# When Knowledge Is Cheap, Insight Is Everything: Jevons Paradox applied to Torah Learning
**作者**: Zohar Atkins
**日期**: 2026-05-12T11:53:59.000Z
**来源**: [https://x.com/ZoharAtkins/status/2054168204658815070](https://x.com/ZoharAtkins/status/2054168204658815070)
---
In 1865, an English economist named William Stanley Jevons published a book that almost no one reads anymore. It was called The Coal Question, and it argued that Britain was about to ruin itself. The argument turned on a paradox so counterintuitive that Jevons spent most of the book establishing it was real before he allowed himself to draw conclusions from it.
The puzzle was this: James Watt’s improved steam engine, patented in 1769 and refined for decades after, was by design a more efficient machine. It produced more mechanical work from less coal. Every reasonable observer predicted that Britain would consume less coal as Watt’s engines spread. The math was airtight: a given amount of coal now produced more useful energy; therefore, to produce a given amount of useful energy, you needed less coal; therefore, total coal consumption would fall. Jevons looked at the numbers and discovered that the opposite had happened. Between Watt’s patent and the publication of The Coal Question, British coal consumption had risen more than tenfold.
What Jevons saw was that efficiency generates demand. Cheaper steam power produced applications for mechanical work that had been economically impossible at the old prices. Railroads became viable. Iron smelting became cheap enough to industrialize. Ocean shipping was reorganized around coal-fired engines. The savings per engine were real, and they were swamped completely by the proliferation of engines.
Jevons drew a darker conclusion than the one he is now remembered for. He thought Britain would exhaust its coal reserves within a century and lose its industrial supremacy as a result. He was wrong about that, because he could not foresee oil. But the underlying observation, that efficiency in an input produces explosive growth in demand for that input, turned out to be one of the most durable findings in economics. It is now called the Jevons paradox, and it shows up everywhere people study the diffusion of new technology. Cheaper lighting lengthened the working day and lit up the night sky. Cheaper computation built an information economy that now consumes more electricity than most countries. The pattern repeats. When the cost of an input collapses, that input becomes the substrate of a world that could not previously exist.
We are about to learn this lesson again. The input this time is knowledge, and the place we are about to learn it first is the Jewish study hall.
For most of human history, access to the great Jewish textual traditions was rationed by labor. To read Maimonides in his own twelfth-century Judeo-Arabic, you needed years of training. To trace a legal argument across the Babylonian Talmud, the Jerusalem Talmud, the medieval commentators, and the early modern halakhic authorities, you needed a library most people never set foot in and a lifetime most people never had. Even after the printing press, even after the digitization of nearly the entire corpus on the open-access platform Sefaria, the actual capacity to make these texts speak to you, to ask them a question and receive a real answer, remained the property of a small guild of trained readers. The cost of consulting the tradition was high, and so consultation was rationed.
LLLms like Yochai and Rav Dicta — along with base models themselves — have collapsed that cost. A teenager in the Israeli desert town of Be’er Sheva with a phone can now ask a question about the meaning of a difficult line in Rashi’s commentary on the Bible and receive, within seconds, an answer that draws on works she has never heard of, in a language she actually speaks, calibrated to her level. The corpus has not changed. The cost of consulting it has fallen by orders of magnitude.
A reasonable person, watching this happen, might predict that the role of the rabbi, the traditional human interface to the corpus, is about to contract. If anyone can ask the tradition a question, who needs the person who used to answer it? This is the wrong prediction, and Jevons explains why. The cost of consulting Torah is collapsing while the demand for what Torah is supposed to produce stands to explode. When consultation gets cheap, the bottleneck migrates somewhere else. It migrates to a place the tradition has been pointing at for fifteen hundred years and almost nobody, including the rabbinic world itself, has organized its life around.
