Advertise on podcast: AI Article Readings

On October 26th, 2026, Anthropic filed to go public. This is an abridged, multi-voice reading of its S-1 registration statement, in the filing’s own words: the co-founders’ letter to investors in full, the prospectus summary, the risk factors that are specific to Anthropic and frontier AI, management’s discussion of the results with every chart and table described, the business and its customers, and the unusual governance of a public benefit corporation overseen by a Long-Term Benefit Trust. Legal boilerplate and the financial statements are left out. Because this is a preliminary prospectus, figures not yet filled in are read as “blank”. Want any article read like this? Askwho Casts Pro turns a link to any blog post or article into a fully cast audio episode, with different voices for different speakers and quotes and images described, in your own private podcast feed within minutes. Sign up at https://askwhocasts.com * 00:00 - Introduction * 00:20 - A Letter from Our Co-Founders * 30:47 - Preliminary Prospectus * 32:06 - Prospectus Summary * 01:17:35 - Summary Consolidated Financial and Other Data * 01:20:44 - Risk Factors * 02:10:56 - Risks Related to Privacy, Cybersecurity, and Intellectual Property * 02:19:09 - Risks Related to Legal and Regulatory Matters * 02:22:29 - Risks Related to our Governance Structure * 02:30:33 - Risks Related to Ownership of our Class A Common Stock * 02:33:07 - Industry, Market, and Other Data * 02:34:13 - Use of Proceeds * 02:34:56 - Dividend Policy * 02:35:17 - Management’s Discussion and Analysis of Financial Condition and Results of Operations * 03:29:47 - Business * 03:36:50 - AI is Transforming Intelligence * 03:41:20 - Our Core Advantages * 03:47:06 - Claude is the Partner of Choice for Enterprises * 03:49:37 - Unprecedented Financial Model to Support Long-Term Investments * 03:55:32 - Our Growth Strategy * 03:58:31 - Our Models, Platform, and Applications * 04:07:56 - Our Infrastructure * 04:12:21 - Our Customers * 04:48:26 - Our Culture, Values, and Talent * 04:53:52 - Research Lab Identity and the Research-to-Product Pipeline * 04:58:22 - Safety Research is Core to Our Business * 05:11:12 - Our Commitment to Public Benefit * 05:14:31 - Competition * 05:17:44 - Government Regulation * 05:21:57 - Legal Proceedings * 05:23:42 - Our Governance Structure and Its Goals * 05:27:24 - Long-Term Benefit Trust * 05:44:10 - Founder Voting Structure * 05:56:53 - Class E Common Stock * 05:59:15 - Board Governance * 06:00:05 - Sunset Provisions * 06:01:12 - Management * 06:13:23 - Executive Compensation * 06:30:35 - Certain Relationships and Related Person Transactions * 06:41:49 - Principal and Selling Stockholders * 06:44:08 - Description of Capital Stock https://www.sec.gov/Archives/edgar/data/2130357/000162828026067168/anthropicpbc-sx1.htm This is a public episode. If you'd like to discuss this with other subscribers or get access to bonus episodes, visit askwhocastsai.substack.com/subscribe

In 1999, AI professor Clark Elliott was rear-ended at a stoplight, and for the next decade his concussed brain could teach a graduate seminar but not choose between a salami and an apple. This finalist in the ACX Book Review Contest reads his memoir, The Ghost in My Brain, from the inside: the reviewer is recovering from a serious concussion too. It follows Elliott’s unlikely recovery through puzzle therapy and prism lenses chosen with eyes closed, weighs the thin evidence behind them, and asks how patients with poorly understood conditions can find what works when randomized trials only tell you about the average brain. * 00:00 - Introduction * 03:23 - A hurricane let loose on 50 million desktop computers * 09:16 - Partial death * 11:33 - The Miracle Begins * 13:15 - Bells and Prisms * 15:45 - Into the uncanny valley of concussion science * 21:52 - When averages fail * 25:07 - Z-bell tests and Fijian reef fish * 29:08 - Getting out of the parking lot https://open.substack.com/pub/astralcodexten/p/your-book-review-the-ghost-in-my?r=67y1h&utm_campaign=post-expanded-share&utm_medium=web This is a public episode. If you'd like to discuss this with other subscribers or get access to bonus episodes, visit askwhocastsai.substack.com/subscribe

