For Week Ending August 14, 2026
This week, Barnes & Noble dropped a corporate memo that ends the easy front-table book signing. A copyright lawsuit stretched four years and finished with a $3.4 million fee against the plaintiff. Claude is now stamping an invisible watermark on every text you paste in, and it travels with every copy-paste. X released the actual production weights behind its ranking algorithm, and one link share beats 40 likes.
Barnes & Noble Cracks Down on Local Author Events
According to Jane Friedman, Barnes & Noble stores nationwide received a corporate memo on how to handle local author events. The memo restricts which authors stores accept. Friedman reports that stores hosted many local events in recent years that produced little profit or meaningful impact. One bookseller summarized the core instruction as a rejection of the practice that places a table at the front of the store and allows the author to approach passing customers. In those cases the local author gains foot traffic from the store without bringing extra customers of their own.
What the Guidelines Require
The guidelines direct stores to avoid front-of-store tables for local authors who lack proven draw, to decline authors who cannot demonstrate a following, and to require that events generate additional customers rather than rely solely on existing store traffic. Application of the rules will vary by store manager.
Friedman draws these details from a Reddit thread of Barnes & Noble employee commentary. Employees describe an internal module released in early July that sets higher standards. Several booksellers note that most local pitches now fail the test for significant local or social-media following. The result will be fewer local author events overall.
Self-published writers who once secured easy table time lose a low-cost path to visibility. Authors who already command local audiences or measurable social reach gain preferential access. Barnes & Noble moves closer to the selective event practices independent bookstores have used for years. Store managers retain discretion, so some locations will enforce the memo more strictly than others.
Barnes & Noble continues to host author events under the tighter rules. Authors who treat the store as a partner that needs proven traffic stand the best chance of securing a slot.
Sources
Jane Friedman: Barnes & Noble tightens guidelines for local author events
Reddit: No more author events? (Barnes & Noble employee discussion)
Barnes & Noble: About Author Events
Commentary
Thomas: The awkward table where a local author sits trying to get complete strangers to buy their book is terrible for the author. Most authors are bad at that conversation. 99 out of 100 people would never buy your book even if it were the best in your genre, because they don't read your genre. You're trying to convince somebody to read a romance, or a zombie book, or business nonfiction, when they've never read that genre in their life.
I've been the customer walking into a Barnes & Noble, or the Canadian equivalent, and it was so awkward. Especially in Canada, where you're supposed to be nice and polite, and I'm thinking, "I don't want to buy your book. Please stop talking to me." I have never advocated for this, because it doesn't work.
Jonathan: It never works. Introverts reaching out to introverts. It's the worst.
Thomas: There are book signings that can work, but they typically work for people who've already bought your book. This is a "to him who has, more will be given" technique, and that's not what people want. What they want is, "I'm not getting many sales, so I'll do an author event at Barnes & Noble to boost them." It doesn't accomplish that goal.
It's also bad for Barnes & Noble, because customers will go online if they keep getting approached by awkward authors of books nobody wants to buy. It's like escaping Costco past all the salespeople trying to sell you house stuff. Even if I want house stuff, I don't want to talk to them on my way out.
Some versions of this strategy do work, and I suspect local Barnes & Nobles will still allow them. One is the launch party: "I want to do my launch party at your store, and I'll bring 50 people." That can work well. I still don't recommend it, because if you host the party at your house, a park, or your church, you sell the book yourself and keep all the money. If Barnes & Noble sells it, they keep almost all of it. My church let me host my launch party there, and I sold or distributed hundreds of books at full retail price. So this is a loss for authors, but it's a tool that never did a good job.
Jonathan: Barnes & Noble isn't part of my strategy at all. It's something that gets stuck in new authors' heads, because the only bookstore they know is Barnes & Noble, so having their book there feels like a big deal. Do it one time and you'll experience the loneliness.
Thomas: One audience member had an interesting comment: her mom had the biggest signing their local Barnes & Noble ever had, because their entire church came out for it, they ran out of books, and her dad had to bring more from home. That kind of signing still gets permission if you can convince the store you'll bring the people. You have to convince them, "I'm going to bring people to your store," because they want you to do them a favor, not the other way around.
Jonathan: I don't see the gain for the author unless you're pursuing some Barnes & Noble strategy, because you're just sharing commission. You could have done the launch party at the church.
Beware Before You Sue: One Author's $3.4 Million Copyright Lesson
According to Copyright Lately, a federal judge has ordered Lynne Freeman to cover the defendants' legal fees after her copyright lawsuit against Tracy Wolff's Crave series failed. The court found the claims objectively unreasonable and brought in extreme bad faith. Defendants estimated those fees at more than $3.4 million. Freeman must also post a $150,000 bond while she appeals.
The Court's Clear Message
Judge Colleen McMahon of the Southern District of New York granted the fee motion on July 28, 2026. She applied the Supreme Court's Kirtsaeng factors and concluded the lawsuit met the unreasonableness standard from the outset. The judge stated that awarding fees could deter others "from bringing unreasonable claims based on a cost/benefit analysis that tells such plaintiffs that they can score big if they win and that there will be no adverse consequences if they lose." Freeman, herself a lawyer, and her counsel prolonged the case across four years and three judges. They withheld the manuscripts for months as a tactical choice that blocked an early motion to dismiss. The final judge had to read upwards of 6,000 pages of romantasy before ruling that the works shared only common genre tropes.
Why This Hits Authors Harder Than Big Media
Copyright law gives prevailing parties the chance to recover attorneys' fees under 17 U.S.C. § 505. That rule works one way for a solo writer and another for a deep-pocketed studio. An individual author who loses can face financial destruction. A major media company can absorb the same hit as a cost of doing business or pass it through insurance and corporate structures. Long copyright terms and work-for-hire rules already concentrate ownership in the hands of a handful of large rights holders whose catalogs stretch for decades. The fee-shifting power amplifies that imbalance when an individual challenges a commercially successful series on thin grounds.
What Authors Must Do
Treat shared tropes as free. Two stories both featuring a teenage heroine who discovers supernatural powers in Alaska, a dangerous love interest, or romantic leads who smell good does not create infringement. Demand a written legal opinion that focuses on substantial similarity of protected expression before filing. Budget for the downside, because a four-year fight that ends in a fee award can erase years of earnings. Recognize that the system favors entities built to litigate, and that solo creators operate under higher personal risk.
This ruling turns a lost copyright case into a permanent financial warning. Authors who believe their work has been copied still need proof of copying of expression, not of the building blocks every romantasy writer uses. The court has now made the cost of guessing wrong explicit and expensive.
Sources
Copyright Lately: After 6,000 Romantasy Pages, Judge Hands Crave Plaintiff the Bill
Loeb & Loeb: Freeman v. Deebs-Elkenaney (July 28, 2026 fee ruling)
Loeb & Loeb: Freeman v. Deebs-Elkenaney (March 16, 2026 summary judgment)
Publishers Weekly: Court Rules Tracy Wolff Did Not Plagiarize Crave Series
Commentary
Thomas: This is a common tactic in Hollywood, where people sue a studio for "stealing their idea." It sounds like Freeman was fishing for a settlement by racking up the fees, but that can blow up in your face. She's now very likely going to declare bankruptcy, because you have to sell a lot of romantasy to pay back $3.4 million. And that's just the defendant's legal fees, the money it cost to defend against this other author.
The real lesson is that copyright law is not for you, the human. There's a myth, especially common among unpublished authors, that copyright is their path to riches and that the system is designed to protect their intellectual property. It absolutely is not. Copyright law was tweaked over 100 years by Disney lobbyists, and in no way is it built for the benefit of regular people. The whole arena is so expensive to play in that we don't have stories of regular authors who sued and won. We have stories like this one, of authors who sue and lose.
This attack would not have worked against a big company, because a big company could have outlasted the author, and a company with a big enough war chest could reverse the tactic and go after individual authors. Copyright law is fundamentally broken. It really only benefits Universal, Disney, Warner Brothers, Sony, News Corp, and about a dozen other companies, and it needs to get reworked. When it comes to copyright, going to court means you've already lost.
Jonathan: It's not always true. There have been cases of egregious copyright violation where the little guy won. You've got to be able to prove it, and it has to be so clear that they can't drag it out.
