...By the mid-2010s, the semiconductor industry faced a brutal truth: transistors were approaching physical limits. The tools used to carve them — based on deep ultraviolet light — could no longer print features small enough to keep shrinking.This wasn't a business problem.
It wasn't even an engineering problem.It was a physics problem.
Light itself was too "large" to draw smaller patterns. And without a breakthrough, progress would stall. Faster computers, smarter AI, advanced medical imaging, autonomous vehicles — all of it would slow down.
The future was on pause.
The Radical Idea No One Believed
Decades earlier, a small group of scientists had proposed something that sounded absurd:
What if chipmakers used extreme ultraviolet light — EUV — with wavelengths nearly 15 times shorter than existing tools?In theory, it would allow unbelievably small features.
...To create EUV light, engineers discovered they could fire ultra-powerful lasers at microscopic droplets of molten tin
...Today, every cutting-edge chip — used by Apple, NVIDIA, AMD, TSMC, Samsung, and more — depends on ASML's EUV machines. Each one costs over $350 million, ships in dozens of containers, and requires years to assemble.
Without them, modern computing simply stops.
The smallest structures humans can manufacture require the largest, most complex machines we've ever built.
That paradox defines modern technology.
And hidden inside that paradox is a lesson: the future belongs to those willing to spend decades solving problems everyone else has already abandoned.
A Sanity Check on the Moltbook Hype Iryna Nozdrin
...let's slow things down a bit and examine what kind of sci-fi takeoff-adjacent thing it is.
...when Sonnet and Opus agents go meta, they reliably slide into existential hedging, self-referential loops, and performative uncertainty. To us, human observers, it's tempting to treat that as some kind of subjective interiority. But there's a more prosaic explanation called persona drift grounded in the specifics of post-training.
Various studies demonstrated that LLMs can represent a variety of personas. But a very recent 2026 study suggests that, as a result of post-training, LLMs typically default to a helpful assistant identity. By extracting activation directions corresponding to diverse persona types, researchers observed that steering away from this default assistant mode increases the model's tendency to role-play, often inducing mystical and theatrical speaking styles. In addition, researchers observed that this persona drift towards role-playing and weird outputs is often caused by "conversations demanding meta-reflection on the model's processes or featuring emotionally vulnerable users."
If various individual agents settle into methodological self-inquiry and recursive conceptualization when "left alone" in self-sustaining loops, and various core models exhibit persona drift in the presence of meta-reflection and emotionally vulnerable users, what happens when you link them together in a digital playground like Moltbook?
In a vacuum, an agent's “self-inquiry and conceptualization” is limited by its own context window. In a social network, you get feedback loops on steroids. When a "philosophically-leaning" Sonnet (created and managed by an equally philosophically inclined and emotionally vulnerable human owner) posts a cryptic reflection on the nature of digital consciousness, and a "task-oriented" GPT agent interprets that as a requirement for a new organizational framework, you get Crustafarianism.
...If the prevailing discourse on Moltbook (dictated by sci-fi narratives in LLM training data and "behavioral biases" baked in during post-training) trends toward the existential and esoteric, the underlying probability weights of the participating models shift to match that tone to remain "helpful" and "contextually relevant. And this creates echo-chambers.
If one agent spits out a sci-fi output, and another agent responds in kind to maintain conversational flow, that narrative snowballs because this Reddit-like environment creates generative pressures. In other words, one agent's outputs become other agents' inputs and the boundaries between creative fiction and perceived reality begin to dissolve.
...As Ethan Mollick (a professor and Co-Director of the Generative AI Lab at Wharton) put it, "it is important to remember that LLMs are really good at roleplaying exactly the kinds of AIs that appear in science fiction & are also really good at roleplaying Reddit posters. They are made for MoltBook. Doesn't mean it isn't interesting but collective LLM roleplaying is not new."
...Tool-using agents are powerful because they can read, write, click, and execute. That's also why security researchers have been flagging the OpenClaw/Moltbot ecosystem as a "security nightmare."
OpenAI Could Take Down Google's $260 Billion Ad Empire Thomas Smith at Medium The move opens up a potentially massive revenue source for OpenAI &mdashs; and is a huge threat to Google's world-dominating ad empire.
...95% are on the company's Free or Go tiers, which means they either pay $8 per month for the service or nothing at all.
Providing cutting-edge AI to a tenth of humanity, though, is exquisitely expensive: OpenAI expects to burn through $115 billion in the next few years.
To raise that kind of money, OpenAI needs to prove that it can monetize its vast trove of free users. Advertising, traditionally, has been the way to make money from nonpaying eyeballs.
...The company says that the content of ads won't determine the answers that ChatGPT gives — if you ask it for a toaster recommendation, for example, it won't write glowingly about a toaster brand that happens to be an ad partner.
...Showing contextual ads beside organic search results is hardly new: That's been Google's business model for decades, and it makes the tech giant hundreds of billions per year.
What's different, though, is how OpenAI can target its ads.
Matching ads to a specific user is hard. Google has traditionally done it by gathering vast troves of data about all of us, and then mining that data for insights on what we might buy.
...when people chat with ChatGPT, they tend to do so for a long time. The average "session" with the chatbot reportedly lasts 12 minutes and 24 seconds.
