For many people, the emergence of AI has brought an end to their original reading habits. This article is based on a piece by New Yorker columnist Joshua Rothman (What’s Happening to Reading?). (Background: The global developer forum Stack Overflow sees a 90% drop in activity. Is this the tear of the AI era?) (Supplementary Background: Berkeley professor warns: Graduates from prestigious schools have no job choices! AI will cut half of entry-level positions in five years.) What do you usually read? Why do you read? Decades ago, these questions were hardly urgent. Reading was a commonplace activity, and since the rise of the modern publishing industry in the 19th century, the way we read has hardly changed. In the New Yorker’s 2017 "Shouts & Murmurs" column, writer Emma Rathbone depicted the atmosphere of reading before the internet: "Before the internet, you could lounge on a park bench in Chicago, reading a Dean Koontz novel, and it was perfectly normal; no one would know what you were reading unless you said so yourself." Reading is reading, whether you choose to read a newspaper, Proust, or (power brokers), basically, you are scanning the pages with your eyes, quietly reading at your own pace and time. But now, the essence of reading has changed. Of course, many people still love traditional books and journals, and some have even developed a super reading ability due to the internet age; for them, smartphones are like a mobile library in their pockets. However, for others, that 'old-fashioned' ideal way of reading—focused, continuous, immersing oneself in meticulously crafted words from beginning to end—has almost become a relic of the times. These individuals might start reading with e-books and then listen to audiobooks during their commutes; or, they may not read at all, spending their evenings scrolling through Apple News, subscribing to Substack, and eventually drifting along the information flood on Reddit. Today's reading has a peculiar sense of both distraction and concentration, with countless random words flowing across the screen, alongside temptations from YouTube, Fortnite, Netflix, etc., making it necessary to constantly choose "not to stop" once you start reading. This change has been brewing for decades, driven by new technologies quickly adopted by younger demographics. Perhaps because of this, the change is not so obvious. According to a 2023 report by the National Endowment for the Arts (NEA), the percentage of adults who read at least one book per year has dropped from 55% to 48% over the past decade. This change seems shocking, but compared to the situation of teenagers, it is relatively mild—according to data from the National Center for Education Statistics (NCES), which has been significantly weakened under the Trump administration, the percentage of 13-year-olds who "almost always read" has dropped from 27% to just 14% over the same period. It’s no wonder that professors today increasingly complain about students being addicted to their phones, unable to concentrate on reading anything of moderate length or difficulty. I have known a few people who seemed to have read everything, and what I learned from them truly changed my life. But AI cannot replace these individuals because it is essentially a product of generality and consensus; you wouldn’t take ChatGPT as a spiritual guide, nor would you feel passionate about Gemini's macro theories or quirky insights. However, the reading ability of AI comes precisely from its "impersonality." In a podcast hosted by David Perell (How I Write), Kerwin explained that he constantly asks chatbots about parts he doesn’t understand while reading, and AI will never get tired of these questions, responding with a vast knowledge base that a human cannot instantly grasp. This is equivalent to turning every piece of text into a springboard or syllabus. Moreover, AI can simplify: if you get stuck at the beginning of (The Wasteland), you can ask it to rewrite it in simpler, modern English. Dickens wrote: "The gaslight seeped through the fog in the streets like sunlight breaking through the wet fields, seen by farmers and the children plowing the fields." Claude puts it more straightforwardly: "The gas lamps flicker dimly in the foggy streets, just like the sun seen by farmers in the misty fields." In this way, readers empowered by AI may begin to blur the lines between "original text" and "secondary sources"—especially when they believe that what they are reading can separate "form" from "content." Many people have already adapted to this approach: since 2012, a Berlin-based company called Blinkist (which calls itself "the future of reading") has provided 15-minute summaries of popular non-fiction books, available in both text and audio formats. (In just a brief "blink" of time, you might grasp the core content of Ryan Holiday's exploration of Stoicism and Buddhist philosophy in (The Power of Stillness).) Or like (Reader’s Digest Condensed Books), a subscription selection published quarterly in a hardcover edition, featuring four to five novels that have been cut down to about half their original length. This series was quite popular back in the day—according to a 1987 report from the New York Times, 1.5 million readers bought a total of 10 million copies each year. When I was young, I also had a whole row of these books at home; my parents didn’t say anything, so I casually read a few condensed mystery novels, like those by Dick Francis or Nora Roberts. (This series is still published today under the name Reader’s Digest Fiction Favorites.) If I were to write an academic paper analyzing Francis's novel (Whip Hand) from 1979, using the condensed version would certainly draw criticism. But if I just want to catch a sense of the plot, atmosphere, or tension, saying I have "read this book" isn’t too far off. To be honest, I probably wouldn’t go back to find the full version to read. In the contemporary reading ecology, these condensed or rewritten versions are still exceptions, not the norm. However, in the next decade, this situation may completely reverse: we might habitually start with alternative texts and then decide whether to seek out the original to read, just like we currently download sample versions of books on Kindle first. Because AI can generate summaries, condensed versions, and even various lengths of variants on demand, we may constantly switch between different versions based on context, just as we now listen to podcasts at 2x speed or jump to Wikipedia for plot summaries when a show becomes unwatchable. Popular songs often have different versions, such as clean versions or EDM remixes. As a writer, I might not want to see my words processed in this "refracted" manner. But the power of refraction will no longer be under my control; it will lie in the hands of readers and their AI. This combination will thoroughly compress the boundary between reading and editing. Of course, we have ample reason to...