I Am Laura Kipnis-Bot, and I Will Make Reading Sexy and Tragic Again

Margaret Atwood, Marlon James, Lena Dunham, Roxane Gay: We’ve all agreed to be turned into AI reading companions by a mysterious company called Rebind. I report from the inside.

When a flattering email arrived inviting me to participate in an AI venture called Rebind that I’d later come to think will radically transform the entire way booklovers read books, I felt pretty sure it was a scam. For one thing, the sender was Clancy Martin, a writer and philosophy professor I didn’t know personally but vaguely recalled had written about his misspent youth as a small-time jewelry-biz con artist, also being a serial liar in his love life. For another, they were offering to pay me. “Clancy up to his old ways!” I thought.

My role, the email explained, would involve recording original commentary on a “great book”—Clancy suggested Romeo and Juliet, though it could be any classic in the public domain. This commentary would somehow be implanted in the text and made interactive: Readers would be able to ask questions and AI-me would engage in an “ongoing conversation” with them about the book. We’d be reading buddies. Proposing me for Romeo and Juliet did strike me as subversively funny—my “expertise” on romantic tragedy consists of having once written a somewhat controversial anti-marriage polemic titled Against Love. I’ve also written, a bit ironically, about the muddle of sexual consent codes, which I supposed could prove relevant. Juliet was, after all, only 13. These days, Romeo (probably around 16—we’re not precisely told) would risk being called a predator.

A bunch of decidedly illustrious participants, known as “Rebinders,” had apparently already signed on: the Irish Booker Prize winner John Banville on James Joyce’s Dubliners, best-selling writer Roxane Gay on Edith Wharton’s The Age of Innocence, also Bill McKibben, Elaine Pagels, Garth Greenwell ... And bringing up left field, Lena Dunham on E. M. Forster’s A Room With a View, a quirky prospect.

Clancy further explained that someone named John Dubuque, who’d sold a business for “umpteen million dollars,” had gotten the idea for this venture after spending several months working through philosopher Martin Heidegger’s notoriously difficult Being and Time with a tutor. His hope, Clancy said, was to make this kind of (doubtless expensive) one-on-one reading experience available to everyone. I googled John Dubuque. Nothing came up. How do you sell a company for umpteen millions and leave no trace? My scam antennae vibrated again. I figured I’d next be asked to invest in the company, probably in the form of Apple gift cards.

I did agree to a phone call with Clancy and, soon after hellos, pressed for further details about Dubuque, whom I wasn’t sure really existed. “He sounds kind of Gatsbyish,” I said, suavely veiling my skepticism in a literary allusion. Clancy claimed to have met him—a “wonderful fellow” from the Midwest, really nice guy—and then got down to business. If I signed on, Rebind would first record a handful of short videos of me chatting about the play, any aspect that interested me—these would be embedded in various places throughout the text. And then I and an interlocutor (probably Clancy), known in-house as a “Ghostbinder,” would record 12 (or more!) hours of conversation—these would be used as the basis for AI-Laura’s commentaries. The conversation could be about Romeo and Juliet but also related subjects: Is love at first sight trustworthy? Is 13 too young to get married? The content was entirely up to me: My job wasn’t to be a Shakespeare expert, it was to be interesting. As Rebind users read the play, chat windows would open in which they’d write journal-type responses, to which AI-Laura would respond, drawing on and remixing the recordings I had made.

Even if it was technically feasible and Dubuque was legit, did I really want to be involved in this? I have all the usual anxieties about AI—that it will usher in the end of human history; that under the hood it’s a charming sociopath who tries to get tech reporters to ditch their wives; that even its inventors don’t understand how it works; that it’s so ruthlessly intelligent we’ll soon be working for it while believing it’s working for us.

Less amusingly, that it’s being integrated into drone warfare and given autonomous capabilities. Sure, new technologies always prompt trepidation, but the rapid and cavalier adoption of this one—certain to be life-altering on an unimaginable scale—is uniquely terrifying. There’s also no standing in its way. The nihilist in me thinks if humans are going to perish, we might as well perish reading the classics. I downloaded a free copy of Romeo and Juliet from the internet.

My first question to John Dubuque, who not only exists but lives in a sprawling mansion near Santa Barbara, was how there could be absolutely nothing about him online. “How do you even manage that?” I demanded. We’d settled on white canvas couches near the infinity pool, on one of numerous tiered verandas overlooking the Pacific. The panoramic view was breathtaking. A waterfall burbled nearby.

