科技纪年/Tech Craft Chronicles

谁说搭代理要花钱还要懂代码？10 分钟、一台免费 VPS、一行命令——完事。网速快到你怀疑人生，唯一的缺点是你会开始嫌弃以前用的那些付费玩意儿。别问，问就是真香。 Who said you need to pay or know coding to set up a proxy? 10 mins, one free VPS, one command—done. Speed so fast you'll question your life choices. Only downside? You'll start looking down on those paid services you used to rely on. Don't ask, just try it. 实操案例 用免费 Linux VPS 一键部署 Argosbx 现代节点，Mac M2 客户端连接，可选 WARP 改善出口。/Deploy a modern Argosbx node on a free Linux VPS and connect from a Mac M2 client, with optional WARP egress. 准备免费 Linux 机器/Prepare a free Linux box 申请云厂商长期免费档 ARM 或 x86 小机，系统选 Ubuntu，记下公网 IP 与 SSH 账号。/Get a long-term free-tier cloud VM running Ubuntu and note the public IP plus SSH login. 用 Mac 连上服务器/SSH in from your Mac 在 Mac 终端执行 ssh 用户名@服务器IP，首次确认指纹后登录。/From Terminal on Mac run ssh user@server-ip and accept the host key once. 打开命令生成器/Open the command generator 浏览器打开 Argosbx 生成器页面，选 VLESS-TCP-Reality 或 VLESS-XHTTP 等客户端兼容协议，端口可留空随机。/Open the Argosbx generator page, pick a cli...

以前想雇个 AI 程序员,钱包先递辞职信。现在终端里住进一位不睡觉的卷王:扔张截图过去,它连猜带写把网页端上来,速度快得像外卖小哥抢单,账单轻得像奶茶找零。配一次,命令行、桌面、编辑器三处通吃。小毛病嘛——偶尔自信过头,写完还嘴硬,你得客串一下代码监工。但比起省下的时间和钞票,这点脾气就像猫掉毛:忍了,真香。 Hiring an AI programmer used to make your wallet file for divorce. Now a sleepless workaholic lives in your terminal: toss it a screenshot and it serves up a working webpage faster than a delivery rider grabbing orders, with a bill lighter than bubble-tea change. Configure once, and it works across CLI, desktop and editor. Quirks? Occasionally overconfident and argues back — you'll moonlight as a code inspector. But like cat hair on a lovable cat, totally worth it. 实操案例 用 DeepSeek 免费额度驱动 Codex CLI,让它看一张 UI 截图自动写出能跑的 Python Flask 网页,再用 CLIProxyAPI 把这套能力包装成本地 API,一个例子串起全部技术要点。/Use DeepSeek's free credits to power Codex CLI, have it look at a UI screenshot and auto-build a runnable Python Flask webpage, then wrap it all into a local API via CLIProxyAPI — one example covering every key point. 1. 安装 Codex CLI/Install Codex CLI...

真正有意思的是：AI 不需要“打开一个浏览器”，它可以直接接管已经存在的页面。只要那个 Chrome 可调试，Codex 就能列出页面、选择页面、导航、点击、输入、截图。换句话说，你负责把门打开，AI 负责进门——有点像谈恋爱，但千万别把钥匙给错人。 The interesting part is that AI does not necessarily need to launch a new browser. If Chrome is debuggable, Codex can work with pages that are already open: list them, select one, navigate, click, type, and take screenshots. You open the door; the AI walks in. Just make sure you gave the key to the right guest. 正确玩法 | The Correct Setup 第 0 步：检查环境 | Step 0: Check the Environment Node.js 必须至少是 22.12.0，并确认已经安装正式版 Google Chrome。 Node.js must be at least 22.12.0, and the official Google Chrome build should be installed. node --version   # 必须 ≥ 22.12.0,不够先 brew install node 第 1 步：启动真正的无痕 Chrome | Step 1: Launch Real Incognito Chrome 核心就是这一招：独立临时 profile + --incognito + 调试端口。 The key combination is: isolated temporary profile + --incognito + remote debugging. # 清掉可能残留的旧调试实例,避免 profile 锁死 pkill -f "remote-debugging-port=9222"; sleep 2 # 无痕 + 独立临时 profile + 调试端...

你有没有想过,自己也能当一回"云管家"?一台远方的小服务器,一块自己的域名,几个 Docker 积木一搭,就有了一个能远程访问自家网络、管自己数据的私人小站。面板一开,节点、用户、证书全在一屏里排排坐,比对着一堆配置文件抓头发优雅多了。小小的不完美嘛——服务器要花钱、关机也可能偷偷计费,像忘了拔插头的电水壶,记得彻底销毁就好。但那种"我的服务我做主"的踏实感,真香。 Ever fancied being your own cloud butler? A little server far away, a domain of your own, a few Docker blocks stacked up — and you've got a private station to reach your own network and own your data. One dashboard lines up nodes, users and certs neatly, way classier than yanking your hair over config files. The tiny catch — servers cost money and can quietly bill you even when off, like a kettle you forgot to unplug, so destroy them fully. But that "my service, my rules" feeling? Totally worth it. 实操案例 在本机 macOS 上,用 Docker 起一个 Remnawave 面板做本地学习环境,再用 Python 脚本调它的 REST API 自动查看节点/用户状态,先把架构吃透、零云费用,将来上 VPS 时无缝迁移。/On a macOS Mac, spin up a Remnawave panel locally with Docker as a learning sandbox, then use a Python script to hit its REST API and auto-check node/user status —...

  Codex 是能在终端读代码、改文件、跑命令的 AI 编程小工；llama.cpp 是本地“发动机”，在你 M2 上免费跑开源模型；CLIProxyAPI 是“翻译台”，把接口调成 Codex 认得的样子；cc-switch 是“档位开关”，把这条免费链路存成 provider 随时切——全程不登录付费账号、不用 key。再让一段精简 Python 脚本自动起服务、生成配置。全开源、零 API 费。唯一的小情绪是：16GB 喂本地模型有点紧，它偶尔慢悠悠憋半天才吐一行💤，重活还得让给云端。 Codex is a terminal AI coding buddy; llama.cpp is the local engine; CLIProxyAPI is the translator; cc-switch is the gear shifter. No paid account, no API key, zero API cost. The only mood swing: 16GB is a tight squeeze, so heavy jobs may still be better left to the cloud. ① 模型怎么选 / Choosing the model M2 16GB 优先： gpt-oss:20b ：首选，20B MoE、约 3.3B 激活、128K context。 Qwen3-Coder ：编程强，Q4_K_M。 Gemma 4 12B ：更省内存。 首选： gpt-oss-20b-mxfp4.gguf ，约 12–13GB。 内存紧时换 gpt-oss-20b-Q4_K_M.gguf ，约 11.6GB。 https://huggingface.co/ggml-org/gpt-oss-20b-GGUF https://huggingface.co/unsloth/gpt-oss-20b-GGUF ② 下载模型 / Download the model 最省事： llama-server -hf ggml-org/gpt-oss-20b-GGUF --jinja -ngl 999 --ctx-size 32768 手动： pip install -U "huggingface_hub[cli]...