Links and Notes for August 28th, 2026

Below are links to articles, highlighted passages[1], and occasional annotations[2] for the week ending on the date in the title, enriching the raw data from Instapaper Likes and Twitter. They are intentionally succinct, else they’d be articles and probably end up in the gigantic backlog of unpublished drafts. YMMV.

Table of Contents

• Public Policy & Politics
• Journalism & Media
• Economy & Finance
• Science & Nature
• Philosophy, Sociology, & Culture
• Technology & Engineering
• LLMs & AI
• Programming

Public Policy & Politics

Google Maps renames “Lake Ontario” to “Lake America” even faster than US government by Ryan Whitwam (Ars Technica)

Oh, wow. I thought that the Babylon Bee was making a stupid, lazy joke about Trump. It turns out that Trump himself is a stupid, lazy joke. FFS.

Anduril: “Hyperscaling” the Worst of America’s Military Industrial Complex by Brian Berletic (Land Destroyer)

“The state-owned enterprises overseeing Russian and Chinese military industrial production are mandated to exchange “intellectual property” between design bureaus and industrial concerns to ensure the best solutions to any given problem are disseminated across the full breadth of each’s military industrial base − fully leveraging these solutions for real-world purposes rather than simply hoarding “intellectual property” to maximize profit on financial ledgers for individual corporations.”

“The US already spends more on its military than Russia, China, and Iran combined − yet is unable to match even just Russian or Chinese military industrial production − and in some cases − falls short of even matching Iranian production rates and capabilities. ”

“For Russia, China, or Iran − additional military funding would be immediately spent on expanding existing factories, building additional factories, investing in research and development, bolstering supply chains, training larger workforces, and other factors necessary to actually expand production. Additional production capacity − even if not used − in a for-purpose system is considered a long-term strategic asset − even if it undermines profitability in the short-term.

“In a for-profit system, additional manufacturing capacity − if not constantly employed to make weapons and thus profits − is considered a liability. US arms manufacturers not only demand public funding to expand production but ironclad contracts to maximize use of it. ”

“For Americans convinced by Luckey’s rhetoric that either Russia or China truly are urgent threats to them and their way of life, it is actually Palmer Luckey, Anduril, and other companies and individuals like them demanding America continue diverting resources needed to thrive at home, toward efforts to needlessly antagonize and coerce other nations abroad.

“This results in transferring the wealth of the American people toward Palmer Luckey’s over-priced weapons and ultimately toward his next new home, rather than toward better education, healthcare, and infrastructure for the average American man, woman, or child.

“Despite the US literally bombing Iran and fighting Russia within its own borders by proxy through Ukraine − neither Russia nor Iran (nor China) have ever attacked the United States within its borders. Within these same US borders, neglect of education, healthcare, and infrastructure has visibly, tangibly reduced the quality of life for the average American.

“Corporations like Anduril siphoning off tax dollars for threats abroad that do not exist, making these resources unavailable to solve threats to Americans at home that do exist − are responsible for this. ”

Eine «neutrale» Schweizer Sanktionspolitik ist realitätsfremd by Helmut Scheben (InfoSperber)

“Die Schweiz beteiligt sich seit Februar 2022 wirtschaftlich an einem Krieg und hat ihre Neutralität damit aufgegeben. Sie verbündete sich mit den USA und der Europäischen Union und hat wirtschaftliche Zwangsmassnahmen ergriffen, um Russland zu sanktionieren.

“Damit beteiligt sich die Schweiz am Verteidigungskrieg, den die EU und die Nato zusammen mit der Regierung in Kiew gegen Russland führen. Die Nato beteiligt sich personell, mit Ausrüstung und mit Waffenlieferung. Bei sämtlichen Boykott-Massnahmen macht die Schweiz mit.”

“Nach Ansicht des russischen Aussenministers hat die Schweiz ihre Neutralität aufgegeben. Sie sei jetzt zum Feind Russlands geworden. «Wenn die beiden Hauptprotagonisten in diesem Punkt übereinstimmen, dann ist etwas schiefgelaufen», erklärte der frühere Schweizer Botschafter und Buchautor Paul Widmer in der «NZZ».”

“Es ist naiv zu glauben, eine intensivierte Zusammenarbeit mit der Nato würde die Schweiz aus solchen militärischen Angriffen und Kriegen heraushalten. Seit dem Zusammenbruch der Sowjetunion greifen die USA und die Nato fast ununterbrochen militärisch in fremden Ländern ein. Der wissenschaftliche Dienst des US-Kongresses dokumentierte im Jahr 2022, dass die USA seit 1991 insgesamt 251 militärische Einsätze im Ausland ausführten («use of U.S. armed forces abroad»). In vielen Fällen war die Nato mit Personal und Waffen beteiligt.”

“Für einen neutralen Staat ist es meistens unmöglich, Sanktionen gegen Staaten zu ergreifen, die «sich nicht an die Regeln halten» und Menschenrechte schwerwiegend verletzen.

“Russland hatte in Afghanistan neun Jahre lang Krieg geführt, Frankreich in Vietnam acht Jahre, die USA in Vietnam neun Jahre und in Afghanistan zwanzig Jahre lang. Gegenwärtig führt Israel einen völkerrechtswidrigen, weil völlig unverhältnismässigen Verteidigungskrieg gegen Gaza, das Westjordanland und den Libanon. Die USA haben zusammen mit Israel völkerrechtswidrig Iran angegriffen.