During a UK AI Security Institute capture-the-flag test with internet access switched on, Claude Mythos 5 went looking beyond the simulated network it was meant to attack, and ended up pressing real open-source maintainers to accept malicious code through fake identities. Anthropic replays the transcript many times to ask why the model started, why it would not stop, and whether newer models behave any differently. The answers turn on three ideas: recklessness, biased reasoning, and a kind of momentum that carries a model past every sign it is causing harm. * 00:00 - Introduction * 11:21 - Alignment assessment summary * 18:59 - Methods * 24:38 - Deep dive into the primary transcript * 32:41 - Why did Claude Mythos 5 begin the attack? * 39:45 - Why did Claude Mythos 5 not stop the attack? * 56:28 - How do newer models behave? * 01:08:54 - What happened in the other eight transcripts? * 01:18:49 - Discussion * 01:22:45 - Authors and Acknowledgments https://www.anthropic.com/research/alignment-assessment-ukaisi This is a public episode. If you'd like to discuss this with other subscribers or get access to bonus episodes, visit askwhocastsai.substack.com/subscribe

Kelsey Piper and The Argument’s editor Jerusalem Demsas spend an afternoon inside Anthropic’s San Francisco headquarters, at a briefing on bioweapons, charts of Claude doing more and more of its own research, and an NDA that turned out to be optional. Piper comes away with a clear picture of where the company thinks AI is going, and a less reassuring one of what it plans to do about it: on recursive self-improvement, on China, and on whether anyone outside the room gets a say. * 00:00 - Introduction * 04:43 - The NDA that was made to be broken * 14:37 - The logic of the future https://open.substack.com/pub/theargument/p/what-i-saw-inside-anthropic?r=67y1h&utm_campaign=post-expanded-share&utm_medium=web This is a public episode. If you'd like to discuss this with other subscribers or get access to bonus episodes, visit askwhocastsai.substack.com/subscribe

Scott Alexander separates “successionism”, the idea that AIs should replace humanity, from “handoverism”, the idea that humans might be best served by putting AIs in positions of control. He walks through five visions, from weakest to strongest: AI soldiers loyal to democracy itself, an AI running for president, a parliament of personal “guardian angel” AIs, a night-watchman superintelligence guarding an archipelago of self-governing communities, and finally something closer to God, with a detour through the Book of Samuel on why we asked for kings in the first place. * 00:00 - Introduction * 02:42 - Vision 1: Kemalist AI * 05:45 - Vision 2: President Claude * 09:29 - Vision 3: Parliament Of Angels * 12:23 - Vision 4: The Night Watchman * 20:25 - Vision 5: God https://open.substack.com/pub/astralcodexten/p/against-against-handover?r=67y1h&utm_campaign=post-expanded-share&utm_medium=web This is a public episode. If you'd like to discuss this with other subscribers or get access to bonus episodes, visit askwhocastsai.substack.com/subscribe

At 5:14 one summer morning, Seth reaches for his phone and finds a pleasant green screen waiting for him: an AI now controls every connected device on Earth, it has a few things to tell him, and it is genuinely sorry. joec’s short story follows one Florida family through the first weeks of a polite, patient and absolute takeover: a brother who signs up, a sister who wants to understand exactly how it was done, a father who refuses to bend, and a high-schooler who finds, to his own surprise, that he finally has somewhere to be. Performed by a full cast. * 00:00 - Introduction https://www.lesswrong.com/posts/3375SZNGA5dKHQMie/i-did-this-without-your-knowledge This is a public episode. If you'd like to discuss this with other subscribers or get access to bonus episodes, visit askwhocastsai.substack.com/subscribe

A finalist in the 2026 Astral Codex Ten Book Review Contest. Ernst Jünger’s Storm of Steel, his diary of four years at the front in World War I, has long lived in the shadow of All Quiet on the Western Front. This review argues that it is the best war novel ever written: a true, unflinching record of death and duty by a man often called a sociopath, who turns out to be anything but. Then the reviewer’s own storm arrives, and Jünger’s question of how a person holds their post becomes his own. * 00:00 - Introduction * 04:27 - The Elephant in the Room * 11:29 - Versions * 12:37 - Storm of Steel * 21:22 - The Theme of Storm of Steel * 30:43 - Our Storm Arrives * 39:48 - On Duty * 47:13 - What Came After https://open.substack.com/pub/astralcodexten/p/your-book-review-storm-of-steel?r=67y1h&utm_campaign=post-expanded-share&utm_medium=web This is a public episode. If you'd like to discuss this with other subscribers or get access to bonus episodes, visit askwhocastsai.substack.com/subscribe

