorge-search` 迁进单仓库并替换一个已经有存量索引的实现：字段名要逐字符一致、mapping 改了必须重建索引、`/readyz` 到底承诺什么，以及 failover 该由谁负责。

Discover the real-world breakdown of costs for running AI in production — from infrastructure to software licensing and more.

从 2014 年起，我开始用现在这个域名写博客，期间使用过 WordPress、emlog、Typecho 等 PHP 类博客框架，静态博客框架 Hexo，以及现在的 Astro。 自从使用 Hexo 之后，博客便不再有后端了。那时所有的书写工作都在 IDE 中完成，每次要写东西，只需在本地执行 git pull 获取最新内容即可。后来 CODING.NET 推出了 WebIDE，我的书写习惯便逐...

I have been pretty focused on automation in the home lab. But there is also mindset changes I needed to make to keep from having to baby sit my lab… The post 7 Weekend Projects That Make Your Home Lab Less Dependent on You appeared first on Virtualization Howto.

Explore the fundamental differences between AI agents and chatbots. Understand their capabilities, deployment strategies, and the impact on modern tech infrastructures.

Learn how to build a custom AI agent using OpenClaw in just 30 minutes with step-by-step guidance and insights.

据 Akamai 报告,2026 年 Web 应用攻击同比增长 40%,其中注入攻击和身份认证失效占比最高。无论你是开发者还是运维,了解 Web 应用常见漏洞都是必修课。本文基于 OWASP Top 10(2021 版),通过真实案例 + 代码讲解,带你建立完整的 Web 安全防护意识。

据 Akamai 报告,2026 年 Web 应用攻击同比增长 40%,其中注入攻击和身份认证失效占比最高。无论你是开发者还是运维,了解 Web 应用常见漏洞都是必修课。本文基于 OWASP Top 10(2021 版),通过真实案例 + 代码讲解,带你建立完整的 Web 安全防护意识。

抑制情绪或试图隐藏我们感受到的情绪，也会被证明是徒劳的。压抑情绪需要自控力，而自控力会随着时间的推移而减弱，最终会让我们不愉快的情绪更强烈或者情绪发泄更反复无常。

PHP still runs a substantial share of the web, and most of it is code somebody else wrote years ago.  Finding a developer willing to own that rather than rebuild it is harder than finding one who can write new features. The providers below solve that problem in genuinely different ways. The gap between them […]

Ollama ships a built in web search and page fetch API you can hand to a model as tools. Here is how to build a local research agent with qwen3:4b that decides what to search, reads the pages, and cites its sources.

Ollama Cloud lets you run frontier sized models by adding a :cloud suffix to the model name. Same commands, same localhost port, no GPU required. Full tutorial with Python, JavaScript, curl and a local plus cloud routing pattern.

Comparing OpenClaw and AutoGen, two open-source AI frameworks, reveals their potential in developing scalable AI agents. Learn about their setups, features, and application environments.

最近xiaoz花了几天时间，做了一个新产品 —— AI Pulse（全球 AI 简报）。它会实时抓取 X 上的全球 AI 动态，重点盯着 OpenAI（ChatGPT）、Anthropic（Claude）、Google Gemini 等官方账号，消息一手可靠。不用刷 X，打开 AI Pulse，就能

Learn how to build an AI agent capable of intelligent web search and summarization using Python and Docker, with a focus on Retrieval-Augmented Generation, web scraping, and API integration.

An AI-generated storyboard is useful input to a Remotion Docker pipeline, but it is not a trusted build artifact. It may contain an unsupported claim, an invalid asset path, or scene code that should never run on a workstation with developer credentials. A safer video pipeline looks more like software delivery. The storyboard becomes a […]

Learn how to extract structured JSON outputs from LLMs like OpenAI, Claude, and Gemini for enhanced data interoperability in AI applications.

The default Kubernetes scheduler has no idea your GPUs are shared between teams. This hands-on lab uses Kueue to add job level admission, per team quotas, cohort borrowing and reclaim preemption, and runs on a laptop with no GPUs required.

Kyverno 1.19 deprecated ClusterPolicy and Policy, with removal coming in v1.20. This hands-on lab covers the new CEL based ValidatingPolicy and MutatingPolicy types, a full old-to-new mapping table, and how to test policies in CI before you migrate.

Ingress-NGINX maintenance ended in March 2026. This hands-on lab walks through converting your Ingress resources with ingress2gateway 1.0 and serving the same traffic through Envoy Gateway and the Kubernetes Gateway API.

早上起床,窗帘自动拉开、咖啡机开始煮咖啡;下班回家,空调提前打开、灯光自动亮起;晚上睡觉,所有电器自动关闭、安防系统启动。这不是科幻电影,这是 Home Assistant 能给你的真实生活。本文手把手教你搭建一套稳定、强大的智能家居系统。

早上起床,窗帘自动拉开、咖啡机开始煮咖啡;下班回家,空调提前打开、灯光自动亮起;晚上睡觉,所有电器自动关闭、安防系统启动。这不是科幻电影,这是 Home Assistant 能给你的真实生活。本文手把手教你搭建一套稳定、强大的智能家居系统。

普通家用轿车，一般新购后6年免检，博主的代步车今年正好第6年，到了需要首次车检的时间。之前没有做过车检，没啥经验，网上大概了解后自行去完成了车检，踩了一些坑，用这篇文章分享下我的整个车检过程。注意：博主在成都地区，为5座普通小型轿车，其它地区或车型不一定适用。随着时间推移，不确保信息是否发生变更，以

Discover the best practices for deploying OpenClaw AI agents to production, covering foundational setups, Dockerization, and Kubernetes scaling strategies.

Discover the essential practices of prompt engineering for optimizing interactions with models like ChatGPT, Claude, and Gemini.

Discover how Claude Code transforms software development through AI-driven multi-file editing, command execution, and streamlined git operations in this in-depth guide.

Prerequisites Python 3.10+ Installs: pip install langchain langchain-openai, pip install langgraph, pip install crewai An API key for your chosen model provider(s) Same task for all three frameworks: an agent that answers a question using one search tool. LangChain from langchain.agents import create_agent from langchain.tools import tool @tool def search(query: str) -> str: """Search the […]

Prerequisites Python 3.10+ An API key from a model provider (OpenAI, Anthropic, Google, etc.) Install packages: pip install -U langchain langchain-openai Step 1: Install & set your API key uv add langchain langchain-openai export OPENAI_API_KEY="your-api-key" Step 2: Define a tool from langchain.tools import tool @tool def check_order_status(order_id: str) -> str: """Look up the shipping status […]

Prerequisites Python 3.10+ Install: pip install -U langchain langchain-openai langchain-community chromadb An OpenAI (or other provider) API key Step 1: Load & split your documents from langchain_community.document_loaders import DirectoryLoader from langchain.text_splitter import RecursiveCharacterTextSplitter docs = DirectoryLoader("./knowledge_base", glob="**/*.md").load() chunks = RecursiveCharacterTextSplitter( chunk_size=800, chunk_overlap=100 ).split_documents(docs) Step 2: Embed & store in a vector database from langchain_openai import […]

Per-token prices collapsed, but agent loops made bills explode. The engineering fix: where tokens actually go, and five levers that take 80-90% off the bill with verified 2026 pricing, code, and a worked example.

每天我们都在和 TCP/IP 打交道:刷网页、看视频、聊微信。但当 HTTP 请求到达服务器时,数据是怎么一层层”剥洋葱”传递的? 本文从最基础的概念讲起,带你穿透协议栈,理解每一次网络通信背后的原理。

每天我们都在和 TCP/IP 打交道:刷网页、看视频、聊微信。但当 HTTP 请求到达服务器时,数据是怎么一层层”剥洋葱”传递的? 本文从最基础的概念讲起,带你穿透协议栈,理解每一次网络通信背后的原理。

过去几年里，有多次提到减肥，那就在这个三伏天减肥提上日程把 三伏天，是一年中最热的时节，也是人体本身代谢最快、燃脂效率最高的黄金时段。都说“冬练三九，夏练三伏”练暂时并没有办法提上日程，但是吃… 还是需要注意。 2026年的三伏天： 初伏：7月15日 — 7月24日 中伏：7月25日 — 8月13日 末伏：8月14日 — 8月23日 抓住一年中阳气最盛的40天，代谢翻倍，但切记莫贪凉、多饮水，不只是为了体重秤上的数字，更是为了那份由内而外的紧致与气韵。 上一次下决心减肥是2018年夏天，一晃8年时间过去了，那时的蝉鸣仿佛还在耳边，转身却已隔了八个夏天。那一次在武汉空中健身花费2万元办理了高级会员并购买了私教课，最终因为长期出差，课没上完也没退费…后有因为疫情，费也没退。 当是在减到10-20斤的时候，又开始缓慢增长… 武汉本身就是碳水天堂，然后北京、上海、广州、深圳、福州、西安 这几个常常去的城市更是收录着丰富的美食打卡地图。2024年更换新工作后，虽然不会再长期的俾昼作夜，晨昏颠倒，三餐不定，但仍就因为岗位性质，工作排得太满，连上厕所都要见缝插针。 身体每天都被困在椅子上，灵魂每天都被困在日程表上。…那日行万步早已成了奢望，每天唯一的“有氧”运动，或许就是加班的尽头，骑上小牛回家的时候。 最终~体重超过 100Kg，减肥势在必行！！！ 减肥，用对方法其实并不难 当然，也胜在坚持，有了2018年那次2万元的教训，这次绝不再乱花钱。 第一步，目前属于大体重，肯定是不适合突然就去运动的，先从吃开始。 减肥的核心公式 消耗的热量 ＞ 摄入的热量 = 体重下降 这个差值就是“能量缺口”（Calorie Deficit）。我们身体不够用的时候，就会去分解库存的脂肪来供能。 摄入（Input），一日三餐，零食、水果、咖啡饮料 每日吃进去的所有东西，都自带热量，把他们相加，就是每日摄入…

Explore how OpenClaw AI agents are set to transform industries in 2025, with real-world applications ranging from customer support to predictive maintenance.

Explore how OpenClaw AI agents are set to transform industries in 2025, with real-world applications ranging from customer support to predictive maintenance.

I have covered a lot of different angles on home lab backups. For the most part though, these are around the hypervisor, virtual machines on a Proxmox server, and persistent… The post I Built a Server That Catches Every Change on My Home Lab Network appeared first on Virtualization Howto.

最近做了一个RSS阅读器产品，叫Feedglot。做了两个月，目前完成度比较高，计划招募一些内测用户。

Explore containerizing AI agent workflows with Docker, leveraging OpenClaw for scalable and efficient AI systems deployment.

Explore containerizing AI agent workflows with Docker, leveraging OpenClaw for scalable and efficient AI systems deployment.

A fresh installation of Proxmox is very complete out of the box and you definitely won’t have any problems with your server just sticking with the default installation. However, there… The post 7 Things I’d Install First on a Brand-New Proxmox Server This Weekend appeared first on Virtualization Howto.

Leverage Claude Code's AI-driven capabilities to streamline and automate your DevOps processes directly from the terminal.

It is amazing to me just how much things have changed just in the last couple of years. Not even more than that. We have seen a major hypervisor basically… The post 8 Home Lab Things I Stopped Doing the Old Way appeared first on Virtualization Howto.

Explore AI's transformative role in enterprises by 2025, examining adoption stats, use cases, and ROI data essential for strategic business decisions.

Explore AI's transformative role in enterprises by 2025, examining adoption stats, use cases, and ROI data essential for strategic business decisions.

I made a smol client-side syntax highlighter that uses the CSS Custom Highlights API called MicroLighter. I’ll talk about why I made it in a bit but first I wanted to… ahem… highlight… some of the features. Zero-dependencies ~2kb minifed+gzipped Uses CSS ::highlight(token-name), not spans Leverages Textmate’s language grammars Human-readable light-dark() themes All languages/grammars loaded on-demand Moves all non-highlight functionality to a custom element At some point I broke syntax highlighting on my Jekyll blog. I’ve used a handful of syntax highlighters over the years (Highlight.js, PrismJS, Rouge, Shiki, etc, etc) and I’ve felt the trade-offs between different client-side and server-side implementations. Faced with picking another, I knew I wanted to explore Bramus’s…

Well, if you have received a message in your Proxmox VE Server version 8 webUI telling you that Proxmox VE 8 is no longer supported, you are not the only… The post Still Running Proxmox VE 8? Here’s What You Need to Do Before End of Life appeared first on Virtualization Howto.

直接上结果，体感非常快，相比deepseek-v4-flash并发成功/请求错误总tok/s输出tok/sTTFT中位/P95延迟中位/P95>60s>100s3264/640384...

Choosing the right AI agent framework is a critical decision for developers. Explore the strengths and challenges of OpenClaw, LangChain, and CrewAI for AI agent development.

When it comes to free and open source firewalls in the home lab, there are a few choices, but many home labbers tend to settle on OPNsense for its wide… The post 9 OPNsense Settings I Change for Every Home Lab Firewall appeared first on Virtualization Howto.

Explore the integration of OpenClaw with local LLMs using Ollama and LM Studio, optimizing AI agent deployment with minimal cloud reliance.

运维岗位竞争激烈,如何从众多候选人中脱颖而出?本文基于我多年招聘运维工程师的经验,从简历优化、技术准备、面试技巧、谈薪策略四个维度,系统讲解如何拿到心仪的 Offer。同时附上 50 道高频面试题,帮你全面备战。

运维岗位竞争激烈,如何从众多候选人中脱颖而出?本文基于我多年招聘运维工程师的经验,从简历优化、技术准备、面试技巧、谈薪策略四个维度,系统讲解如何拿到心仪的 Offer。同时附上 50 道高频面试题,帮你全面备战。

Stuart Robson’s post Solving CSS @layer Ordering in Design Systems with Design Token Metadata, a response to Chris Coyier’s Thinking Horizontally in CSS @layer, came to me at a serendipitous time. You know I love a good blog-and-response, but these posts on the topic of lowering the specificity of component-level tokens and scaling that out was a problem I was having with some of my web components. Their @layer trick didn’t work for me, but here’s what did… The local component variable problem Let’s say we have a global stylesheet with our design system tokens. /* theme.css */ :root { --system-bg: #00f1; --system-fg: #000; --system-font: Inter; sans-serif; } And then we set up our web component with some local component variables that use design system tokens as defaults. <…

最近有一个很火的智能体Pi Agent，它因小巧、轻量、可扩展性强而出圈，于是xiaoz去了解了下Pi Agent基本思路，再根据这些思路，用Golang开发了一个同样小巧、轻量的智能体Zlite，于是下这篇文章分享和记录AI智能开发的思路和一些经验。提示：以下内容是根据自己的开发经验和对Agent

If you have been following my blog lately, I recently installed the announced Proxmox AR64 build on my Raspberry Pi 5 8GB in the home lab. However, with that setup,… The post I Got Proxmox VE 9 Working on Raspberry Pi 5 With NVMe Boot and the Onboard NIC appeared first on Virtualization Howto.

Explore OpenClaw and Semantic Kernel to find the AI framework that aligns with your development needs. Compare capabilities, community support, and practical use cases.

One of the exciting bits about AI and using it with infrastructure is that you can use what is known as MCP servers that allow you to have an easy… The post Could This Be the Easiest Way to Run MCP Servers in Your Home Lab? appeared first on Virtualization Howto.

Build a scalable and resilient multi-agent system with our detailed architecture guide and code examples. Discover the power of autonomous agents working in harmony.

It is that time of year again where school is starting for the kiddos and everyone is talking about back to school. If you have a home lab there is… The post Back to Home Lab School 2026: 7 Learning Projects to Start This Weekend appeared first on Virtualization Howto.

为什么向下螺旋（消极情绪产生更多消极情绪）比向上螺旋（积极情绪产生更多积极情绪）更为常见。人们总是慢步向前，快步后退。

Prerequisites Python 3.10+ Install: pip install -U langchain langgraph pydantic An API key for your model provider Basic agent with a tool from langchain.agents import create_agent from langchain.tools import tool @tool def check_order_status(order_id: str) -> str: """Look up the shipping status for an order ID.""" return f"Order {order_id} is out for delivery." agent = create_agent( […]

Discover what's coming next for OpenClaw, the open-source AI agent framework. Explore development roadmaps, community dynamics, and future enhancements.

环境准备 （1）使用脚本安装docker、docker-compose 1 bash <(wget -qO-

If you are like me, you spend a lot of time behind the wheel of the Linux command line. I usually am connected to Proxmox hosts, Docker servers, kubernetes nodes,… The post 8 Modern Linux Commands I Use Instead of the Old-School Alternatives appeared first on Virtualization Howto.

Explore a comprehensive guide comparing OpenAI, Claude, and Gemini APIs, focusing on integration, customization, and use in AI applications.

Anthropic's Claude Managed Agents gives you hosted infrastructure for running autonomous AI agents, no custom sandbox or agent loop required. Here's how it works and when to use it.

Most of what your coding agent does all day is boring. It reads a file. It runs git status. It greps a directory. It parses a JSON blob, decides the schema is wrong, and retries. It calls a tool, gets a 500, and calls it again. Somewhere in the middle of those forty steps, maybe […]

Picking the perfect tool feels overwhelming sometimes. The market offers dozens of options. Each vendor promises the moon and stars. Marketing materials blur together after a while. Modern IT teams face incredible pressure. They handle thousands of devices across multiple time zones. They juggle procurement, maintenance, and retirement cycles. They answer endless requests from frustrated […]

RAM is one of those things that can disappear in a home lab VERY quickly, especially if you don’t pay attention to how it is allocated across the board. In… The post 7 Proxmox Memory Tweaks That Make Your RAM Go Further appeared first on Virtualization Howto.

No content preview

自学能力的强弱,基本决定了一个程序员能走多远。在这个知识付费的时代,其实互联网上有大量优质、免费的学习资源。本文整理了 15 个我亲身使用、强烈推荐的学习平台,涵盖编程、数据、AI、设计等多个方向。无论你是初学者还是想进阶,都能找到适合的资源。

自学能力的强弱,基本决定了一个程序员能走多远。在这个知识付费的时代,其实互联网上有大量优质、免费的学习资源。本文整理了 15 个我亲身使用、强烈推荐的学习平台,涵盖编程、数据、AI、设计等多个方向。无论你是初学者还是想进阶,都能找到适合的资源。

Learn how to add persistent memory to OpenClaw AI agents for continuous context retention across sessions.

想问一下大家，现在用 AI 写代码时，用的是带界面的 GUI 工具，还是不带界面的 CLI 命令行呢？最近 xiaoz 开发了一款名为 Zacp 的程序，它能够将那些仅支持 CLI 使用的 Agent 接入 WebUI，让你在浏览器中就能流畅使用。目前测试过 Codex cli、Pi Agent、O

During my last vibe check the kids were still in school. Time has flown by. Work-life balance disrupted. Camps attended. Vacations were had. Bank accounts drained. A normal summer routine. Inbound and outbound travels Since the last vibe check I’ve been on a handful of airplanes visiting both work and family. We also had family come to visit. And I logged some hours on Texas highways. May: Another trip to Redmond for work May: A trip to South Dakota to see my nephew Max graduate May: My wife’s parents visiting June: My dad and step-mom visiting June: My kids went to Houston to visit my mom June: My son went to summer camp June: My wife and daughter went to the Hamptons (!!!) July: We all went to San Diego, hung out with our friends in Encinitas, met my wife’s family in Mission, and tagged …

The pager goes off at 02:00. You ask the internal operations agent for the safe rollback steps. It replies instantly, cites a runbook, and sounds certain. Ten minutes later the cluster is stuck, because that runbook was retired two releases ago. Connecting an agent to your docs solves access, not trust. If the source is […]

Explore the foundational role of AI embeddings in transforming complex data into searchable formats, revolutionizing the efficiencies within vast datasets.

