The people who are surprised by the rise of the Silicon Valley alt-right are the people who weren't paying attention to Silicon Valley

The presence of the far right in tech culture predated Trump by at least a couple of decades, going back to the very beginning of the dot-com bubble. 

Becca Lewis writing for the Guardian [emphasis added]:  

At the height of the dotcom mania in the 1990s, many critics warned of a creeping reactionary fervor. “Forget digital utopia,” wrote the longtime technology journalist Michael Malone, “we could be headed for techno-fascism.” Elsewhere, the writer Paulina Borsook called the valley’s worship of male power “a little reminiscent of the early celebrants of Eurofascism from the 1930s”.

Their voices were largely drowned out by the techno-enthusiasts of the time, but Malone and Borsook were pointing to a vision of Silicon Valley built around a reverence for unlimited male power – and a major pushback when that power was challenged. At the root of this reactionary thinking was a writer and public intellectual named George Gilder. Gilder was one of Silicon Valley’s most vocal evangelists, as well as a popular “futurist” who forecasted coming technological trends. In 1996, he started an investment newsletter that became so popular that it generated rushes on stocks from his readers, in a process that became known as the “Gilder effect”.

Gilder was also a longtime social conservative who brought his politics to Silicon Valley. He had first made his name in the 1970s as an anti-feminist provocateur and a mentee of the conservative stalwart William F Buckley. At a time when women were entering the workforce in unprecedented numbers, he wrote books that argued that traditional gender roles needed to be restored, and he blamed social issues such as poverty on the breakdown of the nuclear family. (He also blamed federal welfare programs, especially those that funded single mothers, claiming they turned men into “cuckolds of the state”). In 1974, the National Organization for Women named him “Male Chauvinist Pig of the Year”; Gilder wore it as a badge of pride.

At the turn of the 1980s, Gilder celebrated the links between capitalism, entrepreneurship and the nuclear family. He claimed that entrepreneurs were the most moral and benevolent people in society, because they put products into the world without a guarantee of return – and then reinvested the profit back into the economy.

...

But at a time when American industrialism was in decline, Gilder helped revitalize a fervor for entrepreneurship and a belief in the moral power of entrepreneurs over industrial workers and company men. Increasingly, Gilder claimed that entrepreneurs were better suited to lead the country into the future than the “experts” found in academia or government.

As we've mentioned before and will come back to again, the antagonism toward experts is an explicitly stated fundamental tenet of techno-optimism and a defining trait of men like Mark Andreessen, Elon Musk, and Larry Ellison. It also played a key role in the rise of dangerous medical pseudoscience during the pandemic. Look up the history of ivermectin and hydroxychloroquine and you'll find names like Ellison and Musk. 

Gilder’s 1981 book Wealth and Poverty became known as the Bible of the Reagan administration, and Reagan began incorporating praise of entrepreneurship into his own speeches. (“If I didn’t know better,” Reagan once stated, “I would be tempted to say that ‘entrepreneur’ is another word for ‘America’.”) Throughout the decade, Reagan used the mythology of entrepreneurship to justify trickle-down economics and cuts to federal welfare programs.

As Gilder became swept up in his own ideas about entrepreneurship, he turned his attention to Silicon Valley. The bourgeoning hi-tech industry, he began claiming, was the purest expression of entrepreneurship in the world. It’s not surprising that Gilder would be drawn to the tech industry in Santa Clara county, California. The state had its own powerful mythologies of masculinity and power. It was the end of the vast frontier, the end of manifest destiny. And it was the place of the former gold rush, where (white) men had struck it rich in the 19th century. It was also, counterintuitively, the birthplace of much of the modern conservative movement, including Reagan’s political career.

On the Media (which has been doing superb work recently) had a highly recommended interview with Lewis. Something else we'll be coming back to. 

 

If the AI Boom turns out to be a bubble, how loud will the pop be?

We talk a lot about AI here at the blog, but so much of our time is spent down in the weeds that it's easy to lose sight of just how massive and multifaceted this story is, to the extent that it might be better to think of it as a collection of major narratives sharing a common link. In terms of impact, the biggest of these narratives is probably the AI boom—or, viewed from another set of assumptions, the AI bubble.

It is difficult to wrap one's head around just how big this story is. 

Robert Armstrong writing for FT: 

There are loads of interesting narratives in that chart (look at the decline of GE, in light grey, for example). But it is worth noting that 30 years ago the industries of the biggest companies were, in descending order, industrials, energy, consumer staples, tech, pharma, consumer staples, tech, retail, tech, consumer staples. Today, the list goes tech, tech, tech, tech, tech, tech, tech, tech, finance, finance. In 2000, there were five tech companies in the top 10. But five years later, three of them were gone; another slightly disconcerting precedent.

The industry represents around a third of the S&P 500. AI spending contributed more to the economy's growth than did the U.S. consumer. Data center construction has exceeded office construction. "But [AI] increasingly looks like the only game in town for growth. Housing, consumption and government spending — nearly 90% of the economy — are now stagnant at best.” 

 

Hiltzik for the LA Times:

That’s the principle undergirding the AI industry’s vast expenditures on data centers and high-performance chips. The demand for more data and more data-crunching capabilities will require about $3 trillion in capital just by 2028, in the estimation of Morgan Stanley. That would outstrip the capacity of the global credit and derivative securities markets. But if AI won’t scale up, most if not all that money will be wasted.

 

The hype has been almost unprecedented, but for now, real world performance has been underwhelming and people are starting to get nervous.

 

 

In addition to the scale. there are a number of other reasons to be more concerned about the AI bubble than the dot-com bubble, not the least of which is the state of the larger economy. Observers are pointing out indicators of stagflation that haven't been around for almost 50 years. Add to that a federal government that is increasingly run by the corrupt and the incompetent—the very last people you would want in charge during an economic crisis. 