The bottleneck moves to chiddush, a Hebrew word that means, roughly, the production of genuine new insight from inherited material. The English word innovation gives the wrong flavor. So does originality. Chiddush is closer to seeing something true that nobody saw before, in a text that was already there. It is the act of reading that produces, rather than the act of reading that absorbs. The rabbinic tradition has held for a very long time that this act is what Torah study is for, and that without it, the study hall is not really functioning. The Talmud states this directly: a house of study cannot stand without chiddush (Chagigah 3a). By which the tradition means: without something new being seen, what is happening in the room is no longer Torah, however many books are open on the tables.
It is worth pausing here to note that the rabbinic tradition was not, in the nineteenth century, sitting around waiting for an English economist to explain to it how human desire interacts with supply.
The book of Kohelet, known in English as Ecclesiastes, already observed that the eye is not satisfied with seeing, nor the ear filled with hearing (Kohelet 1:8). The Talmud, in Sukkah 52b, sharpens the principle into a structural law: there is a small organ in a person, the Gemara says, which when starved is satisfied and when satisfied is starved. On the same page, the sage Abaye generalizes: the greater a person, the greater his appetite. The early rabbinic anthology Kohelet Rabbah (1:13) gives the principle its proverbial form: a man with a hundred wants two hundred. The eighteenth-century ethical work Mesillat Yesharim, the standard Jewish text on moral psychology, makes the same observation as a developmental claim: indulgence does not satisfy desire, it expands desire’s capacity (chapters 1 and 13). The nineteenth-century commentator Malbim, glossing the book of Proverbs (21:17), says the same of pleasure as such: attachment enlarges want rather than quieting it.
What the tradition was abundance generates further appetite. This is the same observation Jevons made about coal, transposed from industrial input to human desire. Or rather, Jevons’s observation about coal is the same one Kohelet made about the eye, transposed from human desire to industrial input.
Now we can return to the line from Chagigah 3a: a house of study cannot stand without chiddush (novel insight). The reason a house of study cannot stand without chiddush is precisely what Kohelet identified about the eye and Jevons identified about coal. Abundance generates further appetite. A learner with access to a tradition will be moved, by the structure of desire itself, to want more from it than the tradition has yet given. The only thing capable of meeting that expanded appetite is the production of new insight from inside the tradition.
For most of Jewish history, this collapse was avoided through a brutal economic fact. Producing chiddush required so much preparatory labor (reading the text, the parallel texts, the commentaries on the text, the commentaries on the commentaries) that the appetite was, for most learners, suppressed by the impossibility of feeding it properly. The obligation to produce chiddush, which the Talmud frames as universal, was theoretically binding on every learner and practically binding only on the small number of people who could afford the labor. The practical fact ate the theoretical principle for breakfast, and the tradition, with some discomfort, accommodated the discrepancy by treating chiddush as the elite achievement of a few rather than the obligation of all.
What happens when the practical fact changes?
You are commanded to produce something new.
When the cost of consulting a corpus falls, two things happen at once. Demand for the corpus increases. More people consult it, more often, about more things. That part is obvious. The less obvious part is that the bottleneck migrates.
When books were expensive, the binding constraint on Torah was access to books. When print loosened that, the constraint became literacy. When literacy spread, the constraint became time. When time pressures eased, the constraint became guidance: someone to tell you which page to open and why it mattered. Each loosened constraint revealed the next.
AI loosens the constraint of consultation: the friction of finding, translating, and contextualizing relevant sources. The deeper constraint is chiddush itself: the production of genuine new insight from the material now so easily at hand.
Here Jevons hands off to the French economist Jean-Baptiste Say. Say’s proposition, that supply creates its own demand, was a claim about how new productive capacity reshapes desire. When sewing machines lowered the cost of clothing, people began to imagine wardrobes. When recorded music lowered the cost of hearing a symphony, people discovered that music could accompany every hour of the day. New supply generated demand that had not previously existed because it had not been imaginable.
Cheap consultation of the corpus will do the same to chiddush. When a learner can, in an afternoon, gather every source the medieval commentators cite on a given verse along with their full midrashic and halakhic afterlife, the question that suddenly becomes pressing is what do I see, having seen what they saw? The supply of accessible material generates a demand for synthesis that did not exist when the material was inaccessible. A student who two years ago could spend a semester just locating the relevant texts can now spend that semester actually thinking with them. The chiddush has nowhere to hide.