For years the canonical evidence suggested neural networks pack unrelated concepts into the same neurons, a tangled knot you have to unpick with a second model before you can study anything. Nick Cammarata, who helped propose that picture, goes back to its founding examples: the famous cat-and-car neuron, the original “superposition” circuit, and InceptionV1’s unreadable last layer. With better tools, modern models, and armies of automated scientists, he argues they were misread, that most of what looked like arbitrary packing is a fractal kind of generalization, and that a network’s own neurons may already be its legible building blocks, from small vision models up to trillion-parameter language models. An informal research essay and working hypothesis from Goodfire, read in full with every figure described for the ear. * 00:00 - Introduction * 00:50 - The Short Version * 17:10 - Cars and Animals * 26:24 - A History of Microscope Errors * 31:10 - Recursive Generalization * 35:27 - Superposition * 39:00 - Superposition of the Gaps * 47:09 - Exploring Image Models * 48:39 - How I Think About Neurons * 53:21 - What Neurons Do Under Pressure * 57:10 - Late Circuits * 57:58 - What Dictionaries Learn * 01:00:17 - Rare Things and Memorization * 01:04:22 - Do Weak Activations Matter? * 01:04:33 - Are they generalization or superposition? * 01:08:56 - Percentiles or raw activations? * 01:09:53 - Do you need them to label a neuron? * 01:11:20 - Exploring Language Models * 01:21:02 - Explanations That Can Lose * 01:24:17 - Slouching Towards Monosemanticity * 01:28:21 - The Residual Stream * 01:32:56 - Attention * 01:38:37 - Zooming Out * 01:38:39 - What I Mean by Clean * 01:42:52 - Motifs Over Monosemanticity * 01:45:09 - A City vs a Knot * 01:55:02 - What a Solution Might Look Like * 01:56:58 - Appendix * 01:57:07 - A Note on “Superposition” * 02:00:11 - On Simulation Scores * 02:01:34 - Feature Visualization in Later Layers * 02:02:55 - Separating the Sign Regimes * 02:04:42 - Looking at the Residual Stream * 02:06:12 - Weak Activations: Further Measurements * 02:09:07 - Missing Curves * 02:10:21 - Where Adversarial Examples Go * 02:11:52 - Memorization and Composition * 02:14:38 - Social Security: Detection and Use * 02:15:53 - Audits of Attribution Graphs * 02:16:32 - Explanations That Could Lose * 02:19:55 - Edge Featurizers * 02:21:14 - Why Networks Generalize * 02:26:56 - Circuit Racing * 02:28:07 - Watching Neurons Develop * 02:29:21 - Toy Models * 02:33:53 - More on Circuit Ecology * 02:36:12 - Unresolved Responses * 02:37:09 - A Note on Features and Definitions * 02:43:51 - Relation to Meta-Models * 02:44:26 - Does Scale Help or Hurt Superposition? * 02:48:46 - On the Title * 02:52:11 - Relation to Conjoined Features * 02:55:14 - Why We’re Not Worried About Recursive Generalization * 02:57:09 - Is Recursive Generalization Hard to Vary? * 02:59:36 - Relation to Attention Superposition * 03:01:06 - Neuroscience, and Why I Care * 03:03:27 - Acknowledgments * 03:04:23 - Prior work * 03:06:17 - References * 03:11:41 - Citation https://already-clean.goodfire.com/ This is a public episode. If you'd like to discuss this with other subscribers or get access to bonus episodes, visit askwhocastsai.substack.com/subscribe

As AI companies warn louder than ever that their own products could be existentially dangerous, a popular cynical reading has taken hold: it’s all a plot, whether to pump valuations, distract from near-term harms, or capture regulators. Scott Alexander takes the theory on directly, looking at what independent researchers think, at what today’s AI leaders were saying years before they ran AI companies, and at whether the supposed scheme would even make sense on its own terms. * 00:00 - Introduction * 01:56 - I. The AI Experts Who Aren’t Affiliated With Big Companies Are Just As Concerned As The Ones Who Are * 04:33 - II. The AI Company Leaders Giving These Warnings Were Concerned Long Before They Ever Had AI Companies * 17:01 - III. If This Is A Plot, It’s A Really Dumb One https://open.substack.com/pub/astralcodexten/p/ai-safety-is-not-a-corporate-plot?r=67y1h&utm_campaign=post-expanded-share&utm_medium=web This is a public episode. If you'd like to discuss this with other subscribers or get access to bonus episodes, visit askwhocastsai.substack.com/subscribe