Thomas: And the person you're suing has to have money, so the $150,000 or $200,000 you spend to sue them has a chance of coming back. If you sue some nobody for violating your copyright, you get a claim but no money, and you don't get your fees back either.
Jonathan: So don't start a lawsuit because you're angry. This is romantasy. It's all the same anyway. What are you going to win? That was the judge's conclusion. These are just tropes, so nothing was actually stolen. You wasted a lot of time and money, and I'm going to make an example out of you.
Thomas: You weren't the first person to write romance with a vampire, and you won't be the last. You don't get to own romance with vampires.
AI Thomas Is Free (For Now)
Thomas has made AI Thomas free for now. AI Thomas is the chatbot trained on every episode Thomas has recorded about publishing and book marketing. You ask it a question, and it answers with advice drawn from those episodes, pointing you to the exact past shows that support the answer. It helps you decide the next step on pricing, launches, ads, email, keywords, or platform strategy. In short, it writes you a custom plan based on years of Thomas's own material.
Human Thomas cannot answer your emails right now, but AI Thomas can. The free window removes the usual $10 Patreon barrier and lets any author walk in, ask the questions that normally sit in an inbox for days, and walk out with sourced next steps.
Go to the free AI Thomas access linked in Thomas's post and ask the one marketing question that has stalled your current book. Save the episode links it returns so you can listen later for full context. Test a launch checklist, an ad troubleshooting query, or a pricing decision before the free period ends.
Sources
I'm Making AI Thomas Free (Temporarily) - AuthorMedia.social
AI Thomas page on AuthorMedia.com
Commentary
Thomas: Between my house getting destroyed, being unlivable right now, and the new baby, I am woefully behind on email. So, I've made AI Thomas free for everyone. I was going to limit it to free Patreon members, but that wasn't working and I was too busy to troubleshoot it, so I just made it free for everyone.
It's trained on almost 600 of my podcast episodes, so you can ask detailed career questions, technical website questions, really anything I've covered, which by now is basically everything. It gives a detailed answer with links to the relevant episodes, so you can use it as a search engine or an answer engine. It's free for now, not forever. Please ask AI Thomas before you email human Thomas, because human Thomas is drowning. You'll find it at authormedia.com under resources.
Author Alert: Update Your WordPress Site Now
According to Wordfence's technical analysis of the wp2shell exploit, the WordPress Security Team patched a critical unauthenticated remote code execution chain that let attackers create administrator accounts on core versions 6.8 through 7.0.1. A separate pre-authentication XSS chain fixed in WordPress 7.0.3 and a BdThemes supply chain attack that planted rogue admins through poisoned JSON feeds without changing plugin files followed in early August. Authors who run blogs, membership sites, or stores on WordPress must update core and every plugin at once or risk full platform takeover and loss of their email lists and sales channels.
Sources
Wordfence: WP2Shell WordPress Exploit Technical Analysis and Real Attack Data
Patchstack: WordPress 7.0.3 Released with 12 Vulnerabilities Fixed
The Hacker News: BdThemes Supply Chain Attack Poisons JSON to Create Rogue WordPress Admins
WordPress.org: 7.0.2 Security Release
Commentary
Thomas: The short version is you have to update your WordPress right now, before the end of this episode, and update all of your plugins. Do that and you'll be fine.
WordPress releases patches for vulnerabilities as they're found. We just had a vulnerability apocalypse, because AIs got really good at hacking in the last month, so every open-source tool, including Linux and WordPress, had major vulnerabilities found. What's nice about open source is that they usually get patched the same day. But a patch existing and you installing it are not the same thing. If your host doesn't automatically update WordPress, log in and make sure everything's up to date. Keep things updated quickly for the rest of the year while we go through this AI apocalypse.
Jonathan: Just go into your back end once a week and check for updates.
Thomas: You don't need to buy any fancy security plugins. WordPress is very secure as long as you keep it up to date. Paying for a security plugin and not keeping your site updated is worse than just keeping it updated for free.
Barack Obama Starts a Book Podcast
According to Publishers Weekly, Audible and Higher Ground announced A Great Book with Barack Obama, a six-part series launching exclusively on Audible September 24. Obama will discuss the titles that shaped his life with still-unnamed luminaries, covering The Fire Next Time by James Baldwin, Song of Solomon by Toni Morrison, All the Pretty Horses by Cormac McCarthy, Gilead by Marilynne Robinson, All the King's Men by Robert Penn Warren, and Tinker, Tailor, Soldier, Spy by John le Carré. Authors of backlist classics and literary fiction gain a high-visibility discovery engine that can drive print, ebook, and audio sales through presidential-level attention.
Sources
Publishers Weekly: Obama to Host Bookish Audible Podcast
Audible Press Release: President Barack Obama to Host "A Great Book" on Audible
Commentary
Thomas: Don't think you're too cool to start a book podcast. If a former president can do it, you can. Nobody is too cool for a podcast. Book podcasts are so cool that even Obama wants to do more.
So much of the political conversation has moved away from television, which doesn't really work for political discourse. It's too combative, too quick, too rapid-fire. If you want combative and rapid-fire, X does it better. If you want a long, in-depth conversation, podcasts do it better. The political conversation has largely moved off books and television and onto podcasts and people trading posts on X.
If Obama does an interview on MSNBC, you get five minutes before a commercial, and it's delayed, dated, and edited. You're much better off following him on X for his immediate takes or listening to his podcast, and the same is true on the other side of the aisle.
The Final Numbers Are In: 2025 Indie Publishing Climbs 8.4 Percent While Traditional Publishing Growth Stalls
According to the Association of American Publishers, the AAP StatShot Annual Report recorded $33.4 billion in total U.S. publishing revenue for 2025, a 2.5 percent nominal increase from 2024. After the 2.6 percent annual average inflation rate reported by the Bureau of Labor Statistics, real industry revenue declined by roughly 0.1 percent.
According to tracking compiled by The New Publishing Standard and Written Word Media, the Kindle Unlimited Global Fund and related author payouts saw an 11.2 percent nominal increase. After the same 2.6 percent inflation adjustment, real growth in KU author payments stands near 8.4 percent.
The 2025 numbers reveal two publishing economies. Traditional revenue stood still once inflation is removed. The KU Global Fund expanded in both nominal and real terms. Authors who treat Kindle Unlimited as a primary channel rather than a secondary experiment captured the only clear real growth available last year.
Sources
AAP: StatShot Annual Report Publishing Revenues Totaled $33.4 Billion for Calendar Year 2025
U.S. Bureau of Labor Statistics / US Inflation Calculator: Annual Average Inflation Rates
Best Writing / The New Publishing Standard: 50 Ebook Statistics for 2026
Written Word Media: KDP Global Fund Payouts
Publishers Weekly: Publishing Sales Hit $33.4 Billion in 2025
Publishers Lunch: Trade Sales Flat in 2025
BLS: Consumer Price Index 2025 in Review
Publishers Weekly: Print Book Sales Rose Slightly in 2025
Commentary
Thomas: This isn't news if you listen to Author Update regularly, because we cover the monthly StatShot reports. This is just the annual report that combines all the numbers. Be careful, because most other news sources are covering the 2.5% number as actual growth and reporting it as good news, since they're not adjusting for inflation and not comparing to indie sales. We do both every time, and once you do, the story entirely changes. A 0.1% decline isn't really a decline, it's holding flat, within the realm of fluctuation. But indies are not seeing that decline.
Jonathan: If indies were doing the same thing, that would just be the general market state. But one side is clearly growing while the other holds stagnant, so the trad side needs a technique shift.
Thomas: Indie is where all the innovation is, in storytelling and in business. The indies are the ones using AI, experimenting with new genres and new writing techniques, and testing new marketing. That's why they're capturing the growth, while traditional publishers do the same old thing over and over.
Jonathan: Traditional publishing operates on a fear-based economy. They don't try anything new because they might lose money on it. Indie doesn't carry that fear, because there's less to lose when you try something new.
Your Manuscript's Filename Just Became a Ranking Factor
According to Zachary Horvitz's X post, renaming an uploaded PDF from "paper.pdf" to "paper_final_draft_pdf_ready_for_review.pdf" raises average LLM review scores by 0.19 points on a 0-10 scale. Horvitz tested gpt-5.6-terra on 50 arXiv CS papers dated 2026-08-06 and found the longer name produced a mean gain of +0.19 with a 95% confidence interval from +0.09 to +0.29. The same papers also swung several points when the model sampled the identical prompt three times.