2023 年，我借助 AI 把两百多行 beautifier.js 嵌入 Go、通过 goja 执行，快速做出了 soulteary/nginx-formatter。三年后，为了让它在 CI 中更可靠，我用原生 Go 的 Lexer、Parser、AST 和 Printer 重写了格式化内核。本文记录 v2.0.0 到 v2.2.0 的三步重构：替换内核、支持单文件、用 Cobra 拆出 format、serve、version 子命令，并复盘转义丢失、单引号 JSON、Lua RawBlock 等历史问题的处理结果。

延续之前用 Split DNS 提升跨系统双机开发体验的做法，手工维护 macOS 的 /etc/resolver 配置在设备和项目变多后越来越麻烦：覆盖前要备份、改完要刷新缓存、排错要在 dig、scutil 和 dscacheutil 之间来回切换。本文介绍我为此写的开源小工具 soulteary/splitdns，用来创建、修改和检查 Mac 上的 resolver 配置，并覆盖双机联动与单机使用两种场景。

If you’ve spent any time in developer forums, YouTube, or search bars lately, two names keep surfacing: GitHub Copilot and OpenClaw. They sound like competitors, and plenty of “X vs. Y” articles frame them that way, but they actually solve very different problems. Copilot lives inside your editor and helps you write code. OpenClaw runs […]

Explore the transformative roles of AI agents by 2025, from transportation to finance, fundamentally redefining industries and societal functions.

Explore the transformative roles of AI agents by 2025, from transportation to finance, fundamentally redefining industries and societal functions.

从 Docker Desktop 换到 Podman 后，Linux Rootless Podman 用 pasta 发布 80、443、22、53 这类低位端口时会因权限不足失败，把宿主机端口换成 8080 才能启动。本文介绍我为此写的小工具 soulteary/portmap：让 Rootless Podman 继续发布高位端口，再由宿主机上一个功能单一的进程提供标准端口入口。

两年前用不到三十行的 Dockerfile 把 WordPress、SQLite 和官方镜像拼在一起，两年后这个 Dockerfile 明显变长了。本文继续拆解 soulteary/docker-sqlite-wordpress，随着上游从 WordPress 6.5.2、PHP 8.3、SQLite Integration 2.1.9 升级到 WordPress 7.0.2、PHP 8.5、SQLite Integration 3.0.0-rc.8，SQLite 接入方式依旧靠 db.php 接管数据库层，新增代码主要集中在上游 Monorepo 组装、Rust 原生解析扩展、五种架构的多架构发布、旧数据卷自愈修复，以及后台只读的 SQLite Diagnostics 运行诊断页面。文章同时说明了根目录 Loader、id/ID 结果键兼容修复、挂载卷权限、WAL 备份与并发，以及需要核对实际 SQLite 引擎版本等容易忽略的限制。

If you’ve spent any time searching for Docker or Kubernetes help, you’ve probably noticed the questions repeat. Thousands of developers hit the same handful of walls: containers that can’t talk to the host, services that never get an external IP, pods stuck in CrashLoopBackOff, and Deployments that silently ignore a freshly pushed image. This tutorial […]

Learn how to fine-tune large language models with LoRA, a method that reduces computational costs by focusing on low-rank updates.

No content preview

No content preview

Kubernetes conversations this week span everything from managed-service comparisons and storage volumes to scheduler research and key-management integrations. With so much happening in the ecosystem, it is easy to forget that most engineers are still climbing the same first hill: getting a working cluster, deploying an app, and understanding what actually happens under the hood. […]

Explore how to build AI agents using function calling with OpenAI and Claude, enhancing interaction with large language models to automate tasks effectively.

好久…好久…没有走进电影院完完整整的看过一场电影。 这次借着午休的时间去到摩尔影城看星爷快上映一个月的《功夫女足》后却很失望… 草根组队、街头遭遇，到决战时的“大力金刚腿”和“太极劲”，电影故事框架几乎是和《少林足球》一个模子刻出来的复制品。 全篇都是夸张的表情加上“无厘头”的生硬，困意十足。 踢球就是为了踢球，不带有功利性，才有机会从自身出发做到极致。在工作中一样，目标肯定要有，但不能为了做而做，足够纯粹，才能把优秀转化为本能。 极致，很多时候都是默默的沉浸，专注一件事，无数次忘我的进入心流状态，最后将这件事做到无可附加，抵达本真的形态。此时无须再向任何人和事去证明什么。无须目标，无须对标，自身做到极致和专注，能超越身边的一切。也就印证了电影里的那句 “足球场上最高的境界，不是打败所有人，是心里没有敌人。” 九败一胜，在追求极致的道路上，必定会收获失败，很多人认为败了就是输了，但在追求成功人眼里，失败是宝贵的经验和数据。当我们的目标不再是赢，而是抵达极致的时候，每一次的失败，都是成功的排除了又一种方案，距离成功又进了一步。 在经历了无数次失败后，将清楚的知道自己是谁，有什么能力，能做成什么样子，自己将不再被外界的评价左右。事上练，九次败的永远是那些事，最后的一胜是内心的境界，练就的不再是一门技艺，而是一颗强大的内心，输得起，才能赢。 电影中 踢球就是为了踢球，不再只是一句电影台词，也不是一句的口号，而是经历过无数次失败、无数次自我怀疑、无数次重新站起来之后，依然选择回到球场时，那种极其沉重的坦然。 电影《功夫女足》中的那些对白，句句浅白却充满哲理： “我踢球不是为了赢，我踢球就是为了踢球。” “输不起，你就永远赢不了。” “足球场上最高的境界，不是打败所有人，是心里没有敌人。” “以前总想找个肩膀靠一靠，现在才发现，自己的脊梁骨最硬。” “我们不是等着王子来救的睡美人，我们是自己踢开城堡大门的女战士。” “姑娘，你踢的是球，不是他的情绪。” “原谅别人，不是认输，是放过自己。” “足球是圆的，人生也是。不到最后一秒，你怎么知道滚向哪里？” “什么是功夫？就是当你一无所有时，还能踢出那一脚。” “没了张屠夫，就吃带毛猪？地球照样转，女足照样踢！”

An API integration usually performs efficiently on launch day. Two systems intercommunicate: orders move, data syncs, and everyone signs off. The trouble shows up months later, when transaction volume climbs, more platforms get connected, and the business asks for changes nobody planned for. The scale is the problem: the average organization now runs close to […]

Learn to build a RAG chatbot using LangChain and ChromaDB, combining retrieval-based models with generative AI for highly contextual responses.

工欲善其事,必先利其器。这句话在程序员身上体现得淋漓尽致:选对工具,一天能干完别人三天的活。本文整理了 30 个我日常使用、亲测好用的开发者工具,涵盖代码编辑、终端增强、API 调试、效率提升等方方面面。无论你是前端、后端还是运维,都能从中找到适合自己的工具。

工欲善其事,必先利其器。这句话在程序员身上体现得淋漓尽致:选对工具,一天能干完别人三天的活。本文整理了 30 个我日常使用、亲测好用的开发者工具,涵盖代码编辑、终端增强、API 调试、效率提升等方方面面。无论你是前端、后端还是运维,都能从中找到适合自己的工具。

Learn to build a multimodal AI app that interprets images and text using AI technologies and cloud-native solutions.

Search interest in “Grok CLI” has spiked recently, but a lot of that traffic reflects genuine confusion rather than a single well-understood product. Anyone who types the term into Google right now lands in the middle of a naming collision, a pricing question, and a fresh security controversy, all at once. This post untangles what’s […]

Discover how to set up GitOps with ArgoCD in Kubernetes. This comprehensive tutorial covers installation, configuration, and initial integration steps for seamless deployments.

网关有了（见前篇 LiteLLM），模型接了一堆，下一个问题很自然：这些模型到底会不会干运维的活？拿面试题问"CrashLoopBackOff 怎么排查"是测不出来的，背题谁不会。得来真的：给它...

Explore the intricacies of Agentic AI and how autonomous agents operate to revolutionize decision-making. Learn the foundational concepts and start building simple but powerful autonomous agents.

Dive into AI agent frameworks for 2025: LangChain, CrewAI, AutoGen, and OpenClaw. Explore their capabilities and use cases for building intelligent systems.

Dive into AI agent frameworks for 2025: LangChain, CrewAI, AutoGen, and OpenClaw. Explore their capabilities and use cases for building intelligent systems.

No content preview

Ollama vs GPT Comparison: Which is Best for Developers? Introduction Choosing between Ollama and GPT isn’t really an either/or decision — it’s a question of what you’re optimizing for: cost, privacy, latency, or raw capability. Ollama lets you run open-source models like Llama 3, Mistral, and Gemma directly on your own hardware, while GPT gives […]

Learn how to build a full-stack app efficiently using Claude Code, a powerful AI-driven coding assistant by Anthropic. This step-by-step guide walks through installation, setup, and essential development phases.

2026年7月31日，DeepSeek官方正式发布了V4-Flash正式版。从公开的跑分数据来看，该版本不仅显著优于自家的DeepSeek-V4-Pro预览版，也超越了GLM-5.2。不过，基准测试的分数仅作参考，实际性能仍需通过多场景实测来验证。但Deepseek-V4-Flash价格是真的便宜，

No content preview

There are myriads of Docker containers spread across many docker registries on the Internet. Most of these are on popular registries like Docker Hub, but others are found on other… The post The 9 Checks I Make Before Trusting Any Docker Image appeared first on Virtualization Howto.

Docker has quietly transformed from “the container company” into the security and runtime layer for agentic AI development. If you’ve checked docs.docker.com lately, you’ve probably noticed the homepage no longer leads with Swarm or Compose tutorials, it leads with questions like “Can I run my AI agent in a sandbox?” and “What are Docker Hardened […]

朋友的电商网站用 WordPress 搭建,但打开首页要 5 秒以上,跳出率高得吓人。经过两轮优化后,加载时间压到了 1.2 秒,转化率提升了 35%。本文把这套完整的优化方案分享给你,从服务器配置、WordPress 设置到前端优化,逐步击破性能瓶颈。

朋友的电商网站用 WordPress 搭建,但打开首页要 5 秒以上,跳出率高得吓人。经过两轮优化后,加载时间压到了 1.2 秒,转化率提升了 35%。本文把这套完整的优化方案分享给你,从服务器配置、WordPress 设置到前端优化,逐步击破性能瓶颈。

For the past several months now, I have shifted my home lab to immutable Linux distros like Flatcar which has changed how I think about installing software and things like… The post The systemd Services and Timers I Use in My Immutable Linux Home Lab appeared first on Virtualization Howto.

现在很多海外券商都收紧了注册，导致国内用户开户门槛陡增，不过近期博主实测了美国第一证券（Firstrade）还能通过国内真实资料开户（不需要海外居住地址），这篇文章分享下开户、入金、购买美股的全流程，供各位参考。关于FirstradeFirstrade（美国第一证券）是1985年成立的老牌美国券商，

I have been reflecting on how far things have come in my home lab and ironically, I feel like I am more organized now more than ever before, even though… The post How I Organize My Entire Home Lab So I Can Actually Find Anything appeared first on Virtualization Howto.

I’ve had the pleasure in my career of sitting in tons of different board rooms talking with executives and getting a rundown on different charts and graphs they care most deeply about. As expected, “number chart go up” is always the desired outcome. Number chart go up is good because it makes other number charts go up, specifically stock ticker number. One common refrain I’d hear everywhere I went whether they dealt in pizzas, coffee, coupons, computers, or software licenses went something like this (paraphrasing), “On average customers use our product once per month and if we can double that number then we’d have no more problems.” An over-simplification of course, but they’re desperate to move the number from one… to two. It doesn’t take a business degree to figure that out that doubling…

Reverse proxies are one of the most critical pieces of infrastructure once you stand up a home lab and start learning containers and hosting things in those containers. You quickly… The post The Reverse Proxy Mistakes I Still See in Home Labs (And How I Avoid Them) appeared first on Virtualization Howto.

Termark 是由指针漂移科技工作室打造的新一代跨平台 SSH 客户端。与传统工具不同，它内置的 AI 助手不仅能进行日常对话，更能切实协助完成运维任务，显著提升工作效率。Termark可靠吗？技术成熟如果你了解开源堡垒机 Next Terminal，那对 Termark 的开发者应该不会陌生——

上篇写了LiteLLM，网关是有了，但用起来是个黑盒，谁在调、一次agent任务烧了多少token、中间转了几手，全靠猜。决定上Langfuse，开源的LLM可观测平台，给调用链留个底。本以为d...

Restaurant work asks a lot from people in very little time. New hires absorb recipes, sanitation rules, guest standards, and service steps while shifts keep moving. Training needs structure, pace, and visible follow-through. A learning management platform gives managers a single place to assign lessons, track progress, and address weak spots early. The right choice […]

I am all about monitoring things in my home lab and with DNS being so important, if there is one service to monitor, this is it. DNS is at the… The post I Thought My DNS Was Fine Until I Started Monitoring This appeared first on Virtualization Howto.

CodePen v2 is now released! Because I co-host a podcast with Chris I’ve had an up-close perspective at this monumental effort over the last five years. I know how much work has gone into this release and I applaud the team for hitting the finish line. At the crux of the new experience is the new and improved editor. When you first launch into the v2.0 editor, your first response will probably be “Who moved my cheese!?” New is always uncomfortable, but there’s a fix for getting back to the old familiar UI… Hit ⌘ + U or CTRL + U to hide the new pen sidebar Install the Classic Block Template to remove ~14 lines of HTML (Optional) Okay, now hit ⌘ + U or CTRL + U again, because what’s in the sidebar are all the new features and favorite parts I want to talk about. An entire file system The fir…

I have spent a lot of time experimenting with AI agents in my home lab over the past year. I have connected them to services through MCP servers as you… The post These Downloadable AI Skills Made My Home Lab Agents Much More Useful appeared first on Virtualization Howto.

Discover how to create an AI coding agent to automate writing and testing code, transforming your development workflow.

想拥有一个属于自己的个人博客,但又嫌 WordPress 臃肿、Typecho 功能太少?Hugo 是你的最佳选择:用 Markdown 写文章,一键生成纯静态 HTML,部署到 GitHub Pages 完全免费。本文从环境准备到上线运营,手把手教你搭建一个高性能、可定制的个人博客。

想拥有一个属于自己的个人博客,但又嫌 WordPress 臃肿、Typecho 功能太少?Hugo 是你的最佳选择:用 Markdown 写文章,一键生成纯静态 HTML,部署到 GitHub Pages 完全免费。本文从环境准备到上线运营,手把手教你搭建一个高性能、可定制的个人博客。

One of the biggest frustrations with running docker containers in the home lab is not usually the “big” complex problems. It usually comes down to the small tasks that slowly… The post This Lightweight Docker Tool Solves One of My Biggest Home Lab Annoyances appeared first on Virtualization Howto.

Explore the leading AI coding assistants of 2025, comparing Claude Code, Cursor, and GitHub Copilot for optimal software development workflows.

Explore the leading AI coding assistants of 2025, comparing Claude Code, Cursor, and GitHub Copilot for optimal software development workflows.

Every once in a while I come across an open source project that immediately makes me stop what I’m doing and spin up a test container in the home lab.… The post This Free Docker Dashboard Shows Your Entire Stack in 3D appeared first on Virtualization Howto.

Discover the world of OpenClaw, a promising open-source AI framework, and learn how to set it up and start building your first AI agent.

Discover the world of OpenClaw, a promising open-source AI framework, and learn how to set it up and start building your first AI agent.

AI Models Comparison 2026: Key Insights and Analysis Eighteen months ago, “which AI model is best” had a short answer. In mid-2026, it doesn’t. The frontier has splintered into a dozen credible contenders spanning American labs, Chinese open-weight challengers, and specialized players optimizing for cost, speed, or agentic reasoning rather than raw benchmark bragging rights. […]

Understanding Vibe Coding and Its Impact on Developers The smell of coffee hangs in the air. A developer stares at a blinking cursor. The screen remains empty. Writer’s block hits coders too. The blank page feels intimidating. But something shifts. A new conversation begins. The developer types a simple request. The AI responds instantly. Code […]

Learn how to build a customer support AI agent using Retrieval-Augmented Generation (RAG) to enhance response accuracy and user satisfaction.

Ollama has become one of the most popular ways to run large language models locally, and its official Python library makes it simple to integrate local LLMs directly into your Python applications. In this guide, you’ll learn how to install, configure, and use the Ollama Python library to generate text, build chat applications, stream responses, […]

Cleanup tasks are usually not something that we get overly excited about. However, one thing that I do get excited about is making my lab run better and having fewer… The post 9 Proxmox Cleanup Tasks You Probably Forgot appeared first on Virtualization Howto.

One of the coolest things you can do with Proxmox is spin up LXC Containers (Linux Containers). These are used heavily by those who are self-hosting services in their home… The post The Proxmox LXC Settings I Never Skip Before Deployment appeared first on Virtualization Howto.

Learn to deploy AI models efficiently on Kubernetes using KServe, a serverless inferencing solution for scalable machine learning model management.

Every AI model is only as good as the data it learns from, and a growing share of that data is scraped from the open web—product catalogs, reviews, forums, news, search results, images. But the web has grown hostile to automated collection at exactly the moment AI teams need it most.  Sites throttle, block, and […]

What Is Kimi K3? Kimi K3 is the newest flagship large language model from the Chinese AI lab Moonshot AI, and it has quickly become one of the most talked-about open-weight releases of 2026. The model scales to roughly 2.8 trillion total parameters, making it the largest open-weight model released to date, while still activating […]

Deploy Large Language Model inference at scale on Kubernetes with this in-depth guide, tailored for seamless scalability and maximum efficiency.

Quick Take Click/tap here to sign-up for a Rivian Demo Drive I finally got to test drive the new Rivian R2! Here's what the demo drive was like, some close-up details other reviews have overlooked, and my 10-item wishlist that could tempt even more Tesla drivers to make the switch. Rivian likely has a hit on their hands with the new R2 EV. This is by far the most significant new EV launch this year, and strong sales & delivery numbers will be great for Rivian, and for the overall health of the US EV industry. My test drive experience was quite good. You can sign up for a test drive near you by making a reservation a day or two in advance here. Video Featuring chapter ⏱️ links: 00:00 - Intro at the Rivian store in The Westchester (Mall), White Plains NY 00:39 - R2 close-up details others ha…  ( 36 min )

Explore Istio, Linkerd, and Cilium, three leading Kubernetes service meshes in 2025, analyzing their architectures, features, and practical applications.

Explore Istio, Linkerd, and Cilium, three leading Kubernetes service meshes in 2025, analyzing their architectures, features, and practical applications.

我方图床服务出现突发异常故障，导致部分用户出现图片上传失败、已上传图片无法正常访问、页面空白等问题，对大家的正常使用造成极大困扰，我们对此深表歉意。

我方图床服务出现突发异常故障，导致部分用户出现图片上传失败、已上传图片无法正常访问、页面空白等问题，对大家的正常使用造成极大困扰，我们对此深表歉意。

Discover how Claude Code, an AI coding assistant by Anthropic, compares to GitHub Copilot Pro in terms of pricing, features, and real-world performance. Learn about its unique capabilities and decide which tool best suits your development needs.

Discover how Claude Code, an AI coding assistant by Anthropic, compares to GitHub Copilot Pro in terms of pricing, features, and real-world performance. Learn about its unique capabilities and decide which tool best suits your development needs.

Speed Up Your Local Coding Agent: Running Gemma 4 with Multi-Token Prediction in Ollama Ollama 0.31 introduced a major performance upgrade for Apple Silicon users: multi-token prediction (MTP) for Gemma 4. In practice, this means a coding agent backed by Gemma 4 can generate code noticeably faster without any change in output quality. This tutorial […]

There are so many new tools out there for home lab related activities. So, I have to pick and choose what I decide to test out in my home lab.… The post This New Open Source Tool Completely Changed How I Manage My Home Lab appeared first on Virtualization Howto.

最近在刷X的时候了解到招商银行推出了万事达借记卡，xiaoz作为一个集卡爱好者自然不会错过，这篇文章将记录和分享申请整个过程和消费实测，有需要的朋友可以参考。招行万事达借记卡有什么特别的优势吗？招行万事达借记卡其实并没有明显的优势，如果你已经有VISA、MasterCard外币卡，那么此卡可有可无，

Understanding the Context Layer for Enterprise AI Most of the conversation about implementing enterprise AI centers around model strength and optimization. It’s exciting to talk about powerful AI models and innovative capabilities, but they only play a limited role in ensuring that AI systems succeed in a production setting.  Organizations that move beyond pilots are […]

Explore how to build a customer service bot using OpenClaw, an open-source AI framework, delving into AI agent development and best practices.

