审阅规范，而不是代码

Your team merged 30 PRs yesterday. How many did you actually read?

你的团队昨天合并了 30 个 PRs。你实际阅读了多少个？

Not skim. Read. Understood the tradeoffs, checked the edge cases, thought about how the change interacts with code three directories away.

不是浏览。要阅读。理解权衡，检查边缘情况，思考更改如何与三个目录外的代码交互。

If you're honest, the number is small. Maybe two or three. Maybe zero.

如果你诚实，这个数字很小。也许两三个。也许零。

That’s not a discipline problem. It’s a math problem.

那不是纪律问题。那是一个数学问题。

AI-generated pull requests now account for a huge share of committed code, and teams with high AI adoption are merging dramatically more PRs than they were a year ago. Review time is up, PR volume is up, and AI agents are touching more and more of your stack.

AI 生成的 pull requests 现在占提交代码的巨大份额，高 AI 采用率的团队正在合并大幅更多的 PRs，比一年前多得多。审查时间增加，PR 数量增加，而且 AI 智能体正在越来越多地接触你的技术栈。

You are producing code faster than you can understand it.

你生成代码的速度快于你理解它的速度。

For decades, the constraint on shipping software was writing it. Getting the logic right, getting the types right, getting the tests to pass. Our entire workflow — sprints, code review, CI — was designed around that constraint.

几十年来，发布软件的约束在于编写它。让逻辑正确，让类型正确，让测试通过。我们整个工作流程——sprints、code review、CI——都是围绕那个约束设计的。

That constraint is gone. Writing code is the cheap part now.

那个约束已经消失了。现在编写代码是廉价的部分。

But most teams are still organized around the old bottleneck:

但大多数团队仍然围绕旧的瓶颈组织：

Staring at diffs.
盯着 diffs 看。
Doing line-by-line reviews of code an agent wrote in forty seconds.
对一个 agent 在四十秒内编写的代码进行逐行审查。
Treating the pull request as the primary quality gate.
将 pull request 视为主要质量关卡。

The data is already telling you this doesn’t work:

数据已经在告诉你这不起作用：

AI-generated PRs wait far longer before anyone even picks them up for review.
AI 生成的 PRs 在任何人甚至拿起它们审查之前，需要等待更长时间。
Once review starts, they move faster — but the acceptance rate is dramatically lower than for human-written code.
一旦审查开始，它们移动更快——但验收率比人类编写的代码低得多。
Reviewers have learned to deprioritize them. Nearly 70% of AI PRs get rejected.
审查者已经学会了对它们降低优先级。近 70% 的 AI PR 被拒绝。

That’s not a tooling gap. Reviewers don’t trus...