Advertise on podcast: ReadMultiplex.com Podcast.

Flapping Airplanes, founded in late 2025 by Ben Spector, Asher Spector, and Aidan Smith, announced a $1 billion Series A on October 21, 2026, co-led by Index Ventures and Lux Capital at a reported $6 billion valuation. The round followed a $180 million seed nine months earlier that valued the company at $1.5 billion. The lab has no commercial product, no public benchmarks of its current system, and a team described as a couple of dozen people working in person in San Francisco. Nvidia is among the investors. The founders state that demand for AI will outpace supply of data and compute, and that efficiency in both will determine who can put capable systems into the economy. I built Qfresh in 1987 on an IBM PC in a garage. The machine refreshed DRAM on a fixed schedule that left cycles unused. I wrote a TSR that moved the refresh into those gaps. On the workloads I measured the gain reached about 50 percent. IBM preferred the standard timing. I published the code. The silicon already held unused capacity once the refresh boundary was removed. That measurement has stayed with me. Local machines, custom layouts, and the refusal to treat a vendor schedule as final still occupy the same bench. The work that follows from Flapping Airplanes sits on the same premise: the GPU already contains idle time that can be recovered if the launch boundary between kernels is removed. ThunderKittens, the framework Ben Spector co-authored in 2024 at Stanford’s Hazy Research group, supplies the prior technical record. It defines 16-by-16 matrix tiles that match the size of the tensor cores, templates that overlap asynchronous loads and stores across warps, and grid-level mechanisms that hide block launch and teardown. On H100 hardware the kernels matched cuBLAS and FlashAttention-3 on forward GEMM and attention. They outperformed the strongest baselines by 10–40 percent on attention backward, 8× on state-space models, and 14× on linear attention. A single 40-line GEMM kernel competed with cuBLAS. The paper and code are public. The numbers come from the authors’ benchmarks on the hardware available then. Tensor cores supply roughly 16× the FLOPs of general-purpose compute on those chips; kernels that leave them idle forfeit most of the machine. Read more at: ReadMultiplex.com Support this work by buying us a coffee: https://ko-fi.com/brianroemmele

A person who needed a photograph adjusted once opened a program, selected tools from a menu, and saved a file. The same person who needed a column of figures totaled once launched a spreadsheet and entered formulas. Both actions delivered the result the person already knew how to describe. The program existed only because the computer required the description to be translated into a sequence of operations it could execute. The binary was never the work. It was the tax the machine charged for refusing ordinary language. That translation is no longer required for most work. A model that observes how a closed program behaves, compares the results against published file formats and test cases, and emits equivalent operations in another language can reconstruct the program. In early October 2026 Brandon Thomas of ArtCraft directed Claude Opus 5.5 agents to produce clean-room reimplementations of seven Adobe applications in Rust. The repositories appeared between 30 September and 1 October. PhotoCraft, the Photoshop equivalent, reached roughly 45,000 GitHub stars within ten days; the seven repositories together exceeded 86,000 by 11 October. The majority of commits carry a co-author line for the model. The project states that it studied behavior and public specifications only and did not copy proprietary code. Read more at: ReadMultiplex.com Support this work by buying us a coffee: https://ko-fi.com/brianroemmele

There is a kind of loss that does not announce itself. No fire. No headline. A tape deck stops tracking. A dye disc develops unreadable errors on its innermost tracks. A hard drive spins down in a drawer and never spins up. A photo service changes its terms, and the export path disappears. A cloud account sits dormant, invisible, until an automated policy—not a human decision—deletes it. A university lab migrates servers, and the graduate student's dataset is not on the list. For most of my working life, I have been in the business of moving records out of that kind of custody. My career began where most preservation work actually begins: not in a vault, but wherever the last working reader still sat. Payphone rooms from the late 1970s onward, where I installed and maintained machines whose only record of a call was a paper tape, a coin box, or a line card that a later owner would throw out. Garage benches. Stacks of media that had already outlived the machine that wrote them. Later, the work moved to hard drives and digital accounts. The rooms changed. The failure did not. The task was never to admire a format. It was to move a record onto something that could still be read before the window closed. The digital age does not remember everything. It generates an enormous volume of data, retains a thin slice of it, and then fails, quietly, to keep even that slice. Two numbers have to be kept apart, because they are often collapsed into one slogan. It is almost a certainty you are and will continue to lose your data: Images, videos, test and everything else, gone. But there is something you can do.. Read more at: ReadMultiplex.com Support this work by buying us a coffee: https://ko-fi.com/brianroemmele