AI continues to advance at blazing speeds. One of the biggest developments over the past year has been the introduction of Model Context Protocol, or MCP. MCP servers provide a… The post My Favorite MCP Servers for the Home Lab in 2026 appeared first on Virtualization Howto.

Explore the strengths of Kubeflow and Argo Workflows as you decide the best tool for managing machine learning pipelines on Kubernetes.

Proxmox has a lot of really great features that make it ideal to run your virtualization stack in the home lab. You can take advantage of either virtual machines or… The post 7 Proxmox LXC Commands That Changed How I Manage My Home Lab Containers appeared first on Virtualization Howto.

最近深度使用 Claude Code，体验是真不错，但有个痛点：任务一跑十几分钟，人不可能干等着。起身倒杯水、回个消息，回来发现它早就停在"是否允许执行 Bash"上干等我了，白白空转半天。先是...

Explore scenarios where smaller language models like GPT-2 outperform larger counterparts, providing efficiency in resource-constrained environments.

I have tried out a lot of different volume backups over the years and there are many great tools out there. However, there was a database backup tool that recently… The post This Free Backup Tool Just Added Docker Volume Backups appeared first on Virtualization Howto.

Discover how to extend your AI agents with OpenClaw, leveraging plugins and extensions for flexibility and innovation in AI solutions.

杜老师以前周末经常过得很”散”，一晃就是一天。后来杜老师摸索出 3 种周末过法，今天跟大家分享一下。

杜老师以前周末经常过得很”散”，一晃就是一天。后来杜老师摸索出 3 种周末过法，今天跟大家分享一下。

Bensz GPT-5.6系列模型的社区反馈、基准表现和使用建议 本博客由科研AI Agent实验室BenszResearch强力驱动！如何更快地访问本站？有需要可加电报群获得更多帮助。本博客用什么VPS？创作不易，请支持苯苯！推荐购买本博客的VIP喔，10元/年即可畅享所有VIP专属内容！ 概览 这次升级最实用的变化，不是多了一个“最强模型”，而是终于有了三档可选：Sol 攻坚、Terra 干日常活、Luna 跑可验证的批量任务。 Sol 在终端、代理式编码、网页浏览和电脑操作上确有亮点；但代码基准之间的排序并不一致，暂时还不能简单喊“全面领先”。 Terra 的 API 单价只有 Sol 一半，几项代理测试却咬得很紧。对多数开发者，它可能比旗舰更值得先上手。 105 万 token 是能塞进去的容量，不等于每个细节都能稳定找回；单次输入跨过 27.2 万 token，还会踩到价格台阶。 首发三日的结论很朴素：别急着全切 Sol Ultra。按任务难度 […] Bensz

在《装修日记（4）》中预告过7月搬家，本周五正式搬入未名府新家，周末收拾了两天。从上周末开始，就在紫樾府陆陆续续打包行李，在这里住了一年，东西还真不少。找蓝犀牛搬家，约了最大的厢货。

Explore OpenClaw, an open-source AI agent framework, and understand the broader context of AI agent development through comparisons with leading frameworks.

缘由 我接到中国电信的通报，说我的网站存在安全问题，需要立即处理。我第一反应是网站可能被攻击或者被植入了异常内容。随后我向运营商索要了更详细的信息。 通过运营商反馈的问题域名。让我有些疑惑。这个域名和网站内容并没有解析、部署在电信的服务器上，而是使用腾讯云的 EdgeOne 和 COS 托管的静态内容。讲道理应该是腾讯云找我处理，但是现在反而是运营商先联系了我。 不过，既然已经发现问题，最重要的还...

GLM-5.2 has become one of the most searched models on the Ollama library, and for good reason. Built by Z.ai, it’s positioned as their flagship model for what they call “long-horizon tasks”: large, multi-step engineering work that unfolds over hours rather than minutes. In this tutorial, we’ll walk through what makes GLM-5.2 different, and how […]

Explore Lovable.dev, an AI-powered platform revolutionizing full-stack web app development. Understand its features, technologies, and how it simplifies the creation process.

AI is showing up literally everywhere and massively changing how things are carried out and being done. I am looking at ways that AI can benefit my home lab and… The post I Connected AI to My Proxmox Cluster Using MCP and It Was Better Than I Expected appeared first on Virtualization Howto.

One of the exciting things that I love about running a home lab is discovering new community projects before they become mainstream. Every so often there are projects that appear… The post 7 Open Source Infrastructure Projects I’m Watching Closely Right Now appeared first on Virtualization Howto.

Explore the choice between fine-tuning and RAG for AI applications, focusing on customer support systems. Learn about their setups, advantages, and best-use scenarios.

Starryblu是WoTransfer Pte. Ltd.旗下的一款全球支付App，与熊猫速汇同属一个产品生态——不同的是，熊猫速汇专注国际汇款，而Starryblu更侧重全球账户管理与多币种支付。通过一个账户，用户可以管理多种货币、完成换汇与跨境转账，并支持申请虚拟卡或实体借记卡，满足海外消费、

I have been moving heavily into immutable server operating systems especially for hosting my containerized home lab resources. Flatcar has been my OS of choice lately for container hosts. But… The post The NixOS Tweaks I Wish I’d Started Using Sooner in my Home Lab appeared first on Virtualization Howto.

Explore the architecture of OpenClaw, a new open-source AI agent framework, and its foundational concepts in AI development.

The thing being contained changed. So did the container. I’ve been in the Docker community since 2015. I’ve written more docker run tutorials than I can count, and I still get people in the Collabnix Slack asking me which base image to use for their Spring Boot app. That question is still valid. But it […]

不知道大家有有没有遇到这样一个问题，就是在看海外动作电影的时候没有字幕，效果大打折扣，不过最近xiaoz发现一款宝藏开源软件Caption-Trans，它可以使用 AI 大模型翻译视频字幕，特别为日语优化（作者介绍），至于具体场景，你就自己想象吧。两年前市面上已有类似工具，不过多数需要付费订阅。Ca

I think everyone goes through this shift as they transition their home labs from virtual machines over to containers. You tend to approach containers exactly the same way as you… The post Why I Stopped Treating Docker Containers Like Tiny Virtual Machines appeared first on Virtualization Howto.

Explore how Model Context Protocol (MCP) empowers AI coding tools like Claude Code to enhance productivity and accuracy by providing context-aware code understanding and edits.

There are some home lab projects that teach you a few things, and are great to incorporate. However, there are some projects that looking back on my journey, really taught… The post The Home Lab Projects That Taught Me the Most appeared first on Virtualization Howto.

Explore how to enhance AI agent security with guardrails against prompt injection attacks, ensuring robust and secure AI deployments.

Ampere Skylark 微架构评测¶ 背景¶ Ampere eMAG 采用的是 Ampere Skylark 微架构，虽然是 2018 年的处理器了，但也顺带评测一下。其前身是 AppliedMicro 的 X-Gene 3 微架构，用在 Ampere eMAG 芯片上，用的是 TSMC 16nm FinFET+ 工艺。 官方信息¶ 只找到了 X-Gene 2，即 Skylark 更早一代的信息：AppliedMicro X-Gene2 前端¶ 取指带宽¶ 为了测试实际的 Fetch 宽度，参考 如何测量真正的取指带宽（I-fetch width） - JamesAslan 构造了测试，实验结果如下： 可见取指带宽是每周期四条指令，并且很多时候并不能打满。 测试过程详见测试代码。 L1 ICache¶ 为了测试 L1 ICache 容量，构造一个具有巨大指令 footprint 的循环，由大量的 nop 和最后的分支指令组成。观察在不同 footprint 大小下的 IPC： 可以看到 footprint 在 32 KB 之前时可以达到 4 IPC，这对应了 32KB 的 L1 ICache。 测试过程详见测试代码。 BTB¶ 构造大量的无条件分支指令（B 指令），BTB 需要记录这些指令的目的地址，那么如果分支数量超过了 BTB 的容量，性能会出现明显下降。当把大量 B 指令紧密放置，也就是每 4 字节一条 B 指令时： 可见 Skylark 的 BTB 对分支紧密放置的情况支持不是很好，在 1024 个分支之内的 CPI 达到了 3.5，这一点在较早的处理器上比较常见，比如 Neoverse N1 的 BTB，不过稍微新一些的处理器，随着设计的改进，类似的情况比较少见了。之后到 8192 个分支，CPI 达到了 5。…

杜老师居家办公一年多，最大的感受是”分心比想象中容易得多”。今天杜老师就把自己用了一年多、一直稳定在用的 5 个减少分心的动作分享给大家。

杜老师居家办公一年多，最大的感受是”分心比想象中容易得多”。今天杜老师就把自己用了一年多、一直稳定在用的 5 个减少分心的动作分享给大家。

I see so many tools listed and presented that are said to solve a problem in the Proxmox realm. I have tried many different tools over the past couple of… The post I Tried This New Proxmox Inventory and Auditing Tool and I Was Impressed appeared first on Virtualization Howto.

Gateway API’s Inference Extension was one of the standout topics at recent KubeCon events, tackling a very specific pain point: generic HTTP load balancing algorithms are a poor fit for LLM traffic, where request cost varies wildly and a single long-running generation can pin a backend for seconds. This tutorial sets up the Inference Extension […]

Serving large language models efficiently is one of the dominant themes at recent KubeCon and CloudNativeCon events, and the NVIDIA GPU Operator has become the de-facto way to make GPUs first-class, schedulable resources inside Kubernetes. This tutorial walks through installing the operator and deploying a real LLM inference workload on top of it. What the […]

The Model Context Protocol (MCP) has quickly become one of the most talked-about topics at recent CNCF and KubeCon hallway tracks. It gives AI assistants a standard way to call tools and read live data, and Kubernetes is a perfect target: instead of copy-pasting kubectl output into a chat window, you can let an AI […]

Explore the state of AI agents in 2025, covering the latest market trends, powerful frameworks, and future predictions shaping this transformative technology.

各位老板好，我是xiaoz，几天又过去了，这几天我一刻没闲着，每天都在疯狂迭代ZNote私有笔记，现在给老板汇报下最新支持的AI 检索问答功能。ZNote开源地址：https://github.com/helloxz/znote什么是 AI 检索问答？简单来说，和传统AI对话不同的是，ZNote的A

Apple M2 (Avalanche & Blizzard) 微架构评测¶ 背景¶ 之前分析过 M1 和 M4，趁着机会，也评测一下 M2 的微架构，给出一个从 M1 到 M2 再到 M4 的发展脉络。 官方信息¶ 苹果发布了 Apple Silicon CPU Optimization Guide，包括了一些 M2 的微架构信息。 现有评测¶ 网上已经有针对 Apple M2 微架构的评测和分析，建议阅读： 我们找到了 Windows 电脑续航差的原因！苹果 M2 深度分析 不为人知的角落，Apple M2 的小小努力（其一） Apple M2 Blizzard 微架构评测 (上)：阳春白雪 Apple M2 Blizzard 微架构评测 (中)：阳春白雪 下面分各个模块分别记录官方提供的信息，以及实测的结果。读者可以对照已有的第三方评测理解。官方信息与实测结果一致的数据会加粗。 Benchmark¶ Apple M2 Avalanche/Blizzard 的性能测试结果见 SPEC。 前端¶ 取指带宽¶ Avalanche¶ 为了测试实际的 Fetch 宽度，参考 如何测量真正的取指带宽（I-fetch width） - JamesAslan 构造了测试，实验结果如下： 可以看到每 16 条指令会多一个周期，因此 Avalanche 的前端取指宽度确实是 16 条指令，与 Apple M1 Firestorm 和 Apple M4 P-Core 都相同。 Blizzard¶ 用相同的方式测试 Blizzard，结果如下： 可以看到每 8 条指令会多一个周期，意味着 Blizzard 的前端取指宽度为 8 条指令，和 Apple M1 Icestorm 相同，不过表现在图像上不太一样。 测试过程详见测试代码…

睡前在网上买点破烂，睡醒看看破烂发货了没有，只要快递还在路上，这日子就有盼头。一个人如果没有物欲，那就离抑郁不远了。有些钱看似乱花，其实都是防止自杀而已。一杯奶茶的钱能买20个鸡蛋，烦躁的时候总不能吃20个鸡蛋…一件衣服能买50斤土豆，不开心的时候总不能蒸一锅土豆。日子得有点甜头，不花点小钱取悦自己，就怕花大钱看心理医生。 有些钱看似随手乱花，其实都是必要的精神安慰。珍惜自己的不良嗜好，它有可能是自己热爱生活的主要原因。所以在经济条件允许的情况下，能让你开心的消费从来都不是浪费。 不要为花出去的钱心疼，钱并没有离开你，只是换种方式陪着你… 前段日子挺忙的~ 一些文章的配图都用起了AI~，但总说感觉缺少了些什么…检讨~以后少用AI配图。 小牛还是耐造的 第2辆小牛，2025款NXTUltra大师版，行驶里程已经达到5000公里了，和上一辆NXT玩家版一样基本没啥毛病，是最靠得住的伙伴… 别说~ 这玩意还有点解压~！！！促进大脑分泌多巴胺、内啡肽，这类物质将直接改善情绪、减轻焦虑，舒缓紧绷的神经…… 小牛NXT专用行车记录仪 2026年06月28日小牛NXT的专用行车记录仪发布了… 随手下单到货后第一时间去小牛门店安装，除了WIFI名称和密码无法修改，目前都挺满意的。 价格不便宜，比沃尔沃S60的那套隐形行车记录仪都要贵… 但对比 DJI 大疆~ 成本还是要低得多，只是用于行车安全，每天两点一线，没那么多时间与精力去专门拍摄视频做素材。 艾尔优 X id221，准确型号是 DC1mini，使用wifi直连传输录像文件，12V-120mA，虽然没有电工证，但换算功率：12V × 0.12A = 1.44瓦（W） 应该是很低的，不知道算得对不对。但毕竟标签上的“120mA”通常指最大能提供的电流。实际耗电多少，完全取决于这款小牛NXT行车记录仪的负载需要多少，应该会更加省电。 装车后的外观，不仔细看是看不出来的~ 雷克沙(Lexar) 64GB 633x microsSDXC UHS-I 储存卡，速度100MB/s 应用性能等级为A1 近一年时间小牛电动新车不断，电摩领域继FX风速后，又推出了4299元的NX风速，那40L储物空间，着实羡慕。 后续又出了NX风速ABS版，也和25NXT大师一样有了BSD+RCW但价格上就没啥吸引力了，毕竟空车，电池还得自己配，最关键的是这是电摩，在武汉的三环线内禁行。 再后来，又推出了NX马拉松，72v50ah 大锂电146公里的续航，仅6500元。但该简配的简配，10寸的后轮及电机是真的恶心…… 但……这次……NX马拉松的前大灯绝了。 NX马拉松大灯 45000坎德拉的大灯，在原厂级别算是亮到离谱了，夜骑安全感直接拉满。 最最最关键的是，无须更换继电器，25NXT大师原厂DC直接对插就可以直上，功率提升忽略不计，官方售后订货价格173元…还有保修。 远比自己花费5-600元去改装店买一个，在换继电器，最终大灯是亮了，功率大了，电耗也就增加了及后续一系列问题，要省心得多… 赶紧牛店预订NX马拉松大灯，一周后顺利安装~ 确实亮了好多~ NXTUltra后轮爆胎 订货到收货中间有5天左右时间，恰逢那段日子武汉下雨，这次25NXT大师的后轮被钉子扎了。一个上午的时间后轮胎胎压从2.8将至0.9，到下午直接为0，心理嘀咕着，这次是多大的洞？ 中途下楼查看，果然，一个自攻螺丝钉扎在后轮胎上。 记得上一台NXTPlay在6000多公里时第一次破胎，1公里的路程，推行了近20分钟… 还推了2次，后来发现是双孔无法补，所以最终更换米其林 CITY GRIP SAVER 轮胎。相比第一次的无助，这次知道在小红书上，找上门维修和补胎的。打了老板电话20分钟赶到现场，5分钟完成轮胎修补后离开，收费20元。 用的是胶条… 老板自己都说这玩意，寿命不长，不一定经得起我日常的骑行… 毕竟 NXTUltra 70Km的极速过坑，冲击力还是非常大的。如果是买个菜，确实能管个半年一年的……唉~感觉又要花钱买轮胎的时候要到了。 当然沃尔沃S60也没闲着，也要上来凑个热闹 上个班都这么的不省心~ 至此，25NXTUltra入手258天，行驶5227公里，又已开始惦记 米其林 CITY GRIP SAVER 轮胎……

前言 使用docker部署Prometheus+Grafana+Alertmanager监控系统，并集成钉钉告警 环境准备 （1）使用脚本安装docker、docker-compose <span

If you’ve ever added nvidia.com/gpu: 1 to a pod spec and wondered what actually happens behind the scenes, you’re not alone. GPU allocation is one of the most misunderstood parts of Kubernetes, mostly because it works quite differently from how Kubernetes handles CPU and memory. This is a simple, no-jargon explainer of how it actually […]

GPU nodes are the single most expensive line item in most AI infrastructure bills, often 10-20x the cost of a comparable CPU instance. Leaving them running around the clock “just in case” is one of the fastest ways to burn budget. This tutorial walks through autoscaling GPU node pools in Kubernetes with Karpenter, the CNCF […]

Most Kubernetes clusters running GPU workloads have no visibility into whether those expensive GPUs are actually being used. Industry surveys consistently find average GPU utilization sitting well below 50% in shared clusters, yet teams keep provisioning more hardware because they can’t see where the waste is. This hands-on guide sets up full GPU observability on […]

GPUs are the most expensive line item in most AI infrastructure budgets, yet a single pod rarely saturates a modern GPU like an A100 or H100 for inference or light training workloads. Kubernetes, by default, treats a GPU as a single indivisible resource. This guide walks through the four approaches the CNCF community is actively […]

Explore the capabilities of AI coding assistants like GitHub Copilot and Claude Code. Learn how these tools can transform your development workflow.

忙成狗的日子，连吐槽都没力气，早已电量不足，却还在硬撑，上班、加班、连梦里还在回消息…每天都是两点一线，接触最多的就是这辆小牛电动车和沃尔沃S60了 跨入这个所谓的“新行业”第三年，这段日子里时间长得就像一场没有尽头的道路，攥着满手的待办，在晨昏里不停奔走。春天的花开了…然后又谢了，武汉街边的梧桐叶从嫩绿变深黄… 发现每天都只是匆匆路过，从未停下脚步好好看一眼。日子被拆成了一个又一个的截止日期，一个节点刚过，下一个节点已经在前方等着，连发呆的片刻都带着承重的负罪感。 很久没有和过往的那些朋友们好好的聊过天，很久没有去看过一部电影或电视剧，很久没有在某个午后晒着太阳无所事事的发呆。那些曾经稀松平常的松弛感，在这段漫长的忙碌里，都成了遥不可及的小事，自己像是一直沉在这片忙碌的海水里，浮上来喘口气的间隙都少得可怜，日子漫长却感觉没有尽头，然而还在不停的向前赶。 生活被突如其然的各种琐事填得满满当当，回头看时，只记得一路的奔波，竟想不起有多少个具体的细节，只觉得这段日子，漫长又疲惫。下班后来到位于斗级营的网红山顶烧烤，距离黄鹤楼仅 260 米的天台烧烤店，又见黄鹤楼，山顶的晚风加上滋滋冒油的烤串和满桌的快乐水，都是属于武汉的烟火浪漫……但人均150元的价格，居然没吃饱… 晚上下班走出大厦的时候，整条街都已经安静了下来，白天那些挤那拥挤的人行道，此时只剩下路灯把能将自己的影子拉得很长很长。路边的店铺大多卷上了闸门，只有 24 小时的有家便利店还亮着暖黄的灯，玻璃上蒙着一层雾气，偶尔还能看见店员趴在柜台打盹… 此时一个人、一台车、一条道、一路风景，却能短暂的从喧哗中逃离，回归内心的平静。不同于开车回家，到楼下了不下车还要在车里坐好久，一路吹风、看沿途的夜景、绿植与行人，让视觉和触觉得到放松；晚风和新鲜空气，改善了沉闷情绪，缓解了久坐带来的身心疲惫。 不用等地铁、不用赶公交、不用按导航走，想骑快就吹吹风，想慢就晃着看风景，看到好看的小路就拐进去，路过好吃的小摊就停下来买。没有目的地，也不用赶进度，就顺着风骑，这种不用被安排、不用赶时间的自由感，是这一天里，最难得的快乐。 骑电动自行车同样属于骑行，属于温和的有氧运动，能促进大脑分泌多巴胺、内啡肽，这类物质将直接改善情绪、减轻焦虑，舒缓紧绷的神经，相比剧烈运动，骑行电动自行车的行强度低，不会带来什么身体上的负担。 虽然公司距离家的距离只有8公里，这期间的骑行速度、路线、启停都由自己决定，这种自主掌控感持续抵消工作中那些被动受压的无力感，带来了更多轻松自在的心理体验。 生活不易，起起伏伏都是常事，遇到不顺心的人还有事，不用去钻牛角尖，但可以骑上小牛电动车，驰骋在深夜最最实诚的道路上，它不会催你赶路，只陪着你，把一身的疲惫慢慢走散。 晚风的清凉、夜色的迷人、深夜的宁静都会让自己的内心逐渐趋于平静。 终于… 高考结束了，中考结束了，暑假它又来了~！

畅游阁 抖音刷到畅游阁终于开放了，这不周末就出去看了，可以乘坐电梯直达顶层，电梯很迷小较小，阁楼外视野开阔，可以看到很多爱好者在邕江玩桨板，但是最近频繁上热搜，路过来来往往的船，部分人游出安全水域，接近航道，很危险。 聚餐 这家南宁连锁的烧烤店，很好...