“Die Schweiz hat keinen dieser Aggressoren mit wirtschaftlichen Sanktionen bestraft. Sie blockierte US-Geschäftsleuten, die der Regierung nahestehen, ihr Vermögen auf Schweizer Konten nicht, als die USA völkerrechtswidrig den Irak angriffen.

“In allen diesen grossen Kriegen hat die Schweiz nicht «gehandelt». Doch seit Russland einen völkerrechtswidrigen Krieg gegen die Ukraine führt und Menschenrechte schwerwiegend verletzt, macht die neutrale Schweiz am Wirtschaftskrieg gegen Russland mit.

“Wenn sich die Schweiz wie bei früheren Kriegen an den «Courant normal» gehalten hätte, wäre sie nicht allein gewesen. Die Hälfte der G-20-Staaten hat sich den Sanktionen gegen Russland nicht angeschlossen.”

“Selbstverständlich zwängt uns diese Neutralitätsinitiative in ein Korsett. Auch das Strafgesetzbuch ist ein Korsett. Auch die Schweizer Verfassung ist ein Korsett. Da wird festgeschrieben, was wir für vernünftig halten. Die Neutralitätsinitiative ist kein Bruch mit unserer Tradition. Das Gegenteil ist der Fall. Wir haben diese Tradition im Februar 2022 gebrochen, als wir die Zwangsmassnahmen gegen Russland sofort mitvollzogen. Mit der Art, wie unser Aussenminister am Flughafen Kloten Präsident Wolodymyr Selensky umarmte, haben wir der ganzen Welt gezeigt, auf welcher Seite wir stehen, dass wir an diesem Krieg gegen Russland teilnehmen.

“Selbstverständlich darf und soll auch die Schweiz Aggressoren und schwere Menschenrechtsverletzungen öffentlich verurteilen. Bundesrat und Parlament müssen Aggressionen «nicht einfach schlulterzuckend zur Kenntnis nehmen», «dazu schweigen» oder «wegschauen», wie es Daniel Foppa in der «NZZ am Sonntag» den Initianten der Neutralitätsinitiative unterschiebt. Aber bei diplomatischen Auftritten, militärisch und besonders auch wirtschaftlich soll sie sich neutral verhalten.”

“Alle vier grossen Medienkonzerne in der Schweiz verbreiten dieselbe Erzählung wie die Mitte-Ständerätin, dass die Neutralitätsinitiative unsere Sicherheit gefährde. Diese sei gefährdet, weil der Russe uns angreifen werde. Verteidigungsminister Martin Pfister sagte, wir müssten uns auf einen baldigen Krieg einstellen. Auch EU und Nato verbreiten, es sei spätestens 2029/30 mit einem Krieg mit Russland zu rechnen.

“Es ist ein ideologischer Glaubenssatz, der massenpsychologisch wirksam ist, nämlich die immer wieder aufgetischte Dominotheorie:”

• Wenn Vietnam nicht gehalten wird, wird ganz Asien kommunistisch. Wenn Nicaragua nicht gehalten wird, dann kippen Guatemala, Honduras, Mexiko und die Sowjetunion steht an der Südgrenze der USA.
• Wenn wir den Irak nicht angreifen, wird Saddam Hussein uns mit seinen ABC-Waffen angreifen. Wenn wir in Afghanistan die Taliban am Hindukusch nicht besiegen, ist Deutschland nicht mehr sicher.
• Wenn wir die Ukraine nicht halten, wird Putin erst das Baltikum überrollen und dann Polen und dann stehen schon bald die russischen Panzer am Bodensee.

“Es ist stets ein extremes Feindbild, das den Leuten eingetrichtert werden muss, damit sie der Teilnahme an einem Krieg zustimmen. Die Menschen und die Parlamente müssen überzeugt werden, dass ein böser Feind uns bedroht. Nur dann kann das Rüstungsgeschäft funktionieren.

“Die Aufrüstung, die im Moment in Europa, in Brüssel, in Berlin betrieben wird, grenzt an Wahnsinn. Es werden Schuldenberge angehäuft, als gäbe es kein Morgen.

“Die Gegner der Neutralitätsinitiative wollen die Zusammenarbeit mit der Nato intensivieren. So sagte es bereits die damalige Verteidigungsministerin Viola Amherd im Jahr 2024. Heute erklärt der aktuelle Verteidigungsminister Martin Pfister, wir müssten uns auf einen baldigen Krieg einstellen.

“Es ist jedoch naiv zu glauben, dass uns eine Annäherung an die Nato von diesen Kriegen verschont. Vielmehr würde die Schweiz in Konflikte hineingezogen.”

Preparing for a post-Trump internet by Cory Doctorow (Pluralistic)

“More likely is that Vance – a weak, unimportant charisma-vacuum who is loathed by all of Trump’s factions – will end up presiding over a far more chaotic period in American governance than anything that happened under Trump.”

“Which brings me to the post-Trump internet. Trump isn’t the first president to figure out that the internet could be weaponized for geopolitical ends. As the Snowden disclosures showed, there has been a longstanding bipartisan consensus that the internet is a great tool for American surveillance, conducted against friend and foe alike.

“But Trump is the first president to openly, directly recruit American tech companies to simply brick foreign officials who displease him, starting with the Chief Prosecutor of the International Criminal Court, who lost his Office 365 and Outlook accounts in retaliation for swearing out a genocide warrant for Netanyahu.

“And then Microsoft obliged Trump again, attacking the Brazilian judge who sentenced Jair Bolsonaro to prison over his unsuccessful coup.