For most of Jewish history a serious learner could honorably say, I would produce chiddush if I could, but I cannot. The corpus is too vast, my time is too short, my teachers are too few. That sentence has expired. The obligation has gone from aspirational to operational.
The first objection runs that AI-assisted chiddush is not really chiddush. The model did the work. The human pressed the buttons. To call the output an insight is to debase the word.
This objection depends on a particular theory of what chiddush is. If chiddush is invention, the addition of something new to the tradition from outside it, then yes, scaling production cheapens the currency. The rabbinic tradition has always understood chiddush differently. Chiddush is recovery: the surfacing of structure that was always present in the gift but not yet visible.
The locus classicus is in the Talmud, tractate Menachot 29b. Moses ascends to heaven and finds God tying ornamental crowns onto the letters of the Torah. He asks why. God tells him that one day, generations hence, a man named Akiva ben Yosef will derive heaps and heaps of laws from each one of those tags. Moses asks to see him. God tells him to turn around, and suddenly Moses is sitting in the eighth row of Akiva’s study hall. He cannot follow the discussion. He does not understand what the students are saying. His strength fails him. Then a student asks Akiva where some particular ruling comes from, and Akiva answers: halakha l’Moshe miSinai, a law given to Moses at Sinai. And Moses, the text says, is settled.
Everything turns on what settles him. Akiva’s chiddush, which Moses himself could not have produced, traces back to him. The thing Akiva sees was always there, implicit in the gift, waiting for someone with eyes to surface it.
This is the rabbinic theory of chiddush in compressed form. The Torah has a structure with depth, and chiddush is the surfacing of structure that was always present but not yet visible. The crowns were tied at Sinai. Akiva became their reader. If this is what chiddush is, the worry about scale dissolves. There is no risk of running out of crowns. The tradition has no upper limit on the structure it contains, because the structure is the surplus of revelation over comprehension. Every generation surfaces what its tools allow it to surface. The early rabbinic sages saw one set of crowns; their medieval successors saw another; each set saw what their methods made visible. There is no reason to think we have reached the end of what is recoverable.
A second objection runs deeper. Even granting that chiddush is recovery and that AI can lower the cost of the preparatory work, something disordered remains about a world in which insight becomes cheap. A tradition is more than its insights. Cheap chiddush, the worry goes, produces a population of solo learners typing prompts into a chatbot, each generating private surfacings, none of them building anything together. The corpus survives; the people of the book don’t.
This is the right worry, and the tradition met it on the same page of the Talmud as the Akiva story. In Bava Metzia 85b, the third-century sage Reish Lakish is marking the burial caves of the great rabbis. He locates them all except for the cave of Rabbi Chiya, which keeps eluding him. He breaks down. Did I not analyze Torah like Chiya? A heavenly voice answers: yes, you analyzed like him. You did not spread like him. Chiya, the voice says, sowed flax. From the flax he made nets. With the nets he trapped deer. He fed the meat to orphans and made parchment from the hides. On the parchment he wrote the five books of Moses. He went to towns with no teachers and taught five children five books and six children six orders of the Mishnah, and told them: until I return, teach each other. In this way, he said, I have made it so that Torah will not be forgotten from Israel.
Reish Lakish was a virtuoso of analysis. Chiya was a builder of rooms. The heavenly voice’s verdict ranks them: peerless analysis without spreading is enough to lose your cave. The rabbi whose work survives is the one who builds the room in which others learn.
A different Talmudic debate, in Horayot 14a, stages the same tension as institutional politics. The question is which is greater: Sinai, meaning the master of the corpus, the one who has read everything, or oker harim, the uprooter of mountains, the master of original insight who tears the tradition open. The community votes Sinai, with the slogan that everyone needs the master of wheat (a play on the Aramaic for master of the corpus). The academy, in lived practice, gave its presidency to the mountain-uprooter, the third-century sage Rabbah, for twenty-two years. The official answer and the lived answer diverged. We are about to experience that divergence sharply. The wheat is being technologized. The question of who can do something with it is going to dominate everything.