我在使用的是腾讯云轻量应用服务器，免费快照额度只有2个，每一次都是手动删除旧快照再创建新快照实在太麻烦，问了客服也没有直接设置的功能，但好在提供了API接口来实现。 所以当前使用了1Panel面板的计划任务和腾讯云API来实现：自动识别并删除旧的自动快照，...

A practical guide to function/tool calling with Ollama: how it works, which local models support it, a minimal agent loop, and common pitfalls to avoid when building local AI agents.

Explore Kubernetes Operators, the bridge to automating complex applications in Kubernetes, highlighting their architecture and utility.

任何新奇的信息——出乎意料的新闻故事，或者猫咪做出不可思议行为的短视频——都会引发多巴胺分泌。这在短期内可能让你感觉良好，但随着时间的推移，你可能会沉迷于此，并因此降低智商。

选择简单模式，人生会越来越困难；选择困难模式，人生会越来越简单。

Learn how to run AI Agent securely in a MicroVM using Docker Sandboxes

Learn how to use LangGraph to build advanced, stateful AI agent workflows. This guide covers architecture, node definitions, and state transitions for creating intelligent AI systems.

No content preview

FedRAMP automation platforms promise faster authorizations, lighter evidence workloads, and cleaner continuous-monitoring cycles. This guide compares six leading options and highlights where each one best fits. 1. Vanta: your all-rounder for cloud-native SaaS teams that need FedRAMP plus everything else Vanta is built for teams that want compliance to behave like the rest of their […]

After thousands of real Claude Code sessions in production codebases, the senior engineers who get the most out of Claude Code share a set of non-obvious practices. Here are the top 10, each with working code you can use immediately. Tip 1: Explore First, Plan, Then Code Tip 2: Use Plan Mode for Big Changes […]

One of the most-searched Claude Code topics in 2026 is how to give Claude persistent memory and custom commands. This tutorial covers CLAUDE.md, the memory system, and custom slash commands (skills) – the trio that makes Claude Code truly feel like a team member who knows your project. 1. CLAUDE.md – Give Claude Permanent Project […]

Developers are searching “Claude Code vs Cursor vs GitHub Copilot” more than ever in 2026. Here is a hands-on breakdown of all three – with real commands, pricing, and when to use which. Quick Comparison at a Glance Feature Claude Code Cursor GitHub Copilot Interface Terminal / CLI IDE (VSCode fork) IDE extension Price $20/mo […]

Claude Code is exploding in 2026. Based on current search trends and Google data, here are the 5 hottest Claude Code topics with hands-on code you can run today. Topic 1: Install and Setup Claude Code The #1 most-searched topic. Get Claude Code running in your terminal in 2 minutes. Topic 2: Claude Code + […]

Learn how to run large language models locally with Ollama. Explore comprehensive setup, leveraging Docker and Python environments for optimal performance and privacy.

Learn how to run large language models locally with Ollama. Explore comprehensive setup, leveraging Docker and Python environments for optimal performance and privacy.

Agentic AI is not another shiny object for GTM teams. If you treat it like a chatbot, you get chatbot results. If you embed it into workflows that influence pipeline, coverage, and revenue, then you see lift where it matters. According to BCG, companies embedding agentic AI into core workflows are already seeing measurable gains, […]

I am always on the lookout for new and interesting tools that allow us to manage our Proxmox servers and clusters more efficiently and overall, a little easier. I stumbled… The post I Found a New Way to Manage Proxmox Without Opening the Web UI appeared first on Virtualization Howto.

各位老板好，我是xiaoz，距离上次ZNote更新已经过去3天，这次我把ZNote的安卓APP给造出来了，现在双手奉上，需要的朋友可自取哦。注意：安卓APP需要搭配ZNote服务端一起使用！如果您还不了解ZNote，可以先看下往期回顾：小而美的ZNote笔记软件 —— 支持Docker私有部署和WE

杜老师家里有一只橘白相间的猫，平时最爱跳上书桌”监工”。前几年给猫拍的照片都是糊的，后来摸索出 3 个小技巧，今天跟大家分享一下。

杜老师家里有一只橘白相间的猫，平时最爱跳上书桌”监工”。前几年给猫拍的照片都是糊的，后来摸索出 3 个小技巧，今天跟大家分享一下。

⏱ Reading time: ~13 min | 🗓 Updated: June 2026 What if your AI coding assistant could search your GitHub issues, update your Linear board, query your database, browse documentation, and run code in a sandboxed environment — all within a single session, without you copy-pasting between tools? That’s exactly what the Model Context Protocol […]

⏱ Reading time: ~13 min | 🗓 Updated: June 2026 Every Claude Code user knows you can just type tasks in plain English. But power users know that CLAUDE.md — Claude Code’s project memory file — combined with custom slash commands and a well-tuned settings.json can transform Claude from a capable assistant into a domain […]

⏱ Reading time: ~12 min | 🗓 Updated: June 2026 One of the most powerful — and least understood — features of Claude Code is its multi-agent architecture. While most developers use Claude Code as a single AI session working through tasks sequentially, the platform supports spawning multiple subagents that work in parallel, each with […]

⏱ Reading time: ~14 min | 🗓 Updated: June 2026 The AI coding tool war of 2026 is real — and the three frontrunners are Claude Code, Cursor, and GitHub Copilot. Each has passionate advocates, and each genuinely excels in different scenarios. But which one belongs in your development workflow? Google Trends data for Q2 […]

⏱ Reading time: ~13 min | 🗓 Updated: June 2026 If you’ve been using Claude Code for more than a few days, you’ve likely noticed a pattern: Claude finishes a task, you check the output, verify it manually, run a linter, run your test suite, and then move on. Now imagine all of that happening […]

The question “ChatGPT or Claude?” has become one of the most searched comparisons in AI in 2026. According to Google Trends, “ChatGPT vs Claude” saw breakout interest driven by rising queries like “chatgpt 5.5 vs claude opus 4.7,” “claude mythos” (+2,700%), and “chatgpt codex pricing.” Both ChatGPT (by OpenAI) and Claude (by Anthropic) have rapidly […]

Claude Code has rapidly emerged as one of the most powerful AI-powered coding assistants available to developers today. Launched by Anthropic in early 2025, it peaked in search interest around April 1, 2025, and has maintained consistent global traction — particularly in China, Singapore, South Korea, Taiwan, and Israel. Related trending searches include “Claude Code […]

If you are like me, for the longest time, I just assumed that Docker was very smart in the way it recovers unhealthy containers. And, to be fair, there is… The post I Thought Docker Would Restart Unhealthy Containers. I Was Missing One Tool. appeared first on Virtualization Howto.

今天刚结束低压电工作业的实操考试，趁热把流程记一下，希望对后面要考的同学有点用。 关于培训 低压电工作业属于特种作业，证书由应急管理局颁发，全称「低压电工作业证」，俗称从业证。 按应急管理局特种作业考试要求，得先参加培训并通过考试，才能取证。培训机构要选有特种作业安全技术培训资质的，名单在北京市应急管理局官网上能查到： 培训一共 4 天，每天刷身份证 + 人脸识别，培训记录是后面报考的必要条件。 ...

One of the core monitoring components of my home lab is uptime monitoring. In other words, are my apps, services, web resources, etc up and running? Is anything down? I… The post The Home Lab Services I Never Stop Monitoring (And Why They Matter) appeared first on Virtualization Howto.

2026年6月12日（星期五）下午7:40分，收到了腾讯云流量包用量耗尽的通知，我感到奇怪，每个月的流量顶多不超过50G，怎么会一下子耗尽。 刚准备吃饭的我懵了，这不对劲，马上登录服务器看情况，果然还在刷，一下子就超了，我只能把服务器关机了，腾讯云里面还...

Learn how to build an AI agent from scratch with Python in this detailed, step-by-step guide. From setting up environments to implementing functional neural networks, this comprehensive tutorial covers it all.

各位好，我是xiaoz，距离上次发布ZNote 0.1.0版本，已经过去两天，这2天我一刻也没闲着，现在已经更新到0.3.1版本了，给各位老板汇报下ZNote笔记软件最近更新情况。ZNote官网：https://znote.xphub.dev/ZNote近期更新由于近期更新内容实在是太多了，所以无法

周末两天继续Vibe Coding，在做几样东西：围绕钉钉的个人助理系统：自动维护待办清单和项目的OnePage，自动生成日报和周报Obsidian发布到微信公众号的插件。微信API中转服务：微...

Learn how to run Qwen3.6-35B-A3B (FP8) on the NVIDIA DGX Spark GB10 Grace Blackwell supercomputer using vLLM. Step-by-step guide covering single and dual-node setup, agentic tool calling, FP8 quantization, performance benchmarks, and optimization tips — based on the most-searched thread on the NVIDIA Developer Forums with 28,000+ views.

Explore the orchestration of multiple AI agents using OpenClaw, an open-source framework designed for dynamic agent coordination.

If you spend much time at all watching Proxmox content, you will hear about a lot of the big features, especially with the new releases in versions like Proxmox VE… The post 9 Proxmox Features That Quietly Improved How I Manage My Home Lab appeared first on Virtualization Howto.

The AI computing landscape is evolving at an unprecedented pace, and NVIDIA has once again redefined what’s possible at the desktop level with the NVIDIA DGX Spark. Powered by the cutting-edge GB10 Grace Blackwell Superchip, DGX Spark is the world’s smallest AI supercomputer – a palm-sized powerhouse that brings data center-class AI performance directly to […]

I am all about project-based learning and also all about Linux! If you’re anything like me, the weekend is when you are able to get some of your best home… The post 9 Linux Management Tools Worth Deploying in Your Home Lab This Weekend appeared first on Virtualization Howto.

Brand ambassadors give a company a face in live moments where decisions happen fast and trust forms in seconds. The job mixes clear speech, product fluency, and careful listening so each exchange feels safe and useful. That combination matters more than most teams realize at first. For health and wellness teams, that human contact supports […]

ARM Neoverse V3 (代号 Poseidon) 微架构评测¶ 背景¶ 使用 ARM Neoverse V3 核心的 AWS Graviton 5 最近上线了，相比之前的 Neoverse V2 应该有一些改进，所以测试一下这个微架构在各个方面的表现。 官方信息¶ ARM 关于 Neoverse V3 微架构有如下公开信息： Arm® Neoverse V3 Core Technical Reference Manual Arm Neoverse V3 Software Optimization Guide Neoverse V3 与 Cortex X4 高度相似，这里也列出 Cortex X4 的相关信息： Arm Unveils 2023 Mobile CPU Core Designs: Cortex-X4, A720, and A520 - the Armv9.2 Family Arm Cortex-X4 advances frontiers of CPU performance Arm® Cortex‑X4 Core Technical Reference Manual 下面分模块记录官方信息和实测结果。官方信息与实测结果一致的数据会加粗。 现有评测¶ 网上已经有 Neoverse V3 微架构的评测和分析，建议阅读： 最强 Arm 处理器 AWS Graviton5 架构剖析 Neoverse V3 微架构 下面分各个模块分别记录官方提供的信息，以及实测的结果。读者可以对照已有的第三方评测理解。官方信息与实测结果一致的数据会加粗。 Benchmark¶ Neoverse V3 (AWS Graviton 5) 的性能测试结果见 SPEC。 前端¶ L1 ICache¶ 官方信息：64K…

不是换个 Prompt 就行。危险操作的三层防御、Plan 模式的零代码实现，以及两个只有真正做过才知道的坑。

SPEC CPU 2026 Workload Analysis (FP Rate)¶ 中文版本 Background¶ Following the INT Rate article, this article continues with the workload analysis of SPEC FP 2026 Rate. The test environment is the same as the previous INT Rate article and won't be repeated here. Recommended reading: Evaluating SPEC CPU2026 and SPEC CPU2026: Characterization, Representativeness, and Cross-Suite Comparison SPEC FP 2026 Rate Analysis¶ 709.cactus_r¶ Cactus is a computational framework, used here to solve the Einstein equations in vacuum. Command: cactus ShiftedGaugeWave.par Measured runtime is 103.4s, reftime is 858s, corresponding to 8.30 points. Performance under different compilers and flags: Compiler + Flags Time (s) Score Improvement over GCC 14 -O3 (%) GCC 14 -O3 103.4 8.30 0 GCC …

SPEC CPU 2026 负载特性分析（FP Rate 篇）¶ 本文同步发布到本人的知乎。 English version 背景¶ 继 INT Rate 篇 后，本文继续分析 SPEC FP 2026 Rate 的负载特性。 测试环境与先前的 INT Rate 篇 相同，这里不再赘述。 推荐阅读：Evaluating SPEC CPU2026 和 SPEC CPU2026: Characterization, Representativeness, and Cross-Suite Comparison SPEC FP 2026 Rate 分析¶ 709.cactus_r¶ Cactus 是一个计算框架，这里用它来求解真空中的爱因斯坦方程。命令参数如下： cactus ShiftedGaugeWave.par 实测数据显示，运行时间为 103.4s，reftime 是 858s，对应 8.30 分。不同编译器和编译选项对 709.cactus_r 的优化情况如下： 编译器 + 选项 时间 (s) 分数 相比 GCC 14 -O3 性能提升 (%) GCC 14 -O3 103.4 8.30 0 GCC 14 -O3 -march=native 83.9 10.23 23 GCC 14 -O3 -ffast-math 101.2 8.48 2 GCC 14 -O3 -ljemalloc 100.7 8.52 3 LLVM 22 -O3 94.6 9.07 9 LLVM 22 -O3 -march=native 90.5 9.48 14 可见 -march=native 能提供巨大的性能提升，LLVM 22 在 -O3 下比 GCC 14 快，不过 GCC 14 的 -O3 -marc…

SPEC CPU 2026 Workload Analysis (INT Rate)¶ 中文版本 Background¶ I've been running some benchmarks with SPEC CPU 2026 recently, and plan to do in-depth workload analysis combined with the test results. This article focuses on SPEC INT 2026 Rate workload characteristics. For SPEC FP 2026 Rate analysis, see the FP Rate article. Test environment: CPU is Intel i9-14900K P-Core @ 5.7 GHz, Linux distribution is Debian Trixie, compiler is GCC 14.2.0, default compilation flags are -O3. This CPU can actually boost up to 6.0 GHz, but occasionally fails to boost under single-core workloads for unknown reasons (degradation protection?), specifically manifesting as the CPU core being forced down to 4.7 GHz after running for a while. So I opted for the more reliably achievable 5.7 GHz. Only one physic…

SPEC CPU 2026 负载特性分析（INT Rate 篇）¶ 本文同步发布到本人的知乎。 English version 背景¶ 最近用 SPEC CPU 2026 跑了一些基准测试，打算结合测试结果做一些深入的负载特性分析。本篇主要是分析 SPEC INT 2026 Rate 的负载特性，SPEC FP 2026 Rate 的分析请看 FP Rate 篇。 本文测试环境：CPU 为 Intel i9-14900K P-Core @ 5.7 GHz，Linux 发行版为 Debian Trixie，编译器是 GCC 14.2.0，默认编译选项是 -O3。其实这颗 CPU 最快能 Boost 到 6.0 GHz，但时不时因为未知原因（防缩缸？）在单核负载下也 Boost 不上去，具体表现为每跑一段时间后 CPU 核心就会强制降频到 4.7 GHz。故退而求其次，选择在更容易稳定达到的 5.7 GHz 频率来跑。能稳定跑到 6.0 GHz 的只有那一个物理 P 核，其他 P 核也都能上 5.7 GHz，降频了换一个核心即可。6.0 GHz 下的性能可以参考之前的测试结果：INT 和 FP，基本上，从 5.7 GHz 到 6.0 GHz，性能可以按频率线性放缩。本文可能针对同一个负载给出多个不同的运行时间，这可能是因为多次运行导致的性能波动，也可能是因为部分数字包含了 perf record 的开销，不过误差都很小，可以放心对比。本文所用的脚本已开源到 jiegec/spec2026。 推荐阅读：Evaluating SPEC CPU2026 和 SPEC CPU2026: Characterization, Representativeness, and Cross-Suite Comparison SPEC INT 2026 Rate 分析¶ 706…

SPEC CPU 2026 在其他指令集上的编译¶ SPEC CPU 2026 官方只附带了 aarch64/ppc64le/riscv64/x86_64 指令集的预编译 tools，如果要在其他指令集上使用，就需要首先编译 tools，过程如下： cd /mnt && tar xvf install_archives/tools-src.tar wget -O config.guess 'https://git.savannah.gnu.org/gitweb/?p=config.git;a=blob_plain;f=config.guess;hb=HEAD' wget -O config.sub 'https://git.savannah.gnu.org/gitweb/?p=config.git;a=blob_plain;f=config.sub;hb=HEAD' cp config.* /mnt/tools/src/make-4.2.1/config/ # build tools mkdir -p /mnt/config cd /mnt && echo 'y' | SKIPTOOLSINTRO=1 FORCE_UNSAFE_CONFIGURE=1 MAKEFLAGS=-j16 ./tools/src/buildtools mkdir -p /mnt/config cd /mnt && . ./shrc && packagetools linux-loong64 例如下面是在 LoongArch 上编译 SPEC CPU 2026 的 Dockerfile，假设 SPEC CPU 2026 已经解压到 /mnt： RUN cd /mnt && tar xvf install_archives/tools-src.tar RUN wget -O config.guess 'https://git.savannah.gnu.org/gitweb/?p=config.git;a=blob_plain;f=config.guess;hb=HEAD' RUN wget -O config.sub 'https://git.savannah.gnu.org/gitweb/?p=config.git;a=blob_plain;f=config.sub;hb=HEAD' RUN cp config.* /mnt/tools/src/make-4.2.1/config/ # build tools RUN mkdir -p /mnt/config RUN cd /mnt && echo 'y' | SKIPTOOLSINTRO=1 FORCE_UNSAFE_CONFIGURE=1 MAKEFLAGS=-j16 ./tools/src/buildtools RUN mkdir -p /mnt/config RUN cd /mnt && . ./shrc && packagetools linux-loong64 RUN /mnt/install.sh -f 参考官方文档：Building the SPEC CPU®2026 Toolset。