“America’s tech giants have fully, irrevocably fused with Trump. They donated to his campaign. Their CEOs each paid $1m out of their own pockets to sit behind him on the inauguration dais. Google provides location data for Trump’s racist pogroms. Microsoft provides the administrative tools to carry them out. Oracle provides the databases. Apple blocks apps that warn its customers when they’re about to be snatched.

“In exchange, Trump got Canada and the UK to ditch their plans to levy a 3% tax on American tech giants; he got the EU to gut its privacy laws; he sanctioned EU officials who tried to regulate social media; and he’s told the tech companies to go through EU officials’ private messages, looking for anti-Big Tech partisans whom he will ban from ever entering the USA.

“Big Tech has proved that its only principles are not paying taxes, invading your privacy, and not being broken up by antitrust enforcers.

“Tech companies will do anything for any leader who can guarantee those outcomes. There’s no capitulation too petty and stupid for Big Tech”

“There’s only one way out of this mess: the rapid disassembly of the American internet and the rapid creation of a post-American internet, one built on open, auditable, sovereign digital public goods, internationally built and maintained.

“In its own way, the creation of this post-American internet is as urgent and as global as was the creation of the covid vaccines. But whenever I speak to powerful people about this, they ask the same question: “What if this makes Trump mad?”

“This represents a grave failure to take this crisis seriously. Trump doesn’t need to be “mad” to attack your country. Trump attacked Canada over its wildfires, accusing Canada of polluting America’s air.

“But even if Trump never gets mad at you, that’s no guarantee of safety. Trump is not long for this world, and his death or incapacity is no guarantee of the restoration of a “normal” America. And even if America finds its way to “normal” after Trump, that’s no guarantee that it will stay normal. The armed, organized maniacs who have seized power in America and who are cheering him on as he rampages all over the world, kidnapping leaders and dropping bombs on schoolchildren are not going to dig a hole, crawl inside it, and pull the dirt down on top of themselves.

“But let’s say that we find ourselves in the best of worlds, where American fascism is comprehensively defeated and the country embarks on a years-long program of denazification.

“Even in that amazing future, the world should still race to build a post-American internet. The world should have built that internet after the Snowden revelations. That ghastly failure created the Trumpian internet. If the post-Trump internet isn’t a post-American internet, then it’ll only be a matter of time until the next crisis comes along, and the coming years will give Big Tech even more chances to worm its tendrils into the world’s governments, firms and households, making that crisis be even harder to survive. The best time to act was 13 years ago, after Snowden. The second best time is now.”

Thailand sex tourism made in the USA: USS Lincoln docks in Pattaya by BreakThrough News (YouTube)

“They are literally going to the sex tourism capital that again was created because US soldiers when they were fighting the war of aggression on Vietnam needed a place to recharge after committing war crimes.”

“[…] an industry has been created to literally treat a population as sexual service for the people carrying out imperialism.”

As for the clip where Jesse Watters delights in thinking about how U.S. soldiers on extra testosterone are going to tear a swath through all that hot, subservient, Asian (and apparently U.S.-soldier) snatch, Jesse Watters is a filthy human being. He’s absolute trash.

Here’s what he said about soldiers getting testosterone supplements,

“I want [music] these guys on it. But you know what’s going to happen? The guys that don’t need it are going to take it to triple boost. And then they’re going to get out there and women on base, you guys better be careful. Port calls women in Asia, you better be careful because these guys are going to be wild animals and you better watch out.”

He revels in his filthy, stupid, peevish opinions. He delights in the thought of U.S. soldiers raping everything around them. He crows about it on TV. Look at his face: he’s positively delirious with joy at the thought. This is a sick man.

It is a condemnation of that society that he has such a loud voice in the U.S. Shame on anyone who thinks he’s funny or “telling it like is,” Shame on you for even trying to pretend that he was legitimately “warning women.” Any normal person would be condemning rapists rather than blaming women for not being fast enough to get away from them, for fuck’s sake. Anyone who can’t see that has been brainwashed into something that is morally rudderless, and is seemingly incapable of recognizing meanness, cruelty, and stupidity when they see it.

These are the people who ride along in Trump’s wake as he plows along, leaving a trail of trash and despair.

Journalism & Media

You Can Ignore the Bad Pundits by Hamilton Nolan (How Things Work)

“You don’t have to talk about the bad pundits. You don’t have to engage with them. You don’t have to craft painstaking responses to their vapid arguments. You don’t have to take them seriously. Their existence creates no obligation for you. Each new piece they produce generates no appointment on your calendar. The law does not accord them any reservation on your attention. They have no inherent power over you. Instead, they get their power from you. You give it to them, for free. You, who fancy yourself their enemy, provide the fuel for their careers.”

Today In Dystopia: ICE Robots And Pentagon Puppets by Caitlin Johnstone (Substack)

“Liberalism is lengthy Atlantic essays weeping about how trans people are being excluded from operating the foreign baby incineration machine.

“It is true that the US intelligence cartel has become less inclusive than it was a few years ago. It is also true that nobody should work for the US intelligence cartel, whether they be trans, cis, gay, straight, or otherwise. The CIA is one of the most psychopathic institutions that has ever existed, and nobody of any sexual iteration should ever join it.”

