<?xml version="1.0" encoding="UTF-8"?>
<rss version="2.0" xmlns:atom="http://www.w3.org/2005/Atom" xmlns:dc="http://purl.org/dc/elements/1.1/">
  <channel>
    <title>DEV Community: 王泉伟</title>
    <description>The latest articles on DEV Community by 王泉伟 (@wangquanwei).</description>
    <link>https://dev.to/wangquanwei</link>
    <image>
      <url>https://media2.dev.to/dynamic/image/width=90,height=90,fit=cover,gravity=auto,format=auto/https:%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Fuser%2Fprofile_image%2F4240040%2F3a5105ac-45fd-40ab-9fd2-479a6f7b9fe2.png</url>
      <title>DEV Community: 王泉伟</title>
      <link>https://dev.to/wangquanwei</link>
    </image>
    <atom:link rel="self" type="application/rss+xml" href="https://dev.to/feed/wangquanwei"/>
    <language>en</language>
    <item>
      <title>I Let an AI Agent Fix My Website: 7 Minutes, 4 Tasks, Honest Results</title>
      <dc:creator>王泉伟</dc:creator>
      <pubDate>Thu, 12 Nov 2026 10:52:51 +0000</pubDate>
      <link>https://dev.to/wangquanwei/i-let-an-ai-agent-fix-my-website-7-minutes-4-tasks-honest-results-3k44</link>
      <guid>https://dev.to/wangquanwei/i-let-an-ai-agent-fix-my-website-7-minutes-4-tasks-honest-results-3k44</guid>
      <description>&lt;p&gt;I run an AI tools directory — 450+ tools, a prompt library, and tutorials. After a while, the problems piled up: broken search-engine push, thin tool pages, a dead data collector. I wrote them all down into a 22-item fix list, then handed the 4 most urgent items to Codex, OpenAI's coding agent.&lt;/p&gt;

&lt;p&gt;No hype in this post. Just what happened, including the two times it failed before it even started.&lt;/p&gt;

&lt;h2&gt;
  
  
  The task
&lt;/h2&gt;

&lt;p&gt;The four P0 items:&lt;/p&gt;

&lt;ol&gt;
&lt;li&gt;Search-engine push was broken — new content wasn't being pushed.&lt;/li&gt;
&lt;li&gt;Both sites were missing llms.txt files.&lt;/li&gt;
&lt;li&gt;Tool detail pages were thin — only 27% of listings had complete data.&lt;/li&gt;
&lt;li&gt;The data collector had been stalled for 8 days straight.&lt;/li&gt;
&lt;/ol&gt;

&lt;p&gt;My rules for Codex were simple: the source code is on the server, not in your hands — don't pretend you have it. Don't touch the live site. And don't make things up — verify what you can, and say clearly what you can't.&lt;/p&gt;

&lt;h2&gt;
  
  
  It failed twice before it started
&lt;/h2&gt;

&lt;p&gt;First attempt: Codex wouldn't even enter the working directory. A trust check blocked it, and it sat there for 18 minutes. The problem was my launch setup, not the agent.&lt;/p&gt;

&lt;p&gt;Second attempt: it started, then stopped — a missing component (&lt;code&gt;code-mode-host&lt;/code&gt;) meant it couldn't read files at all. I had to download the matching component from the official release, install it next to the Codex binary, and restart.&lt;/p&gt;

&lt;p&gt;Total time spent on environment setup: over half an hour. Actual task time: 7 minutes.&lt;/p&gt;

&lt;p&gt;Lesson one: an AI coding tool is only as usable as its environment. Permissions, components, and working-directory setup decide whether it can work at all.&lt;/p&gt;

&lt;h2&gt;
  
  
  What it did in 7 minutes
&lt;/h2&gt;

&lt;p&gt;Once running, Codex was disciplined: verify first, then act, then self-check.&lt;/p&gt;

&lt;p&gt;&lt;strong&gt;Verify.&lt;/strong&gt; It checked both sites' &lt;code&gt;/llms.txt&lt;/code&gt; URLs itself and confirmed the 404s. It opened a tool detail page and confirmed the template boilerplate was still there. For the push and collector backends, it couldn't log in — and it said so explicitly instead of pretending it had verified them.&lt;/p&gt;