记一次循环依赖导致的运维小事故¶ 背景¶ 每次没有 UPS 或 UPS 容量不够用的倒闸对于运维来说都是一次鸡飞狗跳。这次很不幸，鸡飞狗跳终于轮到了我，还好花了一个半小时还是解决了。在这里做个简单的复盘。 现象与排查¶ 首先介绍一下现象：停电恢复之后，服务器开起来，但是无法从互联网连接内网的网关（即主网关）。还好，之前考虑到网关的重要性，做了备份，走内网的备用网关顺利进入了内网，然后发现主网关即使 IP 地址是对的，也连不上互联网。虽然通过 keepalived 的配置，主网关和备用网关会通过虚拟 IP 给内网机器提供一个高可用的默认网关，但由于 keepalived 只是检测了机器是否开机，并没有判断能否正常访问互联网，所以 keepalived 总会选择优先级更高的主网关，导致虚拟 IP 指向主网关，因而内网的机器都连不上互联网，还得继续尝试修复主网关。 主网关运行在 ESXi 的虚拟机里，于是进入 ESXi 管理网页，看看它的网络情况。这个虚拟机的网络用的并不是 ESXi 的普通网络，而是通过 vCSA 配置的基于 DS 的 LACP。看了几个不同的 ESXi，发现问题都集中在 LACP 上。而 ESXi 是没法配置 DS 的，所以就先去看了 vCSA。连上 vCSA 以后一看，所有的 ESXi 都掉线了。原来，之前为了方便配置，vCSA 都是通过域名连接 ESXi 的，而域名就需要有 DNS 服务器解析，然而主网关连不上互联网，也就连不上配置好的互联网的 DNS 服务器，于是 vCSA 无法配置 ESXi，然后 ESXi 的部分虚拟机就会断网，正好主网关又在被断网的虚拟机里面。这就形成了一个循环依赖。 既然找到了问题，那就需要打破循环依赖：把主网关在 keepalived 里的优先级调低，让备用网关上位。结果这时候发现一个小的问题：备用网关的 NAT 忽然不工作了。排查了一下，发现是因为 net.ipv4.ip_forward = 1 写在了 /etc/sysctl.conf 文件里，而 Debian 升级 Trixie 以后，这个文件已经不会被应用了，而要把内容写到 /etc/sysctl.d/*.conf 里面去，通过 /usr/lib/systemd/systemd-sysctl --cat-config 来确认是否持久化成功。由于主网关一直工作得很好，备用网关很久都没有做 NAT 了，导致这个问题一直没有被发现。 修好以后，vCSA 就能找回 ESXi 了，然后通过 vCSA 再重新配置一下 ESXi 的 DS LACP 网卡，然后一切就恢复了。 反思¶ 虽然事故解决了，但这个过程中暴露了很多问题： 首先就是循环依赖：ESXi 的 DS LACP 依赖 vCSA，vCSA 依赖 DNS，DNS 依赖主网关，主网关通过 DS LACP 访问互联网。如果没有准备好备用网关，且备用网关恰好没有用 DS LACP 因此不受影响，那么修复起来就更麻烦了。解决循环依赖的办法也很简单，对于一些重要的虚拟机（如网关），它所依赖的功能越简单越好。 备用网关的 NAT 功能在升级 Debian 版本后，因为没有仔细阅读 sysctl 的变化而失效了，升级后缺少对功能的检查。 keepalived 只判断机器是否在线，但没有判断机器是否可以正常连接互联网、承担起网关的职能。 ESXi + vCSA 这类闭源软件，修复起来还是比较痛苦的，很多内部的工作原理并不清楚，可调试性也比较差，以后还是会谨慎选择。

分析 Pi Agent 的扩展系统设计，探讨其在简洁与可定制性之间的平衡，以及 70+ 官方扩展示例背后的设计哲学。

源码揭示 Claude Code 的 6 层渐进式压缩架构，从工具截断、大结果落盘、并行预算、微压缩到全量摘要，解读它如何在 200K 上下文窗口中持续工作而不会"爆掉"  ( 4 min )

Contribute to farzaa/clicky development by creating an account on GitHub.  ( 12 min )

Library for describing the state of a Git checkout. - sebastianbergmann/git-state  ( 5 min )

AI coding assistant skill (Claude Code, Codex, OpenCode, Cursor, Gemini CLI, OpenClaw, Factory Droid, Trae). Turn any folder of code, docs, papers, or images into a queryable knowledge graph - safi...  ( 24 min )

Contribute to glanceway/glanceway-sources development by creating an account on GitHub.  ( 10 min )

深入分析 Claude Code 的 Agent Loop 设计原理，从架构、提示词、工具设计和容错系统四个层面解读它如何实现持续工作的能力  ( 2 min )

A native macOS app for managing scheduled tasks. Built with SwiftUI + SwiftData. Menu bar resident, cron expressions, script execution, and more. - lifedever/TaskTick  ( 8 min )

macOS native ambient voice input & meeting transcription — Apple SpeechAnalyzer + Vision OCR + FluidAudio speaker diarization, fully on-device. - Marvinngg/ambient-voice  ( 9 min )

跨平台M3U8下载器，支持多线程下载、断点续传、边下边播、AES解密、浏览器自动识别地址. Contribute to Liubsyy/M3U8Quicker development by creating an account on GitHub.  ( 11 min )

各厂家 Coding Plan 实际价值对比. Contribute to mahonzhan/awesome-coding-plan development by creating an account on GitHub.  ( 7 min )

为OpenAI Codex打造的桌面伴侣。. Contribute to borawong/AiMaMi development by creating an account on GitHub.  ( 7 min )

一个使用原生Android开发的哔哩哔哩安卓客户端，支持触摸、遥控，以及安卓5. Contribute to cat3399/blbl development by creating an account on GitHub.  ( 6 min )

Android Automation Tool Based on Vision-Language Models - Turbo1123/roubao  ( 27 min )

Modifier+drag to move any window smoothly on macOS, exactly like native speed - XueshiQiao/AnyDrag  ( 7 min )

Browse Mastodon, Bluesky, X, Misskey, Nostr and RSS all in one app. One timeline, all your accounts, cross-post everywhere. - DimensionDev/Flare  ( 10 min )

基于Apple VisionKit 的 macOS 抠图. Contribute to 94xy/xykt development by creating an account on GitHub.  ( 10 min )

A parser for Laravel Blade templates written in Rust. - ryangjchandler/blade-parser.rs  ( 6 min )

Securely synchronize files with your devices on iOS using Syncthing - pixelspark/sushitrain  ( 22 min )

Token-Share: A native macOS menu bar LLM API gateway — translate and stream between OpenAI Chat Completions, OpenAI Responses, and Anthropic Messages protocols locally. - leemysw/token-share  ( 5 min )

Contribute to tforschbach/CodexRemote development by creating an account on GitHub.  ( 16 min )

Laravel HTTP Request to Curl Command. Contribute to jigar-dhulla/laravel-http-to-curl development by creating an account on GitHub.  ( 10 min )

No content preview

No content preview

Build and preview Laravel applications without leaving your browser - aschmelyun/liminal  ( 8 min )

熊猫视频自动化引擎: 用程序员的方式帮你自动化视频创作与发布. Contribute to szhshp/panda-video-generator development by creating an account on GitHub.  ( 20 min )

Use Codex from Claude Code to review code or delegate tasks. - openai/codex-plugin-cc  ( 22 min )

Own your AI. The native macOS harness for AI agents -- any model, persistent memory, autonomous execution, cryptographic identity. Built in Swift. Fully offline. Open source. - osaurus-ai/osaurus  ( 24 min )

PAO is agent-optimized output for PHP testing tools. - nunomaduro/pao  ( 8 min )

Siri Shortcuts Programming Language 🍒. Contribute to electrikmilk/cherri development by creating an account on GitHub.  ( 11 min )

项目概述 Everything Claude Code (ECC) 是一个全面的 AI 智能体性能优化系统,专为 Claude Code、Cursor、OpenCode 等 AI 开发工具设计。该项目经过 10 个月以上的生产环境实战验证,提供了完整的工作流程、智能代理、技能库和自动化钩子。 核心价值 降低 Token […]  ( 10 min )

An open-source remake of the short-lived jetbrains git client - DetachHead/rebased  ( 17 min )

IDE for the Cherri Shortcuts programming language. Contribute to electrikmilk/cherri-macos-app development by creating an account on GitHub.  ( 8 min )

企业微信开放平台命令行工具 — 让人类和 AI Agent 都能在终端中操作企业微信. Contribute to WecomTeam/wecom-cli development by creating an account on GitHub.  ( 5 min )

The official Lark/Feishu CLI tool, maintained by the larksuite team — built for humans and AI Agents. Covers core business domains including Messenger, Docs, Base, Sheets, Calendar, Mail, Tasks, Me...  ( 22 min )

Ghostty-based macOS terminal with vertical tabs and notifications for AI coding agents - manaflow-ai/cmux  ( 22 min )

No content preview

PasteMemo - Smart clipboard manager for macOS. Contribute to lifedever/PasteMemo development by creating an account on GitHub.  ( 13 min )

A beautiful SSH task runner. Contribute to spatie/scotty development by creating an account on GitHub.  ( 8 min )

开源微信 Bot 管理平台 + App 应用市场 | Self-hosted WeChat Bot Platform with App Marketplace | Lark · Slack · Discord · DingTalk · GitHub · Notion · 20+ Apps | AI Tools | 7 Language SDKs - openilink/openilink-hub  ( 24 min )

Terminal games built with the Symfony TUI component - GromNaN/symfony-tui-games  ( 6 min )

Daily GitHub Trending radar with Chinese recommendations and SMTP delivery. - Hazel-Lin/github-trending-radar  ( 30 min )

🔎 Artisan Browse provides an interactive terminal UI for discovering, searching, and executing all Laravel Artisan commands. - joshembling/artisan-browse  ( 11 min )

FireRed-OpenStoryline is an AI video editing agent that transforms manual editing into intention-driven directing through natural language interaction, LLM-powered planning, and precise tool orches...  ( 29 min )

A robust, configuration-driven ETL and data import framework for Laravel. Handles CSV/Excel streaming, queues, validation, and relationships. - zappzerapp/laravel-ingest  ( 14 min )

Fast PHP language server with deep type intelligence. Generics, Laravel, PHPStan annotations. Ready in an instant. - AJenbo/phpantom_lsp  ( 10 min )

⚡️ A Laravel package to prevent circuit breaker. Contribute to algoyounes/circuit-breaker development by creating an account on GitHub.  ( 9 min )

macOS application designed to visually compare folders and files - visualdiffer/visualdiffer  ( 8 min )

macOS application designed to visually compare folders and files - visualdiffer/visualdiffer  ( 8 min )

tiny AI companions that live on your macOS dock. Contribute to ryanstephen/lil-agents development by creating an account on GitHub.  ( 6 min )

Feature-packed native macOS screenshot tool: annotate, auto-redact PII, record GIFs, OCR + translate, scroll capture, beautify, and more. No Electron, no subscription. - sw33tLie/macshot  ( 15 min )

Aria2App is an advanced download manager based on aria2 that can handle remote servers too. - devgianlu/Aria2App  ( 3 min )

No content preview

No content preview

GitNexus: The Zero-Server Code Intelligence Engine - GitNexus is a client-side knowledge graph creator that runs entirely in your browser. Drop in a GitHub repo or ZIP file, and get an intera...  ( 38 min )

CLI tool for inspecting and managing services listening on localhost ports - raskrebs/sonar  ( 18 min )

No content preview

Fast, scriptable CLI for the App Store Connect API. Automate TestFlight, builds, submissions, signing, analytics, screenshots, subscriptions, and more. JSON-first, no interactive prompts - rudrankr...  ( 26 min )

A one-stop video subtitle translation tool. Contribute to cddqssc/Caption-Trans development by creating an account on GitHub.  ( 5 min )

Contribute to tiajinsha/JKVideo development by creating an account on GitHub.  ( 11 min )

🔮 A bot to automatically update and merge GitHub PRs - chdsbd/kodiak  ( 7 min )

🎃 A fast, out-of-the-box terminal built for AI coding. - tw93/Kaku  ( 12 min )

1000UserGuide：对独立开发者和创业者来说，找到前1000个早期用户太关键了。这里精心整理了300多个国内外渠道，适合独立开发者和创业者推广产品的渠道。 - naxiaoduo/1000UserGuide  ( 15 min )

Native macOS App Store Connect tool with MCP. Submit iOS apps to App Store with AI agents - blitzdotdev/blitz-mac  ( 18 min )

纯粹的谷歌浏览器截图插件，支持三种截图模式. Contribute to Bitkoala/KoalaSnap development by creating an account on GitHub.  ( 17 min )

要让Jackson直接使用类的字段（Field） 进行序列化和反序列化，而非默认的getter/setter方法，核心是修改Jackson的属性访问策略：关闭对getter/isGetter/setter的自动检测，开启对所有字段（无论访问修饰符）的检测。 下面分通用Java项目、Spring Boot项目、局部类注解 […]  ( 4 min )

A premium macOS DMG creation tool with live-rendered SwiftUI backgrounds, glassmorphism, and Retina support. - saihgupr/DMGMaker  ( 6 min )

A full-featured download manager — rebuilt from the ground up - AnInsomniacy/motrix-next  ( 17 min )

Remote Control for Codex. Contribute to Emanuele-web04/remodex development by creating an account on GitHub.  ( 37 min )

No content preview

Laravel Pulse card that list all scheduled tasks. Contribute to hosmelq/laravel-pulse-schedule development by creating an account on GitHub.  ( 4 min )

The open-source screen recorder and editor for creating polished walkthroughs, demos, tutorials, and product videos. Add auto-zoom, webcam overlays, cursor animations and more to your screen record...  ( 21 min )

Easily generate Stripe/YouTube looking IDs for your Laravel models. - red-explosion/laravel-sqids  ( 15 min )

中文社区版OpenClaw，同原版保持定期更新，已内置钉钉、企业微信、飞书、QQ以及国内网络环境优化。你的专属个人AI助手。支持所有操作系统和平台。🦞 - jiulingyun/openclaw-cn  ( 10 min )

The PHP AST Viewer is a tool for viewing the Abstract Syntax Tree of PHP code. By visualizing the structure, it helps developers gain a deeper understanding of the code, thus improving code quality...  ( 4 min )

Security scanner built for Laravel, detects misconfigurations, vulnerabilities, and exposed secrets with a beautiful TUI. - Eljakani/ward  ( 34 min )

No content preview

Deck is a modern, native, privacy-first clipboard manager for macOS. - yuzeguitarist/Deck  ( 13 min )

A radial, gesture-first app switcher and file hub for macOS. - yuzeguitarist/Orbit  ( 18 min )

💡 Use the right package manager. Contribute to antfu-collective/ni development by creating an account on GitHub.  ( 21 min )

JavaScript in-page GUI agent. Control web interfaces with natural language. - alibaba/page-agent  ( 9 min )

JavaScript in-page GUI agent. Control web interfaces with natural language. - alibaba/page-agent  ( 9 min )

Get the main content of any page as Markdown. Contribute to kepano/defuddle development by creating an account on GitHub.  ( 23 min )

ClawX is a desktop app that provides a graphical interface for OpenClaw AI agents. It turns CLI-based AI orchestration into a desktop experience without using the terminal. China website is https:/...  ( 35 min )

High-performance N-dimensional arrays for PHP, powered by Rust - phpmlkit/ndarray  ( 14 min )

在macOS中文输入时使用英文标点。. Contribute to vaspike/MacEasySymbol development by creating an account on GitHub.  ( 10 min )

An AI coding agent by JetBrains that ships code from your terminal, IDE, or CI/CD pipeline - powered by any LLM you choose - JetBrains/junie  ( 10 min )

Passphrase generator with Laravel integration, inspired by Bitwarden. Uses the EFF long word list by default with support for custom wordlists. - nicobleiler/php-passphrase  ( 23 min )

ElasticSearch, OpenSearch and DynamoDB NoSQL GUI client for Mac, windows and linux - geek-fun/dockit  ( 11 min )

Easy peer to peer networking for Apple devices. Contribute to EthanLipnik/Loom development by creating an account on GitHub.  ( 21 min )

IronClaw is OpenClaw inspired implementation in Rust focused on privacy and security - nearai/ironclaw  ( 20 min )

A Rust compiler with ownership checking, written in PHP - mrconter1/rustc-php  ( 8 min )

Sub2API-CRS2 一站式开源中转服务，让 Claude、Openai 、Gemini、Antigravity订阅统一接入，支持拼车共享，更高效分摊成本，原生工具无缝使用。 - Wei-Shaw/sub2api  ( 32 min )

Rule Helper. Contribute to sourcetoad/rule-helper-for-laravel development by creating an account on GitHub.  ( 9 min )

核心方案：基于市场温度的动态定投系统 该方案通过 “场外增量控制” 与 “场内全权委托” 相结合，实现低成本积累筹码与高位风险对冲的闭环。 1. 核心数学模型 每月定投的总金额为M，根据当前市场温度T<

mactop - Apple Silicon Monitor Top. Contribute to metaspartan/mactop development by creating an account on GitHub.  ( 25 min )

No content preview

🎙️Voice input and translation app for macOS. Press to talk, release to paste. - hehehai/voxt  ( 37 min )

Petal is a native macOS menu bar app for fast, local-first audio transcription. - Aayush9029/petal  ( 5 min )

MacOptimizer is a system optimization tool designed specifically for macOS, featuring a modern SwiftUI interface with eight core functional modules: - ddlmanus/MacOptimizer  ( 27 min )

No content preview

A terminal-based SSH manager inspired by lazydocker and k9s - Written in go - Adembc/lazyssh  ( 18 min )

No content preview

🎉 Desktop Spotify Web Player Instant Synchronised Lyrics - mantou132/Spotify-Lyrics  ( 9 min )

Contribute to MrPunyapal/pest-annotator-plugin development by creating an account on GitHub.  ( 14 min )

OpenCode mobile client via Telegram: run and monitor AI coding tasks from your phone while everything runs locally on your machine. - grinev/opencode-telegram-bot  ( 20 min )

今天看到网站下方显示的域名已经整整十年了，查看了下域名信息，才发现原来真的已经十年了。 域名信息 域名年龄：10年0个月 注册时间：2016/2/24 当初只是随手玩玩的域名，没想到居然一直坚持了十年。 记得还是2015年，那会儿刚来北京，工作也比较清闲，就想着自己写一个网站。 第一版博客还是用的Java写的，ssh框 […]  ( 3 min )

Transform natural language into shell commands instantly. No dependencies, no complex setup - just type what you want and get the command you need. - matheusml/zsh-ai  ( 10 min )

A lightweight macOS selection translator built with pure Swift 6, featuring on-device Apple Translate for privacy, only 5MB install size and stable ~50MB memory usage. 一款轻量级 macOS 划词翻译工具，纯 Swift 6 ...  ( 12 min )

An open-source userscript manager for Safari. Contribute to quoid/userscripts development by creating an account on GitHub.  ( 33 min )

Contribute to saif0200/notchprompt development by creating an account on GitHub.  ( 6 min )

The Testing Framework. Contribute to php-testo/testo development by creating an account on GitHub.  ( 7 min )

Utility to execute a command as another user. Contribute to sudo-project/sudo development by creating an account on GitHub.  ( 11 min )

A git diff pager based on delta but with a file tree, à la GitHub. - dlvhdr/diffnav  ( 12 min )

Textream is a free macOS teleprompter app for streamers, interviewers, and presenters. It highlights your script in real-time as you speak, displayed in a beautiful Dynamic Island overlay. With ext...  ( 28 min )

Contribute to medvio/move-to-screen development by creating an account on GitHub.  ( 4 min )

GitHub Agentic Workflows. Contribute to github/gh-aw development by creating an account on GitHub.  ( 8 min )

Tiny, Fast, and Deployable anywhere — automate the mundane, unleash your creativity - sipeed/picoclaw  ( 54 min )

Create a Laravel command for your OpenAPI specs. Contribute to spatie/laravel-openapi-cli development by creating an account on GitHub.  ( 7 min )

Take screenshots of web pages. Contribute to spatie/laravel-screenshot development by creating an account on GitHub.  ( 7 min )