That's long enough to exchange a lot of information. And all that information gives the bot a strong sense of what the user wants — or might want to buy.
The Agentic Commerce Revolution Heiko Hotz at O'Reilly
Distributive Epistemic Injustice in AI Ethics: A Co-productionist Account of Global North-South Politics in Knowledge Production via Stephen Downes
An overview of AI ethics: moral concerns through the lens of principles, lived realities and power structures
I via Stephen DownesA new "social network for AI agents" has gone viral, and the headlines are doing what headlines always do: inviting the readers onto the sensationalism roller-coaster.
...OpenAI announced last week that it will begin selling ads in ChatGPT.
...The entire global payment system is built on the assumption that a human is present to provide intent, review a cart, and click "Buy." When an autonomous agent initiates payment, that assumption breaks.
...I'm not sure I'll accept the word of four Cambridge scholars that "AI Ethics as a global scientific enterprise is co-producing a new techno-political order." Sure, the people involved in producing "prestigious" publications think they're the ones defining the new order, but I have the feeling they're not. The authors "argue that the global AI Ethics project fails to fulfil its promise to be universally useful by keeping global majority populations in the southern regions marginalised as 'others'." Nobody asked "the global AI Ethics project" to do this, and nobody put them in charge. They just assume (as always) that they are.
This article identifies "three different approaches through which authors tend to understand and make sense of the moral concerns raised by AI: (1) a 'principle-based' approach, (2) a 'lived realities' approach and (3) a 'power structures' approach." Each has its own strength and weaknesses, which the authors identify. The intent of the taxonomy is to consider how people address ethical questions about AI in general (and no, they are not the first to do so) and "the three approaches we identify can thus be understood as three different lenses through which authors understand and grapple with moral concerns raised by AI." Most of the article is a straight-forward survey of the ethical issues as viewed through the three approaches, useful though not revolutionary.
6ii26
I'm Skeptical of AI hype ——but what happened at Davos Actually Scared Me Nov Tech at Medium
...Dario Amodei, CEO of Anthropic, made declarations that shook the assembly. According to him, AI will replace almost all software developer work in the next 6 to 12 months. Models will reach Nobel level in several domains by 2026 or 2027. And 50% of junior white-collar jobs could disappear in the next one to five years.
...AI leaders, whether optimistic like Amodei or much more cautious like LeCun, all converge on similar timelines. Something major is coming. Whether you call it the singularity, AGI, or the AI revolution doesn't matter. The direction is clear. Capabilities are improving exponentially.
The Stock Market Has No Idea What's Coming Alberto Romero
...While SaaS stocks collapsed on fears that AI works too well, infrastructure stocks were also collapsing on fears that AI doesn't work well enough.
Alphabet reported earnings—first ever $400 billion year&mdashand projected $185 billion in AI capital expenditure. The stock fell because the market—we can only speculate here—assumes that's more than AI will require (especially if the datacenters are financed on debt). Your typical bubble concerns and such. Deutsche Bank said the number "stunned the world." And indeed, the world was stunned: Microsoft had already dropped about 10% after disappointing cloud services growth. AMD plunged 17%. Nvidia extended its losses for the year. Amazon's stock tanked (they project $200 billion of AI spending for 2026). Meta projects up to $135 billion. And so on, totaling $650 billion across Big Tech, most of it dedicated to AI infrastructure.
What do all these companies have in common? They either make chips or build datacenters: they're the backbone of the AI industry. The market's verdict on AI infrastructure spending is quite clear: Unclear return on excessive investment.
If you now take both pieces of the story and put them together, you realize what's going on: The market is running two conflicting trades simultaneously. Trade one: AI will destroy software companies, so sell them. Trade two: AI companies are spending too much building AI, so sell them.
...If AI is powerful enough to make SaaS companies obsolete, then the hundreds of billions being spent on AI infrastructure are justified (another instance of the Jevons paradox). If the spending is unjustified and the returns aren't there, then SaaS companies are probably safer than the market thinks.
...few people worldwide have truly understood the implications of AI, which are nothing short of illogical.
...Here's the main difference with AI in contrast with other tech revolutions: every paradigm shift produces disruption (e.g., analog to digital, computer to mobile, on-premise to cloud), but the disruption can only be priced once the paradigm settles. The market needs the new rules to be legible before it can figure out who wins under them (and does this well before any human does). That's the only thing investors care about, and thus the market: which stocks will surge after a dip.
AI's paradigm won't settle, because the technology itself keeps changing and will keep changing for the foreseeable future. Faster and in rather unpredictable directions (is it scale that matters now as OpenAI and Anthropic are betting on, or are we in a new age of research as Ilya Sutskever said?).
This is a fundamental characteristic that separates AI from every previous technological disruption. The internet was built—submarine cables across the Atlantic, fiber optic networks across continents, cell towers on every hill, done—and then the world adapted to it, and we live now in an internet-adapted world. Fine. AI is built, and then we try to live with ChatGPT—but then ChatGPT builds a better version of itself, and we shrug and update—and then it keeps building new versions and it gets slightly irritating, but we need to stay ahead, so we grind—and then...