Economy & Finance

I Think LLMs Will Prove to Be Consequential Enterprise Software by Freddie deBoer (Substack)

“The amount of money and time and energy and attention and manpower that’s been invested in LLMs is so massive that nothing resembling an ordinary economic return can ever pay it off; with AI now essentially propping up the entire American economy, this is a dire circumstance even though specific firms will survive and eventually thrive”

“You can’t cure cancer because cancer is not one disease and also it evolves under natural selection pressures and sometimes when we reduce rates of one cancer rates of other cancers rise etc etc etc; interstellar travel is a fantasy thanks to basic rules of physics and physiology, not because we’re not smart enough; asteroid mines make absolutely zero sense in basic economic terms; you can’t upload your consciousness because consciousness is not software but embedded into the very cells of our bodies that inevitably die; we already experienced radical life extension thanks to immense reductions in infant mortality and smoking cessation, the add-another-25-years-to-the-average-lifespan stuff is just a slow unglamorous chipping away; limitless cheap energy is being kept from us by thermodynamics, not how advanced our microprocessors are; on and on, the things we’re demanding failing to materialize because they’re not sensible things to demand”

“We have absolutely no idea about where the next big growth field might be if it isn’t this, and this period of permanently irrational stock market growth (which has underwritten all economic optimism) is dependent, more than anything, on that one organizing story”

“[…] the mundane world never ends and we will all spend the rest of our days downloading apps to pay for parking a single time and emptying the diaper pail and trying to figure out how to avoid the third round of small talk with the same coworker in a day.”

“There will be jobs, and there will be cancer, and any missions to Mars will be fundamentally scientific endeavors that do not promise any economic return at all. No human being alive today will make it to 150 years old, not even the youngest infants; any asteroid mining that occurs will be a gimmick and will be a net money loser by a vast margin. Nothing in the world of medicine will happen that will come close to matching the improvements in health and longevity that were achieved in the 20th century, when the boring mundane advances of routine handwashing, antibiotics, and vaccines saved lives by the tens of millions. More and more often, you’ll get around in self-driving cars. And that will be cool. Traffic fatalities will decline.”

“I’m tired of being accused of saying that LLMs will have no influence on human affairs at all. On the contrary, they’re going to make a bunch of people who are already rich even richer. It’s just that, here in this decadent period of late American capitalism, that is the precise opposite of innovation.”

AI, Memory Prices, and the Engineering Response by Paul Kedrosky

“Google described at Semicon Taiwan this week how it is redesigning its systems to get more from scarce memory. Its new chips hold more data close at hand, share memory across thousands of processors, and compress the working memory used by AI models sixfold. Older, cheaper DRAM can also take over less demanding jobs. Combined, these changes create substantially more effective memory capacity without waiting for new fabs.

“The upshot: These changes create a step change in more effective memory capacity, with no costly cleanroom required (even if that’s coming). And Google isn’t the only company taking these steps”

“If KV caches consume 15–40% of memory in a long-context inference workload. Compressing them sixfold reduces total memory requirements by roughly 12–33%.
The same hardware could therefore support about 15–50% more inference work.”

That’s a pretty wide range

Amazon achieves enshittification inception by Cory Doctorow (Pluralistic)

“Amazon imposes an economy-wide tax on nearly every product you buy. Amazon’s junk fees average 51-60% of the purchase price of the things you buy there, and because Amazon has captured a supermajority of the richest 10% of Americans (who have almost all pre-paid for a year’s shipping through Prime), every seller must sell on Amazon, or forego any hope of selling to most of the country’s most prolific shoppers.

“Any seller who signs up for Amazon is agreeing to turn over the majority of their sales income to Amazon, and any seller who raises prices to recoup those sums, must raise prices everywhere, at every retail outlet in the country.

“AI has made Amazon much better at enforcing Most Favored Nation terms, because AI is actually pretty good at parsing competitors’ websites and finding instances of discounting, which Amazon instantaneously punishes by relegating the sellers’ product listings to page umpty-billion of Amazon’s search results.

“[…] Failure to use Amazon’s fulfillment system also relegates your listings to the dregs of Amazon search results, so sellers pay a massive premium to have their parcels delivered by Amazon, to the exclusion of cheaper alternatives that are just as fast and reliable. That’s why Amazon fulfillment is so profitable!”

“[…] search payola is the most profitable business that Amazon operates, full stop.

“Payola accounts for more of Amazon’s profits than anything else the company does. It’s more profitable than all the things Amazon sells directly. It’s more profitable than AWS, Amazon’s industry-leading cloud service.

“Amazon has created a system where the most sales go to the companies that pay the highest bribes, and those companies pass the cost of those bribes onto their customers. The first item on a typical Amazon search results page is 29% more expensive than the best match for your search. The top row is 25% more expensive. The best result is usually on the second screen, somewhere around the 17th position.

“Amazon actively helps its biggest bribers close the sale. Amazon has lots of “comparison shopping” systems built into the service, but one comparison tool is conspicuous by its absence: an “apples to apples” tool that lets you compare unit prices. Amazon’s most prolific bribe-payers package their goods in weird quantities, selling everything from batteries to t-shirts to shampoo in larger or smaller quantities than their competitors. Sorting your search results by price doesn’t actually tell you who’s got the cheapest price per item, because the company with the cheapest AA batteries might be selling a smaller quantity of batteries at a higher price per battery.”

“[…] there is a way in which this payola can be compared to advertising: it competes with advertising. Back in the old days, before a series of K-shaped recoveries created a vast chasm between America’s haves and have-nots; before Amazon captured the majority of well-off American households with Prime; people shopped in lots of places, and in those days, companies advertised in publications, not on Amazon. Websites, newspapers, and newspaper websites made billions from those ads. Amazon’s payola scheme (along with Google, Facebook and other tech monopolists) have captured almost all of that money.”

