Quality non-fiction books are the antithesis of AI slop

My first year of college, I had a work-study job which ended up being one of the most sneakily important intellectual experiences of my life.

大学第一年，我有一份勤工俭学的工作，这最终成为我一生中最重要的智力经历之一。

I was a lowly library shelver, assigned to the shelves labelled A through F section in the Library of Congress filing system: mostly works on religion, philosophy, sociology, and history.

我是一个卑微的图书馆书架，被分配到国会图书馆归档系统中标记为 A 到 F 部分的书架上：主要是关于宗教、哲学、社会学和历史的著作。

I say sneakily important because at first glance, shelving books in a library is super boring.

我之所以说“偷偷摸摸”很重要，是因为乍一看，把书放在图书馆里是非常无聊的。

What it amounts to, physically, is reading the label on a book, then placing it on the shelf where it belongs, repeated around a thousand times per shift.

从物理上讲，这相当于阅读一本书上的标签，然后将其放在书架上所属的位置，每个班次重复大约一千次。

To avoid the tedium, I decided that I would also flip to a random page of every book I shelved and read a random sentence from it.

为了避免乏味，我决定随机翻到我搁置的每本书的一页，并随机阅读其中的句子。

Usually, I would stop there — running aground on some passage by a Hungarian classical music critic or a long-dead statistician of Bolivia’s agricultural development or any number of other things that failed to catch my interest.

通常，我会停在那里——在一位匈牙利古典音乐评论家或一位去世已久的玻利维亚农业发展统计学家或任何其他未能引起我兴趣的事情上搁浅。

But other times — like when I came across a book about Hellenistic mystery cults , or The Education of Henry Adams, or Are Clothes Modern?

但其他时候——比如当我读到一本关于希腊化神秘崇拜的书，或者《亨利·亚当斯的教育》，或者《服装现代吗？》

— I would become so absorbed that I’d make my way through several pages before reluctantly depositing the book back where it belonged.

——我会变得如此全神贯注，以至于我会浏览好几页，然后才不情愿地将书放回原来的地方。

And then, very often, I’d do the same with the books on either side of the one I’d liked.

然后，很多时候，我会对我喜欢的书两边的书做同样的事情。

In retrospect, I learned more at this job than in any formal class I’ve ever taken, because it was a filtered form of auto-didacticism.

回想起来，我在这份工作中学到的东西比我参加过的任何正式课程都多，因为它是一种经过过滤的自我说教形式。

The Library of Congress classification system — and the expert staff of an academic research library — had already sorted and filtered these texts.

国会图书馆的分类系统以及学术研究图书馆的专家人员已经对这些文本进行了分类和过滤。

Not to mention the selection mechanism of the fact that that they had been checked out : had, in other words, found a lasting readership.

更不用说他们已经被检查过的选择机制：换句话说，找到了持久的读者群。

Thus I was not seeing a truly haphazard sampling of books, but a targeted, organized, yet still interestingly randomized sampling of good books.

因此，我看到的并不是真正随意的书籍抽样，而是有针对性的、有组织的、但仍然有趣的随机抽样的好书。

Today, undergraduate students will invariably search on Google when asked to find a source, and the results are so much worse than the old method of going to, say, the GR 830 shelf of a research library (basically, “books that the Ghostbusters would read”) and just looking around.

如今，本科生在被要求寻找来源时总是会在谷歌上搜索，结果比过去的方法要糟糕得多，比如去研究图书馆的 GR 830 书架（基本上是“捉鬼敢死队会读的书”）然后四处看看。

But honestly, even research libraries are not what they used to be.

但老实说，即使是研究图书馆也不再像以前那样了。

I am 41, and I feel like I’ve lived through the peak, and now the decline, of what libraries can be (I still love them, of course — in fact I’m currently writing this in the genealogy section of the Santa Cruz Public Library).

我今年 41 岁，我觉得我已经经历了图书馆的巅峰，现在又经历了衰落（当然，我仍然喜欢它们——事实上，我目前正在圣克鲁斯公共图书馆的家谱部分写这篇文章）。

The browsable open stacks of old are being replaced by Learning Labs and Digital Innovation Hubs and seating areas devoted mostly to socializing and snacking, and increasingly, the delightful, weird old books that I had the opportunity to browse as an undergrad are heading to dumpsters, replaced by e-editions.

