- That AI Thing
- Posts
- weekend ai reads for 2026-08-07
weekend ai reads for 2026-08-07
direct links are available on the web at https://thataithing.beehiiv.com/p/weekend-ai-reads-for-2026-08-07
đ° ABOVE THE FOLD: BOOKS
Writers Agree: AI Isnât a Novelist. So Why Is This Author Interested? â Naomi Alderman is a best-selling, award-winning novelist. She thinks AI writing is terrible. Sheâs fascinated anyway. / The Humanist Review of AI (23 minute read)
AI Companies Are Buying Antique Books, Ingesting Their Contents to Train Models, and Then Destroying Them at Incredible Scale â "The world's best AI training data is setting on a shelf." / Futurism (7 minute read)
Iâm Scared a Stranger Will Call My Novel AI, So I Built GitHub for Words (Meet VellumProof, Formerly WritHub, Lol) / dylanâs blog (6 minute read)
But hereâs where Iâve landed: I could spend forever building a fortress against the worst people, or I could build a good tool for the honest majority and trust that most people arenât the villain of this story. Iâm choosing the second one. Iâd rather bet on the human side of things and be occasionally wrong than treat every writer like a suspect â which is, youâll notice, the exact thing Iâm mad about in the first place.
related, AI Questions Spoil Book & Movie Heat On âCall Me, Iâll Hide The Bodyâ Novel / Deadline (12 minute read)
The End of an Era / Hugh Howey (42 minute read)
I now feel awful for anyone who wants to tell a story post-2026. You will forever be doubted. If you love an em dash the way I do, or you like flowy run-on sentences like I do, if you grew up reading Proust to let the words flow over you in all their glory, if you memorized sonnets so the iambic pentameter would soak into your bones, youâre gonna sound like a bot.
đ» QUOTES OF THE WEEK
But ultimately what matters most is solving the right problems.
I feel like all the real art Iâve made in my life came with some degree of risk and vulnerability, that showing other people the thing I made also gave them a mortifying access to my mind or soul, an access to either understand or misunderstand me as a person (both scary, in different ways).
đ„ FOR EVERYONE
LLM Counter-Defaults â set your own defaults
Your LLM came with settings you didn't choose. It agrees with you, sounds sure, and writes in your place. Each dimension below is one of those factory defaults. Move its slider to tailor the result to your own needs and values, then paste it into your LLM once. Every chat after follows your rules.
useful and simple-to-use, if this matters to you
related, no-ai-slop â Removes 20+ patterns of AI slop from any piece of writing. / petergyang, GitHub (4 minute read)
AI Scammers Are Better at Building Trust Than Humans â Researchers pitted a person against a Claude agent and found that, after a week of texting, the AI chatbot was more effective at creating âexploitable trustâ with others. / Wired (15 minute read)
Welcome To The Resistance: Meet The Workers Dodging (And Sabotaging) Their Employerâs AI Mandates â âWhen I copy-edit writing I take an extra twenty minutes to double check, then say I had Claude copy edit it as well. Everyoneâs really happy with my performanceâ / Aftermath (22 minute read)
related, Make Tech Work for Workers â AI Implementation Bingo Card Generator
Restaurants Are Going Bespoke as AI Slop Grows More Abundant | Eater / Eater (10 minute read)
âFrictionmaxxing,â a term that writer Kathryn Jezer-Morton coined to refer to a rejection of technologies of convenience in favor of slower and more analog processes, has escaped containment, coming to mean the general choice to choose pre-internet ways of creation and problem-solving. In this sense, it can refer to the choice of the handcrafted over the AI-generated.
related, The AI Aesthetic / Jim Nielsenâs Blog (3 minute read)
For example, the âshimmering textâ which in AI land implies a kind of âthinkingâ but is being repurposed to indicate any kind of asynchronous task (thinking, fetching, computing, etc.).
Your AI calorie-tracking app may be off by 345 calories â A meal photo may look accurate to an AI calorie app, but the real total could be hundreds of calories higher. / Science Daily (8 minute read)
đ FOUNDATIONS
TokenTown: How an LLM Works, as a City â A language model laid out as a city, one token at a time
TokenTown is an isometric city where every district is one stage of a transformer language model. A convoy carries a hidden state along the roads: it is cut into tokens at the docks, cast into a vector at the foundry, stamped with its position, then driven around the layer ring (attention, residual, feed-forward, residual) once per layer, before the stadium turns it into a probability distribution and the sampler picks one token. That token drives back up the feedback highway and the whole city runs again.
How to manage the risk of AI agents / Section AI (5 minute read)
It depends on your organization, your customers, and your appetite for things going wrong. But there is a framework I keep coming back to that helps make the decision concrete rather than abstract.
