The Em Dash Isn’t the Tell — Your Comment Is

破折号不是判断标准，你的评论才是

Two weeks ago one of my outdoor cats bit me. She’s fine — healthy, pregnant, and deeply offended that I picked her up, but she needed flea medicine and I needed to confirm the pregnancy. (If anyone wants a kitten, I know a grumpy lady who has some.) My pinky swelled up, and typing went from “mildly error-prone” to “not happening.” So I dictated this post.

两周前，我的一只户外猫咬了我。她没事——很健康，怀着孕，而且对我抱她这件事感到非常不满，但她需要用跳蚤药，我也需要确认她是否怀孕。（如果有人想要小猫，我认识一位脾气暴躁的女士，她那里有一些。）我的小指肿了起来，打字从“轻微容易出错”变成了“根本无法进行”。所以我口述了这篇文章。

If you’ve ever looked at raw voice transcription, you know what that produces: one giant unpunctuated block with half the words wrong. My transcript literally claims “AI needed to put flea medicine on her.” It was me. That’s the kind of thing the AI is cleaning up. The ideas are mine. The argument is mine. The punctuation and clarification belongs to the machine, because the machine is better at punctuation than a transcript is.

如果你看过原始的语音转录文本，你就会知道它会产生什么：一大块没有标点符号的文字，而且有一半的词是错的。我的转录稿里竟然写着“AI 需要给她涂跳蚤药”。其实是我涂的。这就是 AI 正在清理的那类东西。观点是我的，论证是我的。标点符号和润色属于机器，因为机器在处理标点方面比转录软件强得多。

By the rules of the current discourse, you’re now supposed to stop reading. That’s the game, right? “Not reading this if it’s AI-generated.” “It has em dashes — slop.” Let’s deal with the em dash first, since it’s apparently forensic evidence now. You can type one. Shift-Option-hyphen on a Mac. Windows-Shift-hypen on Windows. Writers were littering pages with them for a century before the first transformer shipped. Its little brother the en dash is everywhere too, and nobody has ever accused an en dash of being a robot.

按照当前的舆论规则，你现在应该停止阅读了。这就是游戏规则，对吧？“如果是 AI 生成的我就不读了。”“有破折号——肯定是垃圾内容。”我们先来谈谈破折号，因为它现在显然成了某种“法医证据”。你可以自己打出来。Mac 上是 Shift-Option-连字符，Windows 上是 Windows-Shift-连字符。在第一个 Transformer 模型发布之前，作家们已经用了一个世纪的破折号了。它的“小兄弟”短破折号也随处可见，但从来没有人指责短破折号是机器人写的。

The em dash gets singled out for exactly one reason: it’s a fast, cheap way to judge a piece of writing without engaging with a single idea in it. Zero effort, instant superiority. Remember that phrase — zero effort. It’s coming back. Because real AI slop absolutely exists. Someone fires off one prompt, ships whatever falls out, never reads it, then farms for stars and upvotes. That’s slop — not because a model was involved, but because no human was. Effort is the variable. The tool never was.

破折号之所以被单独挑出来，原因只有一个：这是一种快速、廉价的评判文章的方式，无需深入思考其中的任何观点。零投入，瞬间获得优越感。记住这个词——零投入。它还会再出现的。因为真正的 AI 垃圾内容确实存在。有人发一个提示词，把生成的东西直接发布，看都不看一眼，然后去刷星标和点赞。那是垃圾——不是因为有模型参与，而是因为没有人类参与。投入才是变量，工具从来不是。

Here’s what the other end of the spectrum looks like. Hundreds of hours on a single project. I decide the architecture, the language, how it compiles, how it deploys. I fork the output into a container and let three agents loose on it to tell me what’s wrong, then I go look for myself, then we iterate — an annealing process, hardening the thing pass after pass. Does that catch every bug? No. Neither do you. But the result compiles, runs, and does what it claims, and I’d put its defect rate up against most hand-written code without blinking.

光谱的另一端是这样的：在一个项目上投入数百小时。我决定架构、语言、编译方式和部署方式。我将输出结果分叉到一个容器中，让三个智能体去检查哪里有问题，然后我自己去查看，接着我们进行迭代——这是一个退火过程，一遍又一遍地加固它。这能捕捉到所有 Bug 吗？不能。你写的代码也做不到。但结果是它能编译、能运行、能实现预期的功能，而且我敢毫不犹豫地拿它的缺陷率去和大多数手写代码对比。

That’s how a systems architect thinks: in outcomes. There are a thousand valid ways to write the same program in a dozen languages, and the compiler doesn’t award style points for matching whatever format twenty engineers blessed in a forum thread. Pick a language like Rust that enforces good behavior at compile time and the “but is it safe” argument gets a lot quieter.

这就是系统架构师的思维方式：关注结果。用十几种语言编写同一个程序有上千种有效的方法，编译器不会因为你遵循了论坛里二十个工程师认可的某种格式就给你加分。选择像 Rust 这样在编译时强制执行良好行为的语言，“但这安全吗”的争论就会少很多。

And I’ve earned the right to skip the drudgery. CS and math degree, programming since the day I first touched a computer. I once caught a BGP misconfiguration where a router in the US was advertising and black-holing a huge block of another country’s address space — and then I went and learned routing down to the nitty-gritty bits because of it. I’ve built the complex routes, the load balancers, the certs, the domains, the Apache configs, the backups. I know how the machine works, top to bottom. Which is exactly why I refuse to burn five days deciphering Microsoft’s code-signing docs — written by humans, by the way, and still confusing — when an AI can walk me through it in an afternoon. Knowing how to do things the hard way is what buys you the judgment to skip it.