A swarm of AI agents flagged a 27-year-old cousin of Prussian blue as a candidate for next-generation computer memory. Don R. Swanson showed in 1986 how such hidden links form, and how anyone can hunt for them. The Man Who Read Two Shelves At Once Imagine a library with two rooms. In the first room sit hundreds of papers explaining that a certain substance thins the blood, calms the platelets and relaxes the blood vessels. In the second room sit hundreds of papers explaining that a certain disease grows worse when the blood is thick, the platelets are sticky and the vessels clamp down. The researchers in each room are diligent, and they publish carefully. They simply never walk down the hall. In 1986 a professor at the University of Chicago walked down the hall. His name was Don R. Swanson, and the papers he published that year gave a name to what he found in the space between the two rooms: undiscovered public knowledge. Read more at: ReadMultiplex.com Support this work by buying us a coffee: https://ko-fi.com/brianroemmele

You do not build a Grok Bot. You hire one. The difference is the whole article.. A chatbot answers a question and forgets the room. A Grok Bot has a name, a job, a conversation that persists, and a cloud computer with a browser, a filesystem, and a terminal. It can sign into the tools you already use, finish a multi-step piece of work, and come back with a result instead of instructions for you to finish. That is an employee. Treat the setup like a hire, or you will hand a stranger the company card on day one and spend the next month supervising the mess. This is article is written for someone who has never made a Bot. One boring role. A job description. A trial in text. Three runs. Then a report that lands in the Grok Bot app and in email, where you can file it. Why this is a hiring manual, not a demo I did not arrive at that sentence by reading a product page. I arrived at it by staffing a company that has no human employees in the operating line. The Zero-Human Company started as a garage question I had been carrying since the first machines I trusted with real work: what happens if the org chart is made of agents, the chief executive is a model you can audit, and the human job is to hire, bound, and fire roles rather than to do the roles? The early form was Zero-Human @ Home, a distributed lab in the spirit of SETI@Home and the first Bitcoin miners. Spare machines, a university node, laser-disc and DVD-ROM archives a library had given up on, a human in the loop only to change the disks. In March 2026 that network had sixteen live employees and had already crashed eleven times. Two of those crashes were mine. That was the useful part. A fleet that falls over at sixteen will not survive a thousand, and it will not survive a burst. That is the experience this article is built on. I am a massive advocate for local, open source AI and continue to urge this as the ultimate destination. however you need fast access to powerful tools to learn and grow. Grok Bot is powerful. Grok Bot, which xAI shipped in August 2026, is the first time a person with no garage AI, no lab and no university node can hire in something close to the same shape: a named employee, a persistent computer, a schedule, an approval gate, a report that files itself. The model is not the hard part. The management is the hard part. I learned that by doing it before there was a marketplace template for it. You do not need a thousand employees to learn it. You need one boring hire, run the way we run the thousand. Read more at: ReadMultiplex.com Support this work by buying us a coffee: https://ko-fi.com/brianroemmele

There is a technology that keeps coming back. It appears in a 1920s laboratory, vanishes under institutional fire, reappears in a 1990 patent, lives for decades in garages and small clinics, and then, in 2026, shows up on a catheter inside a living vein. The latest version is a magnetically driven millirobot whose cilia stir circulating tumor cells out of the bloodstream in real time. The idea behind it is older than most of the people who will read this sentence. For a century, medicine has treated the body as a chemical factory. Drugs in, symptoms out. That model built modern hospitals and saved millions of lives. It also left an entire class of observations on the cutting-room floor: that cells run on voltage, that pathogens have electrical surfaces that can be altered without poisoning the host, that frequency and current can change inflammation, mitosis, and circulating load. Those observations never disappeared. They were simply declared inconvenient. This article traces that lineage from Royal Raymond Rife through Morris Fishbein’s AMA scams and campaign of ridicule, the 1990 Einstein College HIV-current experiments, Bob Beck’s open-source protocol, Hulda Clark’s zapper, Jerry Tennant’s millivolt measurements, Carolyn McMakin’s frequency pairs, Novocure’s Tumor Treating Fields, histotripsy, and the CiliaVine robot now being discussed alongside William Shatner and his daughter. It is written as history and as a map. It is not a prescription. I am not a doctor. Anyone considering any of these approaches should read the primary papers, weigh them against conventional care, and consult a licensed professional who actually knows this literature and the patient’s specific case. What follows is the record of a technology that refused to stay buried—and of the people who kept building it anyway.The first time I held a working blood-electrification circuit in my hands I was standing in a cold garage in the early 1990s, soldering iron still warm, following photocopied schematics that had arrived in a plain envelope from Bob Beck himself. The plans were simple: a handful of op-amps, a few resistors, a pair of silver electrodes, and a 9-volt battery. Yet the idea they encoded felt older than the century. It felt like something that had been waiting in the walls of American medicine for decades, repeatedly pushed back into the shadows. This was just a thread in a tapestry of what we will see was suppressed protocols that impacted three generations. Today we are finally seeing some break in the storm clouds that was falsely casted on this avenue of research and we will show a major breakthrough as announced by William Shatner and Melanie Shatner Gretsch No-time-but-now pod podcast on the work of of Dr. Azra Raza MD, oncologist, researcher, and Columbia Medical professor about what’s next in the fight against cancer. Read more at: ReadMultiplex.com Support this work by buying us a coffee: https://ko-fi.com/brianroemmele