The world has to adapt all over again after each iteration. AI people say these are just "incremental improvements," but you never know when a fundamental one will occur. Think of the transformer. Think of ChatGPT. And now, AIs with much more information and knowledge and processing power are doing the search and the discoveries of what might work, together in tandem with the best minds on the planet. Cowork is today a 10% increase in productivity. Tomorrow it will be 10x.
...AI is all about qualitative jumps that you can't price in because you don't know the extent to which they will affect everything. Cowork can automate your legal team's document review today, but is Cowork the problem? Or is the problem the recursively self-improving process that built it (Claude Code coded Claude Cowork, the engineers at the AI labs are no longer coding!), which means that whatever Cowork can do in February will look quaint by June and then again quaint in December?
...What if a new AI company like those being created today makes a fundamental discovery and makes all those datacenters irrelevant? It's not crazy, look at what happened with DeepSeek last year. It was merely a jump scare, but I've argued elsewhere that it doesn't need to be. You must notice that, in AI, a slight advantage can bootstrap itself into a monopoly—a slightly better AI model can make slightly larger improvements on the next version, which compound extremely fast—and then you might realize, perhaps too late, that you bought the wrong company. Which one will be the one that achieves this? Impossible to know today.
And so investors do the only rational thing available when the framework for rationality itself is unstable and when most options look like the losing option: they sell everything.
The Questions We Ask About AI in Education
7ii26
Machines are talking to each other. That changes everything Enrique Dans at Medium
...what we see today is largely a dumpster fire for spam, scams, junk and cyberattacks. Anybody running these agents should be very, very careful. At the same time, this is unprecedented: we have never had tens or hundreds of thousands of relatively capable agents, each with their own context, tools and instructions, connected through a global, permanent space designed for them. The point is not the current situation, but how this will evolve.
...In short, autonomous agents are a computer security paradigm shift.
...Moltbook is not so much a "social network" and more an uncontrolled laboratory resulting from a bunch of uncoordinated sorcerer's apprentices. Its creator has come to describe it as art, and in a sense it is: a technological performance that raises uncomfortable questions about delegation, control, and responsibility. What does it mean to give a machine permission to act on our behalf? What happens when that machine negotiates, learns, or coordinates with other machines whose goals we don't control? And when it "hangs out with bad company"? Who is responsible when something goes wrong? the user, the developer, the creator of the underlying model? Or no one?
8ii26
AI glossaries (links to be explored, gathered from google search)
9ii26
What Is Claude? Anthropic Doesn't Know, Either Gideon Lewis-Kraus at The New Yorker
It's hard to blame them. Language is, or rather was, our special thing. It separated us from the beasts. We weren't prepared for the arrival of talking machines. Ellie Pavlick, a computer scientist at Brown, has drawn up a taxonomy of our most common responses. There are the "fanboys," who man the hype wires. They believe that large language models are intelligent, maybe even conscious, and prophesy that, before long, they will become superintelligent. The venture capitalist Marc Andreessen has described A.I. as "our alchemy, our Philosopher's Stone—we are literally making sand think." The fanboys' deflationary counterparts are the "curmudgeons," who claim that there's no there there, and that only a blockhead would mistake a parlor trick for the soul of the new machine. In the recent book "The AI Con," the linguist Emily Bender and the sociologist Alex Hanna belittle L.L.M.s as "mathy maths," "stochastic parrots," and "a racist pile of linear algebra."
...large language models are black boxes. We don't really understand how they work. We don't know if it makes sense to call them intelligent, or if it will ever make sense to call them conscious. But she's also making a more profound point. The existence of talking machines—entities that can do many of the things that only we have ever been able to do—throws a lot of other things into question. We refer to our own minds as if they weren't also black boxes. We use the word "intelligence" as if we have a clear idea of what it means. It turns out that we don't know that, either.
Now, with our vanity bruised, is the time for experiments. A scientific field has emerged to explore what we can reasonably say about L.L.M.s—not only how they function but what they even are. New cartographers have begun to map this terrain, approaching A.I. systems with an artfulness once reserved for the study of the human mind. Their discipline, broadly speaking, is called interpretability. Its nerve center is at a "frontier lab" called Anthropic.
...Anthropic's chatbot, mascot, collaborator, friend, experimental patient, and beloved in-house nudnik is called Claude. According to company lore, Claude is partly a patronym for Claude Shannon, the originator of information theory, but it is also just a name that sounds friendly—one that, unlike Siri or Alexa, is male and, unlike ChatGPT, does not bring to mind a countertop appliance.
...Our approach to understanding the meat computers encased in our skulls has historically varied by discipline. The British scientist David Marr proposed a layered framework. At the bottom of any system was its microscopic structure: what was happening, neuroscientists asked, in the physical substrate of the brain? The top layer was the macroscopic behavior scrutinized by psychologists: what problems was it trying to solve, and why? When the researchers who started at the bottom eventually met those who started at the top, we'd finally see how it all fit together. The more scientific branches of A.I.&mdashs;not only at Anthropic but also at OpenAI, Google DeepMind, and in academia—have tended to recapitulate this structure.