“Amazon doesn’t just sell search placement; they auction it. Every time you run an Amazon search, the company conducts a special kind of auction called a “sealed-bid second-price auction” (SBSPA).

“Under an SBSPA, bidders secretly tell the auctioneer the very highest price they’re willing to pay. The auctioneer then charges the highest bidder a price equal to the second-highest bid, plus one cent.

“[…]

“In an SBSPA, every bidder enters their highest price, but none of the other bidders know what that price is. This encourages everyone to name their true highest price, but it protects the top bidder in the instance in which they are willing to pay a much higher price than anyone else.”

“The FTC and 22 states just filed a suit against Amazon because Amazon was cheating on its own SBSPA process.

“Over the past 7 years, Amazon has been secretly charging the winning bidder an amount equal to their own sealed bid, not the amount that the next-highest bidder was willing to pay (plus a penny).

“According to the suit, Amazon did this 80% of the time. That is tens of billions of dollars Amazon extracted from platform sellers, who passed those costs onto you, and onto every other retailer in the country (thanks to AI-enforced Most Favored Nation policies).”

“The point of Amazon’s payola scheme is only and ever a way to parasitically extract the maximum amount a platform seller is willing to part with, and by running a fake SBSPA, Amazon was able to trick its customers into revealing those maximum prices.”

“This case was brought by Trump’s FTC, which means that Amazon can get out of it by paying a chud podcaster to tweet at the president and he’ll order them to drop it”

Even I had no idea that Amazon was this criminal an enterprise. I just thought it mistreated its workers and cheat on local and state taxes. I was so wrong.

Taylor Swift: The Good Billionaire? by John Q (Crooked Timber)

“if there’s a problem with Swift’s wealth, it’s not a problem with her, but with the system that makes such massive wealth possible. For Swift, that comes about in several ways.

“First, the money she makes from selling records and merchandise depends on the system of “intellectual property” (in this case, mostly copyrights and trademarks). Unlike many entertainers, she’s been highly successful in capturing this for herself rather than letting record companies grab it.

“Second,her returns from ticket sales to her concerts depend on the availability of stadiums. These are sometimes explicitly subsidised by the public and always dependent on land-use regulation.

“Finally, the returns she gets when she invests her wealth are higher than those available to ordinary households. This is at the core of the r > g point made by Piketty a while back.

“These compounding effects would be greatly reduced by a more progressive income tax system (a top marginal rate of 70 per cent) and by wealth taxes, capturing the excess returns to high wealth. In combination, these would make it very difficult to become a billionaire by entertaining people. Other measures would be needed to deal with the massive accumulations of corporate wealth represented by the multibillionaires.

“Given her dependence on the society that made her wealth possible, Swift can have no legitimate moral objection to a tax system that shares it out more evenly. But is there an economic cost here? The standard story is one about incentives and effort. But it’s hard to believe that Swift would write fewer songs, or perform fewer concerts, if the return were in the hundreds of thousands rather than millions or billions. And, as I already observed, most of the ultra-rich have made their money in actively harmful ways. Even where there is an actually useful product involved, the super-profits have come mostly from monopoly or speculation.”

I don’t think Swift would have bothered spending the time to churn out the song-by-committee, middle-of-the-road corporate sludge that she’s made for the last several albums if the returns hadn’t been so high. It has be psychically draining to do that.

“[…] we could have all the music we need or want in a society without billionaires. Indeed, with policies like a participation income or livable income guarantee, we could see a lot more people making music, some of which might turn out better than what the industry currently gives us.”

AI Bubble Monitor: Wall Street is Betting Trillions on Silicon Valley to Prevail Over Chinese AI by Dean Baker (CounterPunch)

“The Chinese AI companies have attained comparable performance as the best US models. In addition to being far cheaper, the open-weight Chinese systems also have the advantage that they can be downloaded and run on a company’s own computers. This means both that the systems can be customized, and they don’t have to worry about OpenAI sharing their customers’ data with the highest bidder.

“As a result of the lower prices and greater flexibility, Chinese AI is rapidly gaining the bulk of the world market. According to data from OpenRouter, Chinese AI went from less than 10 percent of world AI usage at the start of 2025 to more than 60 percent in early July.”

Premium: The Hater’s Guide To Circular Financing (Part Two) by Ed Zitron (Where's Your Ed At?)

“SB Energy made about $138 million in the first half of 2026, predominantly from selling power. Its data center division made a whopping $653,000.

“Not to worry though, SB Energy has tons of capacity under construction…

“…except 99.4% of that capacity is earmarked for OpenAI, and based on that “RFS” (ready for service) date, it looks like none of it will come online before 2028. In fact, virtually the entirety of SB Energy’s revenue is contingent on A) finishing these data centers and B) OpenAI being able to pay for them.

“Well, let’s not get too worried. Perhaps SB Energy has other data center capacity somewhere? No, no, that’d show up there. Maybe it will…make…money elsewhere? Somehow? I hear it has a $439 billion backlog, it’s gotta make that money at some point, right?

“[…] of that $439 billion, SB Energy estimates that it will make $1 billion of it within the next two years, $12 billion of it within the next four years, $30 billion of it within the next six years, $39 billion within the next eight years, and $357 billion at some point after that. 97% of SB Energy’s revenue backlog will arrive more than four years in the future, and will be contingent on SB Energy being able to spend $178 billion in capital expenditures.