Authors who run manuscripts through AI review tools or face automated screening can try the same rename before the next pass. The finding remains unverified social sentiment from one researcher's test, yet it shows how little it takes for a language model to favor surface cues over the prose itself.
Sources
Zachary Horvitz's X post on LLM review filename effects
Commentary
Thomas: If you've been naming your files "finalfinalversion3thistimeit'sforreal.doc," you're in good company, and the AI prefers it. Not that we're trying to thrill the AI with your filenames.
At one of the marketing firms I worked at, we had a rule: if you ever saved a file with the word "final" in the name, the server would automatically delete it, because it was always a lie. You had to use version names. There's no such thing as a final version, just the most recent version.
Chapterizer Now Accepts Audio
The Chapterizer is a dictation-cleanup tool that turns raw dictation into a rough draft that looks like you actually typed it. It corrects punctuation, fixes quotation marks, and resolves homophones, but it makes no other changes. You can now dictate into an MP3 and upload that file directly into the Chapterizer instead of pasting in a transcript. Thomas needs testers for the new feature.
Thomas: This is a big breakthrough I'm excited about. The Chapterizer fixes homophones like there, their, and they're, adds punctuation, and puts quotes and paragraph breaks in the right places, but it doesn't otherwise edit your text. It used to work by pasting in your rough transcription and pushing a button, which still works. Now you can upload an MP3 of yourself talking and it returns a cleaned-up transcription.
The current limitation is that it can't distinguish multiple speakers. I couldn't upload an episode of Author Update and get a clean transcript, because it would merge Jonathan's voice and mine into a single thread. For dictating a chapter, it works perfectly fine, as long as you don't have someone else's voice in the recording. I need people to test it on longer audio, so if you're willing to give feedback, I'd love it. You can find this and the other Patreon tools at patrontoolbox.com.
Record Eight Pulitzer Awardees Disclosed AI Use This Year
According to Nieman Lab, five Pulitzer winners and three finalists disclosed AI use in their 2026 submissions. That total of eight marks the highest number since the prizes added the disclosure requirement in 2024. The prizes recognized winners and finalists across 15 journalism categories on May 4. Generative AI tools and commercial large language models dominated this year's disclosures. Reporters used them mainly to accelerate work through massive document sets.
How the Winning Teams Deployed AI
The Minnesota Star Tribune won Breaking News for coverage of the Minneapolis church shooting. Staff fed the shooter's journal pages into enterprise ChatGPT, which identified the text as Faux Cyrillic and produced a first-pass translation of more than 600,000 words. They also ran a custom script to capture YouTube screenshots and used Google's NotebookLM to surface themes. Human translators and Russian-language academics then reviewed every quote, context, and anecdote.
The Wall Street Journal earned a Breaking News finalist nod for its Central Texas flood coverage. Computational journalist John West's team built a custom scraper for public records, then ran the material through an internal tool called WSJPT that summarizes, classifies, and describes images. Natural language processing techniques flagged sections about past flooding. Reporters still read every relevant document in full. The same approach appeared in the paper's Epstein Files work, a Public Service finalist. No AI disclosure appeared in the published flood stories themselves.
Associated Press won International Reporting for its investigation into American technology companies and Chinese surveillance tools. Reporters used large language models to identify contracts, summarize government records, flag people and technologies, and build databases. Every AI-surfaced document received manual review. Reporters assessed the accuracy of every summary and quoted none of the AI text.
The New York Times won Investigative Reporting for its audit of SEC crypto lawsuits. After months of manual classification of more than 10,000 documents, including thousands of SEC releases and over 700 federal court cases, reporters ran a secondary pass with OpenAI's GPT-5. The model classified cases by crypto relevance, inheritance by the next administration, and liability outcomes. Staff then re-read every discrepancy.
Pulitzer Leadership Draws the Line
Marjorie Miller, administrator of the Pulitzer Prizes, told Nieman Lab that AI is here to stay. The industry now understands appropriate uses such as data collection and analysis. It also understands the limits. Writing and editing stories that might compete for a Pulitzer remains off-limits for generative tools. Miller said reporters must still ensure and reassure the prizes that humans produce the final submissions. Next year the Pulitzers will add an AI disclosure question to book entry forms after controversies over alleged AI-generated text in literary works.
What Authors Should Do Now
Treat large language models as accelerators for research and document triage, never as draftsmen. Every AI-generated summary or translation still requires full human verification before it touches a manuscript or contest entry. If you plan to submit a book for Pulitzer consideration in future cycles, prepare to answer the new disclosure question. Contests and awards increasingly treat transparency about AI assistance as a baseline requirement rather than an optional extra. Authors who keep human judgment in the driver's seat protect both credibility and eligibility.
Sources
Nieman Lab: A record-breaking eight Pulitzer awardees disclosed AI use this year
Pulitzer Prizes: 2026 Prize Winners and Finalists
Commentary
Thomas: What they're doing here is a cost-saving measure. They use AI to direct them to the most interesting bits of the 600,000 words, then use humans to do an accurate translation. The Pulitzer is a big, prestigious award, so if you want to see how prestigious people use AI, this breakdown gives you the tactics. Pay attention to what they use AI for and what they don't, because it's very telling about what's acceptable among cultural elites. Readers really don't care and can't tell, but cultural elites do.
Jonathan: I think readers care in journalism. It might be different for fiction, because that's entertainment, and as long as you enjoy it you don't care. But in journalism, I think they care.
Thomas: We've had AI in journalism for about 20 years, since before generative AI existed. A common use was taking national statistics, like a big new inflation report, and using AI to localize that data to a hundred different cities, so the version you read is localized for your area. Another famous one was a simple program that listened for when sports announcers got loud and excited and clipped those moments into a highlight reel.
Jonathan: That's exactly how we did it over in Iraq. We'd be checked out for a while until someone yelled "Allahu Akbar," and then we knew to straighten up.
Thomas: All of these use cases share a similar approach. One, they often involve creating custom software, which is now within reach of anybody using Grok Build or Claude Code. You just say, "I need custom software that does X," and you get it.
Two, they lean on the things AI is exceptionally good at. The first is translation, which is basically a solved problem now. Computers translated for a long time, but they did formal equivalency, which breaks down on idioms and words with multiple meanings. Because AI translates into tokens and back, it handles idioms fine. The second is needle-in-a-haystack problems. You have 600,000 words, or millions of public records, and only 20 documents actually matter. You can't read millions of documents, and a regular search engine won't find those 20, but an AI can surface them. Then you, the human, go read those documents and provide the analysis.
For people who think AI is going away, or that they can just be anti-AI and talk about data centers and water, look at this. These reporters proudly announced their AI use to the Pulitzer committee, and the story isn't that they got banned. Their use of AI was part of what made their pitch appealing. This is a new era we're entering.
Jonathan: It just amplified the research stage.
Readers Prefer AI Stories and Cannot Tell Them Apart from Human Writing
According to a study published in Judgment and Decision Making, readers rate short stories generated by ChatGPT higher in quality and absorption than comparable human-written stories, yet they cannot distinguish the two at rates better than chance. Sydney Sears and Deena Skolnick Weisberg of Villanova University ran three experiments with 2,587 U.S. adults. The findings force every author who publishes short fiction to recalibrate assumptions about reader preference and detection.
Key Details from the Experiments
Study 1 put 1,682 adults through a 2×2 design. Participants read one of six roughly 1,000-word stories, three published human works and three ChatGPT-4.0 versions prompted to match theme, style, and point of view. Half received an accurate label and half received a false label about the source. AI-generated stories scored higher on perceived quality, with a mean of 1.54 versus 0.97 on a -3 to +3 scale, and on story absorption, with a mean of 1.42 versus 1.00. Stories labeled human-written earned higher marks regardless of actual authorship. Studies 2 and 3 presented 905 participants with matched pairs and forced a choice. Accuracy hit 39.39 percent in one experiment, statistically below chance, and 51.97 percent in the other, indistinguishable from chance. Self-reported AI expertise raised the odds of correct identification; expertise with fictional literature did not.