No content preview

在AI编程工具百花齐放的当下，Claude Code凭借其强悍的代码能力、灵活的终端交互模式以及深度的开发流程融合能力，成为了开发者提升效率的核心利器。它远不止是一个简单的AI写代码工具，更是重构了人机协作编程的新方式，从基础的代码编写、错误修复，到复杂的项目管理、自动化工作流搭建，都能提供全方位的支持。 本文将从基础 […]  ( 6 min )

Keep your PHPUnit mocks under strict eye of static analysis - rectorphp/mockstan  ( 9 min )

Quick Look extension for highlight source code files on macOS 10.15 and later. - sbarex/SourceCodeSyntaxHighlight  ( 47 min )

macOS Quick Look extension for Markdown files. Contribute to sbarex/QLMarkdown development by creating an account on GitHub.  ( 28 min )

No content preview

No content preview

A PHP documentation testing tool that extracts code blocks from Markdown files, executes them in isolated processes, and verifies their output. - TestFlowLabs/doctest  ( 7 min )

NewsGoat is a terminal-based RSS reader. Contribute to jarv/newsgoat development by creating an account on GitHub.  ( 26 min )

Peekaboo is a macOS CLI & optional MCP server that enables AI agents to capture screenshots of applications, or the entire system, with optional visual question answering through local or remot...  ( 10 min )

📦 A macOS utility to seamlessly migrate applications to external storage and reclaim local disk space.【一款 macOS 工具，无缝迁移应用到外部存储并自动建立链接，释放宝贵的本地空间】 - wzh4869/AppPorts  ( 12 min )

The Laravel AI SDK provides a unified, expressive API for interacting with AI providers such as OpenAI, Anthropic, Gemini, and more. - laravel/ai  ( 5 min )

macOS accessibility inspector for AI-agent workflows. Hover, annotate, copy structured output. - smughead/Loupe  ( 6 min )

Terminal emulator designed for elementary OS. Contribute to elementary/terminal development by creating an account on GitHub.  ( 5 min )

A static analysis tool for GraphQL. Contribute to worksome/graphlint development by creating an account on GitHub.  ( 7 min )

Android client for paqet — Root Required — Ferries Packets Across Forbidden Boundaries. - AliRezaBeigy/paqetNG  ( 8 min )

GitHub Action that in a pull request updates the copyright year(s) in your license file. - FantasticFiasco/action-update-license-year  ( 31 min )

portalDB的是Apollo Portal（配置管理门户）的专属关系型数据库，是支撑Portal前端所有管理操作的核心数据底座——所有通过Portal页面进行的配置管理、权限控制、项目组织等操作的元数据、权限数据、操作日志，都会持久化存储在portalDB中。 同时要注意，Apollo的portalDB和负责存储具 […]  ( 3 min )

org.apache.ibatis.plugin.Interceptor 是 MyBatis 插件体系的核心扩展接口，MyBatis 提供的插件机制完全基于这个接口实现——它允许开发者通过动态代理拦截 MyBatis 核心执行流程中的关键方法，在不修改框架源码的前提下，对 MyBatis 的执行逻辑做自定义增强（如分页 […]  ( 6 min )

A Slipstream Client for android. Contribute to VortexOne404/slipstream-client-android development by creating an account on GitHub.  ( 12 min )

A 5-20x faster experimental Homebrew alternative. Contribute to lucasgelfond/zerobrew development by creating an account on GitHub.  ( 6 min )

A beautiful error page for PHP applications. Contribute to bim-g/hitilafu development by creating an account on GitHub.  ( 9 min )

"Paste once, run once." — Trimmy flattens those multi-line shell snippets you copy so they actually paste and run. - steipete/Trimmy  ( 11 min )

Terminal session manager for AI coding agents. One TUI for Claude, Gemini, OpenCode, Codex, and more. - asheshgoplani/agent-deck  ( 28 min )

Library for filtering files. Contribute to sebastianbergmann/file-filter development by creating an account on GitHub.  ( 3 min )

Open-source macOS screen recording app with auto-zoom, cursor effects, and timeline editing — Screen Studio alternative. - syi0808/screenize  ( 9 min )

Native macOS menu bar app to use your Claude Code & ChatGPT subscriptions with AI coding tools - no API keys needed - automazeio/vibeproxy  ( 14 min )

Your own personal AI assistant. Any OS. Any Platform. The lobster way. 🦞 - GitHub - openclaw/openclaw: Your own personal AI assistant. Any OS. Any Platform. The lobster way. 🦞  ( 62 min )

A Laravel AI wrapper package with multiple provider drivers. - devcbh/laravel-ai-provider  ( 16 min )

Circuit breaker for Laravel queue jobs. Contribute to harris21/laravel-fuse development by creating an account on GitHub.  ( 26 min )

🔍 Fuzzy search for PHP, ported from Fuse.js. Contribute to loilo/Fuse development by creating an account on GitHub.  ( 31 min )

在Java开发的日常场景中，导入语句似乎是绕不开的“小麻烦”。新手会困惑为何String无需导入而List必须手动引入，资深开发者也常会在切换业务场景时，为查找Stream类的包路径、处理Date类的同名冲突而卡顿。传统导入方式的臃肿与繁琐，让不少开发者羡慕Python“开箱即用”的模块机制。而JDK 25正式落地的J […]  ( 4 min )

No content preview

No content preview

最近一直在思考一个问题，AI时代，个人技术博客的出路在哪里？ 有了AI之后，似乎一切技术问题都能迎刃而解。想要了解什么知识，AI总是能给你最全面的回答；想要搭建一个项目，AI能给你生成完整的代码和架构图；项目中遇到什么困难，不用上网到处找解决办法，AI几秒钟就能给你N多个解决方案…… 那么个人技术博客存在的意义在哪里呢 […]  ( 3 min )

No content preview

Multi-platform SDK for integrating GitHub Copilot Agent into apps and services - github/copilot-sdk  ( 14 min )

Laravel package for in-memory driver database Eloquent model connections - waadmawlood/truffle  ( 15 min )

Complete Claude Code configuration collection - agents, skills, hooks, commands, rules, MCPs. Battle-tested configs from an Anthropic hackathon winner. - affaan-m/everything-claude-code  ( 24 min )

IDEA插件：在文件树、行末、JSON、COBOL 显示注释. Contribute to LinWanCen/show-comment development by creating an account on GitHub.  ( 10 min )

No content preview

Meta Server（元数据服务器/元服务器）是负责集中管理元数据、协调分布式系统或整合多源信息的核心组件，核心价值在于统一视图、提升效率与保障一致性，广泛应用于分布式存储、网络服务、游戏等场景。以下从核心定义、核心功能、典型应用场景、关键设计要点及实例展开说明。 一、核心定义 Meta Server的核心是作为“元 […]  ( 3 min )

Organize pages in the sidebar in order to make navigation between pages more comfortable. - aymanalhattami/filament-page-with-sidebar  ( 19 min )

Humanizer 的汉化版本，Claude Code Skills，旨在消除文本中 AI 生成的痕迹。 - op7418/Humanizer-zh  ( 16 min )

中文写作排版风格指南，一套中文和中英文混排的写作和排版标准。Chinese Style Guide, a set of standards for writing and formatting articles in Chinese language and Chinese-English mixed content - RightCapitalHQ/chinese-style-guide  ( 18 min )

PostBot 内容同步助手 一款开源的多平台内容同步分发生产力工具。 支持将文章、笔记、动态、图片、视频、音频等内容，一键同步发布至主流媒体平台。覆盖微信/微博/今日头条/小红书/知乎/百家号/企鹅号/视频号/抖音/快手/哔哩哔哩（B站）等国内主流媒体平台，可轻松扩展兼容 X（Twitter）、Facebook、Instagram、TikTok、YouTube、LinkedIn 等国际媒体...  ( 11 min )

A set of additional rules for phpstan/phpstan. Contribute to guanguans/phpstan-rules development by creating an account on GitHub.  ( 7 min )

🎬 火宝短剧 - 基于AI的一站式短剧生成平台 《一句话生成完整短剧，从剧本到成片全自动化》 Huobao Drama - An AI-Powered End-to-End Short Drama Generator "One Sentence to Complete Drama: Fully Automated from Script to Final Video" ...  ( 34 min )

A modern, enterprise-ready state machine implementation for Laravel Eloquent models. - HPWebdeveloper/laravel-stateflow  ( 51 min )

中文博客琅琊榜，只收录精品独立博客. Contribute to qianguyihao/blog-list development by creating an account on GitHub.  ( 255 min )

No content preview

Contribute to osama-98/laravel-exports development by creating an account on GitHub.  ( 30 min )

在数据库日常运维中，表结构变更（尤其是新增字段）是高频需求。对于数据量几千、几万的小表，一条简单的ALTER TABLE语句便能瞬间完成，几乎不会对业务造成影响。但当表数据量突破千万级甚至上亿级时，直接执行ALTER TABLE就可能成为一场“灾难”——长时间锁表导致读写阻塞、主从延迟飙升、应用连接池耗尽，最终引发订单 […]  ( 5 min )

核心结论先明确：在绝大多数数据库（MySQL/Oracle/PostgreSQL等）中，二者的性能几乎完全一致，因为数据库优化器会将 BETWEEN AND 自动等价转换为 >= + <= 的组合。性能差异的关键不在运算符本身，而在「是否能利用索引」「范围边界的合理性」。 一、底层逻辑：BETWEEN AN […]  ( 3 min )

A cross-platform desktop All-in-One assistant tool for Claude Code, Codex, OpenCode & Gemini CLI. - farion1231/cc-switch  ( 28 min )

Lightweight preview update in your repository. Contribute to TheDragonCode/github-preview-updater development by creating an account on GitHub.  ( 15 min )

PHPStan extension for validating php-language-extensions - DaveLiddament/phpstan-php-language-extensions  ( 6 min )

Contribute to NickSdot/phpstan-phpstorm-error-identifiers development by creating an account on GitHub.  ( 5 min )

Browser automation CLI for AI agents. Contribute to vercel-labs/agent-browser development by creating an account on GitHub.  ( 48 min )

在高并发场景中，线程间的数据交换是核心环节，而队列作为数据缓冲的关键组件，其性能直接决定了整个系统的吞吐能力和延迟表现。传统并发队列在面对超高吞吐需求时，往往受限于锁竞争、内存开销等问题，难以满足金融交易、高频计算等极端场景的要求。Disruptor 作为一款高性能无锁队列框架，通过创新的设计理念和底层优化，成功突破了 […]  ( 4 min )

Monitors a php app and send notifications on certain error/exception/resource-exhausting/custom events - staabm/SysMonitor  ( 5 min )

Free Weather Forecast API for non-commercial use. Contribute to open-meteo/open-meteo development by creating an account on GitHub.  ( 15 min )

SARIF formatter for PHPStan. Contribute to jbelien/phpstan-sarif-formatter development by creating an account on GitHub.  ( 6 min )

cs2pr - Annotate a GitHub Pull Request based on a Checkstyle XML-report within your GitHub Action - staabm/annotate-pull-request-from-checkstyle  ( 15 min )

轻量级微信读书客户端，体积小，页面宽，自动翻页，体验增强 A Weixin Reader Desktop with Auto-turning and Wide-pager - dengcb/weixin-reader-desktop  ( 17 min )

No content preview

install GitHub release binaries from the CLI interactively - redraw/gh-install  ( 3 min )

PSR-7 & PSR-15 middleware to add the Server-Timing header - tuupola/server-timing-middleware  ( 10 min )

在Linux系统中，文件系统的层次结构严格遵循Filesystem Hierarchy Standard（FHS）规范，根目录 / 作为整个文件系统的"基石"，承载着系统运行、软件部署和数据存储的核心功能。每个子目录都有明确的职责划分，理解其背后的设计逻辑和实际用途，是运维工程师高效管理系统、开发者 […]  ( 6 min )

本期共有 39 个项目，包含 C 项目 (2)，C# 项目 (3)，C++ 项目 (1)，Go 项目 (4)，Java 项目 (2)，JavaScript 项目 (5)，Kotlin 项目 (2)，PHP 项目 (1)，Python 项目 (5)，Rust 项目 (2)，Swift 项目 (1)，人工智能 (5)，其它 (6)  ( 10 min )

本期共有 40 个项目，包含 C 项目 (1)，C# 项目 (2)，C++ 项目 (4)，Go 项目 (4)，Java 项目 (2)，JavaScript 项目 (5)，Kotlin 项目 (2)，PHP 项目 (1)，Python 项目 (5)，Rust 项目 (2)，Swift 项目 (2)，人工智能 (5)，其它 (5)  ( 12 min )

昨天刷知乎，看到一个问题：“如何评价「AI就是美国的城投」的观点？” 这个问题下提供了一张图，也就是最近在互联网上被广泛传播的那张描绘AI投资关系的网络图，它展示了科技巨头 Nvidia、OpenAI、Microsoft 和Oracle 之间错综复杂的资金流动和合作，号称“AI永动机”。有人据此质疑，这种完全的“内循环”是否构成了一个巨大的资产泡沫，并将AI比作“美国的城投”，最终可能崩盘。

最近一段时间，老T在知乎上连续看了十几个关于康熙与洪承畴方面的问题以及数百个回答，粗略估计下，累计阅读字数可能超过 30 万字，对于这个可以称之为当代影响最大的“民间发明历史”现象，也有着自己的一点感悟。这里老T不再重复探讨到底康熙是不是洪承畴的儿子这件事，只是就这个事件本身提出一些自己的看法。

Ansible Job Platform addresses key IT automation challenges with a centralized, secure, and scalable solution. It integrates CMDB, security credentials, automation orchestration, and job scheduling, ensuring transparency and compliance.  ( 1 min )

前言：这个系列（以及我的博客）好久不更新了，原因有两个，一个是我在学习用双拼打字，手跟不上脑子，写的东西读起来 […] Continue reading...  ( 4 min )

这个双十一，老T买了两台打印机：一台激光打印机，一台3D打印机。

最近我妈接到了一个自称“全省平安建设满意度调查”的电话，前两次都没接，直到第三次才拿起。她还打电话问我是不是诈骗。作为一名法律工作者，老T对这个事情倒是知情，自己前几年在带小孩去医院检查时，也曾被“调查”过，但毫无疑问，这种民意调查，在国内似乎还是个“稀罕事”，相比经常看到国外新闻中各种民调数据，可以说感知确实不那么深，如果不是自己和家人这两次被“调查”经历，还真不知道国内有这种“民调”。

前几天，一则“肇庆9个村划归佛山”的新闻，在老T朋友圈里刷了屏。不少人一看地图就乐了：这不就是规划了多年的广州新机场所在地吗？一时间，“广州出名、佛山出钱、肇庆出地”的老段子又被翻了出来。

我很早之前就一直在用 Hugo theme Stack主题 写东西。这款主题挺好的，但我觉得它更适合图片比较多的博客。而我现在的网站上面已经积攒了1700多张照片，在Stack主题的处理下，生成网站时会多出 7000到8000个文件！这就导致整个网站的仓库体积特别大，想部署到第三方免费平台（比如Cloudflare Pages）就变得很困难。

通过一个真正好用的免费在线拼图工具，结合公开构建和用户反馈，无需刻意SEO却意外获得高排名。本文分享了这一SEO成功的关键因素。

10月在本博上发的说说数量偏少，主要研究电子邮件更新Hugo博客，以及issue作为hugo博客发布端的可行性，顺便分享10万+微信公众号文章写作经历，以及新的VPS商zouter的使用情况。  ( 1 min )

探索一系列完全免费的在线图片处理工具，包括拼图、压缩、裁剪、头像制作、去水印等功能，无需登录，安全隐私，体验友好，满足各种图片处理需求。

昨天接小孩放学回家路上，女儿问了我一个灵魂问题，为什么我们家都在说普通话？是啊，老T全家人都来自湖南同一个县，但自从到了广东，都自觉说上了普通话，特别是老T自己，来广东18年，也整整说了18年普通话，总感觉哪里怪怪的。

最近老T因为此前使用哪吒探针V0版时所需要的通信域名将要到期，于是选择对探针进行升级。探针这种东西，虽然看起来是“个人畜无害”的小工具，但总体来说安全风险还是不小，万一被攻破，那探针所连接的服务器毫无疑问都会被“一锅端”。为此，老T在升级探针之余，也对这个探针增加了一点小防护。

最近老T翻旧文章，回忆起当年折腾QQ登录的日子，感慨万千。电话这东西，全球随便打，移动的号拨给联通，秒接通，毫无障碍。可聊天软件呢？微信、抖音、钉钉、微博、QQ等，个个都是 “围墙花园”，互相不搭理。那凭啥电话行，聊天软件就不行？ 电话为啥能随便打？ 电话系统能全球互通，表面上看是因为 标准化协议，比如国际电信联盟（ITU）定下的各种规矩，能够在技术上无缝衔接。但老T觉得，这背后应该还有更深层次的原因。 首先，电信运营商，尤其是国内几大运营商，基本都是 国企，比如中国移动、中国联通、中国电信。这些 “国家队”背负社会责任，不是单纯的商业公司。电话作为 基础设施，关系国计民生，必然要求互联互通。老T查了下电话历史，发现早年电话网络刚建时，也有运营商各自为战的情况，比如美国 AT&T 公司就曾禁止过其他电话公司接入其长途网络，但到后来发现 电话不互通，社会成本太高，用户怨声载道，监管层一出手，也就解决了。加上电信行业高度垄断，玩家少，协调起来相对容易，协议一签，网络一连，全球电话网就织成。 再者，电话是典型的 基础设施服务，类似水电煤，服务本质是“连接”，不以内容为核心竞争力。运营商赚的是通话费、流量费，客观上也没什么动力在内容上搞封闭。反过来，互联互通还能扩大用户群，增加收入。所以，技术标准+社会责任+垄断协调+基础设施属性，共同造就了电话的畅通。 老T想起当年2000年左右，家里刚装电话那会，在偏远农村用座机就可以打电话给远方的亲友，移动联通随便拨，完全没想过背后有多复杂。这就是基础设施的魅力：你感觉不到它的存在，但它就是好用。 聊天软件为啥不互通？ 反观网络聊天软件，情况完全不同。微信、抖音、钉钉、微博、QQ等，个个都像独立王国，互不往来。表面上是 技术问题，协议不统一，加密方式各异，强行互通可能出安全漏洞。但老T觉得，根源还是 商业利益和生态封闭。 聊天软件公司靠 用户…

从实际需求出发，落地中大型企业域名证书管理系统。已免费开放，支持私有化部署  ( 1 min )

近年来，静态博客的发布方式层出不穷，但许多方法要么复杂，要么难以长期坚持。本文介绍一种简单高效的方式：通过 GitHub Issue 作为 Hugo 博客的发布端，利用 GitHub Actions 自动将 Issue 转换为 Hugo 内容并部署到 GitHub Pages。这种方法尤其适合喜欢用 GitHub 移动端 App 随时随地发布博客的用户。本教程基于老 T 的实践经验，解决了私有仓库图片下载、标签提取等问题，适合公开和私有仓库。  ( 7 min )

最近几年来，老T在静态博客发布这件事情上也算是走了无数弯路。随便搜下老T博客，关于Hugo的数篇文章，无一不是聚焦这个问题。但细数以前的经历，不管是哪种方法，很少有能让老T坚持一个星期以上还主动愿意持续用下去的。大多数时候用它们都是处于一种迫不得已的”窘境“。如今，终极解决方案来了。

前几天老T在微信群看到有人吐槽：“为啥路由器地址是192.168.1.1这么一串数字，记着费劲，不能弄个简单点的，比如1.2.3.4？” 这问题乍一听挺有意思，老T自己也用过不少路由器，但从来没想过 192.168.1.1 这个数字来源是什么，于是在网上查了查资料，顺便也问了下 AI。不过，在多方查找比对下，发现目前中文互联网上对这个问题的解答，似乎都缺点什么，感觉不是很准确。这里，老T将自己考证的结论总结出来，争取还原出历史的真相。