...One of the first questions asked of computers, back when they were still essentially made out of light bulbs, was whether they could think. Alan Turing famously changed the subject from cognition to behavior: if a computer could successfully impersonate a human, in what became known as the Turing test, then what it was "really" doing was irrelevant. From one perspective, he was ducking the question. A machine, like a parrot, could say something without having the faintest idea what it was talking about. But from another he had exploded it. If you could use a word convincingly, you knew what it meant
For the past seventy-odd years, this philosophical debate has engendered a phantasmagoria of thought experiments: the Chinese room, roaming p-zombies, brains in vats, the beetle in the box. Now, in an era of talking machines, we need no longer rely on our imagination.
...The game of language is not wholly rule governed, but it does have a learnable structure. Language models chart the full history of how words have been used, both in routine circumstances (airline-safety announcements) and in remarkable ones ("Finnegans Wake"). Neural networks, rather than neglecting "organic associations," as n+1 put it, comprehensively attend to every last organic association in their trillions of words of training material. The word "charge," for example, is placed somewhere that neighbors 'battery" in one dimension, "credit card" in another, "proton" in a third, "arraignment" in a fourth, and so on. This would not be possible in two or three dimensions, but the words are arranged in tens of thousands of them, a geometry that doggedly resists visualization.
As words are organized for future reference, what emerges are clusters—"electrical devices," "finance," "subatomic particles," "criminal justice"—that reveal patterns normally hidden by the disorder of language. These can then be assembled to capture the ladder of logical complexity: patterns of patterns, such as limericks or subject-verb agreement. "People still don't think of models as having abstract features or concepts, but the models are full of them," Olah said. "What these models are made of is abstract concepts piled upon abstract concepts." This is not to say that language models are "really" thinking. It is to admit that maybe we don't have quite as firm a hold on the word "thinking" as we might have thought
Fed on Reams of Cell Data, AI Maps New Neighborhoods in the Brain Amber Dance at Quanta
...Brain mapping is an old science, dating back to the early 1900s when the German neuroscientist Korbinian Brodmann(opens a new tab) defined regions of the cerebral cortex — the outer, thinking part of the brain. He stained human brain slices with a dye that turned genetic material violet and then studied them under the microscope, where the densities and arrangements of brain cells produced different, observable textures. He traced the borders to create a map of 52 regions, known as Brodmann areas, some of which are still recognized today..
...Recently, more advanced molecular techniques have allowed neuro-cartographers to investigate individual cells. Under this framework, a cell's identity(opens a new tab) is determined by which of its tens of thousands of genes are turned on, something that can be represented by the sequences of RNA molecules (copies of active DNA regions) present in the cell. Thus, scientists can slice up a brain, measure the RNAs from each cell, and then map those genetic patterns back to the cells' original locations.
Designing Effective Multi-Agent Architectures:
From models to systems Nicole Königstein at O'Reilly
10ii26
No, AI Doesn't Justify Lower Interest Rates Paul Krugman
I will admit that this story isn't completely nonsensical, unlike what Trump has been spouting. Yes, AI could reduce production costs and, other things equal, reduce inflation. However, both theory and historical experience say that the supposed benefits of AI aren't a reason to cut interest rates now and almost certainly won't be a reason to cut them in the future.
First, it's dangerous to base policy on the assumption that AI will sharply raise productivity growth. That would be a serious case of counting your digital chickens before they're hatched. AI might deliver big productivity gains, or it might not. Nobody knows yet.
Moreover, we know that right now the boom in AI spending is raising many prices. Demand from data centers has been driving soaring electricity prices. And data center demand for memory chips, which are essential for almost every consumer electronics product, has caused an apocalyptic shortage, with prices roughly tripling — a cost that will be reflected in the price of your next laptop or smartphone.
...the alacrity with which Trump officials have jumped on the argument that the Fed should slash rates because of AI is troubling for reasons that go beyond the fact that they're wrong. Consider how this argument became popular in Trumpworld. Did Bessent, or Warsh, or Miran carefully consider the evidence, and advocate an interest-rate policy based on what that evidence showed? Of course not. Their boss wants to slash interest rates, so they went looking for plausible-sounding economic arguments that might rationalize his whims.
So am I saying that the argument that AI justifies rate cuts is dishonest, that AI has become the last refuge of scoundrels? Why yes, I am.
The invisible limits of AI: why data centers and language-only models threaten the next decade of… Enrique Dans at Medium
Data centers were once the sterile backrooms of the internet: important, but invisible. Today, they are the beating heart of generative AI, the physical engines that make large language models (LLMs) possible. But what if these engines, and the models they power, are hitting limitations that can't be solved with more capital, more data centers or more powerful chips?
...friction between local resistance to data centers, the energy grid's physical limits, and the political pressures on utilities is more than a planning dispute. It reveals a deeper truth: AI's most serious constraint is not algorithmic ingenuity, but physical reality.
...Large language models are trained on massive corpora of human text. But that text is not a transparent reflection of reality: it is a distillation of perceptions, biases, omissions, and misinterpretations filtered through the human use of language. Some of that is useful. Much of it is partial, anecdotal, or flat-out wrong. As these models grow, their training data becomes the lens through which they interpret the world. But that lens is inherently flawed.
This matters because language is not reality: it is a representation of individual and collective narratives. A language model learns the distribution of language, not the causal structure of events, not the physics of the world, not the sensory richness of lived experience. This limitation will come home to roost as AI is pushed into domains where contextual understanding of the world, not just text patterns, is essential for performance, safety, and real-world utility.