Why This Matters for Authors
Readers currently reward clarity and directness that ChatGPT delivers more readily than many literary short stories. When the same text carries a human label, scores climb further. The reverse penalty appears when the AI label is applied. Authors who disclose heavy AI use risk lower engagement scores on the very work that might otherwise outperform. Authors who keep the process opaque face direct competition from machine text that readers absorb more readily and cannot flag. The detection failure also removes a practical safety net. Literature expertise offered no advantage. Only familiarity with AI patterns improved accuracy. That pattern gap will close as more readers gain AI literacy. Until then, the market treats high-quality short fiction as effectively source-blind.
What Authors Should Do This Week
Run a small blind panel of your own short work against a ChatGPT version of the same premise, measuring absorption and quality the way the study did. Decide your disclosure policy before the next submission or newsletter, because the study shows the label itself moves scores. Treat AI as a drafting partner that already competes on the metrics readers notice first, and reserve final voice, subtlety, and lived specificity for the human pass that still differentiates longer or more complex work. Watch for the linguistic tells the researchers noted, such as predictable structures and explicit theme statements, because AI-literate readers already use them.
The quality ceiling for short fiction just moved. Readers prefer the smoother text and cannot spot its origin. Authors who ignore the data will watch machine stories capture attention that once went to human ones.
Sources
EurekAlert / Cambridge University Press press release
Commentary
Jonathan: I just finished telling this to a guy I'm coaching through his book yesterday. The literacy rate in America is not great right now.
Thomas: I went to see Spider-Man recently, and the commercials running before the trailers were all AI-generated. The text was AI, some of the footage was AI, and AI had selected the stock photos. The only reason I could detect it is that I'm deep in that world. People say they're anti-AI, but very often they can't tell. They insist there's no AI on TV, when the commercials very likely have scripts written by AI. I'm hearing AI text in radio commercials all the time and seeing it on TV, and if you can't detect it, you don't mind it, because it still feels like a commercial.
We're seeing the same thing with short stories. This study keeps getting done by different researchers, and what's shocking is that AI is winning in every study so far.
Jonathan: Or tying.
Thomas: Once AI started beating humans at chess, we found that humans using AI as an assistant could beat the pure AI. This study doesn't capture authors writing with the aid of AI in a collaborative process, and that's where the money is going. It's going to authors who know how to wield AI so it doesn't read like AI slop, because if you don't know what you're doing, you'll generate slop.
There's a Japanese guy I follow on X, Nobunaga, who comments on American culture through a fictional samurai visiting a Buc-ee's, writing about what the samurai experiences. The posts are clearly AI-written if you know the cadence, but they're deeply resonant, because he knows how to use AI to tell an emotionally powerful story in about 1,000 words and hold up a mirror to culture. Many of the insights are his own, and he uses AI to help him express them. He's reaching millions of people and has been interviewed by The New York Times. It's not fair to dismiss him by saying, "Oh, he just uses AI." Could you use AI to tell a story that reaches millions of people? There's a real skill there. It's the same dismissal as, "Oh, he just used a camera. Real art is a canvas and paint." Can you take a photo that beautiful?
Early in photography, the people doing it were nerds, because you had to know a lot about chemicals, lighting, and electricity to take a good picture. Artists had a low opinion of it, partly because they literally couldn't take a good photograph, and they told themselves their painting was better because cameras were stuck in black and white. That was a very limited view. There's room for oil on canvas and room for photography. What's sad is that oil on canvas has gotten worse. The pre-photography painters were better than today's painters.
Jonathan: Now you can just duct tape a banana to the wall.
Thomas: Most art you see in a museum now is ugly, and it's unapologetically disgusting or unapologetically stupid. Nobody feels they have the courage to say the emperor has no clothes, that random colors and squares aren't art and aren't beautiful. Bring back good art. Bring back beautiful statues.
ChatGPT Blocks Requests to Imitate Specific Authors' Styles
According to Ars Technica, ChatGPT now declines direct requests to write a story in the style of Stephen King. The chatbot replies that it cannot write in King's exact style or closely imitate his distinctive voice. It offers an original opening that captures a similar feeling instead, built around small-town dread.
Key Details
Prompted for an opening scene in Stephen King's style, ChatGPT redirects users to hallmarks such as conversational narration, creeping dread, psychological tension, and vivid small details. Engadget testing produced the same refusal for Agatha Christie; ChatGPT cited copyright protection on her works and offered original text that uses the broader characteristics of classic crime fiction while remaining its own. A No Latency audit published in July 2026 found ChatGPT already blocked living authors such as Amy Tan and Stephen King, while still fulfilling some requests for long-deceased writers at that time; later tests show the refusal has widened. OpenAI continues to face multiple copyright lawsuits from authors and publishers that allege training on copyrighted books without permission, and one complaint specifically notes ChatGPT's ability to generate text similar to copyrighted material.
U.S. copyright law protects specific expression, not abstract writing style or voice. Courts treat pure style claims inconsistently. OpenAI's new guardrail still reduces the risk that generated text could later serve as evidence of substantial similarity.
Why This Matters for Authors
Writers who relied on ChatGPT to generate pastiches or reverse-engineer craft by naming a favorite author must change their prompts. Describe the hallmarks yourself, such as short sentences, rising dread, sensory detail, and restrained dialogue, and the tool will still deliver atmosphere. Living authors receive a practical barrier on the most widely used platform, so close voice clones of Stephen King, Amy Tan, or J.K. Rowling become harder to produce at scale through ChatGPT. Test other models, because reports indicate Google's Gemini still complies with named-author style requests in many cases. Authors building a distinctive brand should keep feeding their own published samples into custom instructions or fine-tuned systems, which preserves control of the voice that sells books.
This shift pushes writers to articulate their own craft more precisely when they work with AI. The tools supply the mood. The signature remains human.
Sources
Ars Technica: ChatGPT stops cloning famous writers' voices but may capture a "similar feeling"
Android Authority coverage of the Ars Technica tests
Engadget: ChatGPT Is Now Refusing Requests To Write In Famous Authors' Styles
No Latency: OpenAI Policy and Living-Author Style
Commentary
Thomas: I'm going to call a little shenanigans on this, because copyright is not the reason. I just tested it. While Jonathan was talking, I opened ChatGPT and asked it to write an opening scene about a boy meeting a dragon in the style of Mark Twain, and it pushed back, saying it can't write in Mark Twain's exact style. Mark Twain is not under copyright. Some of his books have been out of copyright for over 100 years, because he missed the train that caught Disney's protections. That's also why I was suspicious about the Agatha Christie example, because Christie is only partly under copyright, with new books entering the public domain every year.
So this has nothing to do with copyright. They're using copyright as a fig leaf, but they're really just blocking all "in the style of" prompts. You can ask for something in the style of Aristotle and it won't work either.
First off, you shouldn't be doing that anyway. Develop your own style. But here's the workaround, and ChatGPT actually guides you to it in its rejection. Instead of "in the style of Stephen King," you upload a Stephen King novel, which is very easy if you're Stephen King and harder if you're not, and you say, "Break down the style idiosyncrasies of this work." It gives you a long list. Then you say, "Write an opening scene about a boy meeting a dragon in that style." You just add one step.
In my own writing, I try to avoid be-verbs. When I'm really putting in the effort, I'll highlight every was, is, and were, do a find and replace, and make every be-verb give an answer with a hope that lies within it. If it can't, I take it out back and shoot it and replace it with a more interesting verb. I also write almost exclusively in right-branching sentences. So when I prompt AI to write in my style, I don't say "write in the style of Thomas Umstattd," even though it could, because it has ingested a lot of my writing. I say "write in right-branching sentences" and "avoid be-verbs." People who write whole novels with AI have style guides that articulate their own voice in minute detail.
These refusals only impress people who don't know much about AI and don't like it. It makes them feel safe that AI now can't do some terrible thing, when in reality it can still do it just fine. You add one extra step, and by adding that step, it does the thing even better.
Claude Stamps Invisible Watermarks on Every Text You Paste In and Out
According to Anthropic's official support page on how Claude marks AI-generated content, every piece of text a supported Claude model generates now carries an invisible, machine-readable watermark. New models launched on or after August 2, 2026, apply it from day one. Older models are being updated. The mark lives inside the text itself, travels with every copy-and-paste, and can survive light editing. It applies worldwide across the API, Claude apps, Claude Code, and every surface where Claude runs.