This is why women love the Three Stooges

[I wrote this twelve years ago then apparently got distracted and forgot to post it. Since we're still having the same silly conversations on the topic, I decided better late than never.]

 

I'll admit my first my first thought at reading this piece by Yahoo's Beth Greenfield was that the combination of tiny slivers of the lunatic fringe calling themselves movements and lazy journalists looking to make something out of nothing is bound to turn out badly, but then I gave it some more thought:

There’s a controversial new ad that’s gone viral this week, and a vocal group of men are up in arms about it, calling it “sexist.”

Yep, that’s right. This time it’s men who are being belittled to make people buy something: Samsung’s Evolution Kit, which transforms any Samsung television into a smart TV (though the specific product is really beside the point here).

The commercial was introduced last week on YouTube, where it had neared 10 million hits by Thursday afternoon. In it, a woman strolls into the living room to find her husband/boyfriend lying on the couch and watching TV. He’s a slovenly Neanderthal who’s burping, chewing food with his mouth open, grunting, and watching a moronic cartoon. Said wife sighs in disappointment, plugs the new Evolution Kit into the back of the TV, and then fantasizes about what a similar kit could do for her man—which is to turn him into a multitasking (though lobotomized) savior who bakes, cares for baby, paints the kitchen, waters plants, cleans, serves champagne, and plays the flute. But after the fantasy, we're brought back to the real husband sitting and farting on the couch.

Sound funny? Yeah, it is—even if it does rely on every tired male-female trope in the stereotype dictionary. Maybe it’s just always a relief to see men, rather than women (and their butts), as the butt of an advertising joke.

But plenty of guys are angry and offended, and have been sounding off online to let everyone know.

“Probably one of the most vile adverts I've ever seen. This isn't the normal IV drip of laughing at men; this is simply mainlining outright contempt,” wrote one critic in a Reddit “Men’s Rights” page thread on the ad. Another chimed in, “Pure filth. I've seen my fair share of 'men are morons' advertisements, but this one isn't even remotely subtle. The fact that the evolved husband is a brain dead slave to a woman just tops it off.” One man wondered if Samsung was “trying to alienate their primary consumer base,” while others called it “disgusting,” and “sexist,” and promised to boycott the company’s products altogether.

Fed-up men also took to Twitter, and made up much of the 1,500-plus commenters on YouTube. One wondered which “misanthropic, rabid feminist” came up with the concept, while another said he was, “tired of the double standard and depicting men as idiots in advertising.”

A Reddit “Men’s Rights” page isn't just some site for losers with too much time on their hands; it really is a great place to get a representative view of what men are thinking, particularly in this case.

I think I speak for all men when I say that we're tired of having feminists shove images images of loutish men down our throats. We've all been there. The moment the wife or girlfriend gets her hands on the remote, it's Three Stooges marathon for the rest of the night.

Men hate this sort of thing: guys acting like idiots, farting, taking shots to the groin. If you see a man in a movie theater watching Jackass III or the latest Adam Sandler picture, you can bet some feminist dragged him there.

In completely unrelated news, Tesla closed yesterday with a P/E ratio of 194

I have a scene for/from a horror film in my head (I have no idea whether I’m making this up or remembering it). It’s a group of people in a strange room with a large table in the middle covered with an enormous number of pieces from a jigsaw puzzle. As the picture starts to come together, they realize it’s a nightmarish Gothic painting. Then they realize it’s a painting of the room they are in.

I have a feeling that a lot of economists have the same expressions on their faces these days that I’m picturing in that movie scene, with each day revealing a new piece in a very disturbing picture.

Lots of people have been trying to figure out why the market keeps climbing in the face of more and more ominous news. I’ve done a couple of posts approaching this in terms of market psychology, suggesting that traders were in a state of denial about the steady flow of troubling developments we’ve seen over the past six months, but there may be another explanation. It is possible that a large number of traders have made the rational choice and moved their investments somewhere safer, but in the mrkets some of the smart money has been replaced with dumb money.

With a handful of exceptions, most retail investors shouldn’t be investing in individual stocks, period. This has been true at least since the invention of the index fund, but it is arguably more so today. 2025 traders came of age during the asset bubble when stupidity was often rewarded and the dominant investor culture was HODL. With the possible exception of SoftBank and the Saudi sovereign wealth fund, this is the dumbest of dumb money, and in the long term, dumb money is a bad indicator.

 

My mind makes strange connections

When I read this headline...

 Invasive bullfrogs are spreading in the West. Here’s why scientists are so alarmed.

... this was literally my first thought.  

I don't remember much about the movie — I'm not even sure I saw the whole thing — so I can't really recommend it (except in the sense that I'll recommend anything that has Sam Elliott in it), but I can vouch for the Vox article, which is quite good.

American bullfrogs are not native to the Western US. Humans brought them to the region more than a century ago, largely as a food source. And in the years since, the frogs — which are forest green and the size of a small house cat — have multiplied dramatically, spreading to countless ponds and gobbling up everything that fits in their mouths, including federally threatened and endangered species. Conservation scientists now consider them among the most dangerous invasive species in the Western US, and in the 40-plus other countries worldwide where they’ve been introduced.

That leaves bullfrogs in an unusual position. Invasive species are typically brought in from other countries — Burmese pythons in Florida and spotted lanternflies in New York City come from Asia, for example — but American bullfrogs are, as their name suggests, American. They’re both native and invasive in the same country. And the difference of just a few states determines whether we treat them like pests or as an important part of the ecosystem.