好拼是一款免费的在线拼图工具，现已上架Chrome和Edge插件商店，支持在线网页、PWA和浏览器插件三种使用方式，简洁方便，欢迎体验。

前些天老T在知乎简单回答了个关于悬索桥的问题，虽然赞同的不少，但评论中质疑者也很多，主要争论的就是悬索桥主缆损坏是否可维修更换。由于老T本身对桥梁建设一窍不通，只能以最基本的文字阅读能力，通过查找严肃的文献资料，争取将这个事情说清。

本文介绍了「好拼」免费在线拼图工具的核心功能和最新更新，包括布局拼图和长图拼接，以及新增的Chrome插件，旨在为用户提供更简单、自然的创作体验。

昨天离开了江西，进入浙江开化县，体验小城生活后，连夜赶回上海。文章记录了从开化到上海的最后一站，分享了自驾经历、花费和房车使用心得。

最近，老T有幸回到自己母校佛山大学，现场聆听风靡网络的罗翔老师过来讲课，感觉内心总有一种莫名的思绪在翻滚，但也不知道该怎么描述，于是，只好想到啥就说点啥，并就此记录下来。 大家都是法盲 从事法律工作十几年来，老T时常感到感到无助，怀疑是不是因为自己能力水平太差，是不是因为自己当年读的大学层次不够，总是出现本领恐慌甚至有点自卑，面对同行的专业素养和能力，经常手足无措。有时候面对新问题不知如何下手，面对困难问题不知从哪切入，面对一些无可奈何的事情，总是内心挣扎。 这次，罗翔老师讲课的地点在佛山大学仙溪校区，也是我大学最开始的地方，在这里，我度过了人生中最为轻松快乐的时光。或许也是因为大学时过的太过安逸、舒适，以至于后来进入社会不断遭受挫折，由“从从容容、游刃有余”，逐渐演变为“匆匆忙忙、连滚带爬”。不过，无论如何还是庆幸自己当年高考选择了佛大，选择了法学这个专业，虽然毕业后磕磕绊绊，四处碰壁，但最终，又回到了佛山，重拾了法律工作。 在罗翔老师的授课中，贯穿整场的一条主线是“我是法盲”“大家都是法盲”。他列举了大量刑事司法实践案例，这些案例中绝大多数都是涉及各类部门法，包括行政法、经济法、自然资源与环境保护法、国际法等等，每当提到这种跨部门法的案例时，他都会不由自主拉长音调，自嘲自己真是个“法盲”。 的确，目前我国法律体系之庞大，毋庸置疑。比如他讲到一些“购买什么品种鹦鹉构成犯罪？” “用弹弓打麻雀在什么时间、什么地点构成犯罪？”“自己网上海淘购买哪些药品会构成犯罪？”“薅联程飞机票羊毛是否构成犯罪？”“无证办广场舞培训班是否构成犯罪？”等等。 事实上别说一般法律工作者，就算专门研究这些领域的学者，也很难在短时间内得出答案，甚至很多时候，研究得出的结论还可能是错误的，因为部门法之庞大，有些时候不仅涉及国家层面规定，还得考虑省级以下甚至县一级行政机构的文件，而研究过程很难…

探索开化县，一个宁静的小城，体验根雕艺术和开化纸的历史，漫步马金溪畔，欣赏自然美景和独特建筑，享受一场远离喧嚣的旅行。

本次沿海自驾的第57天，挑战难度极大的三清山徒步。从外双溪索道开始，历经西海岸、阳光海岸、南清园等景区，感受三清山的壮丽与秀美，体验高空栈道的惊险与震撼。

从鹰潭到上饶的沿海自驾之旅，体验龙潭湖公园的宁静与美丽，品尝当地美食如黄老头小炒黄牛肉，参观上饶博物馆和双塔公园，感受江西的自然与文化魅力。

从江西抚州到鹰潭的自驾之旅，探索了龙虎山的道教文化，游览了上清古镇，品尝了地道的牛骨粉，体验了不一样的江西风情。

本文记录了沿海自驾游第54天在抚州的行程，参观了抚州博物馆、文昌里历史文化街区，品尝了江西特色米粉，体验了抚州的历史文化魅力。

赣州Day53旅行记，探索福寿沟博物馆、赣州古城墙、郁孤台等地标，体验赣南客家文化，记录一家三口的愉快旅程。

最近在 Slate 上看到一个名叫 Giovanni Garcia-Fenech 的作者以《美国吃掉了我的名字》为题，写了一篇很有意思的文章。刚好，老T前些天也简单写了个 10月1日即将实施的K签证对普通人有什么影响 分析可能出现的外国移民现象，不由得开始思考，是否外国人到中国时也会遇到“吃掉名字”这种问题。 美国吃掉了我的名字 这篇《How the Government Ate My Name》文章作者出生于墨西哥，全名有四个名字 Leonel Giovanni García Fenech，这在西班牙语国家是标准起名方法，体现了家族传承和历史文化。 但作者在移民美国后，他的名字因美国官僚系统而被简化，在入籍时去掉一个名字，在驾照和护照上名字出现顺序混乱，甚至被缩写为“Giovanni F. Garcia”。 作者后来从美国移居西班牙时，又因文件上的名字不匹配而无法获取墨西哥护照。作者幽默地描述了这些经历，强调名字作为身份载体的脆弱性，以及在不同国家因为名字变更时所带来的不便。 最让老T感到惊讶的是，在这篇文章的末尾，作者提到的一个小故事： 作者在西班牙的一个小店看到一个来自中国的上了年纪的老板娘，这个老板娘的名字居然叫“Lola”（萝拉），而这个“Lola”名字的由来，仅仅是因为此前这个店里有一个名叫“Lola”的小姑娘在此工作过，有客人经常上门就问“萝拉在哪？”于是这个中国老板娘就在一声一声的询问中，变成了“萝拉”。 其他国家的情况 美国的情况主要是基于英文环境，不支持 ASCII （英文字母及其标点符号）以外的字符，导致 ü、ä、ß 等字母都被替换。对长名字也存在文化抵触，即便是姓名中有三个词，也总是拧巴的将中间那个省略掉或者简写。例如，奥巴马的 Barack Hussein Obama 绝大多数时候都只写 Barack Obama 或者简写成 Barack H.…

在2025暑假沿海自驾游的第52天，作者从梅关古道来到赣州，探索了赣州博物馆和江南宋城，体验了赣州的历史文化和美食，与家人共享旅程的温馨。

在2025年暑假的第五十一天，作者探索了梅关古道，体验了古代交通要道的历史和文化，同时欣赏了赣州和大余县的自然与人文景观。

这个国庆长假，跟往年过的别无二致。最近10来年，老T除了个别年份的国庆长假期间有外出旅游，其余时间大多是自驾车辆往返于湘粤“两个家”之间。 如果一定要说有什么不同，那这次主要区别就是因家里出行人数较多，5座小车坐不下，又单独给家人买了火车票。正是这次买火车票的经历，让老T重新认识了巴士这种交通工具。 曾经的天坑 讲起巴士，老T身上就有一把辛酸史。高中时，从乡里乘车去县城的中巴车上四五次被偷学费、生活费；大学时，从乡里去衡阳火车站的路上也被偷过生活费、手机。就连不经过火车站中转的长途巴士也被骗过。2012年春运，老T花费320元购买了一张衡山到佛山的长途车票，结果车子一到花都，司机就想把乘客赶下车。也就幸好当年老T用的黑莓手机已经有带GPS定位的谷歌地图，最终据理力争下，才让司机把车开进了广州市区。 回看这段经历，老T感觉恍如隔世，2012年至今不过13年。电子地图这种新鲜事物在当时大多数人手机上还未普及，而如今，几乎所有人早已习惯电子地图这种基础设施般的服务所带来便利性。也正是这种便利性，让曾经的很多“坑”都被填平了。 现在是真香 由于老T决定帮家人买火车票时已邻近国庆，只剩两天时间，不管是高铁票还是普铁票，几乎都出售殆尽，只剩一些普铁无座票还有售。没办法，也只能硬着头皮买张广州白云站到长沙火车站的无座票先顶着。毕竟时间也不长，总比超载挤在小车里熬十几个小时堵车要强。 比较头疼的是进出火车站的交通工具选择。两头的火车站离家都挺远：家里到广州白云站有60公里，从长沙火车站回老家有140公里（正常是到衡阳下，但衡阳没票）。以往常规方法是选择两边的汽车站到汽车站，然后再从汽车站转火车站。 但老T在微信上搜索发现，现在的巴士，居然已经不需要去汽车站了，而是可以直接在附近公交站上车，甚至有的巴士还能直接上门接送，跟网约车、顺风车没任何区别。 最关键的是，这种巴士还特别便宜。老T随手一查，发现广州到深圳的这种巴士居然有长期只卖19.9元的，甚至还有卖9.9元的。 看到这，老T不禁对此产生了兴趣，想了解下这背后到底有什么原因。 变化的原因 老T一边刷着各种微信小程序，一边在网上搜罗资料。 发现，这几年城际巴士的“变身”，主要得益于互联网平台的强势介入。 过去，巴士运营大多靠传统的汽车站和固定线路，效率低下，成本高企。像老T老家县城的公交公司，前几年亏损后停运了很长一段时间，顺带逼得县城到长沙的热门线路都被砍了一大半，结局就是越降本增效，越进入恶性循环，最终只能全部推倒从头再来，靠着将农村中巴线路全部整合进去，才重新盘活。 现在，各种互联网平台依托大数据和算法，通过优化线路和调度，能实现点对点甚至上门接送的服务。乘客不用再大老远跑到汽车站排队买票，直接在手机上选好时间和地点，巴士就近上车，省时省力。 更重要的是，价格为什么这么低？老T觉得，这一方面应该是平台之间的“补贴大战”，为了抢占市场份额，往往通过烧钱补贴来吸引用户，比如五六十公里的线路，长期保持在19.9元甚至9.9元这种“白菜价”。另一方面，大巴电动化所带来的运营成本的降低也功不可没。随着电动巴士的普及，油费支出大幅下降，加上共享经济模式，车辆利用率更高，空载率低，自然能把价格压下来。 老T看到一篇报道，说这种城际巴士的低价策略，其实并不意味着不赚钱。连顺风车都有全职开的，能载几十人的大巴车，平摊下来成本确实也很低。 想想看，高铁虽然快，但票价动辄几百；自驾油费加高速费也不少；飞机更别提了。巴士这种“老古董”借着科技东风，摇身一变，竟成了性价比之王。

从丹霞山出发，探访仁化县城与南雄市，体验锦江美景、古镇风情及美食，感受粤港澳大湾区的独特魅力。

这几天在知乎有个炒的火热的话题。国家即将开始颁发的K签证，引发很多人焦虑，是否外国人也要进来抢饭碗？ 老T从自己视角也试着简单回答了一下。 总体来说，我认为这个签证政策对普通人没啥影响。K签讲白了还是个技术移民签证，虽然国家层面还没有制定具体办法，但大概率是按照美国H1B模式的。 为啥现在要招外国人参与搞技术？ 重点还是服务国内制造业产业升级和关联产业需求。最主要是需要有懂第三世界国家语言、习俗、社会规范、消费习惯的人过来参与源头生产。 毫无疑问，美国IT互联网巨头里边大量印度裔员工，对美国科技公司开拓印度市场、理解印度用户需求、以及产品设计中融入某些印度文化元素，起到了重要的催化剂、加速器和桥梁作用。 中国制造想要有全球影响力，不是说在终端市场搞些研发中心就能解决问题，不可避免需要更多具有原生文化背景的外国人参与。 而这类职业生态位，与本国人力资源几乎没有竞争关系。 对普通人有什么影响？ 对普通人而言，K签证的实施几乎不会产生直接冲击。 它并非为了填补国内技术人才的“数量缺口”，而是精准地瞄准了“特定能力缺口”，即中国制造需要深度嵌入新兴市场本土语境（语言、文化、消费心理、社会网络）的稀缺能力。 这类引进人才的核心价值，在于帮助中国制造业和关联产业从源头设计、生产环节就精准锚定海外目标市场的真实需求，减少“水土不服”。这与本土人才在技术研发、生产管理、国内市场开拓等方面的核心能力是高度互补而非竞争的关系。 本质上，K签证是中国产业全球化深化到新阶段，为提升“本地化能力”和“源头创新适配性”而进行的一次战略性人才资源配置优化，其影响在于提升中国制造的全球竞争力与渗透力，而非与普通人争夺就业机会。 会不会引进洋垃圾？ 老T也很理解大多数人对这种问题的想法，动则关联到留学生问题。 但说到底，工作签证的前提是有工作，是国内企业愿意招外国人进来，是老板要真金白银给外国人发工资，需要解决企业内部沟通、管理问题，不可能有哪个老板请个“爹”来上班。如果不能发光发热，或者公司内部排斥导致其无法工作，那自然而然签证也就失效了。

探索丹霞山的核心景区——阳元石与长老峰，体验云崖栈道、九九天梯的惊险，欣赏嘉遁亭与观日亭的壮丽风景，感受大自然的鬼斧神工。  ( 1 min )

探索韶关的美丽风光，从韶阳楼的壮丽景色到丹霞山的自然奇观，感受韶关的历史文化和独特魅力。

在韶关自驾游的第47天，探索了马坝人石峡遗址和百年东街。深入了解马坝人的历史，欣赏韶关的自然美景和文化遗产，感受韶关的独特魅力。

最近受“桦加沙”台风影响，夜里气温不上不下，开空调嫌冷，不开又有点闷。老T想着省点电，就没开空调，靠电风扇吹了一夜。结果，省电不成，倒把自己吹感冒了。 起初就是个普通感冒，流了两天鼻涕。可没想到，鼻涕快消停时，嗅觉突然又没了！ 这已经是我第四次经历嗅觉丧失了。 第一次是2022年底阳了之后；第二次是2023年，感觉像是新冠回潮；第三次是去年冬天，症状也像新冠。但这次，我敢肯定不是新冠，用试纸测了，只有一条线。 这让我忍不住琢磨：搞不好前两次也不是新冠呢？ 因为除了第一次阳了和这次测了，中间那两次我其实都没做检测，当时只当是新冠了。现在想想，真有点细思恐极：难道得过新冠之后，再遇上普通感冒也会触发嗅觉失灵？ 作为医学门外汉，老T只好求助网络，想弄清楚感冒和嗅觉失灵之间到底有没有关联，也希望能给有类似经历的朋友一点参考。 要回答这个问题，得从科学角度切入。老T在 Science 上查到一项研究显示，新冠病毒对嗅觉系统的冲击远超普通感冒。它不像感冒那样主要是鼻塞堵路，而是直接“入侵”嗅觉上皮，也就是鼻子里的那些负责闻味儿的细胞。病毒通过 ACE2 受体进入支持细胞（sustentacular cells），导致炎症风暴，进而影响嗅觉神经元。这不是一次性的损伤，而是可能留下“后遗症”：慢性炎症或神经修复不全，让嗅觉系统变得脆弱。 由于老T看不懂这些具体内容，只能简单联想：新冠就好像一场“地震”，震坏了嗅觉基础结构，后续的小感冒则类似“余震”，轻易就能触发嗅觉失灵。 具体到复发机制，《英国医学杂志》上的 Post-viral olfactory loss and parosmia 文章指出，新冠患者嗅觉失灵的发生率高达 50%，其中许多人会发展为持久型或异嗅（闻到的味道变味儿），感染后6个月仍有27-60%的患者有嗅觉问题，1年后26.5-46%，甚至2年后还有8.3%。 为什么普通感冒会“借力打力”？因为感冒病毒（如鼻病毒）通常只引起暂时性鼻炎，但在前新冠患者身上，它可能放大残留的炎症。譬如，一篇案例研究描述了患者反复嗅觉失灵，可能源于过度炎症、再感染或病毒复制反弹。我的经历就很贴合：每次感冒后，不是鼻塞时就失灵，而是恢复期突然发作，这暗示不是机械堵塞，而是神经信号出了岔子。另一项研究进一步证实，后新冠嗅觉功能障碍与“长新冠”相关，涉及免疫细胞浸润和嗅觉上皮基因表达改变，导致系统对新感染更敏感。 当然，这其实也不是什么大事，绝大多数情况下，这种嗅觉失常现象，都能在一两个星期内逐渐恢复，尤其是年轻患者。 我自己这几天也在试着搞点恢复训练，闻闻柠檬、薄荷这些强烈气味，好像现在已经好很多了。 假如你也遇上类似情况，别慌，记录下来，或许像我一样，找找背后的科学依据，心里就能多几分踏实。 生活本就五味杂陈，闻不着时，就多尝尝回忆的味道吧。 参考文献： Olfactory and gustatory dysfunctions as a clinical presentation of mild-to-moderate forms of the coronavirus disease (COVID-19): A multicenter European study. https://link.springer.com/article/10.1007/s00405-020-05965-1 Post-viral olfactory loss and parosmia. https://pubmed.ncbi.nlm.nih.gov/37841969/ Long-lasting olfactory dysfunction in COVID-19 patients https://link.springer.com/article/10.1007/s00405-021-07153-1 Non-neuronal expression of SARS-CoV-2 entry genes in the olfactory system suggests mechanisms underlying COVID-19-associated anosmia. https://www.science.org/doi/10.1126/sciadv.abc5801 Prognosis and persistence of smell and taste dysfunction in patients with covid-19: meta-analysis with parametric cure modelling of recovery curves https://pubmed.ncbi.nlm.nih.gov/35896188/

