像 Nextdoor 这样覆盖 11 个国家、服务超过 1.1 亿用户的产品，对平台团队提出了诸多要求。对于 Nextdoor 工程主管 Cory Dolphin 而言，Codex 代表着一项根本性转变：“从反复提示智能体，转向结果工程——工程师开始思考他们想要看到的结果，并与智能体协作来实现这一结果。”
这意味着个体工程师能够向上迁移——不再局限于某个特定系统或框架的专家角色，而是能够或多或少地端到端掌控产品体验，甚至跨多个平台。生产力提升如此显著，以至于瓶颈已不再是工程本身，而是关于下一步该构建什么的艰难战略问题。
“Codex 从根本上改变了我们对工程的思考方式，以至于我们无法想象没有它的工程实践。”
——Cory Dolphin，Nextdoor 工程主管
产品工程师可以专注于产品本身
Dolphin 解释道，借助 Codex，“工程师可以花更少的时间思考具体如何构建，而将更多时间用于思考最终成果。”这些成果可能表现为智能体可以构建的截图或视频、特定的性能或测试结果，或者全新的功能创意。
Nextdoor 最近推出了“机会提醒”功能，帮助用户找到附近的服务提供商；借助 Codex，工程师正在驱动产品体验和路线图。例如，一位负责提醒功能的工程师意识到，在地图上显示服务提供商会很有帮助。在过去，这类功能需要移动端、前端和后端工程三个团队协作，甚至可能永远无法从待办事项中脱身。
但有了 Codex，“我们让一位工程师端到端地完成了构建，”Dolphin 解释道，“这意味着他们不仅能更快地推动产品，还能更好地理解实际产品体验以及什么才是正确的发布内容。”
“随着工程师开始向上迁移，他们对自己构建的产品承担了更多责任。你会真正看到个体工程师开始驱动产品。”
——Cory Dolphin，Nextdoor 工程主管
压缩软件工程时间
在处理嵌入式 Rust 数据库和存在严格竞态条件的系统时，Nextdoor 借助 Codex 来调试那些最难复现的问题。团队为智能体提供干净的环境和测试框架，然后利用它解决从 Kubernetes Pod 无法启动到在数据分析中找到正确趋势线等各种问题。
“使用 GPT‑5.4 和 5.5 后，我们看到了令人印象深刻的升级。Codex 在极其持久地尝试找到正确解决方案方面表现出色，能够深入探究一些看似深奥的技术细节，最终找到根本原因，”Dolphin 解释道。
关于 Codex 和 GPT‑5.5 的快速模式，Dolphin 表示：“老实说，团队里很多人都对它上瘾了。当你与正在解决的问题之间拥有快速反馈循环时，作为工程师会感到无比兴奋。”
工程工作变得如此之快，以至于 Dolphin 观察到组织不同部门面临的压力发生了转变。“我们的速度大大提升，瓶颈已不再存在于工程领域。现在真正的问题是，如何识别正确的构建方向和策略——而不再是关于如何实际构建它。”
A product like Nextdoor, which serves over 110 million users across 11 countries, puts many demands on a platform team. For Cory Dolphin, Head of Engineering at, Codex represents an essential shift: “away from iteratively prompting an agent, and towards outcome engineering, where engineers start to think about the result they want to see and work with an agent to engineer that result.”
This means that individual engineers move up the stack—no longer locked up as specialists in a certain system or framework, they’re able to own the product experience more or less end-to-end, even across multiple platforms. Productivity has accelerated so much that the bottleneck is no longer engineering, but rather the hard strategic questions about what to build next.
“Codex has fundamentally changed how we think about engineering, to the point that we can’t even imagine engineering without it.”
—Cory Dolphin, Head of Engineering, Nextdoor
Product engineers can focus on the product
With Codex, “engineers get to spend a lot less time thinking about exactly how they build, and more time thinking about the outcome,” Dolphin explains. That outcome might take the form of screenshots or video that the agent can build towards, a certain performance or test result, or a brand new feature idea.
Nextdoor recently released Opportunity Alerts, which let people find service providers near them; with Codex, engineers are driving the product experience and roadmap. As an example, one engineer working on the alerts realized it would be helpful to show service providers on a map. Historically, that kind of feature would have required collaboration between three teams—mobile, frontend, and backend engineering—and might have never made it out of the backlog.
But with Codex, “we were able to have one engineer build it end to end,” Dolphin explains, “which means not only are they able to drive the product faster, but they’re able to better understand the actual product experience and what the right thing to ship is.”
“As engineers start to shift up the stack, they get to be more responsible for the product that they’re building. You really see individual engineers start to drive products.”
—Cory Dolphin, Head of Engineering, Nextdoor
Compressing software engineering time
Working with embedded Rust databases and systems with tight race conditions, Nextdoor turns to Codex for help debugging the most hard-to-reproduce issues. The team provides the agent with a clean environment and harness for investigation, then uses it for everything from figuring out why Kubernetes pods won’t start, to finding the right trend line in a data analysis.
“With GPT‑5.4 and 5.5, it’s been a really impressive upgrade. We see Codex excel at being extremely persistent and trying to figure out the right solution, diving deep into some seemingly esoteric technical details to arrive at the root cause,” Dolphin explains.
About Fast Mode with Codex and GPT‑5.5, Dolphin says, “I’ve got to be honest, a lot of the team are addicted to it. When you have a quick feedback loop with the problem that you’re working on, the feeling is exhilarating as an engineer.”
Engineering work has gotten so much faster that Dolphin has seen a shift in the pressures on different parts of the organization. “We’re moving so much faster that the bottlenecks are no longer in engineering. It’s really now a question of, how can we identify the right things to build and the right strategy—and less about how we actually build it.”
本文内容采集自官方网站，排版和翻译可能与原页面存在差异。
阅读官方全文