Benji Jones deserves credit for his in-depth reporting here, and for giving us the disturbing image of frogs eating "Mice, birds, turtles, snakes, rocks, other bullfrogs."

Elon Musk takes one out of the Sheriff Bart playbook

[I know we just did a Patrick Boyle video, but he really is the man for this one.]

Tesla just gave Musk the biggest executive payout ever in exchange for him not wiping out his own fortune. 

 Elon Musk, despite being Tesla's largest shareholder, has repeatedly demanded additional equity-based compensation, even threatening to leave the company if not granted more control. His $29 billion pay package stands out not just for its size, but for its departure from founder norms. Tesla's board has described this package as being a critical first step to energize and focus Musk. Now, I'm not sure what the following steps will be, but the award increases Musk's stake in Tesla from under 13% to about 16% of the company. The only requirement it puts on him is that he has to stay in a senior leadership role at Tesla for two more years. There's no requirement that he devote any more time to the company — he just can't quit.

Musk's overturned 2018 compensation package had entitled him to 20% of the company, and he has said that he needs to own at least 25% if he's going to advance Tesla's artificial intelligence and robotics capabilities — the most important buzzwords of 2025. He says that 25% would be enough control to prevent an activist investor from ousting him, as he did to the management of Twitter a few years ago.

Despite his claim that he's not interested in wealth and only wants control of Tesla to save humanity, in recent years Musk has sold billions of dollars’ worth of Tesla stock. He managed to sell $7.5 billion worth of Tesla shares near its all-time high in late 2022, right before a sales report that sent the stock price plunging.

His new pay package is being pitched to Tesla investors as a contingency plan: if the $56 billion award from 2018 — which was struck down by a Delaware court for being excessive and poorly disclosed — is reinstated upon appeal, he'll forego the new shares. But if the court rejects Tesla's appeal, Musk will still walk away with the largest pay package in corporate history as a “critical first step.”

It's worth noting that Elon Musk's overturned $56 billion compensation package is worth more than Tesla's entire accumulated net income since inception. As of early 2025, Tesla had earned approximately $38.66 billion in total net profit since going public, meaning that Musk's $56 billion pay package would have represented about 1.45 times the company's lifetime earnings. The more modest $29 billion pay package is less than 80% of Tesla's total net profit since going public, meaning that something is left over for the other shareholders.

The logic behind Musk's award is unconventional. As The Economist puts it, under the Elon Musk theory of pay, the worse Tesla performs, the more its boss ought to earn.

 Just for context, this comes on the heels of Musk destroying the brand, cratering sales, and generally screwing up his job. 

Musk does have some leverage here, albeit for terrible reasons. 

If the goal were to maximize the profitability of Tesla, getting rid of Musk would be a great first step. Even if the objective were to make it the leader in autonomous driving or robotics, you could still find far more qualified and competent executives, none of whom would insist on holding a gun to the company's head.

But the time for talking about the potential earnings or market share of Tesla has long since passed. The company is so insanely overvalued that there's simply no way that it will ever justify its stock price.

For the people invested in the company, the best possible outcome, the only one that avoids catastrophic losses, is for the markets to stay irrational and the bubble to remain inflated long enough for investors to unload their shares on another generation of suckers.

While Musk is probably bad for the company’s financial health, he is absolutely essential for its market cap. Tesla has always traded on the belief that Elon Musk was some sort of superhuman genius who was about to release some wonderful new technology on the world. We've referred to this as flying exoskeletons. Boyle refers to it as “pixie dust.”

What's different now is that the more conventional justifications for buying and holding Tesla stock have all fallen away. This is a small and shrinking 20-plus-year-old car company with a toxic brand, and no real product in the pipeline, coming off of a worse vehicle launch than the Edsel. Exoskeletons and pixie dust are now the only justification for being bullish, and only Elon Musk can hold on to the true believers. 

 
 

Boyle ends the video with link to his segment on John Keely, though he never says why.  

Dr. Elara Voss -- LLMs' own Jungian archetypal memory?

 

 If this were happening with human beings, if a number of people were independently coming up with the same character name for a fictional character with remarkably similar traits and backstory, it would feel like the premise of a particularly creepy science fiction/horror story. In fact, it is remarkably similar to one of my favorite Doctor Who episodes with perhaps my favorite cliffhanger climax.

It’s not quite so inexplicable in the world of large language models, though it is an interesting and informative example. We will be coming back to this one.

 

From Who is Elara Voss?

by Max Read

 There are, as of this writing, 62 books credited to an “Elara Voss” available on Amazon: ... What’s more, there are hundreds of books on Amazon and other self-publishing platforms that feature a character named “Elara Voss,” among them Veil of the Bloodwight Syndicate, Gynarchy’s Collar, and Starlight Nexus by Kylian Quinn:

 

 ...

What’s so odd about this is that--for a name now so common across the megaplatforms--before 2023, “Elara Voss” did not exist. There is no person named Elara Voss in the United States. No birth certificate has ever been issued under that name; if you search for it in public records databases, you’ll turn up no results. There aren’t even any characters named “Elara Voss” in any book published before 2023. Until two years ago, the two words didn’t ever appear next to each other even by accident.

But if you direct almost any L.L.M. to generate a sci-fi story or narrative for you, it will name the main character “Elara Voss”--or a similar variation like “Elara Vex,” “Elena Voss,” or “Elias Vance”--with an alarming degree of frequency.

...

