① 他说过的话

17 条

开启自主智能体的开发者多写了 741% 的代码，但实际发布的软件只多了 30%。
The developers who turned on autonomous agents wrote 741% more code, but only 30% more software shipped.
—— Laurie Voss · [01:47]

指向原始笔记的链接

该研究指出，评审者一次读超过 400 行代码就基本发现不了缺陷，一小时超过 450 行效率就完全断崖式下跌。
And the study says that reviewers stop finding defects effectively if they try to read more than 400 lines of code in one sitting and their effectiveness completely falls off a cliff if they try to review more than 450 lines of code in an hour.
—— Laurie Voss · [03:54]

指向原始笔记的链接

OpenClaw 的创造者 Peter Steinberger 说，你不应该再去提示编码智能体了，你应该去设计那些提示你的智能体的循环。
Peter Steinberger, who is the creator of OpenClaw, says that you shouldn’t be prompting coding agents anymore, you should be designing the loops that prompt your agents.
—— Laurie Voss · [04:53]

指向原始笔记的链接

这里的赌注是，循环设计可以替代检查。
The bet there is that loop design substitutes for inspection.
—— Laurie Voss · [06:21]

指向原始笔记的链接

在她谈论这件事的文章里，她讲清楚了模型为什么在生成代码上进步得这么快——因为编译器是一个免费的验证器。
In the essay where she was talking about this, she made clear why models got so good so fast at generating code, and that is because a compiler is a free verifier.
—— Laurie Voss · [09:36]

指向原始笔记的链接

所以一个可合并性的基准，如果我们能做出来的话，不仅仅是一种度量，它会立刻成为前沿模型的训练信号——也就是可维护性、范围纪律、回归安全性这些。
So a benchmark of mergeability, if we could make one, wouldn’t only be a measurement, it would immediately become a training signal for the frontier models, which means maintainability, scope, discipline, regression safety.
—— Laurie Voss · [09:59]

指向原始笔记的链接

谁写下了今天的评审标准，谁就是在书写明年的默认模型行为。
And whoever writes today’s review standard is writing next year’s default model behavior.
—— Laurie Voss · [10:18]

指向原始笔记的链接

所以对拉取请求的机器评审，已经是全球最大代码托管平台上的主流默认做法。
So machine review of pull requests is the mainstream default on the world’s largest code host.
—— Laurie Voss · [11:24]

指向原始笔记的链接

多轮这一招不断被重新发现，因为误报正是杀死一个评审员的东西，而且它确实有效。
The multi-pass trick keeps getting rediscovered because false positives are the thing that kill a reviewer, and it actually works.
—— Laurie Voss · [12:30]

指向原始笔记的链接

模型倾向于看着代码说，嗯，这在我看来不错，所以就发布吧——而这恰恰是人类在那种情况下会做的事。
The model tended to look at the code and say, well, that looks good to me, so ship it, which is exactly what a human would do in that situation.
—— Laurie Voss · [12:54]

指向原始笔记的链接

这个实验的教训是：审查并没有消失。它被重建成一个系统，而那个系统是由人类构建的。
Review didn’t disappear is the lesson of this experiment. It got rebuilt as a system and that system is built by humans.
—— Laurie Voss · [18:31]

指向原始笔记的链接

他说，我错了。请、请去读代码。我们有大约六个月的时间尝试不读代码。结局并不好。我们不得不把那个系统的大部分拆掉重做。
He said, I was wrong. Please, please read the code. We tried not reading the code for like six months. It did not end well. We had to rip out and replace large parts of that system.
—— Laurie Voss · [19:01]

指向原始笔记的链接

但正在发生的是，人类的角色并没有消失，它正在向栈的上层移动，可能移动了栈的好几层——从直接检查代码，转到设计和调优那些检查代码的系统，以及设计「好」的定义。
But what’s happening is the human role isn’t disappearing, it is moving up the stack, possibly several levels of the stack, from inspecting the code directly to designing and tuning the systems that inspect the code and designing the definition of good.
—— Laurie Voss · [20:21]

指向原始笔记的链接

所以你把人类从环中拿掉，你不只是失去了一个审查者，你失去了那个难以被欺骗的东西。
So you take the human out of the loop, you don’t just lose a reviewer, you lose the thing that was hard to fool.
—— Laurie Voss · [21:26]

指向原始笔记的链接

所以自动化审查会栽在那些以自信姿态包装的坏代码上，而以自信姿态包装的坏代码恰恰是智能体非常擅长生产的那种代码。
So automated reviews fall for confidently framed bad code, and confidently framed bad code is exactly the kind of code that agents are very good at producing.
—— Laurie Voss · [21:33]

指向原始笔记的链接

一旦合并前的审查全部由机器完成，观察代码实际做了什么就成了最后一个屹立不倒的审查者。
Once the pre-merge review is all machines, watching what the code actually does becomes the last reviewer standing.
—— Laurie Voss · [22:07]

指向原始笔记的链接

而未来几年胜出的团队，不会是生成最多代码的团队，而是那些能够拿出证据说明他们为什么信任自己所发布内容的团队。
And the teams that win the next few years won’t be the ones that generate the most code, they’ll be the ones who can say with evidence why they trust what they shipped.
—— Laurie Voss · [23:15]

指向原始笔记的链接

② 出现在这些集

1 集

• 《代码评审未死：人类从引擎变飞行员》 — 作为嘉宾

③ 他谈到的

点进去有真内容 —— 本页主要出口

代码审查 · 智能体 · 提示词注入 · 基准测试 · 误报 · GitHub · Cursor · OpenAI · Anthropic · Cognition

④ 也在聊「AI 编程」的人