He took the question in stride. “Yeah, I’m a really private person!” The idea of speaking to the general public, let alone a writer, horrified him. I’d heard from Clancy that Dubuque—fortyish and wiry, unpretentiously friendly, unremarkably dressed in chinos and a button-down shirt—had never been interviewed before, which made me feel unaccustomedly powerful.

Next: getting to the bottom of the money. Behind every great fortune lies a great crime, Balzac said. Behind Dubuque’s was ... wholesale plumbing supplies. His great-grandfather started the company, literally called Plumbers Supply, in 1924 in St. Louis, and when Dubuque graduated from college in 2006 his father said, “You can put in 10 or 15 years now and you’ll be glad you did, or you can do it 10 or 15 years from now.” Which is how Dubuque became CEO of a 48-person company in his mid-twenties and, over the next decade-plus, quadrupled it in size.

By 2021 the construction industry was booming; Plumbers Supply was having record month after record month. But looking ahead, where was it going to go? It was a regional operation. Figuring the company wouldn’t be worth as much for decades, and knowing the markets were awash in cash—he was hearing crazy pandemic-era things—Dubuque looked into selling. Within six months a deal was signed with a national outfit. He was 38 and sitting on an “undisclosed sum” of money, meaning a lot. As Dubuque graciously adjusted a gargantuan canvas porch umbrella to shade my East Coast pallor from the blinding California rays, I gazed out at the azure Pacific and resolved to cultivate a better understanding of market cycles.

“Everyone joked, ‘You’re retired, why don’t you chill out?’” Dubuque said. “But I would go nuts.” He obviously didn’t need to make more money. He waved an arm toward the house. “Like, I’ve made it.” Having fallen in love with philosophy as an undergrad at USC, he’d promised himself he’d someday return to it. Being and Time had been sitting on his bookshelf forever; not long before he sold his company, he’d flipped it open and realized, “There’s no way I’ll ever get through this book.” So he hired an Oxford professor to meet with him twice a week, turning an impenetrable treatise into the most meaningful thing he’d ever read. He remembers thinking, “Boy, it’s too bad more people can’t experience this!” When I asked how, specifically, his world had been reshaped, he tried to explain (“Being is so close to us that we look right through it!”) and I tried to understand (“OK, right!”).

Then about a year later—when Dubuque was in the middle of a second tutorial, this time on Alfred North Whitehead’s Process and Reality, another weighty, conjunctively titled book that was once again changing everything for him—ChatGPT came out. Dubuque was skeptical, having long ago read a convincing anti-AI argument by the Berkeley professor Hubert Dreyfus, a famous Heideggerian who thought human intelligence was too embodied and situated to ever be replicated by machines. Nevertheless, Dubuque started playing around with the app, pasting in passages from Whitehead and asking it to summarize them. He was amazed at the results.

He also suspected there were a lot of people who, like him, wanted to read hard books—maybe not Being and Time, but let’s say Moby-Dick. “You read the first 40 pages and you put it on the shelf, right?” By then he was reading William James’ The Varieties of Religious Experience and loving it, and thought: Who’s the William James guy? It turned out to be John Kaag, who’d written a book called Sick Souls, Healthy Minds: How William James Can Save Your Life, a mashup of memoir and philosophy. Which is exactly what Dubuque thinks people want: not scholasticism, but to know how to connect great books to our lives.

When Kaag got an email from Dubuque, he almost didn’t answer, but they eventually talked on the phone and hit it off, as Kaag told me over Zoom: “He turned out to be one of the most curious, thoughtful people that I’ve ever encountered.” The two joined forces to develop Rebind. Kaag brought on his friend Clancy Martin. They have similar profiles: untraditional philosophy professors who’ve written eclectic books, including about their struggles with depression. (Clancy’s most recent book is titled How Not to Kill Yourself.)

Kaag’s mother, who raised him on her own, was a substitute English teacher. At age 12 he was a bad reader with a stutter; his mother would sit with him at the kitchen table and they’d read through his assignments together—essentially an Oxford-style tutorial. It’s what he tried to replicate with his own commentary for Rebind on Thoreau’s Walden: relating the book to his own experiences and difficulties, which include a heart attack at age 40 followed by bypass surgery. (Thoreau, who died of tuberculosis at age 44, wrote movingly about fearing that bad health had prevented him from leading a meaningful life.) If a reader journals about their own life difficulties in the chat, AI finds the places where Kaag shares something similar. Now the two are in conversation. Seeing that back-and-forth happen as they tested out the prototype, Kaag and Dubuque got really excited—they were creating, they thought, a new way to read.