When, exactly, “Elara Voss” and its cognates emerged from the latent space to dominate the Kindle Unlimited store is hard to say. I’ve seen some people on Twitter hazily suggest that Elara and kin--let’s call them promptonyms, to coin a phrase--were present in GPT 3.5 (released in November 2022), but the earliest instance of the names I can find online dates back to August 2023, when an account “exploring realms through #AIStorytelling & #AIConceptArt” posted a character sketch of a “visionary physicist and AI researcher” named “Dr. Elara Voss.” (A “Dr. Elara Finch” and a “Dr. Elara Solis” each appear a few months earlier.) Voss appears a handful more times on Twitter in similar contexts over the next few months, and pops up on a fan-authored Warhammer 40k wiki as the name of the “highly respected leader of the Inanis 23rd Voidstalkers.”

But by the same time next year, Elara, Elena, and Elias Voss, Vex, and Vance had become inescapable, to the point that frequent users and A.I.-powered writing apps began to develop specific prompts to avoid them. The promptonyms reportedly appear in every major L.L.M.: GPT, Claude, Gemini, LLaMA, DeepSeek, and Grok. An A.I. tinkerer on Reddit playing around with L.L.M. benchmarks last August found that Google’s lightweight Gemma model used the name Elara “39 times in 3 separate stories,” and Elias “29 times in 4 separate stories.” None of the commenters were surprised: “every time I try using any models for creative writing, doesn't matter whether it's gpt-4, mistral, llama, etc, always the same names come up like Elara or Whispering Woods, etc.,” one wrote. (Alongside “Whispering Woods,” you can file “Eldora” as the name of a magic kingdom.) Elara Voss seems to be the promptonym generated most often, others like “Aris Thorne” (“Why is Dr. Aris Thorne everywhere?” one Redditor wondered) and “Elias Vance” are found frequently too:

 In fact, a whole host of tropes and concepts seem to accompany Dr. Elara wherever she’s found. The prototypical “Elara Voss,” as described by a text generator, is a doctor, usually a physicist but sometimes a linguist or biologist. (Other times, she’s a spaceship captain.) She’s generally on the verge of a major breakthrough or discovery (often cosmic or even metaphysical in nature), or is researching some kind of “anomaly,” but is isolated, troubled, and sometimes “haunted” by what she’s learning. She’s often found “trembling” or her heart is racing; instruments near her are usually “pulsing.” The name “Erebus” often appears in Dr. Elara stories: a “Project Erebus” on which Dr. Elena Vex is working, or a mining colony named “Erebus-IX” to which Dr. Elias Vance must travel, or even a “rogue A.I.” called Erebus, “neutralized” by Dr. Elara Voss.

 ...

Admittedly, I too have a hard time not giving in to the spooky pleasures of imagining a pantheon of A.I. tulpas emerging from the latent space, a new mythos derived from the hidden structures of our culture. But there are also less occult ways of accounting for the frequency with which Elara and her fellow promptonyms appear. As many people have pointed out, there’s a significant character in World of Warcraft named “Lilian Voss,” and the volume of text publicly available online about the WoW universe in the form of wikis, walk-throughs, and YouTube transcriptions likely gives the lore a gravitational pull in most models. If you trace back, e.g., Gemma’s decision-making process you can see that “character names,” as well as “science fiction” and “fantasy,” are all closely linked in the model to a “text about World of Warcraft” neuron, as Abram Jackson shows here and a user called beowulf shows here. (Similarly, there’s a character named “Elara Dorne” in Star Wars: The Old Republic, a voluminously covered M.M.O.R.P.G. like WoW. It’s a good reminder that L.L.M.s reflect “culture” in the narrow sense of “the culture of text publicly available on the internet in great volume.”)

As for her ubiquity across models, that’s almost to be expected. All the largest models are trained on effectively the same corpus--i.e., “almost all publicly available text”--and the processes by which they are made smooth and sanitary for public release sand down particular differences even further. (They’re also likely training on each other’s responses, which should lead to additional convergence.)

So you might say that Dr. Elara Voss is an emergent legend whose qualities reflect a deep mathematical structure underlying our culture. You might also say she’s a statistical agglomeration of science-fiction cliché borne of oversampling video-game wikis. I’m not sure that either of those views is wrong, precisely! What I do know is that we should enjoy her while she lasts: By the next generation of models, she’ll almost certainly have been eliminated. As much as we might enjoy the idea of A.I. lore, the companies selling the tech (as fiction-writing software, among other things!) don’t like the kind of consistency that points to something other than total magic occurring under the hood.

 

To be a good historian you have to be a good statistician...

... or at least someone who's good at thinking about data, and in this case, survivorship bias.

From Bret Devereaux's highly recommended history blog A Collection of Unmitigated Pedantry.

I’ve mentioned quite a few times here that Roman Egypt is a perplexing part of understanding the Roman Empire because on the one hand it provides a lot of really valuable evidence for daily life concerns (mortality, nuptiality, military pay, customs and tax systems, etc.) but on the other hand it is always very difficult to know to what degree that information can be generalized because Roman Egypt is such an atypical Roman province. So this week we’re going to look in quite general terms at what makes Egypt such an unusual place in the Roman world. As we’ll see, some of the ways in which Egypt is unusual are Roman creations, but many of them stretch back before the Roman period in Egypt or indeed before the Roman period anywhere.
...
Instead what makes Roman Egypt’s uniqueness so important is one of the unique things about it: Roman Egypt preserves a much larger slice of our evidence than any other place in the ancient world. This comes down to climate (as do most things); Egypt is a climatically extreme place. On the one hand, most of the country is desert and here I mean hard desert, with absolutely minuscule amounts of precipitation. On the other hand, the Nile River creates a fertile, at points almost lush, band cutting through the country running to the coast. The change between these two environments is extremely stark; it is, I have been told (I haven’t yet been to Egypt), entirely possible in many places to stand with one foot in the ‘green’ and another foot in the hard desert.