从英德出发，沿北江一路自驾至韶关，途经G240国道和020乡道，欣赏北江画卷与喀斯特地貌的壮丽景色，体验京广铁路的繁忙与北江的宁静。

2025暑期沿海自驾游的第45天，沿北江行驶，途经清远、英德，探索上岳古村，欣赏岭南建筑，体验G240国道的北江画卷美景，记录了一段难忘的自驾旅程。

在佛冈县的上岳古村，探索了具有深厚文化底蕴的岭南古村落，感受了朱熹后裔的历史传承。随后在民安小镇体验了当地生活，等待台风的到来，度过了一个宁静的夜晚。

沿海自驾Day43，探访广州珠江沿岸与沙面岛，欣赏广州塔与海心沙的独特城市天际线，感受沙面岛的历史文化氛围，体验广州的历史与现代交融之美。

从深圳到广州的自驾旅程，雨中游览松山湖，错过华为欧洲小镇，与久违的广州朋友聚会，享受温暖的情谊与旅途中的每一刻。

在深圳体验了一场美妙的早茶，品尝了繁楼的多款广式点心，随后与朋友一起轻松爬上铁仔山，俯瞰深圳的城市风光，感受这座年轻城市的活力与包容。

从 9 月 1 日老T正式开始启用公众号算起，到今天刚好 10 天，这 10 天说短也短，说长也长。短的是，自 2008 年老T开始网络写作以来，这 10 天几乎可以忽略不计。长的是，这 10 天算是老T 17 年写作历程中收获最多的时期，充满了新鲜感和挑战。 微信公众号简要回顾 这 10 天里，老T公众号共更新 10 篇文章，其中 8 篇都是从博客搬到公众号的，只有 2 篇是近期新作。 在 8 篇搬迁的文章中，前 4 篇基本就是直接从博客复制粘贴，后 4 篇经过了大范围修订，几乎相当于重新写了一遍。 10 篇文章共计阅读量 2.2 万，其中最高的约 1 万，最少的 17。 另外，老T也收获了首个赞赏，来自安徽的网友 @Chulixia, 在此表示衷心的感谢！ 微信公众号与博客的区别 本来，老T以为微信公众号只是另一个信息发布平台，在已有文稿的基础上，简单发布即可。但由于微信天然的社交属性，这种想法很快就破灭了。微信公众号和博客看似都是写作平台，但本质上差异巨大，尤其是传播机制、读者互动和写作要求等方面。下面，老T就结合这 10 天的亲身经历，谈谈几点主要区别。 推荐算法与搜索排名 微信公众号的核心是推荐算法，它像一个无形的编辑，总在决定每篇文章能不能被推送到更多人面前。老T感觉这些天来，无时无刻不在被算法左右：文章标题要吸引人、内容要迎合读者口味、甚至发布时间都要卡点。 举个例子，老T的公众号目前总阅读量约 2.2 万，其中 90.4% 都来自算法推荐。如果算法不喜欢老T的文章，它就静静躺在后台，阅读量寥寥无几。 相比之下，博客更像一个安静的图书馆，主要面对搜索引擎。文章一发，就扔到网络大海里，由用户自己去搜索、发现。没有算法的“喜好”来左右，不管写的是专业知识还是个人心得，都只是共享出去，设置好关键词，等着有缘人来找。例如，老T的博客文章，很多都是 10 几年前写的，但通过搜索引擎，至今还能带来流量。 这种区别让写作心态完全不同。在微信上，老T得使劲琢磨“算法会不会推”，而在博客，老T只需问自己“这个内容有没有价值”就行。 订阅数 微信公众号的价值很大程度上取决于粉丝数。粉丝越多，推送就越稳，阅读基数越高。老T这 10 天，粉丝从 0 涨到 100+，主要靠算法推荐拉新。但目前粉丝基数很小，新文章容易石沉大海。 博客的订阅数（比如 RSS 或邮件订阅）则没那么关键。博客读者往往是零散的搜索流量，订阅只是锦上添花。老T的博客订阅者不多，但文章通过 SEO 优化，也能持续吸引新访客。简单说，微信更像是是“粉丝经济”，而博客一直是“内容长尾”，前者靠积累人气，后者靠积累内容。 另外，这里老T再补充另一个经常使用的知乎平台情况，完全是另外一套逻辑的。知乎既不靠粉丝数，也不靠搜索引擎，甚至也可以说不靠算法推荐。因为它默认使用算法将问题推送给所有潜在创作者和读者，创作者只需要沿着知乎推荐的问题进行回答即可获得大量自然流量，也算是另一种“算法平权”。例如，老T最近一个月在知乎的回答几个问题，累计阅读近 40 万，绝大多数阅读量都来源于用户被知乎问题所吸引而点进页面后查看的。不管是作为创作者还是用户，老T都完全不用担心算法或关注数的影响。 写作风格 这 10 天下来，老T感觉微信公众号的生态不适合太长、太专业的写法。读者大多用的是碎片化时间在阅读，刷朋友圈或订阅号时，更喜欢短平快的内容。老T花大力气写了几篇 5000 字以上的长文，比如分析印度强奸率、研究外国人如何看待93阅兵等，但阅读量和推荐最多的，反而是飞牛 NAS 相关的那些 2000 字以内的实用性文章。当然，这也可能因为内容差异，几篇长的文章，公众号后台审核后都明确不给推荐。 博客则更注重专业性和深度。没有什么限制，可以尽情展开，写成万字长文也没关系，因为读者是主动搜索来的，往往有特定需求。老T在博客上写的那些长篇内容，阅读量虽不高，但反馈很专业，读者留言讨论也更深入。微信像快餐，博客像大餐，前者求速，后者求精。 推荐限制的不确定性 微信的运营规范有 23000 多字四五百条内容，说是包罗万象也不夸张。但创作者并不知道这些具体规范的阈值，只有写完提交后，才能从助推审核结果中看出端倪。老T阅读量最少的那几篇文章，至今没搞清到底因为什么原因不给推荐，反正后台通知就一句话：“不符合微信公众平台运营规范”。 博客则没这些烦恼，只要遵纪守法即可，搜索引擎抓取后，更多的由用户自己判断。顶多是 SEO 优化的问题，但至少透明可控。 结语 总之，这10天让老T深刻感受到，微信公众号更像一场算法游戏，需要不断试错、优化。而博客则更适合个性化创作，注重长效价值。当然，两者不是非此即彼，老T会继续双管齐下，保持两边同步更新。希望这些心得对其他创作者有帮助。如果你也有类似经历，欢迎留言交流！

昨天在 QQ 群看到朋友说我码字已经超 100 万。确实，截止目前，Hugo 统计 359 篇文章，已合计 1036977 字。其实这个统计在到八九十万的时候，我还是比较关注，毕竟很快要到 100 万，心里还有点激动。但真等到后边恰好破百万时，又没怎么在意了。这让我想到心理学中的目标达成悖论：人们往往在追求目标时投入巨大情感，真正实现后却可能产生空虚感。 回顾码字经历 这次码字破 100 万，也许是个迟来的“成就”。毕竟，熟悉老T的老朋友，特别是从 2008 年开始一起加入 Blogfans 的老兄弟们，都知道我曾经多次大量删除过以往的文章。按照老T在 2017 年留下的老 Mysql 数据库统计，当时字数就已经有 114 万，只是当年并没有去想过这个问题。而如今留下的这些内容里边，属于 2017 年以前写的，占原本比例应该不足 1/4 。不过，不管怎样，这次破 100 万，对自己来说还是有一定纪念意义。 最初想法 回顾老T 2008 年以来码字的经历，我想正是印证了最初写下的那句 When each sunrise, we start a new. 虽然在当年，这句话更多表达的是新生之意，晨光初启、新途又始。到后来，随着日复一日的坚持更新，这句话逐渐演变为红日再升、新章复启。如今，站在新的起点，我更多的在想，这很可能又是一个破晓之时，需要境界新开了。 1 2 3 4 5 6 7 8 9 10 11 12 --- title: "您好, 世界!" slug: "hello-world" date: 2008-05-21 categories: ["技术"] tags: ["WordPress"] --- 经过十多小时从零开始的学习，我终于开通了自己的第一个网站。 在这里我要特别感谢抓抓，要是没有他在最后一刻发出的消息免费提供了本网站的…  ( 1 min )

2020年，老T曾以这个标题写过一篇简单的 文章，主要是因为当年在百度上搜索这个问题答案时，找到的都是千篇一律洗稿内容，鲜有能够讲清这个问题的文章。 甚至可以说，当时网上大多数关于这个问题的解答都是在瞎扯。有些讨论下边，还各种对提问者挖苦、讥讽，认为提问者丝毫不懂网络常识，但又不给出解决问题的思路办法。由于当年老T也恰好遇到这个问题，便进行一番研究。 现今5年过去，国内网络环境又与此前有了一些不同，千兆宽带已成标配，甚至万兆宽带也开始进入部分家庭，老T觉得有必要重新对这个问题进行梳理。 为何1000M光纤宽带，测速只有300M？ 当年老T遇到这个问题，首要原因就出在路由器上。 在1000M光纤刚开始进入千家万户的时候，由于多数家庭用户还在使用旧型号路由器，有些路由器虽然参数上写着支持 1000Mbps 网口，但那个网速只是供内网传输使用的，内外网之间的速度并不能达到 1000M。 也就是说，在内网配置没有明显问题的情况下，用户将两台电脑同时用网线连接到路由器时，这两台电脑之间的内网传输速度能够达到 1000Mbps。 而当用户通过这台路由器连接外网时，这时由于路由器内部芯片处理能力限制，传输速度就会骤减至 300-500Mbps。 老T 画了个示意图，争取把这个逻辑展示清楚。 5年前，不少家庭用户都是因为这个情况导致测速结果达不到运营商标称的速度。 例如： 网件 R7000 WAN TO LAN 最大速率：931 Mbps  网件 R6300v2 WAN TO LAN 最大速率：806 Mbps  华硕 AC68U WAN TO LAN 最大速率: 754.5 Mbps  TP-LINK TGR1900 WAN TO LAN 最大速率: 631 Mbps 网件 R6100 WAN TO LAN 最大速率: 93.1 Mbps 不过，当年老T 遇到的问题比这个还稍微复杂…  ( 1 min )

纪念中国人民抗日战争暨世界反法西斯战争胜利80周年阅兵仪式在昨日隆重举行。老话说，国之大事，在祀与戎，而和平年代的阅兵式，正是这两件事的完美结合。作为当今世界最受关注的事件，老T在网上冲浪过程中，也看到不少外国人的评价，其中很多都比较有意思，这里也简单跟大家分享一下。 俄罗斯人如何看待93阅兵 老T在日常混迹的多个技术论坛上都看到有人发贴讨论中国的阅兵式，其中最典型的评论就是：“看吧，中国果然没有向俄罗斯提供武器，不然战争局面早就改写了。” 由于在多个平台都看到这种评论，老T也感觉摸不着头脑。只能说，相比西方媒体这些年大肆宣扬的“中国威胁论”，西方民间对俄罗斯的恐惧感，确实是入心入肺了。 由于我们这一代人都在亲身经历中国工业现代化进程，对国内的制造能力早已有充分认识。哪怕按照最朴素的唯物主义思维也知道，一个常住人口和GDP都比不过广东的国家，制造能力显然跟中国有巨大差距。 但西方老百姓长期受媒体偏见影响，对中国的这种能力认识还是存在较大偏差。而此次阅兵展示，一定程度上也算是在帮他们纠偏。 不过，在这过程中，我也很好奇，就是作为这种评论里的另一方，俄罗斯人会如何看待中国阅兵呢？ 老T特意找了些懂俄语的朋友帮忙，也翻了翻俄罗斯两大社交平台 VK.com 和 Pikabu.ru，了解下到底俄罗斯普通网友有何观后感。 “大受震撼”的技术派 主要评论包括： “规模惊人！技术顶尖！” “正步完美，装备像科幻片” “机器狗和核导弹？中国准备好应对一切了” “水下无人机和机器人才是未来战争，中国AI武器领先了” “新型坦克配高超音速导弹…北约想清楚要和谁打了吗？” 照老T说，这反应在我们这边看来也还算实在。我们天天看着“下饺子” 新装备官宣，可能有点习以为常了。但对俄罗斯人来说，这次阅兵式上展示的装备，特别是高超音速、无人机、AI 这些新装备，确实刷新了他们的认知，甚至带点羡慕…

周日下午，老T准备在 PVE 装个 RouterOS 测试，结果发现无法创建虚拟机。具体表现是，在虚拟机创建页面，节点处提示 “Node epson seems to be offline” ，然后 PVE 面板上，节点显示灰色问号，提示状态为 “Unknown”。由此，开始了老T一段跌宕起伏的 Bug 修复过程。 PVE 存储设备配置冲突 遇到 Bug 后，老T第一时间怀疑是不是节点配置冲突。因为节点内的虚拟机本身在正常运行，除了节点上有灰色问号，在节点中还有两个存储设备也同样显示黑色问号。 这两个存储设备实际就是主机中的两个机械硬盘。此前，老T将这两个机械硬盘挂载后，直通给了 100 号虚拟机使用。 但考虑到虚拟机中无法读取硬盘温度，便分离了存储，改用 PCI-E 硬件直通，将 SATA 控制器直通到虚拟机。故此遗留两个过期的存储设备。 于是老T在 PVE 面板中直接删除了两个存储设备，然后刷新，看到问题依旧。 老T在想，是否是因为集群状态还没更新，于是重启并强制刷新集群状态，同时检查存储挂载情况。 1 2 3 systemctl restart pve-cluster systemctl restart corosync pvesm status 果然，存储状态有点问题。PVE 配置中保留了两个 HDD 的 LVM 存储定义，但实际物理磁盘已经直通给 100 号虚拟机不可见。 1 2 3 4 5 6 7 8 root@epson:# pvesm status Command failed with status code 5. command '/sbin/vgscan --ignorelockingfailure --mknodes' failed: exit code 5 Name Type Status Total Used …  ( 4 min )

这是老 T 在 8 月份的说说内容，这里将其一并提取发布，主要包括 Folo RSS 更换域名后的一些问题，医保用药问题，汽车滤芯损毁问题，以及在使用飞牛和群晖过程中遇到的一些问题等。 Github issue 作为博客说说发布页面的模板设置问题 在设置模板过程中，需要留意以下几个问题： 页面构建缓存。可能导致页面内容可能无法更新。 1 2 3 4 5 6 {{ $url := "https://api.github.com/repos/user/moments/issues/1/comments" }} {{ $opts := dict "headers" (dict "User-Agent" "Hugo Static Site Generator") "cache" 300 "cacheKey" (printf "gh-comments-%s" (now.Format "2006-01-02-15:04")) }} 内容排序。 github issue api 输出数据是最新的内容在后边，需要倒过来。 1 2 3 4 {{ with resources.GetRemote $url $opts }} {{ if and .Content (ne .Content "") }} {{ $comments := .Content | transform.Unmarshal (dict "format" "json") }} {{ $sortedComments := sort $comments "created_at" "desc" }} 时间格式。github issue 默认使用 UTC 时间，中国的话，需要在基准上加8个小时。 1 2 3 {{ (.created_at |…  ( 1 min )

虽然还没到 9 月 1 日，但我儿子今天已经提前去学校适应小学生活，随着新的阶段到来，我也想着给本博客换个名称。过程中，找两个小孩帮忙，顺便给他们解释了一下网站域名 Lawtee的含义，然后，不知怎得，两娃就突然开始喊我“老T”。感觉也是天意注定，那我就将本博名字改叫“老T博客”吧。 老T即Lawtee 去年我在选择 Lawtee.com 这个域名时，本来也没想太多，主要是 5 位带 law 三个字母的 com 域名，早已被注册殆尽，而 6 字母中带这三个字母的，也所剩无几，矮个子里边挑高个，最后也就选了这个。 在英文发音中，Law 的美式发音 /lɑː/ 读起来跟 老 字很相近，而 Tee 发音为 /tiː/ 跟字母 T 一模一样。因此，Lawtee 读成 “老T” 倒也显得理所应当。 这次更名前，我也在网上简单搜索了一下，好像叫“老T”这种名字的也不多，那就这样用了。 为什么不再用之前名称 此前，我将本博取名 “法律小茶馆” 其实是按照 Lawtee 意译而来，在一些欧洲国家，Tee 就是茶的意思，跟 Tea 的发音也没啥区别。 但正如目前所现，本博其实大部分内容都跟法律没啥关系，与法律相关主题最多能占到 1/4，强行取个这样的名字，很容易让人误解。 不过，总的来说，由于我长期从事法律工作，这个域名倒也没有继续更换的必要。只是文字介绍需要稍微变动一下，不再着重强调自己的职业，而是依据内容展现方向，改成了“法律、科技和生活”。 毕竟，工作只占日常 1/3 不到的时间，工作之余，这个身份也没有必要随时挂着碍眼了。 未来有何规划 这里还是引用之前更换 Lawtee.com 域名时的想法吧，其实也没什么大的变化。 正如我在 关于 页面中提到： “到 2023 年，自己也已经接近 35 岁，在一些行业都是开始想着退休的年龄，于是回头重构一下博客，想着就这样定型玩下去。” …  ( 2 min )

最近，我家一个智能马桶盖出故障，控制板上，除了水温按钮有反应，其他按钮全失灵，而且重新上电也没任何反应。直觉告诉我，这应该不是单一故障，而是多个问题同时存在所导致。为此，我晚上花了几个小时仔细研究了一下这个产品。 故障表现 这台松下 EKS09 智能马桶盖是我 2019 年买的，迄今已接近 6 年，超出了五年保质期。要么自己修，要么换新的。 考虑到我之后买的两个马桶盖水压普遍偏大，而这个松下的冲水柔和，主要是小孩在用，所以我第一反应还是想通过维修来解决。 此前这个马桶盖曾出现过按键失灵情况，一度所有按钮都没反应，但隔天又好了，应该是排线板接触不良或腐蚀问题。 这次故障更加麻烦点。 控制面板上 8 个按钮。其中 6 个功能控制键中，只有 1 个水温键有反应；测试键方面，长按测试后灯常亮，但没进入测试，并且无论如何无法退出测试，着座键及灯光正常。 在断电重新上电后，只听到电机嗡嗡声，但不出水，冲洗杆也不伸出来。 问题查找和维修过程 由于我此前对智能马桶盖的内部构造没有研究过，直觉上认为这次应该是两个问题。 控制面板失灵； 冲洗杆马达坏了或堵了。 主板 马桶盖外壳拆解倒是比较简单，总体来说，就是卸下螺丝，稍微用力就拆出来了。但真要找到问题所在还得花点功夫。 首先是肉眼观察整个主板，发现已经作了防水处理，肉眼也没见到有元器件烧毁情况。也应证了水温按钮可以使用、上电后电机会嗡嗡出声的情况，总体功能上应该没大的问题。 出水杆 其次，重点是研究这个出水杆的构造。 这个出水杆由两个马达控制，蓝色管进水，白色管输送到出水杆中。 我一开始不知道这个东西原理，以为是出水杆末端马达控制升降，三角位的马达在抽水。 后来才知道，出水杆末端那个马达仅仅是起到旋转开关的作用，控制阀门；而控制升降的是三角位的马达。 但拆下这两个马达后，却也没发现什么异常情况。 在整机上电后，末端马达能够…

Hot off the press is Proxmox Backup Server 4.0. PBS is a great tool to use with your Proxmox VE servers to do fast and easy backups of your VMs and LXCs. It’s very easy to setup, and can be deployed in a variety of forms: bare metal, VM or LXC.  Personally, I’ve deployed Proxmox […]  ( 14 min )

之前在 群晖、迷你主机还是捡垃圾，家用NAS怎么选 提到，我家里 NAS 存储的重要家庭数据已经达到 1000GB，主要就是家中多年积累下来的照片、视频，共计约 20 万文件。虽说存储量也不大，但在迁移过程中，还是感觉很吃力，大量小文件读写，用移动硬盘复制都得几个小时。为了解决这个问题，便于日后轻松上阵，我借鉴之前在 hugo 和 soomal 项目中的经验，对 NAS 文件来了次大瘦身，节约 75% 的空间。 清理重复文件 我以前备份照片视频的模式，基本都是采取每年一到两次的频率，将相机、手机照片视频冷备份到 NAS。大体是选择在春节、国庆长假这种时间节点，以及换手机的时候，一股脑把手机上照片视频通过移动硬盘复制过去。 在这过程中，有一个比较麻烦的事情是，每次我换手机都会将上一台手机的照片视频导入新手机，而上一台手机中很多照片、视频实际上都已经在 NAS 有备份。这种情况下，如果自己备份的时候认真一点，通常会考虑删除一遍重复文件，但很多时候比较懒就没管了。造成 NAS 中实际上有不少文件都是重复版本。 要清理群晖 NAS 中的重复文件，通常有三种方法。 群晖自带的存储空间分析器。打开群晖自带的存储空间分析器，查找“潜在重复文件”，通常会输出一个 csv 文件，列明重复文件。把需要删除的文件导出为 txt，然后在“任务计划”中通过命令 cat /volume3/video/delete.txt | xargs -I {} rm -f "{}" 就能把重复文件删除。 这种方法只适合重复文件数量较少的情况。例如，我的群晖每次只支持查找 5000 个重复文件。如果重复文件显著超过这个数值，就不是很实用了。毕竟，每次查找重复文件都耗时颇多。 第三方应用工具。 群晖第三方应用市场有几款删除文件工具比较实用。例如：dupeGuru fdupes 等。我也下载测试过，但…  ( 2 min )

Many home labs have both Proxmox VE hosts (for running VMs and LXCs) as well as a Synology NAS. If you want to run Proxmox Backup Server on your Synology NAS as a VM, this post is for you. I’ll show you how to install PBS as VM on your NAS, as well as configure […]  ( 17 min )

New to Proxmox Backup Server (PBS) 4.0 is a ‘tech preview’ feature for using S3 buckets as a datastore. While many people associate S3 buckets with Amazon, it is more or less a standard that many cloud providers support. This opens up many possibilities for direct PBS to cloud backups, which is great news. However, […]  ( 18 min )

If you have a Synology NAS and want to use it with Proxmox Backup Server (PBS), you have two primary options for your datastore: NFS or an iSCSI LUN. If you want to use NFS storage, check out my previous article: How To: Setup Synology NFS for Proxmox Backup Server Datastore . However, the NFS datastore […]  ( 18 min )

This refreshed 2025 post covers using a Synology NAS NFS share as a datastore for Proxmox backups. Configuring a NFS share for the proper Proxmox Backup Server permissions can be quite challenging. The permissions requirements for PBS are different from just a basic NFS mount to a Proxmox VE host that you use for VE […]  ( 17 min )

In this article I will walk you through how to install Proxmox Backup Server (PBS) 4.0 inside of a VM running on Proxmox 9.0. You might be wondering why you might want to install PBS as a VM instead of a LXC. Good question! If you want to use iSCSI LUNs as a PBS datastore, […]  ( 24 min )