&lt;p&gt;&lt;strong&gt;Deliver.&lt;/strong&gt; It produced two ready-to-deploy &lt;code&gt;llms.txt&lt;/code&gt; files (I byte-compared them against my drafts — identical), a full report, a cron wrapper, and an operations runbook. Two PHP snippets were marked "needs adaptation" because the real source code wasn't available to it. It didn't claim they were drop-in fixes.&lt;/p&gt;

&lt;p&gt;&lt;strong&gt;Self-check.&lt;/strong&gt; It ran checks on its own output: byte-level comparison of the llms.txt files, syntax checks, timeout and lock tests for the cron wrapper. The PHP snippets couldn't be run — no PHP runtime on the machine — and the report says that plainly.&lt;/p&gt;

&lt;h2&gt;
  
  
  What it did NOT do
&lt;/h2&gt;

&lt;p&gt;Three honest limits:&lt;/p&gt;

&lt;ol&gt;
&lt;li&gt;
&lt;strong&gt;No source code, no direct fix.&lt;/strong&gt; The PHP patches are generic implementations. A developer still has to adapt them to the real codebase.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;No backend access, no backend answers.&lt;/strong&gt; Why did push fail? Why did the collector stall? Those need a human in the server panel.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;No runtime, no proof the code runs.&lt;/strong&gt; Written code and working code are two different things.&lt;/li&gt;
&lt;/ol&gt;

&lt;p&gt;So my score for this run: about 70% done, and it said so about the other 30% instead of faking it.&lt;/p&gt;

&lt;h2&gt;
  
  
  Who Codex is actually for
&lt;/h2&gt;

&lt;p&gt;Best for people with a real project, clear requirements, and a working environment — developers fixing bugs, indie hackers maintaining a site, operators tidying up code. If you only have a problem description and no source code or test environment, start small: let it organize the problem, list what to prepare, and analyze causes first.&lt;/p&gt;

&lt;p&gt;And one rule I won't drop: never treat it as an outsourced developer you don't need to review. AI does the heavy lifting; a human does the acceptance check.&lt;/p&gt;

&lt;h2&gt;
  
  
  Where to find the tools
&lt;/h2&gt;

&lt;p&gt;I keep the directory at &lt;a href="https://www.zcyai.cn" rel="noopener noreferrer"&gt;https://www.zcyai.cn&lt;/a&gt; — 450+ AI tools by use case, plus a prompt library and beginner tutorials. The account-management workspace I mentioned (one place for Codex, Claude, and other AI accounts) is at &lt;a href="https://zh.zcyai.cn" rel="noopener noreferrer"&gt;https://zh.zcyai.cn&lt;/a&gt;.&lt;/p&gt;

&lt;p&gt;If you use Codex, what's the most concrete task you'd hand it? Tell me in the comments — the most upvoted one goes into my next experiment.&lt;/p&gt;

</description>
      <category>ai</category>
      <category>coding</category>
      <category>productivity</category>
    </item>
    <item>
      <title>AI 账号太多管不过来？一个工作台全搞定</title>
      <dc:creator>王泉伟</dc:creator>
      <pubDate>Thu, 12 Nov 2026 05:34:53 +0000</pubDate>
      <link>https://dev.to/wangquanwei/ai-zhang-hao-tai-duo-guan-bu-guo-lai-ge-gong-zuo-tai-quan-gao-ding-80c</link>
      <guid>https://dev.to/wangquanwei/ai-zhang-hao-tai-duo-guan-bu-guo-lai-ge-gong-zuo-tai-quan-gao-ding-80c</guid>
      <description>&lt;p&gt;ChatGPT、Claude、Gemini、Codex……用的 AI 一多，账号就乱了：这个要登录，那个到期了，还有一个用量快用完了。每天光切换账号、找密码，就要花不少时间。&lt;/p&gt;

&lt;p&gt;账号一多，麻烦就来了&lt;/p&gt;

&lt;ul&gt;
&lt;li&gt;来回切换：几个平台来回登，浏览器标签页开一堆，找起来费劲。&lt;/li&gt;
&lt;li&gt;到期没提醒：会员哪天到期、用量还剩多少，全靠自己记，经常过期了才发现。&lt;/li&gt;
&lt;li&gt;会话串了：一个窗口里聊工作，一个窗口里聊学习，切错了就乱套。&lt;/li&gt;
&lt;/ul&gt;

&lt;p&gt;一个工作台，全部管起来&lt;/p&gt;