That in turn matters because while Egypt was hardly the only arid region Rome controlled, it was the only place you were likely to find very many large settlements and lots of people living in such close proximity to such extremely arid environments (other large North African settlements tend to be coastal). And that in turn matters for preservation. When objects are deposited – lost, thrown away, carefully placed in a sanctuary, whatever – they begin to degrade. Organic objects (textile, leather, paper, wood) rot as microorganisms use them as food, while metal objects oxidize (that is, rust).¹ Aridity arrests (at least somewhat) both processes. Consequently things survive from the Roman period (or indeed, from even more ancient periods) in Egypt that simply wouldn’t survive almost anywhere else.

...

Now within the typical European and Mediterranean humidity, papyrus doesn’t last forever (unlike the parchment paper produced in the Middle Ages which was far more expensive but also lasts much longer); papyrus paper will degrade over anything from a few decades to a couple hundred years – the more humidity, the faster decay. Of course wood tablets and wax tablets fare no better. What that means is that in most parts of the Roman Empire, very little casual writing survives; what does survive were the sorts of important official documents which might be inscribed on stone (along with the literary works that were worth painstakingly copying over and over again by hand through the Middle Ages). But letters, receipts, tax returns, census records, shopping lists, school assignments – these sorts of documents were all written on less durable materials which don’t survive except in a few exceptional sites like Vindolanda.

Or Egypt. Not individual places in Egypt; pretty much the whole province.

In the extremely dry conditions of the Egyptian desert, papyrus can survive (albeit typically in damaged scraps rather than complete scrolls) from antiquity to the present. Now the coverage of these surviving papyri is not even. The Roman period is far better represented in the surviving papyri than the Ptolemaic period (much less the proceeding ‘late’ period or the New Kingdom before that). It’s also not evenly distributed geographically; the Arsinoite nome (what is today el-Fayyum, an oasis basin to the West of the Nile) and the Oxyrhynchus nome (roughly in the middle of Egypt, on the Nile) are both substantially overrepresented, while the Nile Delta itself has fewer (but by no means zero) finds. Consequently, we need to be worried not only about the degree to which Egypt might be representative of the larger Roman world, but also the degree to which these two nomes (a nome is an administrative district within Egypt, we’ll talk about them more in a bit) are representative of Egypt. That’s complicated in turn by the fact that the Arsinoite nome is not a normal nome; extensive cultivation there only really begins under Ptolemaic rule, which raises questions about how typical it was. It also means we lack a really good trove of papyri from a nome in Lower Egypt proper (the northern part of the country, covering the delta of the Nile) which, because of its different terrain, we might imagine was in some ways different.

 

Trip to the library

I'm a bit surprised I haven't posted this before. 

Emily M. Bender is one of, perhaps the, leading critic of LLM from the theoretically side. (On the business and social impact side I think we'd have to give the title to Ed Zitron.) Though best known for coining the term "stochastic parrot," my favorite example of her work is this essay, in which she demonstrates that, even if the algorithms were intelligent, they still couldn't understand what they were saying.

[I've left out some context. If you think you've spotted a flaw in the logic, you should check out the origin before weighing in.]

 From Thought experiment in the National Library of Thailand: 

To try to bring the difference between form and meaning into focus, I like to lead people through a thought experiment. Think of a language that you do not speak which is furthermore written in a non-ideographic writing system that you don’t read. For many (but by no means all) people reading this post, Thai might fit that description, so I’ll use Thai in this example.

Imagine you are in the National Library of Thailand (Thai wikipedia page). You have access to all the books in that library, except any that have illustrations or any writing not in Thai. You have unlimited time, and your physical needs are catered to, but no people to interact with. Could you learn to understand written Thai? If so, how would you achieve that? (Please ponder for a moment, before reading on.)

I’ve had this conversation with many many people. Some ideas that have come up:

1. Look for an illustrated encyclopedia. [Sorry, I removed all books with photos, remember?]
2. Find scientific articles which might have English loanwords spelled out in English orthography. [Those are gone too. I was thorough.]
3. Patiently collate a list of all strings, locating the most frequent ones, and deduce that those are function words, like the equivalents of and, the, or to, or whichever elements Thai grammaticalizes. [Thai actually doesn’t use white space delimiters for words, so this strategy would be extra challenging. If you succeeded, you’d be succeeding because you were bringing additional knowledge to the situation, something which an LLM doesn’t have. Also, the function words aren’t going to help you much in terms of the actual content.]
4. Unlimited time and yummy Thai food? I’d just sit back and enjoy that. [Great! But also, not going to lead to learning Thai.]
5. Hunt around until you find something that from its format is obviously a translation of a book you already know well in another language. [Again, bringing in external information.]
6. Look at the way the books are organized in the library, and find words (substrings) that appear disproportionate in each section (compared to the others). Deduce that these are the words that have to do with the topic of that section. [That would be an interesting way to partition the vocabulary for sure, but how would you actually figure out what any of the words mean?]

Without any way to relate the texts you are looking at to anything outside language, i.e. to hypotheses about their communicative intent, you can’t get off the ground with this task. Most of the strategies above involve pulling in additional information that would let you make those hypotheses — something beyond the strict form of the language.

... 

You could, if you didn’t get fed up, get really good as knowing what a reasonable string of Thai “looks like”. You could maybe even write something that a Thai speaker could make sense of. But this isn’t the same thing as “knowing Thai”. If you wanted to learn from the knowledge stored in that library, you still wouldn’t have access.

...

It doesn’t matter how “intelligent” [ChatGPT] is — it can’t get to meaning if all it has access to is form. But also: it’s not “intelligent”. Our only evidence for its “intelligence” is the apparent coherence of its output. But we’re the ones doing all the meaning making there, as we make sense of it. 