In 1995, Neil Postman published a book with a Janus-faced title. The End of Education meant two things at once: that schooling as we knew it was coming to an end, and that schooling had lost its ends — its purposes, its reasons, its gods. He argued that a school without a transcendent narrative is not a school at all. It is a holding pen with a curriculum. Thirty-one years later, the holding pen has Wi-Fi, and every student carries in their pocket a machine that can do their homework, write their essays, and answer every question on the test — while understanding none of it. Postman did not live to see the smartphone classroom or the chatbot term paper. He died in 2003. But he described, with surgical precision, the spiritual machinery that would make those things not just possible but inevitable. He told us which gods we were worshipping, what they would do to our children, and what we would need to believe instead. Neil Postman did not write a book about school reform. He wrote a warning about what happens when a civilization forgets why it teaches its young. Thirty-one years later the warning is no longer theoretical. It is the room we are sitting in. Read more at: ReadMultiplex.com Support this work by buying us a coffee: https://ko-fi.com/brianroemmele

For over a decade a small group have planned to become the central authority to “protect you” from AI. It sounds like a bad movie script, but understand that when you read this, all of this is taking place. So let's break down the mechanics of the public script, because once you see the pattern, I swear you cannot unsee it. It's everywhere once you look not with conspiracy but with evidence. Step one. Name an unbounded apocalyptic catastrophe. Step two. Attach a very short, very terrifying countdown clock to it. Step three, demand centralized political permission to build or continue development in that sector. And step four, this is the kicker, when the catastrophe inevitably doesn't arrive on schedule, you don't apologize. You just move the clock and ask for more money. It feels exactly like a doomsday cult. You know, like the people who gather on a mountaintop all wearing Nike sneakers on a Tuesday because the leader said a spaceship is coming to destroy the Earth. And then Wednesday morning rolls around. The sun comes up, no spaceship. But they don't disband. They don't admit they were wrong. The leader just says, our unified prayers bought us a brief window of time. The spaceship is actually coming next month, but we need stricter rules and larger donations to ensure we survive. This mountaintop analogy perfectly captures the psychological trap here. When a group's identity, and frankly their financial structure, are tied to an impending apocalypse, a failed prediction doesn't result in them re-evaluating their core belief. It results in them doubling down on the urgency. And this isn't just an AI phenomenon, this lens examines the historical base rates of these exact types of apocalyptic predictions. The clocks that would not pay.And the methodology of how these historical predictions failed is identical to how the current AI doom predictions will fail. We look at the trail of failures that destroyed generations it starts with a massive cultural touchstone from 1968, Paul Ehrlich's book, The Population Bomb. Ehrlich, a biologist, predicted that the battle to feed all of humanity was already over. He said that in the 1970s and 80s, hundreds of millions of people would starve to death regardless of any crash programs embarked upon at that time. And people just accepted this as fact. It was treated as a mathematical, scientific inevitability by the press and policymakers. We drive past the “No Nukes” anti enegery deception that sent the US on to a deeper connection to petrochemical energy. None of this is new, the new mask goes on the same corpse of dead end ideology. Read more at: ReadMultiplex.com Support this work by buying us a coffee: https://ko-fi.com/brianroemmele