可浏览的开放书库正在被学习实验室和数字创新中心以及主要用于社交和零食的休息区所取代，而且越来越多地，我作为本科生有机会浏览的令人愉快的、奇怪的旧书正被电子版所取代，被扔进垃圾箱。

But one thing that has remained consistently good throughout my life is the books themselves — non-fiction books, I mean.

但在我一生中一直保持良好状态的一件事是书籍本身——我的意思是非小说类书籍。

Even now, as readership of non-fiction declines amid competition from AI chatbots and podcasts, I feel like we are living through a golden age of the form that rarely gets recognized as such.

即使是现在，随着人工智能聊天机器人和播客的竞争，非小说类小说的读者数量不断下降，我仍然觉得我们正经历着这种形式的黄金时代，但这种形式很少得到认可。

Which is why I set aside some time this summer to create — or, rather, induce Claude Code to create — a free platform for searching in the long tail of high-quality non-fiction books.

这就是为什么我今年夏天留出一些时间来创建——或者更确切地说，引导克劳德·科德创建——一个免费平台，用于搜索高质量非小说类书籍的长尾。

Quality is difficult to define, but it’s been my experience that books that win or achieve the short-list of the major non-fiction prizes are almost always noticeably good, so that was the litmus test I used.

质量很难定义，但根据我的经验，赢得或进入主要非小说类奖项候选名单的书籍几乎总是非常好，所以这是我使用的试金石。

To get started, I counted up all the major non-fiction prizes in the English language.

首先，我统计了所有主要的英语非小说类奖项。

Then I had Claude and GPT-5.6 gather the lists of finalists and winners from various online sources (mostly Wikipedia) and arrange it into a searchable, sortable list.

然后，我让 Claude 和 GPT-5.6 从各种在线资源（主要是维基百科）收集决赛入围者和获奖者名单，并将其排列成可搜索、可排序的列表。

(And before you wonder, yes this is actually free.

（在你想知道之前，是的，这实际上是免费的。

I am paying for the hosting and the API costs entirely because I just want people to find and read more good non-fiction books.)

我支付托管和 API 费用完全是因为我只是希望人们找到并阅读更多优秀的非小说类书籍。）

Though you could consider a paid subscription if you want to support this kind of thing:

如果你想支持这种事情，你可以考虑付费订阅：

There is really nothing “AI” about this aside from the tool that collected the data and coded it, 1 and, crucially, semantic search , which for me is the most appealing of all current AI tools precisely because it offers a straightforward improvement for a workflow and habit that researchers already have: it makes text search work better.

除了收集数据并对其进行编码的工具1以及最重要的是语义搜索之外，这实际上并没有什么“人工智能”，对我来说，语义搜索是当前所有人工智能工具中最有吸引力的，因为它为研究人员已有的工作流程和习惯提供了直接的改进：它使文本搜索工作得更好。

So for instance, you can search simple phrases like “modern France” or “social history” or the like, but you can also search things like “classic biographies that are surprisingly weird,” and an embedding model pulls from the 6,500 or so titles to surface some:

例如，您可以搜索“现代法国”或“社会历史”等简单短语，但您也可以搜索“令人惊讶的怪异经典传记”之类的内容，嵌入模型会从 6,500 个左右的标题中提取一些内容：

Sometimes the “choices” that the search makes are a bit baffling, but that is precisely why I like it: the idea is to recapture some of that feeling of a random walk through a well-tended garden that made my library shelving job so rewarding.

有时，搜索做出的“选择”有点令人困惑，但这正是我喜欢它的原因：这个想法是重新捕捉在精心照料的花园中随意漫步的感觉，这种感觉使我的图书馆书架工作如此有意义。

I find it tends to be best for finding “books like.” For instance I found Stefan Zweig’s memoir of pre-war Vienna, The World of Yesterday , to be deeply moving (even before I learned that he committed suicide, in Brazil in 1942, immediately after completing it).