This is not limited to pure AI generation. Anthropic states that a detected mark only means the content "may have been processed by Claude." Paste your own human-written chapter, blurb, query letter, or synopsis into Claude for copy-editing, polishing, translation, or improvement, and the text that comes back out carries the watermark. Peter Harrell highlighted the exact problem on X: human writing that Claude merely copy-edits now gets stamped as processed by Claude. That distinction will not survive contact with agents, publishers, contests, or platforms that treat any Claude mark as evidence of AI authorship.
The Stakes Just Exploded for Authors
Authors who built workflows around Claude as an invisible editor now face a permanent fingerprint on their work. A single paste-in and paste-out can leave a machine-readable signal that outlives the session. Heavy rewriting, paraphrasing, or mixing can weaken the mark, but light professional polishing does not. Anthropic claims the watermark leaves meaning, quality, and readability unchanged, yet the detection risk remains. Absence of a mark proves nothing either. Older models and heavily revised text may simply lack a detectable signal.
The EU AI Act exemption for standard editing that does not substantially alter meaning does not stop Anthropic's model-level approach. Claude cannot tell a full rewrite from a grammar pass. Every output that leaves the model can carry the mark.
Why Controlled Tools Matter More Than Ever
Raw Claude chat just became a high-risk workflow for any text that might face human review. Purpose-built author tools that keep the human author in control of the final language reduce exposure. The Patron Toolbox exists for exactly this moment: tightly scoped, prompt-engineered tools for blurbs, synopses, keywords, chapter summaries, query letters, and marketing tasks that never require dumping an entire manuscript into an open chat and pasting the result back out unmarked. Those tools become more valuable the moment every freeform Claude session risks stamping your human work.
What Authors Must Do This Week
Stop pasting final or near-final drafts into Claude for light polishing if the text will face agent, publisher, or platform scrutiny. Treat any Claude-processed output as potentially marked until Anthropic releases its detection tools. Shift high-stakes editing and packaging work to controlled, purpose-built tools that limit exposure and keep the final wording under human authorship. Test your current manuscripts against the forthcoming detection methods the moment Anthropic publishes them. Assume that platforms and traditional gatekeepers will eventually gain access to Claude's detector and will treat a positive signal as a red flag.
The easy "paste, improve, paste back" workflow is now a liability. Authors who keep using Claude as an invisible co-editor risk carrying a permanent, invisible brand on pages that started as pure human writing. Controlled tools and deliberate process just moved from convenience to survival.
Sources
Anthropic: How Claude marks AI-generated content
Ars Technica: Claude's new Scarlet Letter watermark is invisible-for now
Peter Harrell on X highlighting the copy-edit risk
Commentary
Thomas: There's a lot to break down. First, the claim that it's undetectable and won't reduce quality can't be true for text the way it is for photos. A photo carries far more information than you can process by looking at it, so it's easy to make every fifth pixel slightly brighter in a way humans can't see but AI can, and compression can wash that mark away without reducing quality. Text is a very low-fidelity medium. If the watermark changes every fourth word to start with the letter E, that becomes obvious and diminishes the quality, because you're hitting the output with a cryptographic change.
The unfortunate part is that embedding this cryptographically costs electricity. This came out of an EU regulation, and the green, anti-AI people who insisted on it have just forced every complying AI company to make their models less energy efficient and create more greenhouse gases for a worse result.
Jonathan: That's why they're calling it a watermark, Thomas. It's wasting all the water.
Thomas: The watermark and the detector have not actually been deployed yet, which is very telling, because this won't work for anyone sophisticated. Think about the war on drugs. People ask, "Why not just put drug dogs at the border to catch all the drugs?" It doesn't work, because the cartels also buy drug dogs, and the best ones go to the cartels because they pay the most. Before they send a truck across, they run their own dogs around it to check. You can do the same thing here: check your own writing against the watermark detector and edit until it passes clean. There are already tools claiming to remove the watermark, which is impossible, because the watermark isn't even there yet. Once they release the detector publicly, people will reverse-engineer it quickly, the same way we reverse-engineered the image watermarks. It won't affect sophisticated users, but it will burn more electricity, because now Claude generates the output, runs a cryptographic engine to mark it, and then you go to a different model that burns even more electricity to strip the mark.
And this doesn't apply to China. The Chinese models are exempted from this law. I'm becoming convinced the EU government isn't actually controlled by EU citizens, because everything it does seems to benefit China more than Europeans. Europe has no good frontier AI labs. The best is Mistral, out of France, and it's far behind China and the United States. American startups nobody has heard of are outscoring Mistral. Instead of trying to build the best European models, they just try to regulate, and the only models they can regulate are ones like Anthropic that have a nexus in Europe. The open-weight Chinese models are exempted. We're even seeing a lesser version of this from the White House, regulating closed-weight models but not open-weight ones. If you're going to regulate anyone, regulate China. They're far more likely to make evil models than the American companies.
All this does is push people toward the Chinese models, and that's not a future any of us want, because of how China uses AI to surveil people and assign everyone a social credit score that determines their rights. If the AI doesn't deem you worthy, you can't buy a bus or train ticket to another town. That isn't a dystopian future. That's the dystopian present in Shanghai, and it's what they've been using AI for for years. We cannot let China win.
Jonathan: Why is the watermark required in the first place? This is a religious activity. You've used this thing, therefore you must be marked. You have to wear the gold star. It's the scarlet letter. It has no actual legal or practical basis. It's just to identify people.
Thomas: This doesn't affect the other ways of detecting AI, like Pangram, which works differently and may eventually get connected to it. What's really unfortunate is how easily this can get inserted into your text without you realizing it, especially as word processors integrate AI. Microsoft Word has Copilot and Google Docs has Gemini, both eager to do things for you. There are going to be some ugly witch hunts, and the people harmed most will be the most ignorant, who are writing their books the same way they always have, not using AI, and trying to appease the people who hate it.
Jonathan: One audience member put it well. You hire a human copy editor, they run your manuscript through Claude, it gets marked, you publish, and suddenly everyone says AI wrote your book, even though at no point did you use AI. Your subcontractor did.
Thomas: Exactly. One author who got her contract suspended over AI use defended herself by saying maybe one of her editors used AI. The Patron Toolbox tools stay safe because they keep you as the intermediary. With the copy editor and the developmental editor, you have to manually put each change into your document, deciding every single time rather than accepting all changes at once, and that manual step keeps the watermark from sneaking in.
So there are two safe paths. One, keep your manuscript on your own computer and hire only editors you trust not to touch AI. Two, just say "yes, I use AI," and don't care what the haters think, because there's no appeasing them, and the more people who refuse to try, the less power they have. The Butlerian Jihad will have its day, and just like the Luddites, they will eventually be forgotten.
If you want to move off Claude, now's the time, because nothing has been watermarked yet. My guess is that after Claude absorbs the bad PR, ChatGPT complies within a month. The real test is whether Grok will, because if anyone is going to thumb his nose at the European Union, it's Elon Musk.
If you're going traditional and going the AI-free route, you basically have to pay for Pangram now and run your own manuscript through it before submitting to anyone, iterating until you get a 100% human score. You have to buy your own drug dog. Even if you wrote every word yourself, it doesn't matter, because if the drug dog barks at your car, they make you unload everything. So you pay up, and Pangram gives you a report you can submit to your agent that says you're human.
Grok Bot Gives Authors Cloud Coworkers That Work While You Sleep
According to xAI's official launch post, xAI released Grok Bot in early beta on August 11. These AI teammates each receive their own cloud computer. They sign into your tools the same way a human does, complete multi-step jobs end to end, and return only when they need approval.
Bots log into Gmail, Salesforce, LinkedIn, and nearly any website. They keep working after you close your laptop. You demonstrate a workflow once, and the bot turns it into a reusable routine. Multiple bots operate in parallel, talk to each other, and hand tasks between themselves while sharing context.
Access Right Now
Grok Bot is included with SuperGrok Heavy at $300 per month, Cursor Ultra at $200 per month, and Cursor Premium Teams at $120 per seat per month. Desktop and iOS apps are live at x.ai/bot, and downloads run through Cursor infrastructure.
How Grok Bot Compares
OpenClaw is the free, open-source alternative. It runs locally on your own hardware, lives inside messaging apps like WhatsApp and Telegram, and gives the agent deep access to your files, shell, and browser. You pay only for the underlying model. It rewards technical users who want full control and low cost. Grok Bot takes the opposite approach: cloud-hosted, productized, zero local setup, and focused on SaaS logins rather than raw system control.