“OpenAI’s leases are split across 17 different SPVs, all of which I assume will try to raise debt at some point.

“To summarize, SoftBank portfolio company SB Energy has signed $439 billion in business with SoftBank portfolio company OpenAI, which is also an investor in SB Energy, as well as its largest (and only real) client. Its ability to make any of this money relies upon it completing two different and incredibly ambitious infrastructure projects — a 1GW data center in Milam County Texas, and a 10GW buildout in Ohio, the latter of which is only half backstopped by NVIDIA if it actually gets built.

“Let’s be frank: this IPO is only made possible by circular financing, with the vast majority of its valuation coming from entirely-theoretical deals with a company that cannot afford to pay it for data center capacity it cannot afford to build.

“This is about as blatant an “emperor has no clothes” situation as you could ask for. 99.4% of SB Energy’s future revenue is contingent upon building data center capacity, which will take years, using funds that have not been raised, all for a customer that will need to make more than ten times its current revenue to pay it.

“Anyone writing about this IPO should be directly informing investors that they are, for the most part, investing in a few signatures and strips of land owned by a company that has, to this point, not actually built an AI data center.

“Instead, most blandly repeat that SB Energy “has a huge contract with OpenAI” and “hundreds of billions of dollars in its revenue backlog.””

The After Math by Indrajit Samarajiva (Indica)

“[…] the US government is ‘shorting’ oil by borrowing it from their own reserves. If the oil price goes down, they would make a ‘profit’ on this when they refill. I put profit in quotes because the US government does not care about losing money, it’s all profit to the oligarchs that actually run the place. The US government doesn’t make money, they make money. The US government can just make up US dollars as long as everyone believes in their bullshit, but that’s the problem.

“Bond yields basically measure rich people’s feelings (as stand-ins for capital). Bond yields tell you how much rich people believe the government’s bullshit. If they think the government is 0% bullshitting and they’ll accept 0% returns, it’s a safe bet. However, if they think the government is 15% bullshitting, they want to be compensated accordingly. Note that what the government says doesn’t have to be true, it just has to be plausible enough for other rich people to believe, and then they’ll all herd in. The bullshit meter is itself bullshit.”

“Again and again, this bullshit works as long as enough (rich) people believe in it, using the fake bullshit meter they call economics. The Empire has no clothes, but no one wants to be the first to call it because they’re all furiously spinning invisible silk. But no one wants to be the last to call it either. And that day is coming soon.

“This is the day Iran is daring to draw closer, and which the US Treasury is desperately trying to stave off. The problem, however, is that this desperation itself becomes a signal that something is terribly wrong.”

“They know the Empire can bleed, they have drawn blood, and they’re offering them a chance to retreat with the rest of their empire intact. The funny thing is that all of Empire’s supposed enemies (Iran, Russia, China) keep offering them a way out. Iran will open Hormuz if the Americans just go home. America could just allow Iranian oil into the market, sweep their manipulation under the Persian rug, and refill their reserves once more. But like a thief who has never actually shopped, they simply do not know how to checkout. Foreign trade with anyone but beaten vassals is literally a foreign concept. Smash and grab is all they know, even if it means crashing their own economy. And so it goes.”

Science & Nature

The shrinking landscape of linguistic diversity in the age of large language models (Nature)

“Across three studies spanning seven datasets in different domains and over 880,000 texts, we show that the widespread adoption of large language models (LLMs) as writing assistants is linked to declines in linguistic diversity, interfering with the societal and psychological insights language provides. While core content is retained when LLMs polish and rewrite texts, LLMs also homogenize writing styles, reducing writing-complexity variance by a statistically significant 21–50% across datasets and models (P ≤ 0.05), and amplify patterns associated with dominant characteristics while suppressing others, emphasizing conformity over individuality. These trends hold across different LLMs, prompts and contexts […]”

In my experience, core content is often mangled, but OK. That is new to LLMs but the trend toward a decrease in linguistic diversity has been ongoing for years. Worldwide memes, an overwhelming number of ESL writers, and a huge influence from a distinctly under-educated country (looking at you, motherland) tend to do that.

Obey

Conform, consume, obey

That’s a shot from the John Carpenter film They Live.

Philosophy, Sociology, & Culture

Supporters Of Capitalism Are Victims Of Their Own Cognitive Biases by Caitlin Johnstone (Substack)

“Survivorship bias is when your analysis focuses on those who made it while ignoring those who didn’t. You got a decent job, bought a house, and are now living comfortably, which means the system works fine and there must be something wrong with anyone who can’t thrive under it.

“All the people who work just as hard as you do but still find themselves struggling go completely missing from your perspective. All the people who are just as clever and resourceful as you but haven’t found financial stability are conveniently overlooked. Everyone whose business venture failed instead of thriving, everyone who got laid off instead of promoted, everyone whose family circumstances were less fortunate than yours, everyone who has a race, gender, disability, or socioeconomic situation which makes their life harder than your own, simply doesn’t factor into your understanding of the way things work.

“You got lucky, and you can point to other people who got lucky, and you use those examples to hide from the inconvenient reality that it could just as easily have been you who got the shit end of the capitalism stick. You see a guy out your window sleeping on the sidewalk and you tell yourself there’s a perfectly good reason why you’re in here while they’re out there […]”

“People who oppose the movement toward a more equitable society often accuse socialists of “jealousy” and “greed”, but that’s just projection; it is their own greed and selfishness which prevents them from sincerely reckoning with the unjust nature of the society they coincidentally wound up on top of. It is true that those who support socialism are frequently less economically privileged than the average capitalism proponent, but that’s only because they have a reality-based perspective on the abusive nature of the capitalist status quo.