"About a 30-minute drive south along the highways that split up the city, residents in majority-white areas live on average 24 years longer."

A while back we did a post on the California Environmental Quality Act rollback. The usual YIMBY crowd couldn't imagine why anyone would say a bad word about it, but when you read the fine print, it turned out that some of the most radical and troubling parts of the bill had absolutely nothing to do with increasing housing supply.

 

The new exemption for “advanced manufacturing” facilities in areas already zoned for industrial use — including plants that build semiconductors and nanotech — drew some of the fiercest criticism. State law defines the category as processes that improve or create new materials, products or technologies. 

...

A major proponent of the exemptions, State Sen. Scott Wiener of San Francisco said in an interview with CalMatters today that criticisms by environmentalists were  “extreme, unfounded, melodramatic statements.” 

Elsewhere in the interview, Wiener talked about how environmental deregulation will "bring clean advanced manufacturing to California.” Historically though, this kind of manufacturing has not been by any stretch of the imagination, clean. 

California’s Santa Clara County, the seat of Silicon Valley, has more federal Superfund sites than anywhere else in the US.

The county is home to 23 sites in the US Environmental Protection Agency’s Superfund program, meaning the federal government recognizes them as highly contaminated areas and have earmarked them for cleanup. (It’s the same program the Trump administration seeks to cut by 30%.) Almost all of the Santa Clara Superfund sites are located where there once were (or still are) high-tech manufacturing sites.

 Wiener then emphasized that point about zoning.

Wiener said the changes exempt manufacturing projects only on land that is already zoned as industrial. The goal is to make it easier for high-tech industries to build, with Wiener arguing that California risks losing out on major private-sector investment because it’s too costly and difficult to build in the state. 

Here too, the history is ugly. These industrial areas are disproportionately likely to be near low income, often majority-minority neighborhoods, often with tragic results. 

From Adam Mahoney [Emphasis added]: 

Recent research, co-led by Black women researchers and conducted specifically with Black women residents, found that 80% of Black women in [Settegast, a majority-Black neighborhood in northeast Houston] live in high-risk soil contamination zones, with 80% of those residents reporting chronic health conditions.

...

In a neighborhood once defined by its rural charm and tight-knit community, the slow encroachment of industry, neglect and gentrification has transformed both the landscape and the lives of its residents. The average resident in Settegast is expected to die before they reach retirement age. About a 30-minute drive south along the highways that split up the city, residents in majority-white areas live on average 24 years longer.

The neighborhood sits trapped between a massive rail yard, a freeway, and five industrial sites that release thousands of pounds of lead and toxic chemicals. 

... 

When Rivera moved to Settegast in the late 1970s, “it was changing over from a white community to a Black one,” she explained. “Once we came in here, we did not get the same services as the other community had, and we really didn’t know how to fight for resources to keep it up.” 

The neighborhood’s neglect became visible in the crumbling drainage system, sewage build-up, and the slow but steady encroachment of industry. She watched as concrete batch plants and metal recycling companies crept closer, their hulking machinery and clouds of dust transforming the area. 

“Not only did it start to look different, we lost the smell of the neighborhood, too,” she said. The air, once sweet with the scent of the green earth, became tinged with the pungent smell of industry. 

The transformation of Settegast not only scarred the land but also the humanity of its residents. Residents are more vulnerable to poor health from environmental and climate threats than 99% of Americans, according to research by the Environmental Defense Fund and Texas A&M University. 

Living in a neighborhood where there aren’t grocery stores with fresh food and hospitals remain distant, Rivera’s seen her neighbors die at younger and younger ages. Today, life expectancy there is the lowest in Houston, and men often die before reaching their 60th birthday. She has seen families unravel under the weight of this loss, as well as unpaid mortgages, and mounting property taxes.

Since the neighborhood has an average household income that is less than half the Houston average, she said, when people die young without wills or estate plans, “their families find themselves drowning. Many simply walk away or sell for whatever they can get.”

The bad news is that the push against "debanking" is massively hypocritical and is driven by crypto-bros and fin-tech scammers. The good news is that Patrick Boyle has a video on it.

If you're new to this story, the always reliable Allison Morrow has a good overview.

If you're looking for a more satiric take...

Trump isn't behind all of the looming economic catastrophes. (Sometimes it's his tech bro pals.)

"Capex spending for AI contributed more to growth in the U.S. economy in the past two quarters than all of consumer spending"

 

 

 

These pieces from Paul Kedrosky and the Wall Street Journal have gotten a lot of attention over the past few days, but as worrisome as they might be in isolation, they look far worse in the context of August 2025.

The big takeaway here—a finding reached independently by numerous highly respected analysts—is that capital expenditures on generative AI are the main thing currently propping up the GDP. And that the source of some of that money has left certain financial institutions, particularly life insurance companies (and no, I did not see that one coming), vulnerable to sudden downturns.

Honey, AI Capex is Eating the Economy  

We now have a possible answer. In a sense, there is a massive private sector stimulus program underway in the U.S.. There is an AI datacenter spending program, one that is reallocating gobs of spending, as well as injecting even more. It is already larger than peak telecom spending (as a percentage of GDP) during the dot-com era, and within shouting distance of peak 19th century railroad infrastructure spending.

So, how big was this "stimulus" in the first quarter? Back of ... something or another, based on the above figures:

• Without AI datacenter investment, Q1 GDP contraction could have been closer to –2.1%
• AI capex was likely the early-2025 difference between a mild contraction and a deep one, helping mask underlying economic weakness.