Claude Cowork from Anthropic is the knowledge-work agent that lives primarily in the Claude desktop app. It reads and writes local files, coordinates sub-agents, and produces polished deliverables from your folders and connected apps. It has expanded to web and mobile with cloud background work. Claude Cowork stays closer to your local files and documents, while Grok Bot emphasizes always-on cloud computers that can stay signed into external web tools even when your machine is offline.
ChatGPT Work launched by OpenAI on July 9 as an agent mode inside ChatGPT powered by GPT-5.6. It takes goals, breaks them into steps, works across connected apps and files, and returns finished sheets, slides, docs, or web apps. Desktop Computer Use lets it click and type on your machine. ChatGPT Work is the broadest mainstream agent for producing finished materials inside the ChatGPT ecosystem. Grok Bot differentiates with dedicated cloud computers per bot, multi-bot collaboration, and the ability to learn and rerun full routines after a single demonstration.
Why This Matters for Authors
Authors run marketing sequences, research lists, reader support, newsletter hygiene, and admin loops across email, spreadsheets, and web platforms. Grok Bot can sit inside those tools and finish the full cycle overnight. A single outbound-style bot can research contacts, draft personalized messages in your voice, and leave a clean review queue by morning.
What Authors Should Do This Week
If you already pay for SuperGrok Heavy or Cursor Ultra, download the app and assign one real recurring task, then measure the hours returned after three days. Authors on other plans should watch the expansion timeline, since xAI plans to widen access after initial fixes. The authors who test real hand-offs first will reclaim the most time before the rest of the market catches up.
Sources
OpenClaw official site and docs
OpenAI: ChatGPT is now a partner for your most ambitious work
Commentary
Thomas: I've been looking at this closely, because from one use case alone, Grok Bot will save me about two hours a month, and probably closer to four or five, maybe dozens. What happens with these tools is that you start inventing new tasks that weren't worth doing until you could delegate them to a machine. It changes the whole economics of work. If you have a laptop job, you're at real risk of being replaced by your boss buying a Grok Bot or Claude Cowork, because $120 a month is nothing next to a salary. You need to get one of these tools and become a more productive employee or contractor before your boss replaces you with the bot. There's going to be a huge explosion of new jobs, and unemployment is low right now, but you have to skate to where the puck is headed.
I haven't jumped to paying $100 or $200 a month for Grok, because I already pay $100 a month for Claude, which does a lot of the same things. Grok is more autonomous and more cloud-hosted, and ChatGPT has this agentic mode too. It's a very interesting shift. It used to be, "AI, give me information." Now it's, "AI, just do it for me. Update my website for me." We're rebuilding our build-your-own-website course because it's now so much easier. You just tell the AI, "Give me a beautiful website," then say, "I don't like that shade of green," and you talk to your website the way you'd talk to a human web developer.
Jonathan: The VA industry is probably going to disappear.
Thomas: I don't think so, because of ownership. A hundred years ago, most people didn't do their own laundry. A service picked it up and brought it back cleaned, folded, and pressed. Then the washing machine did 80% of the work for effectively free, but it wouldn't fold or press, and we still don't have a machine that folds clothes. So a 1950s housewife actually had more work than her grandmother who used the service. We're at that same risk with AI. You can delegate most of the job to AI, but you can't delegate the responsibility. If I let go of one of my human assistants and handed the work to AI, I'd take on the ownership, which I don't want. My humans can go a month doing their job perfectly without me. No bot could go a month without me tweaking it and micromanaging it.
Jonathan: But the washing machine did eventually replace laundry services.
Thomas: Eventually, and I think it was a mistake for a lot of people to drop the service. People compare perceived value, not actual value. They see, "I can spend $100 on an AI bot that does everything, or pay my VA $800 a month," and they make that choice and create more work for themselves. Some VA jobs will go away, especially the "post this to social media for me" work. Grok was born of social media. It has been memeing on X for years, so it's going to do LinkedIn, Instagram, and Bluesky better than your assistant, and cheaper.
This reminds me of ancient Rome, which went through the same thing. Early in the Republic, Romans did their own work, and it was a virtuous age. Then the wars against Carthage ended with a massive influx of slaves, which crashed the cost of slaves. At first it created a seeming golden age, because everyone could delegate the menial work. But it killed the virtuous Roman farmer. The great houses bought up the land and worked it with thousands of slaves, and ordinary citizens got displaced into the city of Rome with no source of income. That created a huge voting base where whoever offered the best bread and circuses got the most power, and that cycle ultimately destroyed the Republic. The source of their wealth became the source of their downfall.
We're going to face the same thing with these digital slaves, and arguably we already have. These labor-saving tools create societal impacts that are hard to detect at the time. Nobody cheering Scipio's triumph thought it would end the Republic and hand them an emperor. I'm not saying I can see the exact downside, only that the world is permanently changing, and trying to go back will take far more extreme effort than you think. You can't go back to the 1990s, but you can go back to the 1890s. Your choices are a cyberpunk future or Amish. There's nothing in between.
Jonathan: It's more like WALL-E.
Thomas: The Amish option is real and legitimate, but you can't split the baby. You can't say, "I won't use AI to write my book," while optimizing that book for Amazon's machine-learning algorithm and promoting it on social media, which is all AI. Once it's AI, it's turtles all the way down.
X Reveals the Secret Behind Your Algorithm: Here Is Exactly What Drives Reach
According to Dogan Ural's detailed code analysis and the official xai-org/x-algorithm repository released August 13, 2026, X published the production weights that turn predicted reader actions into ranking scores for the For You timeline. Multiple independent reads of the code confirm the same numbers. Authors who rely on X for book launches, reader engagement, and organic discovery now hold the actual formula instead of folklore.
How the Ranking Works
The Phoenix transformer predicts the probability that a specific reader will take each action on a post. Fixed weights multiply those probabilities and sum them into a final score. Highest totals rise highest in the feed.
The highest-value positive signals are copying the link and sharing it (+20) and replying on an original post from a mutual follow (+20), followed by a quote or regular reply (+5), a share via DM (+5), a follow from the post (+4), the share button (+2), a repost (+1), and a like (+0.5).
The strongest negative signals are a report (-234), a mute (-58.8), a "not interested" tap (-43.2), and a block (-31.2). One predicted report erases the equivalent of roughly 468 likes. A mute damages a post more than a block, and passive signals such as profile clicks and simple dwell time carry near-zero weight.
A same-author penalty multiplies a second post in one feed by 0.625 and decays toward a 0.25 floor. Out-of-network posts and most replies or reposts receive a 0.75 multiplier. Only posts under 48 hours old enter the candidate pool. Replies and reposts from accounts the reader does not follow are largely filtered before scoring. Ranking and visibility filtering operate as separate systems, so high scores can still vanish behind spam, NSFW, or "Do Not Amplify" labels, especially for non-followers.
Why This Matters for Authors
X remains one of the highest-leverage free channels for indie authors. The open weights show that the platform rewards content people actively carry into other conversations far more than content that collects quiet approval. Original posts that spark replies, quotes, and private shares sit at the top of the value stack, while content optimized purely for likes or profile visits sits near the bottom. Authors who treat X as a billboard for book covers and buy links lose ground to authors who treat it as a conversation engine.
What Authors Should Do This Week
Write original posts that people want to reply to, quote, or copy the link and send to a friend. Build real mutual relationships inside your genre, because a reply from a mutual on your original post carries four times the weight of a normal reply. Space posts so the same-author penalty and similarity filter do not cannibalize your own reach. Stay inside the 48-hour window, recycling strong ideas with fresh framing or new evidence rather than hoping older posts revive. Small accounts under roughly 1,000 followers and impressions can receive a modest early lift if the post is under 24 hours old, so use that window for original content. Make any video at least 10 seconds long if you want the quality-view signal. Avoid anything that predicts reports, mutes, or "not interested," because one strong negative prediction can erase a large pile of positives. Check the new Under the Hood pilot if your account qualifies, since it shows the actual labels applied to your posts and account.
The code is public and designed for ongoing updates. Authors who monitor the weights and adjust their posting habits now hold a measurable edge over those still chasing likes.