&lt;p&gt;解决办法是找个账号池工具，把所有 AI 账号放在一个工作台里统一管：哪个账号什么状态、用量还剩多少、什么时候重置，一眼就能看到；切换账号点一下就行，每个账号独立会话，互不干扰。&lt;/p&gt;

&lt;p&gt;装纯研习社账号池就是干这个的，官方下载在 zh.zcyai.cn，Windows 和 macOS 都能用，本地存储，账号信息存在自己电脑上。&lt;/p&gt;

&lt;p&gt;会员代充也在做&lt;/p&gt;

&lt;p&gt;GPT、Claude、Google、Grok 的会员代充也在做，需要的朋友可以私信咨询。账号管好了，代充有着落，平时用 AI 就顺了。&lt;/p&gt;

&lt;p&gt;工具再多，也要管得住。先把账号理顺，再谈效率。&lt;/p&gt;

</description>
      <category>ai</category>
      <category>tools</category>
      <category>productivity</category>
    </item>
    <item>
      <title>AI 工具太多不会选？这份导航帮你一次找齐</title>
      <dc:creator>王泉伟</dc:creator>
      <pubDate>Thu, 12 Nov 2026 01:11:54 +0000</pubDate>
      <link>https://dev.to/wangquanwei/ai-gong-ju-tai-duo-bu-hui-xuan-zhe-fen-dao-hang-bang-ni-ci-zhao-qi-5amn</link>
      <guid>https://dev.to/wangquanwei/ai-gong-ju-tai-duo-bu-hui-xuan-zhe-fen-dao-hang-bang-ni-ci-zhao-qi-5amn</guid>
      <description>&lt;p&gt;AI 工具每天都在出新，找工具的时间，有时候比用工具的时间还长。收藏夹里躺了几十个链接，用的时候一个都想不起来。&lt;/p&gt;

&lt;p&gt;挑工具其实就三个标准：先看场景，再看价格，最后看口碑。先想清楚你要解决什么问题，再去找对应的工具，十分钟就能定下来。&lt;/p&gt;

&lt;h2&gt;
  
  
  按场景找，最快
&lt;/h2&gt;

&lt;ul&gt;
&lt;li&gt;写东西：写文章、改文案、做总结，找 AI 写作类的工具，先看支不支持中文，输出能不能直接用。&lt;/li&gt;
&lt;li&gt;画图做图：做海报、配图、头像，找 AI 绘画类的，重点看风格多不多、出图快不快。&lt;/li&gt;
&lt;li&gt;做视频：剪辑、配音、生成短片，找 AI 视频类的，先试免费额度够不够一次成片。&lt;/li&gt;
&lt;li&gt;办公提效：做 PPT、整理会议纪要、处理表格，找办公类的，能直接套模板的最省事。&lt;/li&gt;
&lt;li&gt;写代码：补全代码、查报错、写注释，找编程类的，支持的语言越全越好。&lt;/li&gt;
&lt;li&gt;学习研究：查资料、读论文、做调研，找学习研究类的，带来源、能核实的优先。&lt;/li&gt;
&lt;/ul&gt;

&lt;h2&gt;
  
  
  一个导航站，全都整理好了
&lt;/h2&gt;

&lt;p&gt;一个个搜太费时间，我平时用的是 &lt;a href="https://www.zcyai.cn" rel="noopener noreferrer"&gt;https://www.zcyai.cn&lt;/a&gt; 这个 AI 工具导航，481 多个工具都按分类整理好了，写作、绘画、视频、办公、编程、学习，想找哪个点分类就行。站里还有提示词库和入门教程，刚接触 AI 的朋友可以先从教程看起，照着做一遍就会了。&lt;/p&gt;

&lt;p&gt;工具每天都在更新，导航站也在收新工具，收藏一个就够了，用的时候直接搜，不用再到处翻。&lt;/p&gt;

&lt;h2&gt;
  
  
  给新手的三句话
&lt;/h2&gt;

&lt;p&gt;第一，先定场景再找工具，不要反过来。第二，先试免费版，好用再考虑付费。第三，同一个需求试两个工具，对比一下输出，十分钟就能分出高下。&lt;/p&gt;

&lt;p&gt;工具是拿来用的，不是拿来收藏的。找齐了，就去干活。&lt;/p&gt;

</description>
    </item>
  </channel>
</rss>