我发现它最适合寻找“类似的书籍”。例如，我发现斯特凡·茨威格 (Stefan Zweig) 的战前维也纳回忆录《昨天的世界》非常感人（甚至在我得知他于 1942 年在巴西自杀之前，就在完成该回忆录之后）。

A search for a books like it using semantic search in the corpus immediately yields some titles that seem promising but which I’d never heard of before:

在语料库中使用语义搜索来搜索类似的书籍会立即产生一些看起来很有前途但我以前从未听说过的书名：

Once I had gathered all this book-related data, it became a fun experiment to make some data visualizations with it, including fun oddities like this display of roughly 5,000 books from the corpus arranged by color (it would be interesting to plot this by decade, to see whether the same graying effect we see in cars over the past few decades is active in book covers, too).

一旦我收集了所有与书籍相关的数据，用它进行一些数据可视化就变成了一个有趣的实验，其中包括一些有趣的奇怪现象，例如按颜色排列的语料库中大约 5,000 本书的显示（按十年绘制此图会很有趣，看看过去几十年我们在汽车中看到的相同灰色效果是否也活跃在书籍封面中）。

More useful, perhaps (since I’ve never seen this plotted anywhere else), is this chart and accompanying ranking which allows you to explore which imprints and publishers have fared best when it comes to non-fiction book awards over the past century.

也许更有用的（因为我从未在其他地方看到过这样的图表）是这张图表和随附的排名，它可以让您探索哪些出版社和出版商在过去一个世纪的非小说类图书奖项方面表现最好。

And this, in turn, got me thinking about the past and future of nonfiction as a cultural force.

这反过来又让我思考非小说类作为一种文化力量的过去和未来。

For instance, here is a chart of all the non-fiction book prizes which I sampled for this project.

例如，这是我为本项目抽样的所有非小说类图书奖项的图表。

I was surprised to learn that even the august, renowned Pulitzer Prize for nonfiction was actually relatively recently instituted, beginning in 1962.

我惊讶地发现，就连著名的普利策非小说类奖项实际上也是最近才设立的，始于 1962 年。

Throughout the 70s, 80s and 90s, the number of prizes increases, until we reach a peak in 2014, and then, in 2020, the beginning of what may be a slow decline:

在整个 70 年代、80 年代和 90 年代，奖项数量不断增加，直到 2014 年达到顶峰，然后在 2020 年开始缓慢下降：

And yet, maybe not.

然而，也许不是。

What most struck me as I began using my own tool to find new books to read was how consistently good the long tail of non-fiction from the past few decades is.

当我开始使用自己的工具寻找新书阅读时，最令我震惊的是，过去几十年来非小说类小说的长尾始终如一。

You can pick a book more or less at random from this list and end up with something extraordinary and original — not because it’s a hidden gem or forgotten, since obviously these books are on the list by virtue of having been celebrated and praised.

你可以或多或少地从这个列表中随机挑选一本书，最终得到一些非凡和原创的东西——不是因为它是隐藏的宝石或被遗忘，因为显然这些书是由于受到庆祝和赞扬而出现在列表中的。

But a book that won enormous praise in newspapers and among literary intelligentsia or scholars in the early 1990s, say — like, for instance, David Levering Lewis’s acute biography of W.E.B.

但这本书在 20 世纪 90 年代初在报纸和文学知识分子或学者中赢得了巨大的赞誉，比如大卫·利维林·刘易斯 (David Levering Lewis) 的尖锐传记《W.E.B.》。

Du Bois , which I’m currently reading — is not exactly the sort of thing that Amazon is likely to recommend, as it’s out of print and currently at 1 million+ in the sales rankings.

我目前正在读的《杜波依斯》（Du Bois）并不完全是亚马逊可能会推荐的那种东西，因为它已经绝版，而且目前的销量排名已超过 100 万份。

Yet there it is on the list, ranked near the top ten of all books because it won no less than four major prizes when it was published back in 1993.

然而它却出现在了榜单上，在所有书籍中排名接近前十，因为它在 1993 年出版时赢得了至少四项大奖。

And I can personally attest that you can buy it used for ~$4 and it’s really good.

我个人可以证明，你可以花大约 4 美元购买它，而且它真的很棒。

While writing this post, I got interested in the bigger question of when the golden age of non-fiction began and why.

