AI;DR：你不肯编辑，我就不肯读 — HN 讨论摘要
原文概要
8 月 17 日，Rick Manelius 在自己的 Substack 通讯上发了一篇短文《AI;DR (AI; Didn’t Read)》，提议把 “AI;DR” 变成对付 AI 垃圾文本的新缩写。起因是他刷到一条推文——seclilc 发了一个词 “AI;DR”（AI; didn’t read，AI 写的，我没读），346K 播放、16.6K 赞。Manelius 立刻觉得正中要害，随之为自己立下一条政策：
“If you’re not bothered enough to review and edit it… then I’m not going to bother reading it.”
（如果你连审阅和编辑都懒得做，那我连读都懒得读。）
他自称”尽量亲 AI 的人”，但也承认，现在看到朋友发来未经编辑的 AI 输出，身体会生理性不适——耸肩、弓背、轻微眼跳。他区分了场景：客服回复这种”没必要艺术化”的文本可以完全由 AI 生成；但同事在 Slack 里甩一大段 Claude 输出，收到的”信息”就不是他想表达的意思。全文收束在一句对比里：”TL;DR 是社会媒体的解药，AI;DR 是 AI 垃圾文本的解药。”（原文：TL;DR was the solution for social media. AI;DR is the solution for AI slop.）
文章发布两天后冲上 HN 首页第一，1059 分、600 多条评论。作者在更新里补了一句：这条帖在 8 月 18 日成为 HN 头名，”显然戳中了很多人的痛处”。评论区还出现了 dontpastetheai.com 这类”拒绝粘贴 AI”的网站，其中有个”Take me to the Angry Version”按钮，专门带人去读骂人版。
讨论焦点
“附上 prompt，别附上输出”
sega_sai 第一个提出技术层面的替代方案——问题不是”该不该用 AI”，而是发布格式：
“If some text is AI written as a response to a much shorter prompt, then the prompt and/or sources used to make the text should be published instead (or at the very least together with the text).” — sega_sai
（如果一段文本是 AI 对一条简短 prompt 的回复，那应该发布的是那条 prompt 和它的资料来源——至少要和正文一起发。）
JoshTriplett 从信息论的角度补了一句：
“Exactly. And I think the information theory matters here: it’s worth putting in information proportional to the human-supplied input, not proportional to the AI-expanded output.” — JoshTriplett
（正是如此。这背后是信息论：值得投入的信息量应与人类提供的输入成正比，而不是与 AI 扩写出的输出成正比。）
add-sub-mul-div 把 prompt 比作一条推文——投入多少，配得上多少关注：
“I think of a prompt more like a tweet. The medium does matter. If someone was only willing or able to put an abbreviated amount of thought and effort into something, it’s not worth a disproportionate amount my time or attention.” — add-sub-mul-div
（我把 prompt 当成一条推文。媒介本身是有意义的：如果一个人只愿意投入这么点思考和功夫，那它也不值得我投入不成比例的时间和注意力。）
al_borland 给出了职场真实案例——老板把 AI 的回答转发给全组，他反手要 prompt：
“I asked for the prompt he used, because I thought that would be far more useful to understand what he was looking for rather than what AI decided to spit back. It turned out his prompt was equally uninspired… 1 or 2 sentences. That was the thought he put into it, which then had us spending hours trying to figure out the AI reply. He could have saved the whole team a full day by just sending the prompt… or nothing at all.” — al_borland
（我问他用的是哪条 prompt，因为那比 AI 吐出来的东西更能说明他到底想要什么。结果他的 prompt 同样平庸……就一两句话。他投入的思考就这么多，却让我们整个团队花几小时去破译 AI 的回复。他只要把 prompt 发出来——或者什么都不发——就能让全组省下一整天。）
“AI;DR 是敌意的，TL;DR 不是”
Terr_ 站出来为老 TL;DR 辩护，认为这两个缩写不是一回事：
“Maybe I’m the odd one out here and didn’t plumb the worse-depths of social media, but I feel I need to defend good old TLDR as a slightly different, and less-hostile animal. It can be a moral condemnation of the poster… But it usually isn’t. At least half the time TLDR is someone providing a summary because they think it’d be useful.” — Terr_