It comes down to two variables: blast radius and reversibility.
GPU Management: Why Idle GPUs Are the New Grounded Aircraft â Utilization, not intelligence, is the next real constraint in AI. / HuggingFace blog (14 minute read)
AI Can Run â Can your phone or Mac run AI?
Yes â if you pick the right model. See what runs locally on your exact phone, how fast, and which app to use.
đ FOR LEADERS
Starbucks bet big on an AI tool; 9 months later, it pulled the plug â The exclusive inside story of how a promising AI system flamed out, as told by dozens of Starbucks employees and the tech startup behind it. Plus, what leaders everywhere should learn from the experience. / Fast Company (18 minute read)
What is really happening to jobs? Separating AI hype from reality / Stanford Institute for Economic Policy Research (23 minute read)
related, AIâs real threat to jobs could be lower pay / Axios (5 minute read)
AIâs Execution Problem â The defining challenge of the AI era is no longer invention alone. It is executionâthe ability to translate intelligence into durable changes in how a business actually runs. / Time (8 minute read)
related (1), The new enterprise AI conversation leaders are having right now / UiPath (10 minute read)
Forward-thinking leaders are building governance into the work instead of adding it later, measuring what changes after AI is deployed, and treating production as the point where the learning really begins.
related (2), AI Leadership: Turning Investment into Value â The leadership practices that turn AI ambition into enterprise value / Accenture (8 minute read)
Engineering management after the cost of code collapsed / Karim Jedda (14 minute read)
đ FOR EDUCATORS
The University After AI â The uneasy marriage between research and teaching is coming undone. / The Chronicle of Higher Education (21 minute read)
On plagiarism â All Human Life Is an Imitation / Terry Eagleton, Unherd (10 minute read)
When I was a university teacher in the pre-AI epoch, a student presented me with an essay which was entirely plagiarized from a celebrated critic except for the outstandingly intelligent bits, which had been deleted. With becoming modesty, the student said he knew he wasnât up to much intellectually speaking and so dumbed down the essay he was hijacking.
Anthropic Launches Claude for Teachers. Why Some Critics Are Concerned / Ed Week (11 minute read)
related, Sally the Robot Was Coming to a New York School. Then the Plug Was Pulled. â Students helped design the A.I.-powered creation as a young female with dark hair and an upbeat personality. Then came the outrage. / New York Times (6 minute read)
Where Did All the Computer-Science Professors Go? â AI companies are stripping universities of their best researchers. / The Atlantic (10 minute read)
đ FOR TECHNOLOGISTS
Open Source Must Be Fun (Or It Will Die) / Mike McQuaid (8 minute read)
Instead: stop doing the parts of your project that are not fun. Maybe thatâs stable releases. Maybe thatâs release notes. Maybe thatâs backwards compatibility. Whatever it is: again, you donât owe anyone anything. You can do OSS however you want to do it.
Nothing Works and Everyone Is Euphoric / Piotr Chmolowski (5 minute read)
Software may have been better back in the day, but that was mainly because it was much simpler. Since then, we have kept coming up with new abstractions, new frontend frameworks, and more infrastructure complexity. The bar for âuser experienceâ has kept rising, but everything has become increasingly fragile.
Speaking to AI Agents like Cavemen Saves 65% of Tokens. We Test. / Jetbrains blog (7 minute read)
Advertised saving: 65%. Measured saving: 8.5%.
đ FOR FUN
Pomegranate â Sci-Fi Short Film / Gossip Goblin, YouTube (27 minute video)
After centuries of simulated wars, wives, and wonders, a jaded aristocrat visits his neighborhood dreamspool dealer in search of something, anything, that still feels real.
using human voice actors improves the outcome significantly
HINT Home Intelligence â Human expertise. An app that knows your home. Together for the first time.
Co-founded by Martha Stewart
best review: âterrible. looks like Al slop. I left more confused than I was before. what do i need to know my soil quality for im not a farmerâ
YoFont â Describe a font. Get the whole typeface.
đ§ż AI-ADJACENT
How Private Equity Lied to Destroy Your Childhood / Goldstream Jay, YouTube (17 minute read)
From the tragic downfall of Toys âRâ Us to the recent bankruptcy of Red Lobster, we unmask how these iconic brands were systematically destroyed at the hands of predatory Private Equity firms. We will break down complex financial engineering scamsâlike the Leveraged Buyout (LBO) and the Sale-Leaseback trapâinto an entertaining, cinematic investigation showing how modern wealth is extracted at our expense.
related, our favorite newsletter is Worse on Purpose, which explains why your mattresses, pet care, appliances, cookware, shoes, etc. got worse â and where to still get good ones; spoiler, the answer to why things are worse is usually P.E.
â