在写这篇文章时，我对一个更大的问题感兴趣：非小说类的黄金时代何时开始以及为何开始。

I suspect it has much to do with the rise of those old-school open stack research libraries, whose origins I wrote about here:

我怀疑这与那些老式开放堆栈研究图书馆的兴起有很大关系，我在这里写了它们的起源：

“Where are all the books?” I asked the campus bookstore clerk earlier this year, as I was gearing up to teach in person for the first time since the COVID-19 shutdowns and my paternity leave.

“所有的书都在哪里？”今年早些时候，我询问了校园书店的店员，当时我正准备自 COVID-19 停课和陪产假以来第一次亲自授课。

It’s true that the basic blueprint of these institutions is an 18th and 19th century development — but the post-war era radically transformed the ways that libraries and archives produced new knowledge, for a range of reasons that I will dig into more in a future post.

诚然，这些机构的基本蓝图是 18 和 19 世纪的发展，但战后时代从根本上改变了图书馆和档案馆产生新知识的方式，原因有很多，我将在以后的文章中深入探讨。

It seems to me that a surprising number of them are related to technological and social change:

在我看来，其中有相当多的问题与技术和社会变革有关：

• The jet plane allowed writers and researchers to travel to multiple continents to research books — the sort of opportunity previously available only to the ultra-wealthy.

• 喷气式飞机使作家和研究人员能够前往多个大陆研究书籍——以前只有超级富豪才有这种机会。

• The erosion of restrictions around class, race, and gender made formerly elite spaces like rare book libraries more widely accessible, and the same process also opened up new questions and research leads (for instance, it is striking how rarely biographers before ~1965 or so dug into the sexuality of their subjects).

• 围绕阶级、种族和性别的限制的侵蚀使得以前的精英空间，如善本图书馆，变得更容易进入，同样的过程也提出了新的问题和研究线索（例如，令人惊讶的是，在大约 1965 年之前，传记作家很少深入研究他们的主题的性行为）。

• Proto-digital and early digital technologies like the Library of Congress classification system and the related MARC (machine-readable cataloguing) standard, developed in the late 1960s, made it much easier to sort and classify books.

• 原始数字技术和早期数字技术，如 20 世纪 60 年代末开发的美国国会图书馆分类系统和相关的 MARC（机器可读编目）标准，使图书的分类和分类变得更加容易。

Crucially, they also made it easier to fact check sources and create high quality endnotes.

至关重要的是，它们还使事实核查来源和创建高质量尾注变得更加容易。

• The advent of broadcast news, oddball TV interview shows ( Dick Cavett!

• 广播新闻、奇怪的电视采访节目的出现（迪克·卡维特！

), and the book-to-Hollywood pipeline created new incentives for authors and new platforms for making their work visible.

），而书籍到好莱坞的渠道为作者创造了新的激励措施，并为让他们的作品被曝光创造了新的平台。

• Word processors and early computers?

• 文字处理器和早期计算机？

I’m still unsure whether these appreciably altered the quality of non-fiction writing, but I think it’s possible.

我仍然不确定这些是否明显改变了非小说类写作的质量，但我认为这是可能的。

Certainly (moving into the 2000s) Wikipedia and Google Books/Hathi Trust have been enormously helpful for me and others in my generation.

当然（进入 2000 年代）维基百科和 Google Books/Hathi Trust 对我和我这一代的其他人有很大帮助。

My own entirely subjective opinion, based on a whole lot of skimming in a whole lot of library books, is that non-fiction writing quality noticeably improved across the whole twentieth century and probably reached a peak around the 1980s to early 2000s.

根据对大量图书馆书籍的大量浏览，我自己完全主观的观点是，非小说类写作质量在整个 20 世纪显着提高，并可能在 1980 年代至 2000 年代初达到顶峰。

Whether it is now declining is, again, a topic for another post — though I’d be curious to hear what you think, dear reader, both about this question and about the Book Prize Index.

它现在是否正在下降，再次是另一篇文章的主题——尽管我很想听听你的想法，亲爱的读者，对这个问题和图书奖指数的看法。

• “Mill’s life has as much to teach here as his arguments.