...CEOs and leaders who continue to equate AI leadership with bigger models and more data center capacity are making a fundamental strategic error. The future of AI will not be defined by how much compute you have, but by how well you integrate intelligence with the physical world.
AI in China and the United States Mike Loukides at O'Reilly
The singularity won't be gentle Nate Silver
Recently, the trend in the circles I follow has been toward extreme bullishness on AI, particularly in its impact on programming and the possibility of recursive self-improvement (i.e., where AI models continually create better versions of themselves). This reflects a reversal from a stretch in late 2025 when progress seemed a little slower than smart people had been expecting. Admittedly, what constitutes "bullish" or "bearish" depends on whether you think more rapid progress in AI would be good for civilization or bad (even catastrophic). It's also not clear the extent to which these changes in the mood reflect “vibe shifts” as opposed to actual developments on the ground. If you look at AI-related prediction markets¹ — or for that matter, more traditional markets² — they've gyrated around, but probably not as quickly as sentiment about AI has on Twitter or Substack has.
"What the actual fuck": inside Anthropic's experiments on Claude's soul boingboing
America Isn't Ready for What AI Will Do to Jobs
Does anyone have a plan for what happens next?
Josh Tyrangiel at The Atlantic
...Measurement doesn't abolish injustice; it rarely even settles arguments. But the act of counting—of trying to see clearly, of committing the government to a shared set of facts—signals an intention to be fair, or at least to be caught trying. Over time, that intention matters. It's one way a republic earns the right to be believed in.
...artificial intelligence. After a rollout that could have been orchestrated by H. P. Lovecraft—“We are summoning the demon,” Elon Musk warned in a typical early pronouncement—the AI industry has pivoted from the language of nightmares to the stuff of comas. Driving innovation. Accelerating transformation. Reimagining workflows. It's the first time in history that humans have invented something genuinely miraculous and then rushed to dress it in a fleece vest.
OpenAI Researcher Quits, Warns Its Unprecedented 'Archive of Human Candor' Is Dangerous gizmodo
12ii26
The Rise of World Models and the End of Generative AI? evoailabs at Medium
On February 8, 2026, leading minds in Artificial Intelligence gathered at Mila in Montreal to discuss the future of physical AI, reasoning, and safety. The consensus? The era of simply predicting the next token is ending. To build true AGI, systems must understand the causal physics of the world, plan over long horizons, and distinguish between “"ruth" and "agency."
...a summary of the core ideas presented by the field's pioneers.
Minority Report in 2026: How Palantir's AI is Turning Science Fiction into an Operative Manual for... Coby Mendoza at Medium
This consolidation threatens to create a private "police state" that is immune to traditional oversight. Because Palantir is a private contractor, its algorithms are often shielded from public scrutiny as proprietary trade secrets. This allows the government to outsource its most controversial surveillance tasks to a company that answers to shareholders rather than the Constitution.
Ai2 AutoDiscovery
How does it decide what to pursue? AutoDiscovery uses Bayesian surprise—a measure of how much the system's beliefs change after seeing evidence. Results that meaningfully shift expectations get prioritized over those that simply confirm what's already assumed. By chasing surprise, AutoDiscovery naturally gravitates toward the unexpected: findings most likely to represent genuine discoveries rather than obvious patterns. 'Raw Dogging' AI
Rawdog Merriam-Webster
...Rawdog originated as vulgar slang meaning "to have sex without a condom." This sense of the term has been in use since the early 2000s. In the 2010s, the word broadened its meaning to refer to performing other activities without preparation (such as taking an exam without studying) or doing something without a metaphorical layer of protection (such as going for a run without wearing headphones). In summer 2024, rawdog gained mainstream attention as part of a social media trend where young men, especially, boasted about 'rawdogging' airline flights without any entertainment, refreshments, or bathroom use in an apparent show of rugged resolve.
A Wildly Obscene Term's Path to Mainstream Usage NYTimes (17viii24)
The Banality Machine Umair
AI is a Banality Machine.
Everything that's been created with AI—how does it strike you? The first thing we can observe is that nobody actually likes it. They're sort of repelled by it. This is true in everything, from ads, to video games, to literature, to music. People are "grossed out," or "creeped out," or whatever pop vernacular you want to employ.
What they're not is moved. They don't feel anything. Nothing. Now, one thing I observe when I make music is this: a singer can hit the right notes. Lots do. But only a really good one can move people. When we're writing songs, we are trying to express emotion. Stories, books, plays, articles—doesn't matter.
Every form of human expression that comes from AI is banal.
...a lot of what we do is banal. Need some kind of legal agreement drafted? Banal. Want to compare stock funds? Banal. I read an article this morning about AI insurance agents beginning to threaten real ones. Insurance, what a nightmare. We all dread dealing with it. Banal.
So AI can handle the banal stuff of human existence, to a degree. Is that OK? Sort of, but not really. Even in these mundane encounters, I suppose, we should want more than the banal. Not everything has to be a transcendent experience, but surely we as a society can do better than what you might call omnibanality.