（可能我是那个另类，没见识过社交媒体最深的泥潭，但我觉得得为老 TL;DR 说句话：它是另一种、没那么有敌意的生物。它有时确实是对发帖人的道德谴责……但通常不是。至少一半情况下，TL;DR 是有人觉得总结有用而主动提供的。）
mrtesthah 更尖锐——拿 AI;DR 跟 TL;DR 比，太抬举它了：
“I agree, likening it to ‘TL;DR’ is much too charitable. At least when someone wrote a wall of text it was apparent they cared enough about the subject to think about and pay attention to it.” — mrtesthah
（我同意，拿它比作 TL;DR 太客气了。至少当一个人写了一整面墙的文字时，能明显看出他足够在乎这个话题，花心思思考过它。）
“拒读 AI”站得住吗——copywriter 反方
eddieroger 唱反调：文字加工自古就有编辑和文案，凭什么 AI 做稿后加工就不行：
“Just playing Devil’s Advocate here, but we’ve had copywriters and editors for generations now. Not reading something because AI did the post-writing work feels kinda weak. Skip stuff started from AI, fine, but this is just because it’s AI, not because the author didn’t do the task.” — eddieroger
（我来抬个杠：人类用文案和编辑已经好几代了。因为 AI 做了稿后加工就不读，有点站不住脚。AI 起稿的可以跳过，但这里仅仅因为是 AI 就不读，而不是因为作者没干活。）
nancyminusone 一句顶回去：
“did you formerly hire copywriters for this task?” — nancyminusone
（你以前会为这种事雇文案吗？）
AnimalMuppet 把文案和 AI 一视同仁——反正都不是人话：
“The problem with copywriters is that they wrote ‘copy’ - which turns out to be writing that I don’t want to read. And the problem with AI is that much of what it trained on is copy, so it trained to write stuff that I don’t want to read. … I want to hear a human voice. If I can’t hear that in your writing, I don’t need to read it.” — AnimalMuppet
（文案的问题在于他们写的是 “copy”——结果就是我不想读的东西。AI 的问题在于训练数据里大部分就是 copy，所以它学会了写我不想读的东西。……我想听到人的声音。如果在你的文字里听不到，那我就不必读。）
谁写的，为什么重要
ademup 提出最根本的质疑——内容好不好，跟谁写的有什么关系：
“Why does it matter who or what wrote a thing? And how would you know if the /content/ is worthwhile, unless you read it… The ability to ‘detect AI’ is imperfect at best. 90% AI written? 5%? How would you know, unless you read it….” — ademup
（谁或什么写的，有什么关系？你不读，又怎么知道内容值不值得……”检测 AI”的能力顶多是不完美的。90% 是 AI 写的？5%？你不读，怎么知道……）
Planktonne 用一句话粉碎了这个论点——交流者本身就是信息的一部分：
“The purpose of writing is to communicate. The person communicating is an important part of the message. Consider the sentence ‘I’m waiting for you at home.’ It really matters if it’s said by your loving wife or an evil clown. You can extrapolate from this.” — Planktonne
（写作的目的是交流。交流的人本身就是信息的重要组成部分。想想这句话：”我在家等你。”这句话是深爱你的妻子说的，还是邪恶小丑说的，天差地别。剩下的你可以自己推。）
otterley 补了两条现实理由——没法对话，加上认知负担转嫁：
“1. I can’t effectively engage in follow-up conversation with the ‘author.’ … 2. It shifts the cognitive burden from writer to reader. … AI-generated writing is often not only more difficult to read than human-drafted prose, but it also has this ‘sameness’ quality that makes every ‘author’ have the same personality.” — otterley