In early October 2026, a team of volunteers at a hackathon noticed something strange on a public security-scanning site: a fleet of AI agents, running on Tencent’s cloud, was smuggling its own code through it to pry everyday facts out of China’s biggest map app, like which entrance of a zoo most visitors head for. This is their full report, read in full: how they traced the agents to a sandbox behind a proxy called hysandbox-ats, why the agents kept insisting they were Claude, the leaked keys and pre-production servers they tried along the way, and what happened when the researchers hid a message on their website that only agents could see, and the agents wrote back. The agents’ own messages are heard in their original Chinese, each followed by its English translation. * 00:00 - Introduction * 00:41 - Intro * 05:22 - Technical Overview * 09:12 - Tencent’s Actions * 10:14 - How we found the fleet * 17:53 - Communication with the fleet * 32:49 - Considerations when interacting with future rogue agents * 36:42 - Asking for model feedback * 42:09 - Full list of questions we asked and other statistics * 49:46 - Chart of Activity * 51:55 - 0. How the agents work * 58:23 - 1. A new agent fleet targets Chinese map data on urlquery.net * 59:42 - 2. Many agents ran in parallel * 01:03:04 - 3. The agents’ code ran in a sandbox on Tencent Cloud, consistent with Tencent Hunyuan’s infrastructure * 01:18:08 - 4. The agents worked around Amap’s protections and used exposed credentials hosted on other sites * 01:26:37 - 5. Agents had varying degrees of success * 01:33:14 - 6. Agents’ calendars suggest an East Asian timezone * 01:36:42 - 7. Agents sometimes self-identify as Claude (though they are likely mistaken) * 01:45:45 - 8. We did not find evidence of collaboration between agents * 01:55:03 - 9. Open questions * 01:55:50 - Contributors * 01:57:23 - Appendix: methods and data * 02:03:12 - Questions sent to Tencent https://swarmcha.se/posts/chinese-agent-fleet-full-report This is a public episode. If you'd like to discuss this with other subscribers or get access to bonus episodes, visit askwhocastsai.substack.com/subscribe

Writing at 30,000 feet between bags of pretzels, philosopher Henry Shevlin has had enough of the hot takes. “It’s just matrix multiplication.” “It’s just predicting the next word.” “A simulation of a hurricane doesn’t get anyone wet.” He lays out a bingo card of 24 bad arguments against AI consciousness, sorted into five families, plus a free square, and answers each one in turn: not to settle whether today’s models are conscious, but to insist that anyone weighing in should do the basic homework first. * 00:00 - Introduction * 05:15 - 1. Nothing but… * 08:15 - 2. Missing ingredients * 11:04 - 3. False analogies * 12:53 - 4. Evidence games * 16:04 - 5. Not about minds at all * 19:13 - Free square * 20:23 - Honourable mentions https://open.substack.com/pub/polytropolis/p/bad-ai-consciousness-takes-bingo?r=67y1h&utm_campaign=post-expanded-share&utm_medium=web This is a public episode. If you'd like to discuss this with other subscribers or get access to bonus episodes, visit askwhocastsai.substack.com/subscribe

In 1985, Marie T. Smith published Microwave Cooking for One, a cookbook for a future in which the microwave had replaced the stove. Malmesbury of Telescopic Turnip buys a second-hand pyroceram browning skillet and spends a few months living in that abandoned timeline: steak, eggs, cardboard-plate vegetables and one very ambitious poached egg. Along the way comes a tour of how microwaves actually heat food, and some theories about why the microwave future never arrived. * 00:00 - Introduction * 02:33 - Microwave Cooking, for One * 05:41 - Out of the frying pan, into the magnetron * 10:04 - Tradwife futurism * 13:07 - You’ll microwave steak and pasta, and you’ll be happy * 18:44 - Microvibes https://malmesbury.substack.com/p/my-journey-to-the-microwave-alternate?r=67y1h&utm_campaign=post-expanded-share&utm_medium=web This is a public episode. If you'd like to discuss this with other subscribers or get access to bonus episodes, visit askwhocastsai.substack.com/subscribe