• “密尔的一生和他的论点一样值得我们学习。

When he suffered that early breakdown, in 1826, it was because he lost faith in the pursuit of utilitarianism, his family creed.

1826 年，他早年遭遇崩溃，是因为他对追求功利主义（他的家族信条）失去了信心。

But he had no one he felt he could talk to about his crisis.

但他觉得没有人可以倾诉他的危机。

His recovery, he said, came in part from reading Wordsworth.

他说，他的康复部分归功于阅读华兹华斯。

It came, too, from rejecting the psychological picture in which his father had raised him, a form of associationism that treated the mind as a mechanism for managing pleasure and pain, just as his father’s utilitarian ethics treated morality as a matter of maximizing the surplus of pleasure over pain.

它也来自于拒绝他父亲养育他的心理图景，一种联想主义的形式，将心灵视为管理快乐和痛苦的机制，就像他父亲的功利主义伦理将道德视为最大化快乐战胜痛苦的问题一样。

As he wrote in On Liberty , human nature is ‘not a machine to be built after a model,’ but a living thing that must ‘grow and develop itself on all sides.’” — Kwame Anthony Appiah on AI and John Stuart Mill , whose Autobiography is one of the books that caught my attention in the open stacks.

正如他在《论自由》中所写，人性“不是一台按照模型建造的机器”，而是一种必须“在各个方面自我成长和发展”的生物。”——夸梅·安东尼·阿皮亚（Kwame Anthony Appiah）谈人工智能和约翰·斯图尔特·米尔（John Stuart Mill），他的自传是在开放书库中引起我注意的书之一。

• Mercifully, Pangram rates the Appiah essay as 100% human written…

• 幸运的是，Pangram 将 Appiah 文章评为 100% 人工撰写……

one can never tell these days.

如今，人们永远无法分辨。

But apparently you can tell, sort of, on Substack — as I was preparing this post, I noticed this newly-added feature :

但显然你可以在 Substack 上看出——当我准备这篇文章时，我注意到这个新添加的功能：

I have been skeptical in the past about software that claims to be able to detect AI writing, but I have to say, Pangram feels different.

我过去一直对声称能够检测人工智能书写的软件持怀疑态度，但我不得不说，Pangram 给人的感觉不同。

I’ve tested it and it’s dismayingly effective — dismaying because, as I wrote about here , a lot of the writing people seem to like online these days is coming up as 100% AI.

我已经测试过它，它的效果令人沮丧——令人沮丧是因为，正如我在这里所写的，现在很多人们喜欢的在线写作都是 100% 人工智能。

I’m glad Substack added this feature and I hope it pops up elsewhere, e.g., it would be interesting to see it automatically applied to the output of major news websites and magazines.

我很高兴 Substack 添加了这个功能，我希望它能在其他地方出现，例如，看到它自动应用于主要新闻网站和杂志的输出会很有趣。

Very cool!

非常酷！

You should see if you're able to replicate your bookshelving experience by maybe having an option on the platform where someone can 'roll the dice' as it were and get returned 1-5 sentences from one of these books at random (or maybe be allowed to filter the dice roll sentence they get by topic or time period), and if it intrigues them maybe a link to the book page proper?

你应该看看你是否能够通过在平台上有一个选项来复制你的书架体验，在这个选项中，有人可以“掷骰子”，并随机返回其中一本书中的 1-5 个句子（或者可能允许按主题或时间段过滤他们获得的掷骰子句子），如果这引起了他们的兴趣，也许是一个到书页的链接？

For those who asked for a "shelf" feature where you can simulate browsing around on the open stacks, here you go!

对于那些需要“书架”功能（可以模拟在打开的书架上浏览）的人来说，就在这里！

It was surprisingly easy to make since I had already collected Library of Congress cataloguing info for about a third of the corpus - it currently only shows that portion of the books but it's still fun I think: https://book-prize-index.vercel.app/fun/library-of-congress-shelf

制作起来非常容易，因为我已经收集了国会图书馆大约三分之一的语料库的编目信息 - 目前它只显示了这部分书籍，但我认为它仍然很有趣：https://book-prize-index.vercel.app/fun/library-of-congress-shelf