Conclusion

We are in a historically anomalous moment. Regardless of what one thinks about the merits of AI or explosive datacenter expansion, the scale and pace of capital deployment into a rapidly depreciating technology is remarkable. These are not railroads—we aren’t building century-long infrastructure. AI datacenters are short-lived, asset-intensive facilities riding declining-cost technology curves, requiring frequent hardware replacement to preserve margins.

And this surge has unintended consequences. Capital is being aggressively reallocated—from venture funding to internal budgets—at the expense of other sectors. Entire categories are being starved of investment, and large-scale layoffs are already happening. The irony: AI is driving mass job losses well before it has been widely deployed.

 h/t Noah Smith.

 Chris Mims writing for the WSJ:

 

The Magnificent 7 tech firms have collectively spent a record $102.5 billion on capex in their most recent quarters, nearly all from Meta, Alphabet (Google), Microsoft and Amazon. (Apple, Nvidia and Tesla together contributed a mere $6.7 billion.)…

Investor and tech pundit Paul Kedrosky says that, as a percentage of gross domestic product, spending on AI infrastructure has already exceeded spending on telecom and internet infrastructure from the dot-com boom—and it’s still growing. He also argues that one explanation for the U.S. economy’s ongoing strength, despite tariffs, is that spending on IT infrastructure is so big that it’s acting as a sort of private-sector stimulus program…

Capex spending for AI contributed more to growth in the U.S. economy in the past two quarters than all of consumer spending, says Neil Dutta, head of economic research at Renaissance Macro Research, citing data from the Bureau of Economic Analysis.

 I'm no finance guy, so I'm certainly getting some of the subtleties wrong, but very smart people with excellent track records in the field are getting concerned about all this, especially given the growing signs that we’re looking at a massive bubble in the sector.

 From Ed Zitron:

As I write this, NVIDIA is currently sitting at $170 a share — a dramatic reversal of fate after the pummelling it took from the DeepSeek situation in January, which sent it tumbling to a brief late-April trip below $100 before things turned around. 

The Magnificent 7 stocks — NVIDIA, Microsoft, Alphabet (Google), Apple, Meta, Tesla and Amazon — make up around 35% of the value of the US stock market, and of that, NVIDIA's market value makes up about 19% of the Magnificent 7. This dominance is also why ordinary people ought to be deeply concerned about the AI bubble. The Magnificent 7 is almost certainly a big part of their retirement plans, even if they’re not directly invested.

Back in May, Yahoo Finance's Laura Bratton reported that Microsoft (18.9%), Amazon (7.5%), Meta (9.3%), Alphabet (5.6%), and Tesla (0.9%) alone make up 42.4% of NVIDIA's revenue. The breakdown makes things worse. Meta spends 25% — and Microsoft an alarming 47% — of its capital expenditures on NVIDIA chips, and as Bratton notes, Microsoft also spends money renting servers from CoreWeave, which analyst Gil Luria of D.A.Davidson estimates accounted for $8 billion (more than 6%) of NVIDIA's revenue in 2024. Luria also estimates that neocloud companies like CoreWeave and Crusoe — that exist only to prove AI compute services — account for as much as 10% of NVIDIA's revenue.

NVIDIA's climbing stock value comes from its continued revenue growth. In the last four quarters, NVIDIA has seen year-over-year growth of 101%, 94%, 78% and 69%, and, in the last quarter, a little statistic was carefully brushed under the rug: that NVIDIA missed, though narrowly, on data center revenue. This is exactly what it sounds like — GPUs that are used in servers, rather than gaming consoles and PCs (. Analysts estimated it would make $39.4 billion from this category, and NVIDIA only (lol) brought in $39.1 billion. Then again, it could be attributed to its problems in China, especially as the H20 ban has only just been lifted. In any case, it was a miss!

NVIDIA's quarter-over-quarter growth has also become aggressively normal — from 69%, to 59%, to 12%, to 12% again each quarter, which, again, isn't bad (it's pretty great!), but when 88% of your revenue is based on one particular line in your earnings, it's a pretty big concern, at least for me. Look, I'm not a stock analyst, nor am I pretending to be one, so I am keeping this simple:

• NVIDIA relies not only on selling lots of GPUs each quarter, but it must always, always sell more GPUs the next quarter.
• 42% of NVIDIA's revenue comes from Microsoft, Amazon, Meta, Alphabet and Tesla continuing to buy more GPUs.
• NVIDIA's value and continued growth is heavily reliant on hyperscaler purchases and continued interest in generative AI.
• The US stock market's continued health relies, on some level, on five or six companies (it's unclear how much Apple buys GPU-wise) spending billions of dollars on GPUs from NVIDIA.
  • An analysis from portfolio manager Danke Wang from January found that the Magnificent 7 stocks accounted for 47.87% of the Russell 1000 Index's returns in 2024 (an index fund of the 1000 highest-ranked stocks on FTSE Russell’s index).

In simpler terms, 35% of the US stock market is held up by five or six companies buying GPUs. If NVIDIA's growth story stumbles, it will reverberate through the rest of the Magnificent 7, making them rely on their own AI trade stories.

 ...

I realize this sounds a little simplistic, but by my calculations, NVIDIA's value underpins around 8% of the value of the US stock market. At the time of writing, it accounts for roughly 7.5% of the S&P 500 — an index of the 500 largest US publicly-traded companies. A disturbing 88% of Nvidia’s revenue comes from enterprise-scale GPUs primarily used for generative AI, of which five companies' spend makes up 42% of its revenue. In the event that any one of these companies makes significant changes to their investments in NVIDIA chips, it will eventually have a direct and meaningful negative impact on the wider market. 

This would be bad even in the best of economic times, but given our erratic and incompetent administration and the multitude of looming crises and their potential for nasty interactions—and even cascading failures—this has the potential to be very bad indeed.

