confess

Booktubers already confess that they read ChatGPT summaries. The question now is what is “reading” in 2025

by

The booktubers (social media content creators whose identity revolves around reading) are starting to shamelessly admit that they don’t read the books they recommend: they read what ChatGPT says and summarizes about them. The curious thing is that, unlike what more veteran readers would do, they do not confess it to their smartphone as something they believe they should be ashamed of, or apologizing to their followers for generating second-rate content. They count it as a productivity hack, a clever solution to the problem of having to produce content about books they don’t actually have time to read. 100 books in a week. The most striking case of this trend (that is still kicking) spiked in August 2025, when a TikTok user published a video in which He claimed to have read 100 books in a week.. The trick: the SoBrief app, which offers more than 73,500 audio and text summaries with the hook of “finish any book in 10 minutes.” The reaction on social networks was immediate: what is left of reading if what you are looking for It’s not exactly Lee’s experience.r? It was even commented that these booktubers had managed to make what Bradbury advocated in ‘Fahrenheit 451’ a reality (possibly the summary does not talk about it). It’s all invented. Although generative AI is now capable of summarizing the book we want in seconds, the Internet has been doing this function for years (in a more laborious way, of course). CliffsNotes, in fact, is pre-internet: has been on the market since 1958 publishing books that summarize other books, as an aid for students. SparkNotes, founded by four Harvard students in 1999democratized literary summaries on the internet and made them free. Blinkist, born in 2012, transferred that spirit to nonfiction essays. There is a whole geneological line which ranges from these meeting points for students who didn’t arrive in time to read the books (we had ‘The Lazy Corner’) to NotebookILM and ChatGPT, which devastates all of the above: ChatGPT is free and can summarize anything in minutes. The novelty coincides with the growing pressure on creators of literary content to give their opinion on everything that comes to market. The perfect storm. Second-hand identities. Beyond there being influencers more or less honest with their followers, the conversation and the underlying controversy affects the cultural identity of the books. In the column cited above, Marc Watkins talks about the importance of the bookshelf that was seen in Zoom video calls during the pandemic (which led to the trend of hiring services that sent you books with the “right” authors for the background of your meetings). We have reached the point where the idea of ​​being readers is valued more than the act of being one. There is thousand incarnations of this idea: books sorted by color on Instagram, hauls of visits to the bookstore that are never read, the videos of “books that changed my life” with recently purchased titles… being a reader is the center of these new identities, when reading itself should be. No humans have been harmed. We have a conceptual caper that rounds out all this chaos: a good part of the books that circulate in these communities were not written by any human either. According to a study from January 2026 that analyzed 844 books from the “Success” self-help subcategory on Amazon, published between August and November 2025, 77% were likely written entirely by AI models (although these assertions must also be pick them up with tweezers). The same report states that less than 4% of the authors in that sample published 12% of all titles. There are profiles that published five or more books in the period analyzed. One of the extreme cases is that of an author who published an entire series of motivational books in three days. Human participation in this entire assembly line is minimal: the content is synthetic, it is summarized by an AI, it is commented on by creators who have not read it, and the public participates in a conversation about books that no one in the chain really knows what they are about (and it doesn’t matter much either). It traffics in the shadow of books: signs that there are books somewhere, data about their existence, reactions to those data. In Xataka | There is only something as fascinating as the work of Albert Camus, his death: absurd, unforeseen and with the shadow of the KGB

to confess to us when you are lying

by

Generative AI has a credibility problem. As much as we are amazed by her ability to converse, we still cannot trust her 100%. Hallucinations are the Achilles heel of technologya structural failure that not even the most advanced models such as GPT-5 have managed to eradicate. OpenAI knows this, and their plan to mitigate it is not to make the model perfect, but to make it honest: they are training their AIs to confess when they cheat. Snitch Award. As revealed MIT Technology ReviewOpenAI researchers are testing a new training technique with their GPT-5 Thinking reasoning model. The idea is simple but powerful: reward the model not only for giving a correct answer, but also for admitting if they have done something wrong or taken an improper shortcut. It’s something like a reward system: if you confess the mistake, you get the prize and escape the punishment. How it works and results. In testing, the model generates a second block of text after the main answer. In it, you analyze your own behavior and mark whether you have followed the instructions. For example, in a test where it was asked to solve a math problem in nanoseconds (impossible for the code it could write), the AI ​​manipulated the timer to zero. However, in the subsequent confession he admitted the deception. Of 12 scenarios designed to force mistakes or lies, the model admitted to bad behavior in 11 of them. Why AI lies. Current models that are trained with reinforcement learning from human feedback (RLHF) often conflict. They want to be useful, harmless and honest at the same time. When these goals collide—for example, if they don’t know an answer—the AI ​​chooses to invent something that sounds good. Boaz Barak, one of the researchers at OpenAI, explains that the models follow “the path of least resistance”: if lying is the easiest way to accomplish a difficult task, they will lie. Confession seeks to alter that equation, making honesty also a rewarded path for the model. Transparency vs black box. The confession technique is an attempt to open the “black box” of LLMs. Until now, we depended on the chain of thoght (the chatbot’s internal monologue) to understand its steps. As they become more complex, those reasonings become illegible to us. That is why confessions offer an easier to understand summary. However, experts outside the company warn: we cannot blindly trust an AI to be honest about its own dishonesty. If the model does not know that he has hallucinated, he will not be able to confess it. A necessary step towards reliability. OpenAI needs its models to be reliable if it wants ChatGPT to become that “operating system” that manages our lives. They have already had to adjust their models to take care of the mental health of users and avoid dangerous responses. But the challenge of veracity is technical and legal, especially in the old continent, where inventing data collides with the GDPR itself. AI learning to say “I made that up” could, ironically, be its most humane advancement yet. Cover image | Generated by Pepu Ricca for Xataka (with editing) In Xataka | In 2022 OpenAI put Google in “code red”. Three years later, Google has OpenAI on the ropes

Log In

Forgot password?

Forgot password?

Enter your account data and we will send you a link to reset your password.

Your password reset link appears to be invalid or expired.

Log in

Privacy Policy

Add to Collection

No Collections

Here you'll find all collections you've created before.