AI编码
pub.towardsai.net ·
@Codebook Fusion
实测Claude Code与Codex：真实项目对比，结果出乎意料
作者用两周时间在三个真实代码库上对比测试Claude Code和Codex，分享了实际体验和结果，但未提供具体数据。
AI推荐理由 个人测试对比Claude Code和Codex，基于真实项目，但来源为Medium个人博客，缺乏具体数据和可验证性，主观性强。对开发者选择工具有一定参考，但信息密度一般。
60
正文
仅限会员故事
Claude
编程
人工智能
机器学习
技术
我在真实项目上测试了Claude Code与Codex。结果让我惊讶
两周时间，三个真实代码库，两个AI编程代理。以下是Claude Code和Codex正面交锋时实际发生的情况。
Codebook Fusion
9分钟阅读
·
1天前
--
3
收听
分享
我差点没写这篇文章。每次我打开新的终端标签页进行下一次比较时，我都告诉自己我已经知道答案了。Anthropic的粉丝会选择Claude Code。OpenAI的粉丝会选择Codex。没有人会因为再多一篇博客文章而改变主意。
这里是朋友链接 https://medium.com/towards-artificial-intelligence/i-tested-claude-code-vs-codex-on-real-projects-the-winner-surprised-me-1f9dce72a143?sk=29a9e4fd57b674b0da5117f350299263
按回车或点击查看全尺寸图片
图片由ChatGPT生成
然后我实际上对同样的混乱、真实的代码库运行了这两个工具两周。一个几乎没有测试的遗留内部API。一个充满属性钻探、没人想碰的React仪表板。一个需要抓取数据且永不崩溃的周末脚本。我预期会赢的工具并没有在所有地方都赢。而我准备不屑一顾的工具，却因为与基准图表无关的原因，在我的工作流程中赢得了永久的位置。
如果你在2026年试图在Claude Code和Codex之间做决定，这里是简短版本：没有普遍的赢家……
查看原文（中文正文为 AI 翻译）
Member-only story
Claude
Coding
Artificial Intelligence
Machine Learning
Technology
I Tested Claude Code vs Codex on Real Projects. The Winner Surprised Me
Two weeks, three real codebases, two AI coding agents. Here is what actually happened when Claude Code and Codex went head to head.
Codebook Fusion
9 min read
·
1 day ago
--
3
Listen
Share
I almost did not write this article. Every time I opened a new terminal tab to run the next comparison, I told myself I already knew the answer. Anthropic fans would pick Claude Code. OpenAI fans would pick Codex. Nobody’s mind was going to change based on one more blog post.
HERE IS THE FRIENDS LINK https://medium.com/towards-artificial-intelligence/i-tested-claude-code-vs-codex-on-real-projects-the-winner-surprised-me-1f9dce72a143?sk=29a9e4fd57b674b0da5117f350299263
Press enter or click to view image in full size
Image generated by ChatGPT
Then I actually ran both tools against the same messy, real world codebases for two weeks. A legacy internal API with barely any tests. A React dashboard full of prop drilling nobody wanted to touch. A weekend script that needed to scrape data and never crash. The tool I expected to win did not win everywhere. And the tool I was ready to dismiss earned a permanent spot in my workflow for reasons that had nothing to do with the benchmark charts.
If you are trying to decide between Claude Code and Codex in 2026, here is the short version: there is no universal winner…