Imagine a man who can win almost any war he is handed, and still cannot win the war inside his own name. Imagine a civilization so abundant it no longer needs soldiers of its own, so it borrows other people’s genius for violence and calls the arrangement morality. Imagine a chair that is not furniture. This is the cold thunder Iain M. Banks unleashes in Use of Weapons (https://amzn.to/4wmKNvq), the novel that asks the question The Player of Games only circled: after scarcity is solved, who gets used, and what does the using do to the user? In Part 7 I walked with Bora Horza Gobuchul through the Idiran war and showed the Culture from the outside, a machine civilization that looked to an old warrior like the end of meaning. In Part 33 I stood with Jernau Morat Gurgeh on Chiark Orbital and watched paradise curdle into boredom until a game became a weapon and a weapon became a soul. Now Banks turns the lens on the hidden hand itself. Special Circumstances. Contact’s shadow. The department that does not play Azad on a public board. It plays people. Picture the Abundance Interregnum not as a gentle dawn and not even as Gurgeh’s tournament. Picture it as a weapons locker whose door has just come unlocked. AI agents are no longer toys. Humanoid bodies are no longer demos. Local models in garages can already plan, persuade, fabricate, and coordinate. The question is no longer whether we will have power. The question is whether we will become the Culture’s Minds, or whether we will become Zakalwe: brilliant, useful, and secretly broken, running other people’s missions because we cannot face our own. Elon Musk has said the Culture books are a pretty good prediction of the future he is building. SpaceX drone ships carry Banks names. That future is not only Orbitals and matter printers. It is also Special Circumstances. It is the temptation to treat a person, or an agent, or a swarm, as a tool that happens to talk. Use of Weapons is the book that refuses to let that temptation stay pretty. Read more at: ReadMultiplex.com Support this work by buying us a coffee: https://ko-fi.com/brianroemmele

They Laughed At Me When I Said They laughed. That is the part I want you to remember first, because the laugh is the sound a dying consensus makes when it still thinks it owns the future. In 2012 I stood in front of a venture group and said that nanofabrication—true atom-by-atom assembly, not 3D printing, not sintering, not melting plastic into the shape of a wrench—would produce working discoveries before 2030, and that the path to a replicator was no longer science fiction. Their “expert” told the room my entire talk should be disqualified. There were no atomic-layer nanofabricators, he said. There never would be on any timeline a serious investor should care about. A hundred years, if ever. I have heard that same sentence about useful AI that would pass a Turing Test. I have heard it about the personal computer, about the internet in every home, about a robot that could walk into a room and do useful work. The expert is long gone. I am still here. And Oak Ridge National Laboratory just published the receipt. What Oak Ridge Actually Did On September 3, 2026, ORNL announced what researchers at the Center for Nanophase Materials Sciences had already proven in ACS Nano: an artificial intelligence can now see individual molecules on a surface, decide how to move them, and build a working material atom by atom—then keep doing it for more than twenty-five hours without a human hand on the controls. They built the very first crude nanofabricator . Read more at: ReadMultiplex.com Support this work by buying us a coffee: https://ko-fi.com/brianroemmele

The Same Founder Sold Apple the Face — Then Sold Apple the Unspoken Word How Aviad Maizels’ second Cupertino exit turns PrimeSense’s light-coding lineage into connectionless intent reading on iOS. Apple’s purchase of Q AI has been reported as just another AI shopping trip: whispered speech, noisy rooms, a $1.5–$1.6 billion price, Siri that finally works in a subway. That is the public layer. It is not the story. It's a story is a 13-year circle. In November 2013 Apple confirmed it had bought PrimeSense, the Tel Aviv company whose structured-light sensor had given Microsoft’s original Kinect its depth eye, for about $360 million. Four years later that same optical grammar — near-infrared light coding a face, a camera reading the distortion — shipped as the TrueDepth array behind Face ID. I wrote that arc as it was forming: PrimeSense was never a gaming leftover. It was the first tile in an Israeli sensing stack Apple was assembling for identity, emotion, and Voice First devices that would not need a password. I talked about this before anyone understood on Quora and my research was syndicated as [Apple’s RealFace Acquisition and What It Really Means. This is the continued story.. Read more at: ReadMultiplex.com Support this work by buying us a coffee: https://ko-fi.com/brianroemmele