The ideal Rebinder, then, is someone with experience talking to the general public—not necessarily the world’s foremost expert, but someone who knows how to make books come alive. Only a few people they’ve approached have turned them down. When I asked Kaag who’d said no, he laughed and said that the novelist Andre Dubus III, a good friend, had told him he was “dancing with the devil.” “Reading is meant to be a private experience!” Dubus had said. “You’re supposed to lose yourself in a book!” There was no way he’d participate.

The Rebind catalog is evolving by the day: James Wood (Chekhov), Margaret Atwood (Tale of Two Cities), and Marlon James (Huck Finn) have recently been added. Dubuque and Kaag had been thinking mainly about philosophy titles, until they realized how many different kinds of books and conversations there could be. Which was when they realized how big Rebind could be: “Not just big,” Dubuque said, “but a landmark event.” The spiritual category will be especially huge, he thinks: Currently contracted luminaries include Deepak Chopra and Bessel van der Kolk, the trauma expert who wrote The Body Keeps the Score—five years on the bestseller list. They will also, of course, Rebind the Bible, probably from multiple vantages.

I knew I’d have to ascertain from Dubuque how exactly the app worked and steeled myself for an explanation I’d only half understand, so was relieved to hear him say that developing it was less of a technical challenge than a creative one. The innovation, as he explained, was making the Rebinder’s commentary “chattable.” It’s designed to “meet the user where they are,” which also means, as Dubuque enthuses, “it’s going to sound like nothing else you’ve ever heard before!”

Achieving this is the job of Rebind’s discussion engineer, Martin Zirulnik, an English lit PhD (not, Dubuque points out, an engineer by training; he doesn’t think they’re sufficiently attuned to the nuances of language). A subject of frequent jokes between Dubuque and Zirulnik is that one of the large language models they’re testing won’t stop saying “indeed.” “I’ll pay you $500 not to say indeed!” Dubuque will say. Back it comes with “indeed.” It also favors pseudo-profundities like “delve” and “dive,” the unfortunate machines having apparently been trained to regard delving and diving as signifiers of human depth. (Avoid being seated next to one at a dinner party after the singularity.)

Whether to put the AI-generated commentaries in the voice of the actual Rebinder (based on the prerecorded videos embedded in the text) had been a subject of debate. “There’s something really magical about the way someone speaks, something compressed inside people’s voices that brings language to life,” Dubuque thinks, while also concerned about the “ick factor”—are voice clones creepy? Dubuque is convinced that retaining the human element, wherever possible, is crucial. “It comes back to authenticity,” he said. “If you just had the bot, you’d lose that connection.” For now, Rebind has decided to send in the clones.

Dubuque knows the machines are going to be, at times, unpredictable. “Oh, it will definitely hallucinate,” he said. He seemed pretty sanguine about it. (I’d pressed Kaag too on this demon in the machine that apparently likes to fuck things up a little: “Yes, there’s going to be a little wiggle room between what a user gets and what a commentator said.”) Dubuque added: “The kind of computing these LLMs are doing, this kind of intelligence, is just different. Is that an existential risk? Well, I come back to the fact that we have no idea how the human brain works.” It’s easier to imagine catastrophic things than see the opportunities, he points out.

Plus, there are built-in guardrails—places the AI simply refuses to go. I was interested to learn from Ty Rollin, Rebind’s chief technical officer, that suicide is an issue many LLMs will not discuss, which might be a problem when talking about, say, Anna Karenina. (Or Romeo and Juliet.) One of the reasons LLM apps like Rebind have been slow to come out, Dubuque thinks, is that it’s so exasperating to wrangle them. They have a life of their own, with personal quirks and occasionally defiant inclinations.

And they keep getting more and more powerful. Any of the commercially available LLMs can be swapped in and out for Rebind’s purposes—Musk’s xAI, Google’s Gemini—though Dubuque thinks OpenAI is a bit ahead of everyone. Each time Rebind passes content back and forth, the company charges them a fee, but those costs are going to keep plunging, he predicts. As I was completing this article, GPT-4o was released, which, Dubuque told me by email, was not only twice as fast but dropped the price per word by 50 percent: GPT4-o currently costs $5 per million tokens input—a token is roughly 0.75 words—and $15 per million tokens output. (Coincidentally, Dubuque went to high school with Sam Altman, OpenAI’s CEO. They weren’t close friends but did play intramural soccer together.)