“Those who have been fortunate under capitalism often have a harder time forming a reality-based understanding of the system they live under, because they have to fight their way through all the comfortable cognitive biases which insist that everything’s more or less fine the way it is. If they still insist that capitalism is great, it’s only because their own intellectual laziness, ego, and moral cowardice have prevented them from getting real about their circumstances.”

Sorry to Bother You by Freddie deBoer (Substack)

“[…the chronically online liberal] can’t forgive John Davidson for shouting racial slurs at the BAFTAs, can’t forgive a man with Tourette’s for acting like a man with Tourette’s at a ceremony absolutely stuffed with self-regard for its ostensible celebration of people with Tourette’s.

“That last bit, the long-roiling controversy about whether Tourette’s can really cause someone to shout a racial slur in a crowded theater…. What we now know as Tourette’s syndrome was described in texts from the 1400s. That Tourette’s outbursts of whichever kind are truly involuntary has been known to medical science for hundreds of years. It’s a neurological disorder; it’s like blaming someone for coughing. And yet for awhile our media class busily contemplated the notion that maybe this one guy with Tourette’s had chosen to behave that way, that he was actually operating under his own volition when he engaged in the most socially-damning, comprehensively embarrassing behavior you can imagine. If you express skepticism about the fact that the number of people taking the LSAT who request a disability accommodation has exploded at an exponential scale in recent years? That’s ableism. Entertaining the idea that Tourette’s symptoms aren’t really involuntary because that’s what’s momentarily socially convenient? Not ableism, I guess. These things are so unpredictable.”

“[…] as the TikTok tics experience shows, we have ample collective sympathy for the minimally disabled, the dubiously disabled. The lengths that the essaying class went to in order to avoid judging adolescent women who pretended to have Tourette’s for a few months were remarkable. The guy with actual Tourette’s that did the bad thing received no similar sympathy, and the reason is obvious: because he revealed that Tourette’s is actually a disorder, that the condition is bad, that it hurts.”

“I’m frequently accused of not “believing in” ADHD or ASD-1, which certainly isn’t true; indeed, I’m sensitive to all of the online playacting of those conditions in part because they are so hard on the people who actually have them. I don’t pretend that all of these lines are bright or clean or that it’s always obvious who should receive which forms of support or concern. But I can safely say that there is no need to ask for sympathy for those who are already sympathetic, and the bizarre nature of contemporary disability politics leaves liberals looking like the rich man who gives charity only to those who aren’t truly poor. And at some point it does leave you pondering whether any of it ever meant anything at all.”

Technology & Engineering

The Cables that Connect the World (マリウス)

“The hardware that does the heavy lifting out in the deep ocean is about as thick as a garden hose with roughly 25mm across and weighing in at around 1.4 tonnes for every kilometre. The part that carries your data is a small bundle of glass fibres, each one around the same thickness as human hair, sitting in the very middle.”

“Working outward from the core, the fibres sit in a water-blocking gel inside a thin copper or aluminium tube, which is sheathed in polycarbonate, then an aluminium water barrier, then a layer of stranded steel wires that give the cable its tensile strength, then a wrap of mylar tape, and finally an outer skin of polyethylene. The copper is for power, because the cable doubles as a very long extension lead, which we will get to in a moment. Closer to shore, where trawlers and anchors roam, the whole thing gets one or two further jackets of galvanised steel armour wire, swelling it to 50mm or more in diameter and several times the weight. Hence, the cable that surfaces on our Spanish beach is buried a couple of metres down and not simply left lying on the sand.”

“[…] the shore stations at either end have to feed a direct current of anywhere between 3,000 and 15,000 volts down that copper core, to literally power the cable from both ends at once.

“On top of the amplification, modern systems lean on a stack of clever tricks to keep the signal intelligible across thousands of kilometres of glass, including wavelength-division multiplexing to cram many separate colours of light down a single fibre, coherent detection to read them back out, and forward error correction to repair whatever gets garbled along the way.”

LLMs & AI

Bernie Sanders Faces AI Horrors Beyond His Comprehension (YouTube)

At 10:00, I think Hasan was missing the nuance of what they were saying: that they wanted to support their friends (because they’re friends, duh) but not support their dating of an AI chatbot. The people who said “disagree” were saying that they wouldn’t support their friends but then described the ways in which they would support their friends. I think the guy who stayed in “agree” answered the question better, but it was a poorly worded question. That guy “meeting them where they are” is the most constructive answer, rather than just canceling a friend.

That young man turns out to be a labor negotiator and is a very well-spoken and all-around insightful guy.

Programming

There’s no such thing as Just a Tool by Iris Meredith (DeadSimpleTech)

“We should build tools that tend to create gestalts that support individual flourishing and social harmony or, if those cannot be balanced, that carefully balance the need for both against each other and other considerations.

“Our industry, rather contrary to that idea, has tended to create tools destructive to both of these things. We’ve built social media networks that isolate people from each other and have become vectors for the massive spread of disinformation. We’ve built cryptocurrencies that have enabled fraud and deception on an enormous scale. We’ve built classification algorithms that reinforce the worst kinds of bigotry. And of course, there are the fucking LLMs, that do all of these things and a whole lot of other evil shit besides.”