There is a small parable in the kabbalistic tradition. A man from the mountains comes down to the city, eats bread for the first time, and asks what it is. He is told it is made from wheat. He returns to his mountain, gathers raw wheat, eats it dry by the handful, and goes home convinced he has tasted what the city tastes. He has eaten the input. He has missed the thing made when the input is ground and kneaded and salted and left to rise and finally exposed to fire. Bread lives only in the baking.
AI is the new master of wheat. It is unusually capable at producing the relevant source, the right translation, the missing citation. Information was never the goal. A learner who consumes AI-generated wheat by the handful, who asks ChatGPT for the meaning of the weekly Torah portion and stops there, has eaten dry grain and gone home thinking she has tasted Torah. She has tasted the input. The tradition lives in what gets made from it.
The vocation of the learner in the age of cheap wheat is to become a baker: to take the now-abundant raw material and turn it into something a human can eat. Chiddush that surfaces structure. Communities that perpetuate the learning. Students who teach each other when the teacher is gone. This is true for the rabbi and it is true for the nineteen-year-old in Be’er Sheva, because the obligation was always universal in principle (Chagigah 3a). The economics restricted it. The economics just changed.
Jevons’ coal built a new England, more dependent on coal than the old one and more capable of using it. Cheap knowledge will build a new Torah world, more dependent on the corpus than the old one and more capable of working it. The question is whether the people inside that world will recognize the work for what it is and pick up the rolling pin.
I write from my perspective as a rabbi and teacher of Torah, but the arguments I make here well apply to academic knowledge production, too.