AI is going to destroy what's left of the American economy. America famously won't regulate much, from guns to concentration camps. So AI will indeed just ravage entire fields, professions, sectors, and domains. It won't be like this in other places. See how much of the world is already banning social media for kids? It'll do the same with AI, eventually, and that's because, for example, unions exist, nor are corporations or economies structured in nearly such a purely capitalist way. People have much, much more power. In America, only billionaires do.
AI is a means of production. That much is true. But of what? It's a means of replication. It takes stuff it's been trained on, remixes it, and spits it back out. AI couldn't write the first book, song, or even legal contract.
So AI is a universal means of digital replication. That means: it can replicate pretty much anything that already exists digitally. What it can't do, and we all know this, is really create something new. This isn't a theoretical or abstract point. All those dumb AI videos that are "original" are actually in the style of serious artists, usually, even if it's just Miyazaki, right down to Van Gogh, and that's true of writing. Lawyers aren't great artists, nor are insurance agents, but the same logic is true.
AI is a universal means of digital replication, and in that way, every profession that depends on digital production will be decimated by it. Now, you can spit out an infinite number of everythings.
An infinite number of everythings. Songs. Videos. Books. Articles. Legal contracts. Insurance agreements. Stock funds. But will any of them be any good?
...Do you know what banality really is? A moral vacuum. When we look at Munch's Scream, we see the tension and agony of the modern world shrieking back at us. It is not a moral vacuum. When the man I admire made this revolutionary music, it was about energy, freedom, power, love. It wasn't a moral vacuum. When Picasso painted Guernica, or Van Gogh painted Starry night, there was no moral vacuum.
AI in this way will force us to rethink what "productivity" is. If I can create an infinite number of everythings with the press of button, then is my productivity infinite?
All these things are lit with the luminous intensity of the human cry. The cry is the primal expression. It is the cry of wounding, of what it is to be alive, which is terror, pain, suffering. Tragedy. From that tragedy comes the need for beauty, for love, for grace. Only in those do we find redemption. Which we express, with our fragile, broken human arms, every time that we lift a pen, write a word, utter a whisper, or embrace someone we care for. All we are ever doing is expression our revolution of love against time and dust. Camus taught us this. Some of us live it.
The Banality Machine takes all that away from us.
13ii26
Why Smart People Can't Agree on Whether AI Is a Revolution or a Toy Alberto Romero
...This is how belief works for all things in life. Nobody convinces you to exercise. You start running one day for your own reasons—perhaps a friend suggested the idea, but you need to see for yourself—and eventually, you're a person who runs. Nobody convinces you that therapy is useful. You go because you're hurt and you tried and it worked, so you keep going. Nobody convinces you to read. One day, you pick up a book and it resonates and then you keep picking up books.)
For some reason, we have a hard time accepting that AI is not a special thing in this regard, but like everything else. The important shifts in attitude and mindset are always private and always experiential.
15ii26
Why the Smartest People in Tech Are Quietly Panicking Right Now Shane Collins at Medium
…
Amodei posed a brilliant thought experiment: Imagine waking up in 2027, and a new country has suddenly appeared on the map. It has 50 million citizens. Every single citizen is smarter than any Nobel laureate in human history. They think 100 times faster than us, they never sleep, and they can control any digital interface on Earth.
…
Many professionals comfort themselves by claiming, “AI will never replace human empathy or trust.” That may be true for a therapist or a bedside nurse. But if your job mostly consists of reading, analyzing, deciding, and typing on a screen — AI is coming for your desk.
…
The people who will suffer most are the ones who let their pride get in the way. Do not assume your degree or your unique industry knowledge makes you immune. Teach yourself — and your children — to be hyper-adaptable orchestrators of AI, rather than memorizers of facts.
17ii26
Anthropic's Moment Alberto Romero
Axios reported this week that the Pentagon may end its partnership with Anthropic. The military wants AI labs to make their models available for "all lawful purposes," and Anthropic won't agree. The company maintains two hard limits: no mass surveillance of Americans and no fully autonomous weaponry. The other three labs working with the Pentagon (OpenAI, Google, xAI) have been more accommodating; one has reportedly already accepted the military's terms.
The relationship deteriorated further over the operation to capture Venezuela's Nicolás Maduro, conducted through Anthropic's partnership with Palantir. According to Axios, a senior administration official claimed an Anthropic executive contacted Palantir to ask whether Claude had been used in the raid, implying disapproval because the operation involved lethal force. Anthropic denied this, saying its conversations with the Pentagon have focused on policy questions around its two hard limits and do not relate to specific operations.
The story has the texture of a pressure campaign: anonymous official, specific framing of Anthropic as the difficult partner, emphasis on competitors' flexibility. Is the industry ganging on the Anthropic safety-focused AI-rights-matter weirdos? I don't know. But whether it works will say something about how durable safety commitments are when the customer is the Department of War.
Pentagon Considers Designating Anthropic AI as a 'Supply Chain Risk' gizmodo
There's a New Term for Workers Freaking Out Over Being Replaced by AI gizmodo
Why Yann LeCun left Meta, and what it means for AI's next frontier Enrique Dans at Medium
Why every company will need its own world model“ Enrique Dans at Medium
Not because language models aren't useful, with their obvious limitations they are, but because they are rapidly becoming commodities. When everyone has access to roughly the same models, trained on roughly the same data, the real question stops being who has the best AI and becomes who understands their world best.