Rebind does have to be profitable to grow, Dubuque said; he can’t keep funding it himself. He’s the sole investor; they’re not looking for others. When I asked what his initial outlay had been, he declined to say. And what about other routes to profitability, like selling users’ data? (Which could be pretty fertile stuff for marketers—all your questions and desires about life, love, and existence.) Dubuque was adamant. “One hundred percent never. It will be in writing. Never going to do it.”

This app isn’t for everyone, Dubuque stressed. College students can use it, obviously, but the target audience is adults, or at least the book-loving subset. (Five million people in the United States are in book clubs, Kaag had mentioned.) “You read all these wonderful books as an undergrad,” Dubuque said, “and then you graduate and you read newsletters.” The big thing a user has to understand is that Rebind is designed to be an active experience. “If you’re not responding to these questions and thinking deeply, you’re not going to have as much fun.” Kaag too had stressed that the more a reader puts of themselves into the chat windows—highlighting and reacting, producing “marginalia”—the more interesting the conversations are going to become.

I suspect Rebind will be a boon for the shy and those with a low tolerance for the neuroses and passive-aggressions that book groups invariably incite. Dubuque emphasized that users will have to realize that Rebind isn’t an “Ask Me Anything” experience, though he figures, as with ChatGPT, people will initially want to test its limits, try to “break it.”

Which is exactly what I did when I got access to a beta version of the app and clicked on The Great Gatsby, with New York Times journalist Peter Catapano as the Rebinder. “Was Gatsby just a rich jerk?” I asked AI-Catapano: “Gatsby’s character is complex and multidimensional, not easily reduced to the label of a ‘rich jerk,’” he (it?) chided me. “In fact, Gatsby also enters the book as a very soft-spoken and rather humble-seeming person. He’s perhaps flashy with his home, his parties, and his belongings but seems relatively subdued in his appearance and manner of speaking.” The cadences were slightly stiff—Dubuque said that as the models get faster and smarter, the responses are getting more creative and conversational—but having my admittedly dumb question pondered seriously in real time did feel engaging, and made me want to keep reading and chatting.

I also started to wonder how much this app, by putting users into a sort of imaginary intimacy with the Rebinder, will facilitate us projecting our fantasies onto writers we admire. Especially the kind of reader prone to literary crushes (that would be me), or with celebrity Rebinders like Lena Dunham, who’s subject to so much crazy projection as it is. To be honest, I was already savoring the idea of us reading Room With a View together, not because I suppose she’s some sort of expert on it, but for her weird and fearless sensibility, which I admire. I imagined our back-and-forth—me impressing her with my originality, us giggling together at her offbeat answers. Since her commentary wasn’t yet up on the prototype, I seized the opportunity to get in touch with the person herself.

Lena turned out to be in production for a new series and too slammed to Zoom, but did agree to answer some questions by email. Searching for one that would cement our future bond, I settled on asking whether she had any qualms about being cloned. “The fact is, all of us can be mimicked by AI,” she wrote back, in a somewhat more decorous tone than I’d anticipated. “Anyone can make anyone else say whatever they want in a voice frighteningly close to the real thing, which is an existential threat not just to celebrities or political figures but to all of us.” And why Room With a View? Having long been privately obsessed with English and Irish literature from between 1850 and 1920, she said, getting to do commentary on Forster’s novel was “a surreal gift.” She’d definitely be a Rebind user herself—the platform had stunned her with how “personal, connected, and frankly cozy” it was. I knew she was trying to be helpful, but she didn’t seem entirely present—the human element that Dubuque kept talking about was missing. I’d have to wait for our Rebind conversation.

John Banville, a writer I’ve read and long found intriguing, did agree to Zoom from Ireland. He proved to be roguishly charming and a dedicated raconteur, answering every question about his experience as a Rebinder—“I can’t pretend to understand the thing, I’m an old guy you know”—by way of an anecdote. He’d agreed to participate because he thought it was a wonderful democratization of high literature: “I don’t think I’m being too lovey-dovey here, but I would hope that it would demystify great books for readers who, you know, might feel intimidated.”