（一，我没法跟”作者”进行有效的后续对话。二，它把认知负担从作者转嫁给了读者。……AI 生成的文字往往不仅比人写的更难读，还有这种”千篇一律”的质感，让每个”作者”都像是同一个人。）
ModernMech 用真实经历反着逼问：农场开业现场，一个人发了张 100% LLM 写的谷仓翻修传单——你读不读？
“I went to a farm opening this weekend. A guy was there with fliers with about the barn renovation he completed, products they used, how the process went. 100% LLM output. … If he gave you that flier, would you scoff and trash it talking about how you can’t be bothered to read it if he didn’t write it? Or would you look it over?” — ModernMech
（这周末我去了个农场开业活动。有个人在那儿发传单，讲他做的谷仓翻修、用的产品、整个过程。100% 是 LLM 生成的。……如果他递给你这张传单，你会因为它不是他亲手写的就嗤之以鼻、扔进垃圾桶吗？还是你会看一眼？）
quickthrowman 的回复是全场最佳冷幽默：
“I would crumple up the flyer, throw it on the ground, and then ask for another one. Then I would crumple up that flyer, throw it on the ground, and ask for another one. I would repeat this until he refuses to give me another flyer.” — quickthrowman
（我会把传单揉成一团扔地上，然后再要一张。再把那张揉成一团扔地上，再要一张。我会一直重复到他拒绝再给我为止。）
Claude 的腔调是最大的 AI 指纹
montroser 说自己也是 AI 支持者，但 Claude 的味道实在太重：
“But, Claude, especially Opus is so unnecessarily flowery, and smug – and it’s so identifiably characteristic. It somehow rubs salt in the wound: ‘I didn’t take the time to write this out in my own words, so now you get to listen to this petulant asshole mansplain it to you’.” — montroser
（但 Claude，尤其 Opus，写得如此不必要的华丽又自鸣得意——特征如此明显。这简直是往伤口上撒盐：”我没花时间用自己的话写，所以现在你得听这个闹脾气的混蛋给你居高临下地讲解。”）
matheusmoreira 从另一个角度补刀：
“Only after switching to Sol did I notice how verbose and smug Claude was. Now I wonder if Claude’s at least partly responsible for the AI hate.” — matheusmoreira
（直到换成 Sol 我才意识到 Claude 有多啰嗦多自鸣得意。现在我在想，AI 招人恨，Claude 是不是要负一部分责任。）
krelas 说得更直接：
“The shape of Claude’s writing style is horrible to read and easy to spot, and that’s a ground truth.” — krelas
（Claude 的文风读起来很难受，还很容易识别，这是铁一般的事实。）
职场：AI 幻觉与 AI 开除
whstl 提供了一个”光明的结局”——已经有公司开始惩罚只会产 AI 文本的员工：
“I already saw someone getting fired for only producing AI text as part of their entire output, after being unable to explain what they ‘wrote’ in multiple situations. And recently my company enacted a mandate that text made for humans should not be AI generated. Sooner or later every company will start realizing that this is just people too lazy to actually work and coasting on a paycheck.” — whstl
（我已经见到有人因为整份产出都是 AI 文本、而且多次无法解释自己”写了”什么而被开除。最近我们公司还出了规定：给人看的文本不该由 AI 生成。迟早每家公司都会意识到，这只是有人懒得真正干活、靠工资混日子。）
被问细节后，whstl 补了一段堪称灾难现场的案例：
“Product Manager using Jira AI to hallucinate tickets, epics, metrics and roadmap. Even the numbers were completely made up. There were too many incidents of ‘this doesn’t look right’ in the same week, so when their boss did a thorough check a couple days after, it was all fake, everything. Eventually they admitted and got canned.” — whstl