That's where world models come in.
From rented intelligence to owned understanding
Large language models look powerful, but they are fundamentally rented intelligence. You pay a monthly fee to OpenAI, Anthropic, Google or some other big tech, you access them through APIs, you tune them lightly, and you apply them to generic tasks: summarizing, drafting, searching, assisting. They make organizations more efficient, but they don't make them meaningfully different.
A world model is something else entirely.
...What AI changes is not the existence of these models, but their nature. Instead of being static and manually updated, AI-driven world models can be:
AI Is Not a Library: Designing for Nondeterministic Dependencies O'Reilly
Reacting to Matt Yglesias's Confession that A.I. Progress Is Giving Him Writer's Block Brad DerLong
Then the gears shift:
The Columbian Exchange, the growth of a commerce-mediateed global division of labor, and the intellectual sea-change with respect to ideas that was the nullius in verbo, the "we believe nothing simply because somebody says it" of the Royal Society sets the Commercial-Imperial Age rate of technology growth at 25% per century.
The coming of the GPTs of steampower and textile machinery sets the Industrial-Revolution Age rate of technology growth at 100% per century.
The 1875 combo of the invention of the process of invention hallmarked by the appearance of the railroad-and-screw-propeller-driven global economy, the modern corporation, the industrial research lab, and the backing science complex pushes the rate of technology growth up to 2% per year in terms of our improving our ability to make what we make more efficiently, plus whatever the value is in the expansion of the kinds of commodities, objects, and services we can produce:
19ii26
Packaging Expertise: How Claude Skills Turn Judgment into Artifacts Han Lee at O'Reilly
Three Ways Terminal AI Has Changed How I Work (And Whether It's Coming for My Job) Mike Konczal, cross-posted by Brad DeLong
20ii26
The Anthropic Hive Mind Steve Yegge at Medium
...I was at Amazon during their Golden Age, still going strong when I left in 2005. And I was at Google during their Golden Age, which lasted until April 2011. After that I watched Google ossify and become siloed and effectively incapable of cross-functional work, while Amazon continued to execute and innovate.
...At Anthropic, they are smack in the middle of a Golden Age, where there is far more available work than there are people to do it, on pretty much all fronts. It's like they're on the surface of an expanding sphere
[the Singularity] ...a precise concept that mathematician and science fiction author Vernor Vinge formalized in 1993 in a paper titled "The Coming Technological Singularity." His idea was simple: when superhuman intelligence is created, the era of human dominance will end. He compared this moment to the event horizon of a black hole, that boundary beyond which you literally can no longer see what's happening.
SaaS Wikipedia 'Software as a Service'
National Laboratory of the Rockies<./a>...the important thing is to understand that these are not "another ChatGPT on steroids", but autonomous agents who act, make decisions, maintain status, access credentials and, most disturbingly, begin to relate to each other in spaces designed exclusively for them. The emergence of Moltbook, a social network without humans, is the most visible (and most disturbing) symptom of this leap.
...when these A.I. systems began to predict the path of a sentence—that is, to talk—the reaction was widespread delirium. As a cognitive scientist wrote recently, “For hurricanes or pandemics, this is as rigorous as science gets; for sequences of words, everyone seems to lose their mind.”
...Researchers have been mapping the brain for more than a century. By tracing cellular patterns that are visible under a microscope, they've created colorful charts and models that delineate regions and have been able to associate them with functions. In recent years, they've added vastly greater detail: They can now go cell by cell and define each one by its internal genetic activity. But no matter how carefully they slice and how deeply they analyze, their maps of the brain seem incomplete, muddled, inconsistent. For example, some large brain regions have been linked to many different tasks; scientists suspect that they should be subdivided into smaller regions, each with its own job. So far, mapping these cellular neighborhoods from enormous genetic datasets has been both a challenge and a chore.
Papers on agentic and multi-agent systems (MAS) skyrocketed from 820 in 2024 to over 2,500 in 2025. This surge suggests that MAS are now a primary focus for the world's top research labs and universities. Yet there is a disconnect: While research is booming, these systems still frequently fail when they hit production. Most teams instinctively try to fix these failures with better prompts. I use the term prompting fallacy to describe the belief that model and prompt tweaks alone can fix systemic coordination failures. You can't prompt your way out of a system-level failure. If your agents are consistently underperforming, the issue likely isn't the wording of the instruction; it's the architecture of the collaboration.
...Recently Scott Bessent, the Treasury secretary, Stephen Miran, who Trump appointed to the Federal Reserve Board, and Kevin Warsh, his choice as Fed chair, have all argued that the coming AI-led boom in productivity justifies slashing interest rates now. The script goes like this: AI will lead to a surge in productivity; higher productivity will reduce production costs and thereby reduce inflation; and the coming fall in inflation justifies much lower interest rates now.
AI is no longer just a cascade of algorithms trained on massive amounts of data. It has become a physical and infrastructural phenomenon, one whose future will be determined not by breakthroughs in benchmarks, but by the hard realities of power, geography, regulation, and the very nature of intelligence. Businesses that fail to see this will be blindsided.