If you enjoyed this piece, you might enjoy Alexandria - our great books library with an AI tutor in the margins.
## 相关链接
- [nader dabit reposted](https://x.com/dabit3)
- [Zohar Atkins](https://x.com/ZoharAtkins)
- [@ZoharAtkins](https://x.com/ZoharAtkins)
- [983K](https://x.com/ZoharAtkins/status/2054168204658815070/analytics)
- [Yochai](https://yochai.lightningstudios.ai/)
- [Rav Dicta](https://rav.dicta.org.il/)
- [Alexandria](http://www.alexandria.wiki/)
- [Upgrade to Premium](https://x.com/i/premium_sign_up)
- [7:53 PM · May 12, 2026](https://x.com/ZoharAtkins/status/2054168204658815070)
- [983.1K Views](https://x.com/ZoharAtkins/status/2054168204658815070/analytics)
- [View quotes](https://x.com/ZoharAtkins/status/2054168204658815070/quotes)
---
*导出时间: 2026/5/14 10:26:32*
---
## 中文翻译
# 当知识变得廉价，洞见便是一切：杰文斯悖论应用于托拉学习
**作者**： Zohar Atkins
**日期**： 2026-05-12T11:53:59.000Z
**来源**： [https://x.com/ZoharAtkins/status/2054168204658815070](https://x.com/ZoharAtkins/status/2054168204658815070)
---
1865年，一位名叫威廉·斯坦利·杰文斯的英国经济学家出版了一本书，如今几乎无人问津。这本书名为《煤炭问题》，它论证英国即将自我毁灭。其论证基于一个如此违反直觉的悖论，以至于杰文斯在书中花了大部分篇幅先确立这个悖论确实存在，然后才允许自己从中得出结论。
悖论是这样的：詹姆斯·瓦特改良的蒸汽机于1769年获得专利，此后又经过数十年的完善，其设计本身就是一台更高效的机器。它用更少的煤产生更多的机械功。每一位理性的观察者都预测，随着瓦特蒸汽机的普及，英国的煤炭消耗量将会减少。这个数学推理无懈可击：一定量的煤现在能产生更多有用的能量；因此，要产生一定量的有用能量，所需的煤更少；因此，煤炭总消耗量将会下降。杰文斯查看了数据，发现实际情况恰恰相反。从瓦特获得专利到《煤炭问题》出版之间，英国的煤炭消耗量增长了十多倍。
杰文斯所看到的是，效率会催生需求。更廉价的蒸汽动力催生出了在旧价格下经济上不可能实现的机械应用。铁路变得可行。铁冶炼变得足够廉价，足以实现工业化。远洋航运围绕燃煤发动机进行了重组。每台发动机节省的成本是真实存在的，但完全被发动机数量的激增所淹没。
杰文斯得出的结论比他如今为人所知的那个更为阴暗。他认为英国将在一个世纪内耗尽其煤炭储量，并因此失去其工业霸主地位。在这一点上他错了，因为他无法预见石油。但他的根本观察——某种投入品效率的提高会导致对该投入品需求的爆发式增长——却被证明是经济学中最持久的发现之一。它如今被称为杰文斯悖论，出现在人们研究新技术扩散的各个领域。更廉价的照明延长了工作日，照亮了夜空。更廉价的计算催生了一个信息经济，它如今消耗的电力超过大多数国家。这一模式不断重演。当某种投入品的成本骤然崩塌时，这种投入品就成为一个此前不可能存在的世界的基石。
我们即将再次学习这一课。这一次的投入品是知识，而我们最先学到这一课的地方，将是犹太经堂。