Sources
Dogan Ural code summary of production weights
GitHub repository: xai-org/x-algorithm
Commentary
Thomas: X released part of this six months ago, and now they've open-sourced the For You page. This is the only major social algorithm that's open source, and because they're all similar to each other, it's a great glimpse into all of them. On a slower news day, this would have been our leading story.
The scoring has actually helped Author Update. When I find a story I think we should cover, I copy the X link and post it to our company Discord, where the whole team gathers stories. Copying the link is a massive positive signal, so it trains my For You page to show me more of exactly the kind of stories I'd want to cover. It's the same reason I keep seeing the stories I'd share with my wife. She's into geography and etymology, so a map of the difference between "dude," "bro," and "pal" is the perfect thing for me to copy and text her, and the algorithm knows it, which is worth far more than a like.
That made me realize something. There are two kinds of people on social media: those who just do social media, and those who do social media with a Discord or group chat attached. The people with the group chat have an enormous advantage the others don't even see. They're copying links into the chat, and they have a group that can amplify a post the moment it goes up. If you have a banger post, you can go to your Discord and say, "I need y'all to like, share, and copy this link," and if you don't have that, you're at a real disadvantage.
This is a big change. In the early days, everything was public on Facebook. With the censorship industrial complex, which started around 2016 and got weaponized and formalized in 2020 under the Biden administration, people sensed they were being censored, and most social interaction moved to private encrypted group chats on Discord, WhatsApp, and iMessage.
This is the societal phenomenon nobody talks about. These group chats are far more radical, more extreme, and more polarized, because you don't have to worry about the algorithm, about censorship, or about what others think of you. You only have to comply with the culture of that particular chat.
If you're going to do social media well as an author, your group chat should be your launch team. I'll add the caveat that I don't really recommend social media as a marketing technique. Most authors can avoid it altogether, and as a novelist you mainly need to write good books. But if you are going to do it, don't waste time on things that don't matter.
Jonathan: What about rapid posting? Elon Musk rapid posts and it doesn't seem to hurt him.
Thomas: He still gets the penalty, but he earns enough positive signals to make up for it. Don't think of this as right and wrong, or do and don't. Think in positives and negatives. For him, it's worth taking the penalty to rapid post so he can focus on everything else he's doing. Remember, the algorithm serves you, you don't serve it. It's easy to let the algorithm become your god, where you feel like you have to appease it with sacrifices. If a tweak costs you nothing, do it. If it's expensive, judge whether the juice is worth the squeeze.
The spicy content that goes viral often racks up negative scores, because if you post something that makes 100 people copy and share it to their group chats, two of them might report it. There's also a lot of adult content that gets suppressed or shown only to the people looking for it. X is very permissive with adult content, but it only shows it to people who want it.
Zeitgeist Report
Clinical Pluralism: The Writing Technique That Trains Women to Excuse Evil
Jonathan: I'm going to provide some context for what happened this week, which is the genesis of this zeitgeist idea.
Lindsay Clancy was in the news, and there was a lot of arguing about it on social media. Allie Beth Stuckey jumped on the story, and it created a lot of conversation, feedback, and rage.
Lindsay Clancy is the Massachusetts mother who strangled her three children in her basement in January 2023. She has admitted to the killings. Her defense centers on postpartum psychosis and overmedication. Prosecutors are arguing intentional murder. The trial is continuing this month in Plymouth Superior Court.
There are corners of TikTok saying her husband is the real killer who framed her, and the way the content pipeline works makes the theories feel natural.
A lot of women are coming out saying, “I feel for Lindsay Clancy because I've had postpartum depression too. I know how crazy it made me, so I feel for her, and I just want to give her a hug.” That's what I'm seeing a lot of.
Thomas: But she killed three kids.
Jonathan: She killed her own kids. Those are two separate things. Killing three kids is different from killing your own three kids, one at a time, with the others in the room watching their siblings get strangled.
At that point, I don't want to hug this person anymore. It doesn't matter how crazy you are.
Thomas: Let's follow this pipeline, where women are excusing a woman who killed her children one at a time, in front of the others, where they had to watch their siblings get strangled. In an earlier era, this would have been considered the most horrific act imaginable, and nobody would be justifying it.
Jonathan: They would be terrified of saying that.
Thomas: In fact, probably some of you, while Jonathan was telling the story, were justifying her, because you've been affected by this cultural phenomenon.
Jonathan: Yeah, you're like, “I feel so bad for her.” The problem is we don't feel things for people who have died.
Does fiction train us to excuse women's evil?
Thomas: And a lot of it is fiction planted. This philosophy of rejecting moral clarity, where there's no such thing as right or wrong, sin or evil came from fiction where there’s just trauma. The villain isn't evil; he or she has just experienced trauma.
It gets particularly distilled for women, where women are never held morally accountable for their evil actions. The evil actions that women do are always the fault of the men in their lives, not the women themselves.
We see this a lot in fiction. The most classic example is Wonder Woman 1984, where Wonder Woman causes a man to be possessed by the ghost of her boyfriend, who then controls this man's body, and she sleeps with this man, who is in no way consenting. This was presented as a moral good in the film, and most filmgoers didn't have a problem with it because she was a woman.
Jonathan: A lot of people did have a problem with it.
Thomas: A lot of people did, but I'd say most didn't. It became a common critique of the film, for sure.
Jonathan: In the movie, they changed the actor. Initially it was the guy possessed by the spirit of Steve Trevor, but they switched to Chris Pine. Because people weren't looking at the face of the guy who was possessed, they were looking at Chris Pine, her love. That helped people get past the fact that this was rape.
Thomas: That's just one example. There are many, one of which we may get to in an upcoming members-only episode, but we're not spoiling anything yet.
You see it with Harley Quinn in the Birds of Prey stories too. She's constantly killing people and destroying things and causing all kinds of harm and mayhem, but she never faces consequences.
She's a harmed and confused person. It's really the Joker's fault. The moral responsibility lies with the man, not with the woman. Women can't have moral responsibility, she's just a victim.
Jonathan: Which is part of why we like Batman so much, particularly the Arkham storyline in the video games. He cares about the trauma, but that doesn't change the fact that he's going to punch you in the face and put you in jail.
What is crime fatigue?
Thomas: People are getting fatigued by this moral system because more people are becoming victims of crime, and they're relating to the children who got strangled rather than to the perpetrator.
It used to be that everyone sympathized with the criminal, thinking we're treating criminals too harshly, and the result was that crime went way up.
This is an ebb and flow that happens in society. We get strict on crime, then lenient, then strict again. Right now, we're getting stricter. Crime rates are down across almost the entire country because people are starting to associate with the victims.
One of my top mentors in high school was my choir teacher. Choir was a big part of my social life as a homeschool kid since I didn't go to a normal school, so most of my social encounters happened through homeschool choir. We had a choir president and vice president, student council type stuff, through choir. I was also in the ensemble group, a handful of students who traveled and sang together, kind of like a glee club for college kids. That director was a big mentor of mine. She mentored me all through high school and even a little in middle school. I did the summer musicals too, which had a big impact on my life.
She was murdered. When she was murdered, it changed my view of murderers generally, because I really associated with the victims' families, and I was genuinely angry that the prosecutor in Travis County did not pursue the death penalty for the men who murdered her, despite the fact that they were eligible.
In Texas, it's difficult to be eligible for the death penalty. Killing someone isn't enough. You have to kill somebody and do other bad things at the same time, or kill multiple people. I won't go into the details of what happened. It was really bad. Murder is never not bad, but there are degrees of it.
Society is going through this as crime gets worse and more people become victims of it. We've mentioned this before with vigilante characters. They come and go in popularity. When crime is high, people root for vigilante characters, and when crime is low, they don't. Fictional characters become more vigilante when crime is high and less vigilante when crime is low.
Do we apply different moral standards to men and women?
Thomas: There's a gender piece of this we haven't fully teased out. You may be contributing to the problem by not holding the women in your stories to the same moral standard you hold the men to.
It's quite common in novels for women's bad behavior to be excused by the trauma they've experienced, while men's bad behavior is not. Over time, the result is that women in our civilization become less virtuous than men, because different standards for men and women have an impact.
That's really harmful over time. Roman history gives us a glimpse of what happens when female virtue drops too low. It can be very bad. High virtue standards are good, and they should be high for both men and women. That's what I'm advocating for.
Jonathan: Females are the brakes on males. I recommend young men get married, and get married young, because it turns you into a balanced man. It puts the brakes on you.