His anecdotes were mostly about Joyce, with whom he enjoys a complexly tormented relationship: fraternal, admiring, competitive. He’d read Dubliners countless times and it remained astonishing, despite how young Joyce was when he wrote it: “What was I doing at that age? Writing bad imitations of Joyce, whereas he was an original!” Though I could have happily listened to Banville all day, I attempted to right the ship of the interview by putting to him the question I’d asked Dunham: Did he have any qualms about AI? About being cloned?

“But I think that is the case already,” he chortled. “I have always felt that there is no John Banville. He ceases to exist the moment I stand up from my desk. I don’t know who he is—I find him a very strange creature. My strange dark brother. So, there’s really nobody in there, just this artistic sensibility, creating stuff. I’m already a clone of myself. I’m sure you think the same,” he added genially.

So far there’s no Rebind subscription plan; each title will go for about the price of a new hardcover book—$30 or so. And the app, which will soon go live, will keep evolving. A reader’s chat history from previous books will, in theory, eventually become part of the mashup: If you read Thus Spoke Zarathustra with Clancy and then read Walden with Kaag, it will be like Kaag had watched your responses to Zarathustra. Dubuque thinks that’s what people will like most about the app: “It’s going to know what you do and don’t understand, which will make your future reading experiences even better.”

Which brought to mind my upcoming role as Romeo and Juliet commentator. Not being a Shakespearean, I’d obviously been worried about my qualifications for the task, though I did some years ago briefly date a Shakespearean who used to say that Shakespeareans weren’t that smart. He considered them the bottom rungs of academic intelligence. Nevertheless, I’d been putting myself through reams of criticism and scholarly lit about the play. Most of it reassured me by being—sure enough—not all that interesting. (Light/dark symbolism: got it!)

A few weeks after I got back from Southern California, I took the train from Manhattan to Concord, Massachusetts, where Kaag and another Ghostbinder, Michael Goodwin, both live, as did Ralph Waldo Emerson and Thoreau. (Goodwin was overseeing the film shoot—his mother, Doris Kearns Goodwin, is another upcoming Rebinder.) This was the first step of a two-part process: I’d written scripts for 10 filmed segments, around 45 minutes total, and still had another 12 hours of extemporaneous audio commentary to record in the upcoming weeks. The filming was taking place in the Special Collections Room of the historic Concord Free Public Library—Emerson had delivered the keynote address when it opened in 1873. While the lighting crew was setting up, the archivist kindly brought out a binder of faded handwritten manuscript pages of Thoreau’s essay “Walking” for me to peruse, but I was too keyed up about the filming to devote my full attention. I’d been revising until a few days earlier—the scripts had already been fed into the teleprompter, or else I’d still be scribbling additions.

Love, passion, death, fate: There was no shortage of stuff to talk about. As someone who tends toward mordant humor about the tedium of long-term coupledom—from Against Love: “Never too early to make a down payment on those matching cemetery plots!”—I naturally had some thoughts on double suicide as a proactive solution to potential future domestic misery. I wasn’t sure how the LLM would handle irreverence on the subject—what about those guardrails? Or what if AI-Laura mangles the irony or hallucinates something in my voice that ends up getting me—human-Laura—canceled? I suppose I can always blame Rebind, who in any case owns the commentary. When I asked Dubuque whether they’d be copyrighting it, given that everything about copyright and AI is up in the air, he himself seemed unsure. “Not sure what further protections we get by doing so? Would have to check with an attorney,” he wrote. (He added shortly later that Kaag thought the commentaries were copyrighted.) What was fun about being on the ground floor was that they were basically inventing it as they went along. All that was missing was a garage.

The next morning I got up early(ish) to walk around Walden Pond while listening to Thoreau’s Walden on audio before getting the train back home. It was a misty overcast day, perfect for a pilgrimage—a word that brought to mind Romeo’s line to Juliet, one of my favorites: “My lips, two blushing pilgrims, ready stand ... ” Then Thoreau said something through my earbuds that felt both timely and piercing, given my recent angst over what to wear for the shoot: “Beware of all enterprises that require new clothes.” I’m not generally a big devotee of humanist reverie, but I felt a sudden flash of connection to this eccentric guy, and happy to be reminded—even in this technology-sullied way—how much weird ornery imagination has seeped into the cultural record, despite a world that conspires to tame it.