在人类历史的大部分时间里，接触伟大的犹太文本传统的机会是被劳动所配给的。要用迈蒙尼德本人的十二世纪犹大-阿拉伯语阅读他的著作，你需要多年的训练。要在巴比伦塔木德、耶路撒冷塔木德、中世纪评注家和近代早期哈拉哈权威之间追溯一个法律论证，你需要一座大多数人从未踏足过的图书馆，以及一段大多数人从未拥有过的人生。即使在印刷机发明之后，即使在开放获取平台Sefaria上将几乎整个文本库数字化之后，让这些文本真正对你说话的能力——向它们提问并获得真实答案的能力——仍然只属于一小群受过训练的读者的行会。咨询传统的成本很高，因此咨询是被配给的。
像Yochai和Rav Dicta这样的犹太法学大语言模型（LLLms），以及基础模型本身，已经摧毁了这一成本。以色列沙漠城市贝尔谢巴的一个青少年，拿着手机，现在就可以就拉希《圣经》评注中一句难解之文的含义提问，并在几秒钟内得到一个答案，这个答案援引自她从未听说过的著作，用她真正使用的语言表述，并校准到她的水平。文本库没有变。咨询它的成本下降了几个数量级。
一个理性的人，目睹这一切发生，可能会预测拉比——传统上人类与文本库之间的接口——的角色即将萎缩。如果任何人都可以向传统提问，谁还需要那个曾经负责回答问题的人？这是一个错误的预测，而杰文斯解释了原因。咨询托拉的成本正在崩塌，而对托拉本应产出的东西的需求却即将爆炸。当咨询变得廉价时，瓶颈就会迁移到别处。它会迁移到一个传统已经指向了1500年、却几乎没有人——包括拉比界自身——围绕它来组织生活的地方。
瓶颈转移到“希杜什”（chiddush）上，这是一个希伯来语词，大意是从传承下来的材料中产生真正的新洞见。英语单词“创新”（innovation）传达的意味不对，“独创性”（originality）也是如此。希杜什更接近于在一部早已存在的文本中，看到以前没有人见过的真实之物。它是产生意义的阅读行为，而非吸收性的阅读行为。拉比传统长期坚持认为，这种行为正是托拉学习的目的所在，没有它，经堂就名不副实。塔木德直接陈述了这一点：经堂没有希杜什就无法屹立（Chagigah 3a）。传统的意思是：如果没有新的洞见被看见，那么房间里发生的事情就不再是托拉，无论桌上摊开着多少本书。
值得在此稍作停顿，指出拉比传统在十九世纪时，并不是闲坐着等待一位英国经济学家来向它解释人类欲望如何与供给相互作用的。
《传道书》（Kohelet）早已观察到：眼看，看不饱；耳听，听不足（传道书 1:8）。《塔木德》在 Sukkah 52b 将这一原则提升为一条结构性规律：革马拉（Gemara）说，人身上有一个小小的器官，它饥饿时便是满足的，满足时便是饥饿的。在同一页上，贤者阿巴耶（Abaye）将其推而广之：人越大，他的胃口越大。早期的拉比文集《传道书大注释》（Kohelet Rabbah，1:13）以谚语的形式表述了这一原则：有一百的人想要二百。十八世纪的伦理著作《正直者之路》（Mesillat Yesharim），即犹太道德心理学的标准文本，以发展性的论断提出了同样的观察：放纵并不能满足欲望，而是扩大了欲望的容量（第 1 章和第 13 章）。十九世纪的评注家马尔比姆（Malbim）在注解《箴言》（21:17）时，对快乐本身也说了同样的话：依恋使欲望膨胀，而非使其平息。
传统所看到的，是充裕会催生更多的胃口。这与杰文斯关于煤炭的观察如出一辙，只是从工业投入转置到了人类欲望上。或者更确切地说，杰文斯关于煤炭的观察，正是柯赫列特（Kohelet）关于眼睛的观察，只是从人类欲望转置到了工业投入上。
现在我们可以回到 Chagigah 3a 的那句话：学习之所（house of study）没有 chiddush（新颖的洞见）便无法立足。学习之所没有 chiddush 便无法立足的原因，恰恰就是柯赫列特在眼睛上、杰文斯在煤炭上所识别出的道理：充裕会催生更多的胃口。一个能够接触传统的学习者，会被欲望结构本身所驱动，想要从传统中获得比传统迄今已给予的更多的东西。唯一能够满足这种扩张了的胃口的，就是从传统内部生产出新的洞见。
在犹太历史的大部分时间里，这种坍塌被一个残酷的经济事实所避免。生产 chiddush 需要如此多的准备工作（阅读文本、平行文本、对文本的评注、对评注的评注），以至于对大多数学习者而言，这种胃口因无法得到恰当的喂养而被抑制了。生产 chiddush 的义务，《塔木德》将其设定为普遍性的，在理论上对每个学习者都有约束力，而在实践中只对少数负担得起这种劳动的人具有约束力。实践中的事实把理论上的原则打得落花流水，而传统则带着某种不安，通过把 chiddush 视为少数人的精英成就而非所有人的义务，来容纳这一落差。
当实践中的事实发生改变时，会发生什么？
你被命令去生产某种新的东西。