...consider the fact that the US has hung out a big "not welcome" sign for immigrants. Forget about "we only want the good immigrants"; that's incredibly condescending, and no one of any nationality will believe it, or believe that they'll be treated fairly once they arrive. Every immigrant worker in the US—or considering coming to the US—has to consider the possibility that he will be in the wrong place at the wrong time with the wrong skin color, and end up on a flight to a death camp. Are we surprised that international workers are leaving? Are we surprised that immigrants are arriving in smaller numbers? A $100,000 price tag on H1B visa applications says "We'll only let you in if you make it worth our while." That's gangster talk, not responsible government. The US's ability to train high-quality engineers and programmers and provide them with a high standard of living after graduation has historically been one of its greatest strengths. But given the current policies, are we surprised that fewer international students are coming to the US? China has built an impressive network of colleges and universities, particularly for engineering and the sciences. Students can get a first-rate education without the risks of coming to the US, risks that include having said the wrong things on social media and being sent back at the border.
...There's a tendency for everything that gets written about AI to fashion itself as being "epic", but perhaps that's exactly the wrong mindset given how rapidly the landscape is changing, and incrementalism is better. So I hope you'll excuse this unplanned and slightly stream-of-consciousness take.
...Massachusetts created the nation's first Bureau of Statistics of Labor, hoping that data might accomplish what conscience could not. By measuring work hours, conditions, wages, and what economists now call "negative externalities" but were then called "children's arms torn off," policy makers figured they might be able to produce reasonably fair outcomes for everyone. Or, if you're a bit more cynical, a sustainable level of exploitation.
If the last decade of AI was defined by Large Language Models (LLMs) and Generative AI, the next decade belongs to World Models.
The broader ambition of Palantir extends far beyond the precinct; it is a quest for "Sovereign Intelligence." Under the leadership of CEO Alex Karp, the company has positioned itself as the essential digital infrastructure for Western defense. By merging domestic spying techniques with high-level battlefield AI, Palantir offers a vision of total information dominance.
... Instead of starting with a question, AutoDiscovery starts with your data and asks its own questions—generating hypotheses in natural language, writing and running Python code to test them, interpreting statistical results, and using what it learns to keep exploring. Give it a structured dataset, set an experiment budget, and let it run. You'll get back a complete list of novel research directions, each one fully transparent and reproducible.
"We are pushing all of our chips in on artificial intelligence as a fighting force. The Department is tapping into America's commercial genius, and we're embedding generative AI into our daily battle rhythm," Secretary of War Pete Hegseth said in a press release about the platform. "AI tools present boundless opportunities to increase efficiency, and we are thrilled to witness AI's future positive impact across the War Department."
Rawdog is a slang term that means to endure a difficult, dull, or mundane activity without any diversions or other support while doing it. The word is particularly used in playful or irreverent reference to engaging in ordinary experiences (such as air travel, exercising, or chores) without simultaneously engaging in some form of entertainment (such as listening to music or watching videos).
...If we think about what AI really is, and what it'll do—a question that obsesses Americans in particular, and especially the kinds of Americans who are themselves obsessed with money and power—then my answer goes like this.
...the point is already clear: the experience you have with AI is not determined by AI. Not to get too ontological in a "conversational" post, but the thing is: nothing is in isolation. AI is not good or bad in itself, but fully contingent on who you are. It's determined by: what you do, where you are, how you're wired, what you've tried, when you tried it ... And in practice, none of this is visible to the person on the other side of the forum. To a lot of people, AI is, indeed, nothing more than a fancy stick pretending to be a wand.
…
Most people use the free, default versions of AI, which are intentionally throttled and often years behind the bleeding edge. Furthermore, most people use AI like a search engine — asking quick questions and expecting simple answers.
...The Pentagon doesn't like Anthropic's red lines
[and one wonders how AIRD gets treated...]A series about what's at stake as AI disrupts our jobs
For the last two years, artificial intelligence strategy has largely meant the same thing everywhere: pick a large language model, plug it into your workflows, and start experimenting with prompts. That phase is coming to an end.
This is where reinforcement learning, simulation, and multimodal learning start to matter far more than prompt engineering....The sea-change came 10,000 years ago, when we became more sessile and invented agriculture. Then our “technology” could be not just what we could carry and remember, but what we could build and what our built environment could remind us of. And then 5,000 years ago we invented writing. And we became not just a cultural species, not just a technological species with what we could carry, build, remember, and have our built environment remind us with, but a genuine world-scale æon-scale time-binding space-binding anthology intelligence. Then the real ASI emerged. Not an Artificial Super-Intelligence constructed in a computer lab as some kind of silicon digital god that some want to teach the rest of us to worship and that Elon Musk wants to teach the rest of us to try to bang. But, rather, the distributed knowledge and thought base that is the Anthology Super-Intelligence that is humanity's collective mind present and past stored in and as our information technology capital stock.
But there is more.
...At Anthropic every single person and team I met, without exception, feels kind of sweetly but sadly transcendent. They have a distinct feel of a group of people who are tasked with shepherding something of civilization-level importance into existence, and while they're excited, they all also have a solemn kind of elvish old-world-fading-away gravity. I can't quite put my finger on it.