我已经赢得了跳过苦差事的权利。我有计算机科学和数学学位，从接触电脑的第一天起就开始编程。我曾经发现过一个 BGP 配置错误，当时美国的一台路由器在广播并黑洞化了另一个国家的一大块地址空间——正因如此，我深入学习了路由的每一个细节。我构建过复杂的路由、负载均衡器、证书、域名、Apache 配置和备份。我从上到下了解机器是如何工作的。这正是为什么我拒绝浪费五天时间去解读微软的代码签名文档（顺便说一句，那是人类写的，而且依然令人困惑），而 AI 可以在一个下午就帮我搞定。知道如何用困难的方式做事，才能让你拥有跳过它的判断力。

While we’re at it, spare me the purity test from people whose files open with a wall of #include statements. You didn’t read the source of your dependencies. You didn’t audit your framework. You took the shortcut the entire industry agreed to call normal — which is fine, that’s what libraries are for. Just don’t pretend your shortcuts are craftsmanship and mine are cheating.

顺便说一句，别再对我搞什么纯洁性测试了，尤其是那些文件开头就有一大堆 #include 语句的人。你并没有阅读你所依赖库的源码，也没有审计过你的框架。你走了整个行业都默认为“正常”的捷径——这没问题，库就是为此而生的。只是别假装你的捷径是匠心，而我的就是作弊。

Accountability is the real line, so let’s draw it. My name is on the repo, therefore the code is mine — every line, however it got there. Find a hole, a bug, a security issue? Bring me a substantive argument and I guarantee I will fix it. Probably with AI. Definitely with more scrutiny than before. That’s what accountability means: answering for what you ship. It has nothing to do with which tools touched the file.

责任才是真正的界限，所以让我们划清它。仓库上有我的名字，所以代码就是我的——每一行都是，无论它是怎么来的。发现漏洞、Bug 或安全问题？给我一个实质性的论据，我保证会修复它。很可能会用 AI 修复。绝对会比以前更严谨地审查。这就是责任的含义：为你发布的东西负责。这与哪些工具触碰了文件毫无关系。

Which brings me to my actual hypothesis, formed over years of watching this play out. The drive-by “AI slop” comment is almost never a preference. A preference sounds like “I write everything by hand because X, Y, Z” — reasons, an argument, something to engage with. And honestly? Good for you. Some circles call that artisan craftsmanship, and it’s a legitimate, even beautiful practice. But that’s not what the drive-by is. The drive-by is a status move. It exists to knock down someone who figured out something you haven’t, so you never have to feel the gap.

这引出了我多年观察后形成的真正假设。那种随口抛出的“AI 垃圾”评论几乎从来不是一种偏好。偏好听起来应该是：“我手写一切是因为 X、Y、Z”——有理由、有论点，是可以交流的东西。老实说？那很好。有些圈子称之为工匠精神，这是一种合法甚至美丽的实践。但那种随口的攻击不是这样。那是一种地位博弈。它的存在是为了打压那些掌握了你还没掌握的东西的人，这样你就永远不必感受到差距。

Underneath it is fear — and the fear is rational. If you spent ten years accumulating knowledge that a model now reproduces on demand, that’s terrifying. Watching your expertise depreciate hurts. But sneering at AI-built projects for another month won’t un-obsolete a single thing you know. Blame-and-deflect is the oldest move we have. Humans have been generating slop since we learned to speak, and blaming someone else for your own stagnation is the classic of the genre. “AI slop” is just the newest sticker on it.

其背后是恐惧——而且这种恐惧是合理的。如果你花了十年积累的知识，现在模型随手就能复现，那确实很可怕。看着自己的专业知识贬值是很痛苦的。但再嘲笑 AI 构建的项目一个月，也不会让你过时的知识变得有用。指责和推卸是我们最古老的手段。人类自从学会说话以来就在制造垃圾，而把自己的停滞不前归咎于他人是这类行为的经典。“AI 垃圾”只是贴在上面的最新标签而已。

The art version is the same argument. If I draw a kick-ass owl by hand and have AI color it, and it moves you — I did my job. Eliciting emotion is the job. And if someone types one lazy prompt and happens to get lucky? Not craft. Same zero-effort laziness as the comment dismissing it. Funny how the two ends of that fight mirror each other. And no, the labs don’t get a pass either. Anthropic paid $1.5 billion to settle over the pirated books in its training pipeline — a speeding ticket at that scale — while its CEO has spent years telling the world how dangerous this technology is. Meanwhile, by their…

艺术领域也是同样的道理。如果我亲手画了一只很酷的猫头鹰，然后让 AI 上色，而它打动了你——那我就完成了我的工作。激发情感就是工作本身。如果有人输入一个懒惰的提示词却碰巧运气好？那不是技艺。这和那些随口贬低它的评论一样，都是零投入的懒惰。有趣的是，这场争论的两端竟然如此相似。而且，那些实验室也别想逃脱责任。Anthropic 为其训练流水线中盗版书籍的问题支付了 15 亿美元的和解金——对于那个规模来说，这只是一张超速罚单——而其 CEO 多年来一直在告诉世界这项技术有多危险。与此同时，根据他们的……