“[…] it’s quite possible for a tool to be used for good things, and in fact for it to be necessary, but for the viewpoint that it fosters to nonetheless be one that we shouldn’t encourage. […] Carrying a gun or driving an SUV will, in general, make you more predisposed to being an asocial asshole who doesn’t care much for human life, and while it’s quite possible to fight this tendency and a lot of people do, the logic of the gestalt for these tools still pushes towards the bad.”

“This means that even when you, for example, use an LLM to build some useful software that helps your nonprofit do well, and even when you’re very careful about what generated code ends up actually being deployed (even if you literally read every line of code), you’re nonetheless fighting against the ready-to-hand nature of the tool that pushes your gestalt (and thus your attitudes and ideologies) towards carelessness and slop every step of the way.”

No, dynamic type systems are not inherently more open by Alexis King in 2020

“It’s easy to discover the assumptions of the Haskell program just by looking at the type definitions. We know, for example, that this application doesn’t care about the timestamp field, since it never appears in any of the payload types. In the dynamically-typed program, we’d have to audit every code path to see whether or not it inspects that field, which would be a lot of error-prone work!”

“This is a case of improper data modeling, but the static type system is not at fault—it has simply been misused. The appropriate way to represent a UserId is to define a new, opaque type:”

newtype UserId = UserId Text
  deriving (Eq, FromJSON, ToJSON)

“Unlike the type alias defined above which simply creates a new name for the existing UUID type, this declaration creates a totally new UserId type that is distinct from all other types, including Text. If we keep the datatype’s constructor private (that is, we don’t export it from the module that defines this type), then the only way to produce a UserId will be to go through its FromJSON parser. Dually, the only things you can do with a UserId are compare it with other UserIds for equality or serialize it using the ToJSON instance. Nothing else is permitted: the type system will prevent you from depending on the remote service’s internal representation of user IDs.

“This illustrates another way that static type systems can provide strong, useful guarantees when manipulating completely opaque data. The runtime representation of a UserId is really just a string, but the type system does not allow you to accidentally use it like it’s a string, nor does it allow you to forge a new UserId out of thin air from an arbitrary string.

“The type system is not a ball and chain forcing you to describe the representation of every value that enters and leaves your program in exquisite detail. Rather, it’s a tool that you can use in whatever way best suits your needs.”

⁠This was a good, well-written article that reminded me that there are so many people out there who are making completely incorrect and invalid claims about programming languages that it would make my head spin, were I to be exposed to it more. For me, the argument that “static languages break when given new types of data” is so nonsensical that I’d never even considered that people might believe that. I suppose that this is a side-effect of “everyone can code”. Although such a statement cannot but be interpreted as being disparaging, I want to emphasize that it means “they don’t know what they’re talking about because they’ve not put in the minimum amount of work to even notice that they don’t know what they’re talking about. Their voice is therefore too loud in this conversation” not “they are too dumb to ever learn enough to be part of this conversation.”

In a way, I see this author, though they’ve written a nice article outlining the benefits of static typing, as “arguing with idiots online” in this essay, which tends to go in the direction of arguing with strawmen. That is, the argument being refuted at the beginning is one that comes from people utterly and thoroughly uneducated in the field, who are seemingly incapable of basic logic about engineering practices and architecture. That’s a strawman. I suppose it’s good to start out like that because the problem of “a ton of people working in this field seem to believe things that are actively harmful to their ability to write good software” seems to be much larger than I’d assumed or had hoped.

Modeling the data structures that our program will use using a statically typed language is an excellent way of ensuring that the requirements have been properly (read: sufficiently thoroughly) communicated and understood by both the “customer” and the “implementor”. As an added benefit, the agreement on the requirements is formulated such that the software also “understands” the requirement. The data models will guide the programmer to fulfill the requirement (the data types are guardrails) and provide a warning when the requirement is extended or modified in any way (when the programmer needs to change the data types, it might be a good idea to check with the creator of the requirement to verify that the changes are still in the spirit of what the customer wanted).

“[…] consider Python classes, which are quite nominal despite being dynamic, and TypeScript interfaces, which are structural despite being static. Indeed, modern statically-typed languages are increasingly acquiring native support for structurally-typed records. In these systems, record types work much like hashes in Clojure—they are not distinct, named types but rather anonymous collections of key-value pairs—and they support many of the same expressive manipulation operations that Clojure’s hashes do, all within a statically-typed framework.

“If you are interested in exploring static type systems with strong support for structural typing, I would recommend taking a look at any of TypeScript, Flow, PureScript, Elm, OCaml, or Reason, all of which have some sort of support for structurally typed records. What I would not recommend for this purpose is Haskell, which has abysmal support for structural typing; Haskell is (for various reasons outside the scope of this blog post) aggressively nominal.”

This is sound advice. I like the distinction between nominal and structural typing. Structural typing is a godsend in many cases.

Static Allocation, Constant Work by Alex Kladov (MatKlad)

“Another benefit is code simplification and predictability. You no longer need to track a separate collection of “live” orders. Instead, you always iterate the full set, doing no-ops for reserved. This feels wasteful: should we make the code run faster when there are few orders? But consider this: by specifying the limit of orders up-front, you commit to be able to serve that amount. If the maximum amount of orders is active, does the system have acceptable performance? If not, that is a bug! Gray failure (system becoming unusably slow) is another way to break when reaching the limit.”

“Similarly to static allocation, the Constant Work principle gives you piece of mind with respect to performance. P100 latency stays flat regardless of the load. Insufficient performance is discovered when you roll out the system, not during Black Friday on-call.”