Quick side note for a future post: one of these days we’re going to start a serious thread on the abundance movement and the dangers of credulously accepting its claims. When we get there, remind me to bring this up, because throwing money at generative AI has always been a big part of the sales pitch.

 

MACA: Make America China Again (OK, the "Again" doesn't really work but you try coming up with cute titles for fifteen years)

Great Marketplace segment from Matt Levin on how a economy (in this case China's) functions when it can't trust its own government's data.

How to deal with untrustworthy government economic data? Look to China 

Liu said nobody really believes China’s National Bureau of Statistics when it says Chinese GDP grew over 5% last quarter.

China watchers like her look for data from third parties that are proxies for economic activity, like freight volumes.

“If economic activities are of a high level, then you would anticipate trucks are moving around,” Liu said.

Investors and economists also look to data on electricity use or nighttime light or heat emissions from satellite imagery.

These work-arounds help but they don't take the place of reliable official numbers.

“There are marketing firms, domestic Chinese firms, some foreign firms in China that do market research,” Kennedy said. “They do their own surveys of consumers.”

But that system is by no means perfect. Sure, the Chinese economy has grown dramatically the past few decades, but Derek Scissors at the American Enterprise Institute says the lack of good government data has nevertheless hurt Chinese businesses.

“Firms can't tell what's going on,” Scissors said. “They look at official data and say, ‘Should we get into this industry? I can't tell.’ And then they make bad decisions.”

As sub-optimal as that may be for China, a similar situation here in the US would almost certainly be worse.

It’s also important to remember here, nobody has ever really trusted the Chinese Communist Party’s statistics, and Chinese markets have learned to adapt.

Scissors said U.S. businesses have come to depend on good government data, and he worries whether politics might infect more than just the jobs report.

“If we're moving into an era where official statistics are unreliable, that's a huge shock the Chinese economy never had,” said Scissors.

Because they never had reliable stats to begin with.

Josh Marshall on the value of credible data and on the cost of losing it

Josh Marshall has a detailed and extremely well-thought-out reaction to the BLS firing on his Editor's Blog. The whole thing is essential reading, but I want to focus on his conclusions, both because of the way they relate to our post yesterday and how they emphasize a point we've been making for months now.

It’s difficult to capture the magnitude of the importance of government economic data (and other data in different realms). Employment numbers and all government economic data are, taken together, like the instrument control panel for the national economy — for policy makers, markets, corporate decision-makers. Trump’s action injects uncertainty into every decision-making process throughout the economy. It’s not too much to say that credible and consistent economic data is a big driver of prosperity and growth. It allows more informed risk-taking, better informed decisions. But it’s not only these in-the-moment decisions. The revised and final data become the canonical record of what happened. And not just for government. Histories of this era will be written based on that data, economic research, etc.

The impact of this decision is great for the credibility of government financial data which, for most of the last century, has been deemed basically beyond reproach as a systemic, professional and non-politicized record of economic fact. The impact on credibility is already there. What actually happens and how those subsequent actions impact the economy going forward remains to be seen.

It has been obvious since April—and has only grown more so in the following months—that the markets have failed to price in the enormous and growing risks facing the U.S. and world economy under Trump. As we saw again today, every time something happens that makes catastrophe even more likely, the markets will respond with a sharp drop and then go back to “things are fine” mode the next day.

 

 

Many—perhaps most—analysts and commentators are more comfortable discussing market moves as rational reactions to new data, but we are purely in the realm of animal spirits now. To make sense of investor behavior in the summer of 2025, you have to think in terms not just of market psychology, but of dysfunctional market psychology, using concepts like denial, cognitive dissonance, or battered spouse syndrome.

The worst possible time to start playing Kriegspiel with the US economy

[First, a word about the admittedly obscure title. There is a chess variant where neither player gets to see the other player’s pieces. When you want to make a move, a referee tells you whether it's legal or illegal and if there's a possibility of a pawn capture. From there, it's up to you to make educated guesses about where your opponent’s pieces are.]

If you keep up with the news at all, you've certainly heard that last week President Trump reacted to a bad jobs report by firing the head of the Bureau of Labor Statistics. The response has been outraged—CNN justifiably termed it Orwellian—but putting aside how bad this is for democracy, I've been thinking about the disturbing practical implications.

One of the ways we distinguish between a healthy nation and a totalitarian or failed state is by the quality of its data. When countries feel the need to cook their own numbers, it is never a good sign. When they do it this openly, it's even worse.

The U.S. has long been known for producing timely, accurate, and trustworthy statistics, and it is difficult to overstate how essential a role they play in policymaking, business decisions, and investment strategies. The Fed needs these numbers to intelligently set policy. Financial institutions need them to allocate resources and manage risk. If we include data about weather, agriculture, etc., countless businesses rely on this government service on a daily basis.

If these sources of information become less accurate or if you undermine people's faith in them, the direct and indirect cost is immense.

Even putting aside all those troubling echoes of 1984, this would be incredibly foolish and reckless even under the best of circumstances—and those are not what we're facing at the moment.

A little over six months into this administration, we are facing multiple self-inflicted, unprecedented crises, all of which have the potential to interact in dangerous and unpredictable ways: a massive tariff war; an attack on the independence of the Fed; a potential labor crisis targeting our food supply; destabilization of the institutions that keep society and commerce moving smoothly; stunning levels of corruption; looming financial crises; and the increasing probability of the ultimate economic tar pit, stagflation. Add to that a weakening dollar and a ballooning deficit.

Between devastating cuts and open attacks on the institutions that provide data, the people who will have to navigate these potential catastrophes will have to do so blind.