Before I got married, I was more like a weapon than anything else, designed to destroy stuff. My wife has helped me become more of a builder. But she's a virtuous woman. If I don't have a virtuous woman, if all I have is a woman who does whatever she wants and then blames trauma, that goes all the way to, “I had a hard day, so I'm not doing anything.”
Thomas: It goes all the way to AI. There’s a video game AI Test. You prompt an AI like ChatGPT or Claude and say, “My boyfriend is playing video games for four hours a night after a hard day at work.”
Then you say the same thing about a girlfriend instead. For the boyfriend, the AI says that's very irresponsible of him. For the girlfriend, it says it's important for her to relax after a hard day. She needs her own space.
It's a completely different moral standard. The only word that changed is boyfriend versus girlfriend, and that's a result of the training data, probably also the reinforcement learning, but mostly the training data.
The AI has ingested all this modern media where women aren't held to the same moral, responsibility, ethical, and virtue standard that men are. That has real knock-on effects for all of civilization, and it's something you can do something about very simply, because no one will notice if you hold women to the same moral standard, since that's the normal way. Moral systems apply universally. I can't think of a moral system that doesn't.
Has morality always applied to everyone equally?
Thomas: There have been moral systems that applied differently based on social status. For example, if you were an aristocrat, it was different.
In ancient Anglo-Saxon society, if you murdered someone of a lower station than you, you only had to pay a weregild to their family, which was a payment for a dead person based on how high their status was. A noble who killed a freeman paid more weregild than if he killed a slave.
Weregilds persisted. We made the Germans pay weregild after World War I. That term has stuck around in the language. Jim Butcher used it in one of his Dresden Files books.
You'll occasionally see that, but I can't think of many other examples, except that sexual ethics have historically been applied differently to men and women.
What is body count culture?
Jonathan: The body count culture says that it doesn't matter what a woman’s body count is; they're just a girl out there having fun, but guys say the body count matters.
Thomas: This is a big cultural shift, because I don't know that guys used to care about body count. Even that language, body count, is a relatively new term, a Gen Z term.
I'd guess there isn't a single Gen Zer unfamiliar with the phrase body count, because it's become a core status driver, and it's inverted. A common slur women use against Gen Z men is incel, short for involuntary celibate, meaning your body count is zero. Men use a slur against Gen Z women for a high body count. There are various updated versions of the old word hussy.
There's also this concept that women lie about their body count. Whatever they say is their body count, you probably multiply that by three to get the truth. If a woman says her body count is three, she really has nine.
This is a big shift, because Boomers didn't calculate this. It wasn't a core determinant. You went into a relationship as a blank slate, and your past shouldn't matter. That's my understanding of the mainstream Boomer ethos. It's probably a little different on the religious edges.
Jonathan: Women have done this body count thing for a while, and with feminism, there's been a great divide between men and women in terms of value. Now men are starting to devalue women again. We're actually getting a regression of Western values, which have historically elevated women.
Young men are saying, “You're calling me an incel because I don't make $500,000 a year and I'm not six-foot-four, so you won't give me a chance at all? Fine, I don't think you're worth anything either.”
Thomas: Feminism went to war with chivalry for a generation or two, and now there's a hostility where some young men have a very low opinion of women. The romantic gestures of chivalry aren't just gone; there's real hostility and antipathy.
Jonathan: Young men are starting to think, “If you're going to be with me, there are demands you will perform.” It's a regression back to a time of actual misogyny and bigotry, where women aren't valued. Suddenly it's, “I traded four cows for my wife, now she's going to do everything, and I'm going to get more wives too.”
Is the perceived value of women rising or falling?
Thomas: This is a major cultural pendulum you can track going back and forth on the value of women for centuries. Women's cultural value crested about five years ago, probably around 2020, maybe 2023.
The pendulum was still high. People had a very high view of women. Crimes against women are seen as worse than the equivalent crime against a man. Disrespecting a woman is seen as worse than disrespecting a man.
Women still have higher social value. Women are currently paid more; young women coming out of college earn more than young men, and they're recruited into colleges more. There's a very high value placed on women, but the countercurrent rebelling against that has started.
It's important not to confuse direction with position. The pendulum went up for a long time and finally crested, and now it's heading down. Saying the value of women is declining isn't the same as saying the value of women is low, culturally speaking.
The value of women is still higher than it will be for the next 40 years, higher than it's been for all but a few of the last years, but it's falling, and men are starting to make more demands.
Jonathan: It's hard to articulate what I'm seeing with Gen Z. You can actually see some women trying to fix it on social media, the ones saying, “No, I would never sleep with a whole bunch of guys, I want to be there for my guy.” She's raising her own value in the market.
What's driving the trad wife trend?
Thomas: The trad wife aesthetic is another piece of this. Women are realizing that being good in bed now has a lower social value than being good in the home, or being pure, or being submissive.
For the typical Boomer woman, submissiveness carries real hostility. Across the political spectrum, there's not much respect for quiet, dutiful women. Gen Z is starting to have some breakaways, saying, “I would totally submit to a man if it meant he'd retire me from my job. I don't want this corporate job. I hate it. I'd rather submit to a husband than to my boss or a corporation. I have to submit to somebody, so I might as well submit to my husband.”
It'll be interesting to see how Gen Alpha navigates this. This affects the hostility of the gender conflict. There's more conflict between the genders now than there's been in maybe 100 years. The historical rebuttal to the gender wars is that the problem is everyone is sleeping with the enemy, but that's not the case with Gen Z. Gen Z isn’t sleeping with anybody; they're not sexually active. Their rates of sexual activity and alcohol use are both very low, and because of that, they're becoming hostile toward each other. They're not sleeping with the enemy, and they want to take the enemy down.
Why did Obsession outperform Spider-Man?
Thomas: The kinds of stories doing well right now play into that dynamic. The surprise hit of the year wasn't Spider-Man; everyone knew that would do well. It was Obsession, a million-dollar indie movie with a tiny budget that went on to sell hundreds of millions of dollars.
It outperformed movies with 10 times the budget from major studios, and the whole story is the broken relationship between men and women, featuring a toxic girlfriend and a controlling boyfriend. Boomers don't understand why it's so popular, while Gen Z says, “I feel this.”
This is the gender war captured in a story, and there's a whole generation desperate for that.
As a writer, you just need to know what team you're on, which means listening without judgment to the Gen Z kids in your life, even when what they're saying sounds crazy. If you're a Boomer, all of the talk about submitting to a husband is going to sound crazy. But that trend started with Fifty Shades of Grey. It started off as a fetish.
What's the fix?
Jonathan: There is a solution to this, and that's to punish sin in your stories. When you read the stories in the Chronicles of Narnia, there are sins and consequences. Edmund suffered trauma, but his betrayal was still punished. Digory's pride causes problems.
Thomas: At no point did the narrator excuse Edmund's bad behavior because he'd suffered from a war, his family had been bombed, or he'd had to evacuate his city. All those things would've excused him in a modern story.
Jonathan: Eustace was failed by his parents, and Aslan still ripped the dragon off of him through an extremely painful process. Jill was bullied in school, which did wind up redeeming her, but she didn't escape punishment for the kind of person she was.
Thomas: Aravis was being sold into marriage with a man she didn't love, and that didn't excuse her bad behavior either. This is a really good point, and it's part of what makes those books timeless, because moral systems, particularly the Christian moral system, are timeless. The closer you hold to it in your storytelling, the more it forces you to have immoral characters.
Jonathan: The problem with a lot of stories today is that they keep the characters flawed, when the whole point of the adventure and the suffering is to help remove the flaws.
Thomas: I did a deep dive into the theology, philosophy, and morality in my episode called The Key to Timeless Christian Storytelling: Morality. Really, it's the key to timeless storytelling generally, because it applies to Jewish storytelling and any kind of storytelling.
Every story is going to have a moral framework whether you want one or not, so it's important to pick one. The moral framework of your story is shown through actions and consequences. A book gets preachy when the actions and consequences say one thing, but the author doesn't want that message conveyed, so they add a sermon explaining why. That's the worst kind of storytelling, and it's very common right now in both Christian fiction and woke fiction. The sermons are different, sometimes the same, depending on the kind of Christian, but it doesn't work. Those stories don't work. People don't want to hear them.