当查阅一个语料库的成本下降时，两件事同时发生。对语料库的需求增加。更多的人、更频繁地、就更多的事情查阅它。这部分是显而易见的。不那么显而易见的部分是：瓶颈发生了迁移。
当书籍昂贵时，托拉（Torah）的限制性因素是能否接触到书籍。当印刷术松动这一限制后，限制变成了识字能力。当识字能力普及后，限制变成了时间。当时间压力缓解后，限制变成了指导：有人告诉你该翻开哪一页，以及为什么那一页重要。每一个被松动的限制都揭示出下一个限制。
AI 松动了“查阅”这一限制：查找、翻译相关文献并将其置于语境中的摩擦力。更深层的限制是 chiddush 本身：从如今唾手可得的材料中生产出真正新颖的洞见。
在此，杰文斯把接力棒交给了法国经济学家让-巴蒂斯特·萨伊（Jean-Baptiste Say）。萨伊的命题——供给创造其自身的需求——是一个关于新生产能力如何重塑欲望的主张。当缝纫机降低了服装的成本时，人们开始想象整个衣橱。当录制的音乐降低了聆听交响乐的成本时，人们发现音乐可以伴随一天中的每一个小时。新的供给催生了此前并不存在的需求，因为在想象到之前，这种需求根本无法被想象。
对语料库的廉价查阅，将对 chiddush 产生同样的作用。当一个学习者能够在一个下午之内，收集起中世纪评注家就某一节经文所引用的全部文献，连同它们在米大示（midrash）和哈拉卡（halakhic）中的全部后世影响，突然变得紧迫的问题便是：在看到了他们所见之后，我看到了什么？可获取材料的供给催生了一种对综合整理的需求，而这种需求在材料不可获取时并不存在。两年前，一个学生可能要用一个学期才刚刚找齐相关文本，而如今他可以用那个学期真正地与这些文本进行思考。chiddush 再无处遁形。
在犹太历史的大部分时间里，一个认真的学习者可以理直气壮地说：如果能够的话，我会生产 chiddush，但我不能。语料库太浩瀚，我的时间太短，我的老师太少。这句话已经失效了。这一义务已从愿望层面的诉求转变为操作层面的要求。
第一种反对意见认为，借助 AI 的 chiddush 并不是真正的 chiddush。模型做了工作，人类只是按了按钮。把这种输出称为洞见，是对这个词的贬低。
这种反对意见依赖于一种关于 chiddush 究竟为何物的特定理论。如果 chiddush 是发明，是从传统之外向传统之中添加某种新东西，那么是的，规模化生产会贬低这种货币。而拉比传统对 chiddush 的理解始终不同。chiddush 是发现（recovery）：让一直存在于这份馈赠之中、却尚未被看见的结构浮现出来。
经典出处位于《塔木德·米纳霍特》29b。摩西升到天上，看见上帝在给妥拉的字母系上装饰性的冠冕。他问为什么。上帝告诉他，许多世代之后，将有一个人名叫阿基瓦·本·约瑟，会从每一个这样的小记号中推导出海量的律法。摩西要求见见他。上帝让他转过身去，忽然间摩西发现自己坐在阿基瓦经学堂的第八排。他听不懂讨论，听不懂学生们在说什么。他的力气渐渐衰竭。然后，一个学生问阿基瓦某条具体裁决出自何处，阿基瓦回答：halakha l'Moshe miSinai——这是在西奈山传给摩西的律法。经文说，摩西由此得到了安慰。
一切都取决于是什么让他得到了安慰。阿基瓦的chiddush（新解），是摩西自己也想不出来的洞见，却能追溯到他身上。阿基瓦所看见的东西，一直都在那里，蕴含在这份礼物之中，等待着有眼力的人把它显现出来。
这就是拉比传统关于chiddush理论的浓缩形态。妥拉具有一种纵深结构，而chiddush就是让一直存在但尚未显形的结构浮出水面。那些冠冕是在西奈山系上的。阿基瓦成为了它们的读者。如果chiddush是这样一回事，那么关于规模的担忧便烟消云散了。不存在冠冕用尽的风险。传统对其所包含的结构没有上限，因为这种结构正是启示超出理解的那部分盈余。每一代人都把他们工具所能企及的东西显现出来。早期的拉比贤哲看见了一组冠冕；他们的中世纪后继者看见了另一组；每一代人看见的都是他们方法所能使之可见的东西。我们没有理由认为可被发掘的东西已经穷尽。
第二种反对意见更为深刻。即便承认chiddush是发掘，也承认AI能够降低前期准备工作的成本，一个洞见变得廉价的世界，依然有某种失序之处。一个传统不仅仅是它的洞见之和。廉价的chiddush，这种担忧是，会造就一批独自学习的个体，他们向聊天机器人输入提示词，各自生成私人的显现，却没有人在共同建造任何东西。文本库留存了下来；而“那书之民”却不复存在。
