Codex的精准度和对细节的关注在正确设置后是*惊人的*。
最近，我一直在使用Codex开发一款塔防游戏，部分原因是为了学习游戏开发的工作原理，部分原因是想看看仅用Codex能做到什么，而不进行任何手动编码。我已经有了我的AGENTS.md和CODESTYLE.md，以及六个其他的全大写md文件等，并且正在进行一些重构，以保持代码库的整洁和文件大小的低廉等等。
然后我在我最新重构的ExecPlan中看到这个：
---
# 观察
- 观察：重构后，屏幕截图在基线正确重新捕获后像素完全相同。
证据：sha256sum screenshots/before-implementation-x.png screenshots/after-implementation-x.png报告在前后对比的1、2和3对之间的哈希值匹配。
---
这真是太疯狂了！我从来没有告诉Codex对应用程序的前后屏幕截图进行<i>SHA比较</i>，但我在我的PLANS.md中确实有指示，要求在游戏的网页应用程序上拍摄前后屏幕截图，以确保我们避免前端回归（它使用GPT-Image-2进行分析）。因此，对于不影响前端的更改，当然在游戏开始时在<i>相同时间戳</i>拍摄的屏幕截图之间不应该有任何不同。
但进行显式的SHA比较——这真是……我从来没有想过的事情。太疯狂了。
查看原文
Lately I've been working on a Tower Defense game with Codex, in part to learn how game development works and in part to see how far I can get using <i>just</i> Codex, no manual coding at all. I've got my AGENTS md & my CODESTYLE md & six other ALLCAPS md files etc, and am working on some refactoring to keep the codebase clean & file sizes low, etc.<p>And then I see this in the ExecPlan for my latest refactor:<p>---<p># Observations<p>- Observation: The refactor made the screenshots pixel-identical after the baseline was recaptured correctly.<p>Evidence: sha256sum screenshots/before-implementation-x.png screenshots/after-implementation-x.png reported matching hashes for before/after pairs 1, 2, and 3.<p>---<p>Which is crazy! I've never told Codex to do an <i>sha compare</i> on before/after screenshots of the app, but I do have instructions in my PLANS.md to take before & after screenshots of the webapp for the game to make sure we avoid frontend regressions (it uses GPT-Image-2 for analysis). So for non-frontend impacting changes, of course nothing should be different between screenshots taken at <i>identical timestamps</i> into the game start.<p>But doing an explicit SHA compare - that's just...not something I would've <i>ever</i> thought of. Wild.