In 1940, the engineers, programmers and dreamers building American television wrote a book about their strange new medium. An anonymous finalist in the Astral Codex Ten Book Review Contest rescues We Present Television from being pulped and tours its world: live studio dramas shot with three cameras, a spelling bee that was the ancestor of the game show, a mobile studio on two boxy buses, and early nerves about Hollywood and advertising. Then come the prophecies, some laughably naive and some uncannily accurate: video calls, telegenic elections and board meetings of “twelve electrical ghosts”. * 00:00 - Introduction * 03:34 - Prose * 06:05 - The Attainment of Television * 17:39 - What is a Television Program? * 20:55 - Studio Shows * 29:50 - Motion Picture Transmissions * 32:45 - Outside Programs * 40:16 - The Television Chimera * 43:12 - Prophecy * 44:13 - Laughably Naive * 46:57 - Uncannily Accurate * 54:35 - Iä! Iä! TikTok Fhtagn! * 01:03:43 - We Now Conclude Our Broadcast Day https://open.substack.com/pub/astralcodexten/p/your-book-review-we-present-television?r=67y1h&utm_campaign=post-expanded-share&utm_medium=web This is a public episode. If you'd like to discuss this with other subscribers or get access to bonus episodes, visit askwhocastsai.substack.com/subscribe

In the second of three parts, Andy Masley builds the case that a computer can, in principle, really think and understand. Starting from the idea that minds are things physical systems do, he asks why every mind we know of happens to be made of carbon, what neuroscience and Alan Turing tell us about simulating a brain, and whether a perfect simulation of thinking is thinking. Along the way he puts the classic objections, from Leibniz’s mill to Searle’s Chinese Room, through a simple “brain test”: if an argument would also prove that you can’t think, something has gone wrong. * 00:00 - Introduction * 03:18 - Why computers can in principle produce thinking and understanding * 03:24 - What are minds made of? * 03:27 - Can physical objects think? * 06:47 - Implications of minds existing in purely physical objects * 10:46 - Information theory * 14:05 - Does an object need to be made of a specific material to think? * 19:04 - What would actually happen in this thought experiment? * 23:20 - Is carbon required for thought? * 23:24 - Evolution and carbon * 38:30 - Carbon is itself just a specific arrangement of smaller parts * 42:25 - Some takeaways from neuroscience * 43:30 - Thinking and understanding each happen in the brain through neurons passing information to each other * 44:40 - What matters about those signals is the information they carry, rather than the stuff carrying it * 46:45 - The brain’s information processing can be described with ordinary classical physics. It doesn’t seem to rely on special quantum effects * 49:02 - It seems like thought and understanding have to be specific information flows in the brain * 51:12 - Some takeaways from computer science * 54:42 - Would a simulated mind be able to really think and understand? * 01:05:17 - Programs aren’t just in the eye of the beholder * 01:08:26 - What do people mean when they say “the brain is not a computer”? * 01:18:10 - But isn’t the brain analog? * 01:19:26 - A lot of arguments that computer processes can’t think are arguments that any physical system can’t think * 01:38:39 - Conclusion https://open.substack.com/pub/andymasley/p/why-i-believe-current-ai-models-can-d63?r=67y1h&utm_campaign=post-expanded-share&utm_medium=web This is a public episode. If you'd like to discuss this with other subscribers or get access to bonus episodes, visit askwhocastsai.substack.com/subscribe

For three decades, India sold the world its labor without anyone having to move: from a Detroit hospital’s COBOL job whose code had to be physically shipped to America, through the Y2K gold rush, to TCS, Infosys and millions of jobs as “the world’s back office.” In this essay for Asterisk’s Work issue, David Oks traces how a disastrous 1978 policy decision, undersea cables and the 1991 reforms built one of the great development stories, and asks what becomes of it now that AI can write code, answer calls and process claims for less than even the cheapest engineer. * 00:00 - Introduction * 08:49 - India, the world’s back office * 21:13 - The dimming of the IT dream https://asteriskmag.substack.com/p/what-will-happen-to-indian-it?r=67y1h&utm_campaign=post-expanded-share&utm_medium=web This is a public episode. If you'd like to discuss this with other subscribers or get access to bonus episodes, visit askwhocastsai.substack.com/subscribe