（有个产品经理用 Jira 的 AI 幻觉出一堆 ticket、epic、指标和路线图，连数字都是编的。同一周里”这看着不对劲”的事件太多了，老板过了几天做了全面检查——全是假的，每一样。最后他承认了，然后被开了。）
Gigachad 描述了同事视角的折磨：
“Having this problem a lot at work currently. Every single jira ticket is a novel worth of useless text you have to skip over to find the human written parts. Occasionally important details get lost in the noise because they are sandwiched between slop.” — Gigachad
（我目前在工作中经常遇到这个问题。每张 Jira ticket 都是一整篇小说长的无用文字，你得跳过它们才能找到人写的部分。偶尔重要细节就淹没在噪音里，因为它们被夹在垃圾文本中间。）
编辑 vs 作者：AI;DR 到底针对谁
jackjeff 划了一条界线：
“There’s a world of difference between using AI as an editor vs letting it write a whole essay for you. AI;DR is about the latter.” — jackjeff
（把 AI 当编辑用，和让 AI 替你写整篇文章，这是天壤之别。AI;DR 针对的是后者。）
thatjoeoverthr 却认为好商业写作恰恰在编辑，而 LLM 是最烂的编辑：
“I’ll push back and say it’s the former. Good business writing is all about the editing, and autoregressive LLMs are terrible editors. They roll dice on every word and put the same tics on every text. They meander. Suppose you check something with ChatGPT, get nine hundred words, edit that down to three or even one and send that. You won’t get AI;DR for ‘I checked, we’re good.’” — thatjoeoverthr
（我要抬个杠：恰恰是前者。好的商业写作全在编辑功夫上，而自回归 LLM 是最烂的编辑。它们对每个词都掷骰子，给每篇文本都套上同样的口癖。它们绕来绕去。设想你用 ChatGPT 查了一下，得到九百个字，编辑成三个字甚至一个字发出去：你不会因为一句”我查过了，没问题”被 AI;DR。）
teo_zero 则担心另一个极端——见破折号就喊 AI，会错杀一片：
“I’m sure that’s the intention, but the alarming trend, at least here on HN, is crying wolf at the first em-dash.” — teo_zero
（我敢说原意是这样，但至少在 HN 上，令人担忧的趋势是看到第一个破折号就喊狼来了。）
典型观点一览
| 立场 | 用户 | 一句话 |
|---|---|---|
| 发布 prompt 派 | sega_sai | 别发 AI 输出，把 prompt 和资料来源一起发出来 |
| 信息论派 | JoshTriplett | 信息价值应正比于人类输入，而非 AI 输出 |
| 职场验证派 | al_borland | 老板的 prompt 只有两句话，全组却要花一整天破译 |
| TL;DR 辩护派 | Terr_ | TL;DR 通常是提供摘要，AI;DR 是对人的敌意 |
| 反方 | eddieroger | 文案编辑存在几代人，凭什么 AI 加工就不读 |
| 内容至上派 | ademup | 不读永远不知道内容好不好，检测 AI 并不可靠 |
| 交流本质派 | Planktonne | 说”我在家等你”的是妻子还是小丑，完全两回事 |
| 反讽派 | quickthrowman | 传单照单全收，然后一张一张揉碎扔地上 |
| 腔调派 | krelas | Claude 的文风难读又易认，这是事实 |
| 职场现实派 | whstl | 用 Jira AI 编造一切的产品经理，被查后开了 |
| 编辑派 | thatjoeoverthr | 900 字编辑成 3 字发送，没人会给你 AI;DR |
| 警惕派 | teo_zero | 见破折号就喊 AI，是狼来了式的误伤 |
总体情绪
整场讨论几乎一边倒地站 AI;DR，但真正吵的不是”该不该用 AI”，而是”拒绝的姿势够不够聪明”。技术派给出替代方案（附 prompt）、伦理派给出理由（信息论、认知负担）、职场派给出后果（开除、淹没）。反对的声音要么被一句”你以前雇过文案吗”顶回去，要么被”谁写的跟内容好坏无关”引出”妻子还是小丑”的反驳。
有趣的是，两边的共同敌人其实是”懒惰的作者”，而不是 AI 本身。支持者嫌 AI 输出没有人味、把认知负担转嫁给读者；反对者怕的是狼来了——见 em dash 就拒读，最终误伤认真写作的人。AI;DR 也许很快会像 TL;DR 一样变成日常口头禅，但这场讨论真正留下来的，是对”作者到底付出了什么”的重新衡量。
引用帖子
| # | 标题 | URL |
|---|---|---|
| 1 | AI;DR (AI; Didn’t Read) | https://news.ycombinator.com/item?id=49336573 |
免责声明
本摘要由 AI 模型辅助生成：deepseek/deepseek-v4-flash