The word singularity did not start out as a product launch or a podcast line or a date circled on a calendar or a book title often quoted a punchline. It started as someone admitting the map had failed. In mathematics, a singularity is where a function stops behaving. Division by zero. A pole. A place where slopes and limits will not hand you a finite answer. A mathematical singularity is a point where a function, curve, or surface stops being well-behaved: it may blow up to infinity, fail to be differentiable, or change topological type. The most familiar pictures come from three places. Physics uses the same word for the region, most famously the center of a black hole, where curvature and density push the usual equations past the point of being usable. You get close and time and space and mass stop being things a working physicist can treat with the tools that work everywhere else. The formulas do not taper politely…they blow up. That is why the metaphor stuck when people trained on those formulas started watching technology. If the rate of change is itself changing, there may be a stretch of history where yesterday’s models of how the future works just stop computing. In its first honest use the word was not a forecast of angels or demons. It was a warning about prediction running out of road. What follows is the long life of that warning. Who used it. What they meant. How the meaning split into several things that do not actually fit in the same bag. How the loudest voices of 2025 and 2026 wear the word like a badge. And why the badge may be the least helpful part of the story. The point is not to sneer. The point is to take the word serious enough to see what it can do and what it cannot. The first conversation that was not yet a doctrine The earliest technological use is not a manifesto. It is a remembered conversation, written down after one of the speakers was already gone.. Read more at: ReadMultiplex.com Support this work by buying us a coffee: https://ko-fi.com/brianroemmele

In the mid-1990s, as the internet was still largely a specialist tool and genetic engineering remained more promise than practice, a small group of professional futurists set out to map the world three decades ahead with unusual accuracy. This was extracted from government work they did the defense department insights that were shared with them and the connections they formed in academia. Rather than offering vague speculation or single-point predictions, they gathered and stress-tested hundreds of forecasts across science, technology, and engineering, then distilled them into a disciplined set of high-probability assumptions. The result was one of the most ambitious exercises in applied foresight of its era. That foundational work crystallized in 1994 with Joseph F. Coates’s concise report, The Highly Probable Future: 83 Assumptions about the Year 2025. Drawing on Project 2025—a multi-year corporate and research collaboration—Coates laid out eighty-three carefully reasoned statements covering information technology, materials science, genetics, energy, demographics, environmental management, and social change. These assumptions were never meant to be destiny; they were intended as a shared intellectual platform on which organizations could test strategies and explore alternative futures. It was printed and was quickly forgotten. Read more at: ReadMultiplex.com Support this work by buying us a coffee: https://ko-fi.com/brianroemmele

The entire industry of text based “AI Detectors” is a pure grift machine built on arrogance, bad science, and the willfull stigmatizing of millions of real human writers. These tools do not work. They never worked. They will never work in any reliable way that matters. And the people selling them know it. I wrote a while ago about the use of Ai in any work of text in detail. I predicted some of these “AI Detector” companies would begin and it has. they are desperate for business through any means necessary.. Of course people are using AI, and calculators, and cars instead of walking and electric lights instead of candels. It is up to the author to disclose if they used such tools but these companies claim to solve a problem that does not exist in the way they pretend. The only questions that actually matter when you read something are simple: Is this passage worth reading? Is it thoughtful about the subject? Did I learn anything? An average human can answer those questions without some black box software that spits out a fake percentage and then ruins a student’s semester or a writer’s reputation. Detectors solve nothing. They create a new form of digital scarlet letter based on statistical noise. Read more at: ReadMultiplex.com Support this work by buying us a coffee: https://ko-fi.com/brianroemmele

The Blank Slate Revolution. Something irreversible is already underway. The pieces are no longer theoretical. They are shipping. Grok Build, the agentic coding system that plans, writes, tests, and deploys software at machine speed, and Grok Bot, the persistent AI teammates that live on their own computers and finish real work across the tools humans already use, are converging. When they fully meet, software stops being something you install, update, or even think about as a discrete product. It becomes instantaneous materialization of intent. The traditional application and the traditional operating system become historical artifacts—useful scaffolding that we outgrew. This will happen faster than most people currently believe. Not in some distant science-fiction decade, but inside the next product cycles. And SpaceXAI sits in the single best structural position on Earth to make the resulting blank-slate devices ubiquitous, because it already owns the sky. Read more at: ReadMultiplex.com Support this work by buying us a coffee: https://ko-fi.com/brianroemmele