这个担忧是切中要害的，而传统也正是在塔木德中与阿基瓦故事同一页的地方回应了它。在《塔木德·巴瓦·梅齐亚》85b中，三世纪的贤哲雷什·拉基什正在为大拉比们标记安葬的洞穴。他找到了所有的洞穴，唯独拉比希亚的洞穴一直找不到。他崩溃了：我难道没有像希亚那样剖析妥拉吗？一个天上的声音回答：是的，你像他那样剖析了，但你没有像他那样传播。那声音说，希亚种过亚麻。他用亚麻做了网。用网他捕到了鹿。他把鹿肉给孤儿吃，用鹿皮做成了羊皮纸。在羊皮纸上他写下了摩西五经。他前往没有教师的城镇，教五个孩子五卷书，教六个孩子六部密西拿，并告诉他们：在我回来之前，你们要彼此教导。他就这样说：我做到了让妥拉不会从以色列中被遗忘。
雷什·拉基什是剖析的大师。希亚是建造房间的人。天上声音的评判给他们排了序：无人能及的剖析若没有传播，就足以让你失去自己的洞穴。作品能够留存的拉比，是那个建造出供他人学习的房间的人。
塔木德中另一场辩论，在《塔木德·霍拉约特》14a，把同样的张力以机构政治的形式呈现出来。问题是哪一个更伟大：西奈，指文本库之主，读过一切的人；还是oker harim，移山者，把传统撕开的原创洞见之主。社群投票给了西奈，口号是人人需要“小麦之主”（这是对阿拉姆语“文本库之主”的文字游戏）。而在实际运作中，学院把院长之位给了移山者——三世纪的贤哲拉瓦——长达二十二年。官方答案与现实答案出现了分歧。我们即将尖锐地经历这种分歧。小麦正在被技术化。谁能用它做出点什么，这个问题将主导一切。
卡巴拉传统中有一个小寓言。一个山里人下山进城，第一次吃面包，问这是什么。他被告知是用小麦做的。他回到山上，收集生小麦，一把一把干吃着，回家时深信自己已尝到了城里人尝到的滋味。他吃下了原料，却错过了当原料被研磨、揉捏、加盐、发酵、最终置于火中之后所成就的东西。面包只存在于烘烤之中。
AI是新的小麦之主。它格外擅长产出相关的原始文献、恰当的翻译、缺失的引注。但信息从来都不是目标。一个一把一把吞食AI生成小麦的学习者，一个向ChatGPT询问每周妥拉段落的意义然后就此止步的人，已经吃下了干麦粒，回家时还以为自己尝到了妥拉的味道。她尝到的只是原料。传统活在于由它所成就的东西之中。
在廉价小麦时代，学习者的天职是成为一名面包师：把如今唾手可得的原材料，变成人可以消化的东西。显现出结构的chiddush。延续学习的社群。在教师离去后彼此教导的学生。这对拉比是如此，对贝尔谢巴那个十九岁的年轻人也是如此，因为这一义务在原则上从来都是普遍的（《塔木德·哈吉加》3a）。只是经济条件限制了它。而经济条件刚刚改变了。
杰文斯笔下的煤炭建造了一个新的英格兰，比旧英格兰更依赖煤炭，也更有能力使用煤炭。廉价的知识将建造一个新的妥拉世界，比旧世界更依赖文本库，也更有能力运用文本库。问题是，身处那个世界之中的人们，能否认清这份工作的本质，并拿起擀面杖。
我是以拉比和妥拉教师的视角来写作的，但我在这里提出的论点同样适用于学术知识生产。
如果你喜欢这篇文章，你可能会喜欢 Alexandria——我们的经典著作图书馆，页边配有 AI 导师。
## 相关链接
- [nader dabit 转发了](https://x.com/dabit3)
- [Zohar Atkins](https://x.com/ZoharAtkins)
- [@ZoharAtkins](https://x.com/ZoharAtkins)
- [983K](https://x.com/ZoharAtkins/status/2054168204658815070/analytics)
- [Yochai](https://yochai.lightningstudios.ai/)
- [Rav Dicta](https://rav.dicta.org.il/)
- [Alexandria](http://www.alexandria.wiki/)
- [升级至高级版](https://x.com/i/premium_sign_up)
- [2026年5月12日 晚上7:53](https://x.com/ZoharAtkins/status/2054168204658815070)
- [983.1K 次浏览](https://x.com/ZoharAtkins/status/2054168204658815070/analytics)
- [查看引用](https://x.com/ZoharAtkins/status/2054168204658815070/quotes)
---
*导出时间: 2026/5/14 10:26:32*
