“They collapsed together, a tangle of sweat-slicked limbs and racing hearts.”
That line is from Daggermouth, the 2026 BookTok bestseller. Word for word, it also shows up in another book. Stony Brook computer scientist Tuhin Chakrabarty says that’s the kind of thing that gives AI away. In a Nature interview out today, he explains his method. His team hunts for long, rare strings of words that turn up verbatim in earlier books. They compared 200 books Pangram flagged as heavily AI-written, 200 self-published bestsellers and 200 prize-listed literary novels. Literary fiction borrowed the least. Self-published genre borrowed more. The AI books borrowed the most. Back in July The Atlantic used his data to say a big chunk of Daggermouth was machine-made. Pangram flagged more than half of it. The author, H. M. Wolfe, says through her lawyer that’s “wholly untrue.” Simon & Schuster and Tor Bramble are standing by her.
So, case closed? No. A detector flagged it, and a researcher checked the detector with… another detector. The detectors are AI. Surprise. What you get is a score and a pile of matching phrases, and the person on the other end still says “I wrote it.” Same place we’ve been all month.
To be fair, Chakrabarty says a few things I agree with. Brilliant writers have used AI. There’s no stigma in using it, he says, the stigma is in using it and hiding it. And no shit, that’s the whole thing.
But look at what his data actually shows. The stuff AI copies best is the formulaic stuff: tropes, clichés, the hackneyed bits every romance has. That’s not a story about AI. It’s a story about effort. If all you brought was a two-minute thought and a prompt, you get back the average of everything that came before. That’s slop, and it’d be slop if a tired human wrote it on a deadline too.
He also pictures the market splitting three ways: pure human, hybrid, and mass-market AI. Maybe. I’d split it differently. There’s work somebody actually put themselves into, and there’s work somebody pumped out to make a buck.
So here’s the little world. Circle is being built for people who put themselves into the work. Members request assignments. Reviews are structured and prompted and submitted right on the site. They guide reviewers through various lenses and encourage real feelings, not polite nods. Writers can revise and resubmit. The credit you earn can be spent on your own work or on someone else’s. On AI, Circle doesn’t try to make the law. We don’t draw up complex formulas to define what’s allowed and what isn’t. We ask our members, once, before they submit their first work, to be honest about if and how they use AI. It’s part of their agreement to be a member. Honesty. That’s what we ask. It’s your word and reputation. We rely on our users to decide how they personally feel about AI writing or production. One place we do not allow AI is in the review of someone else’s work. We are not interested in how AI feels about it. You can go get AI opinion from an AI chatbot. What we want from a review is honest and informed human feedback.
And if being the best is why you write, honestly, the Circle will help you, but you are not the writer we built the system for. We want writer’s that want to be their best. It might seem a subtle difference at first glance, but it’s not. Technically, AI will likely beat you someday if it doesn’t already. That won’t matter to the kind of writer’s we want, because they write for the love of it. For the need to do it. Because they must. In the Circle, that’s who reads and comments on your efforts. If that’s you, join us and earn your satchel.
Registration is open at gapoz.com. No fees. Membership to the Circle is limited and subject to approval. Satchels are earned, not given away. Come meet people, make some friends, and grow your craft together.
Your name. Your word. Your work.
Also on the beat
- Swedish publisher Natur och Kultur hasn’t decided whether to publish Thélyson Orélien’s novel and wants a clearer answer on AI, and Norstedts says the case “will affect all publishers” (Sweden Herald).
- In South Korea, a “No Books, No AI” appeal from 163 publishers and organizations and 333 authors, translators and publishing professionals demands consent and pay after reports that Anthropic bought about 414,000 Korean books (Publishing Perspectives).
- The Authors Guild has joined a global protest against a reported Australian plan that would let AI companies train on authors’ work without a license (Publishers Weekly).
- The European Writers’ Council and the translators’ group CEATL won’t sign the EU’s AI Act transparency code, calling it “technically useless for the book sector” (Coral Wire).
- YouTube says Shorts microdramas passed 6.5 billion views in the first half of 2026, helped by AI subtitle localization and new Shorts series tools. Its pitch: “Write a script. Grab your phone.” (YouTube Blog).
How this was made
AI assisted research and a first draft of this Circle View. Richard Crossley reviewed, edited, and is accountable for the judgment and the final take.
