Abundance and China

Could America pursue an abundance agenda without the threat of the PRC? And can podcasters change the world?

To discuss, ChinaTalk interviewed Ezra Klein and Derek Thompson, who need no introduction, as well as Dan Wang, who has written all those beautiful annual letters and is back in the US as a research fellow at Kotkin’s Hoover History Lab. He has an excellent book called Breakneck coming out this August, but we’re saving that show for a little later this year.

Today, our conversation covers…

• The use of China as a rhetorical device in US domestic discourse,

• Oversimplified aspects of Chinese development, and why the bipartisan consensus surrounding Beijing might fail to produce a coherent strategy,

• The abundance agenda and technocratic vs prophetic strategies for policy change,

• How to conceptualize political actors complexly, including unions, corporations, and environmental groups,

• The value of podcasting and strategies for positively impacting the modern media environment.

Ezra Klein: ... Going back to at least the 2010s, probably before, I’ve begun to really notice this feeling in American politics that they can build and we can’t. This became a pathway through which different kinds of bipartisan legislation that would not otherwise have been bipartisan began to emerge.

The re-emergence of industrial policy in America is 100% about China. Take China out of the equation, and there is no re-emergence of American industrial policy. It’s reasonable from the American perspective, when you’re trying to understand American politics, to understand China as an American political object, because that’s what it actually is in our discourse.

American policymakers don’t understand China at all. Most of what they think about it has a high chance of proving to be dangerously misguided. Dan will be much more expert here than I will, but I’m very skeptical of the bipartisan consensus that has emerged. Nevertheless, it’s completely trackable that China exerts a force on American politics. It has reshaped the American political consensus, often in ways that operate in the shadows because they don’t become part of the major partisan fights of modern American politics. (...)

Dan Wang: I would always be the first person to put my hand up to say I know nothing about what’s going on in China. That is always true...

China is very messy. That is always my first proposition about China — it is very big, and many things are true about China all at the same time. They are a country that claims to be pursuing “socialism with Chinese characteristics,” which is still one of the most wonderful political science terms ever.

What sort of socialism is this? In my view, this is one of the most right-wing regimes in the world. A country that would make any American conservative salivate in terms of its immigration restrictions, its incredible amount of manufacturing prowess, and its enforcement of very traditional gender roles in which men have to be very macho and women have to bear children.

China is all of these things. It is also a place where there are really wonderful bike paths, specifically in Shanghai. This year, Shanghai has completed around 500 parks. By 2030, they want to create 500 more parks. It is a country that is getting better and getting worse all at the same time.

Ezra Klein: This goes back to this idea of envy — the degree to which the right envies China is fascinating. It doesn’t just want to compete with it or beat it. It’s not just afraid of it. What it wants is to be more like it.

America’s politicians are so obsessed with trying to take manufacturing back from China, which I don’t think they have a well-thought-through approach to doing, that they look quite ready to give up America’s financial power. They seem to have reconceived of dollar dominance, which used to be called the “exorbitant privilege” because we got so many advantages from it, as some sort of terrible weakness that has hollowed out our industrial base and that we need to shatter.

Throughout history, being the power that controls the money flows has proven to be an extraordinary lever of control. But it has been recast in current New Right thinking as a sort of feminized decadence — something that “not real” countries and “not real” powers do, a distraction from the “real economy” and the “real work” of making things.

I’m not against bringing back manufacturing. I support the CHIPS Act. There are many aspects of manufacturing that I would like to bring back. But we can become so envious that it becomes hard to see our own advantages and strengths, and then make serious policy built on what we are doing well. That strikes me as one of the profound weaknesses of Washington’s approach to policymaking. It is so obsessed with what we are not doing well that it seems ready to set fire to what we are doing well.

Dan Wang: Edward Luttwak has this term “great state autism,” which he created regarding the US thinking about the Soviet Union. There is certainly an aspect, once you are a “superpower,” of becoming obsessed with the other party. You have to choose your enemies very carefully because you will end up looking quite a lot like them.

I wonder in which way the US is actually quite mimetic in thinking about how to be like the other superpower. In my sense, China — after the 2008 financial crisis, or perhaps after 2012 when Xi came into power — Beijing decided it does not really want to look too much like the US, which has been driven by Wall Street on one coast and Silicon Valley on the other in terms of economic growth.

Rather, Beijing has this purely mercantilist view, which would be recognizable to anyone in the 18th century, which is, “Let’s just make a ton of products. That is our source of power, that is our source of advantage.” (...)

I definitely want to defend the dulcet tones of both Ezra and Derek, but as an amateur member of the community of China watchers, there are debates that aren’t easily resolved. For example, a question I would pose to US policymakers would be: Do you judge it is in America’s interest that China is richer, or is America better off if China is poorer? Having that answer would help structure many subsequent policy choices.

There is debate within the China community about how expansionist China is. They certainly want Taiwan — no question there. But is the next step that they want to take Vietnam, Philippines, as well as Japan? People are extensively debating this. When we can answer these more technocratic questions and reach some agreement, many things become easier.

This isn’t about Ezra’s show, but in the US there aren’t many experts really trying to debate and resolve these questions. In my field studying Chinese technology development and manufacturing, policymakers frequently use the laziest trope that China got where it is totally through stealing. This is easily disprovable, yet we hear it all the time. As long as we can’t move beyond these tropes, it becomes much more difficult to resolve even the harder questions.

Ezra Klein: ... My views are actually quite weak on many of these things. There are areas where I have very strong views about how America should build more and faster. A big portion of the book Derek and I wrote is fundamentally motivated, as we say at the end, by competition with China. We believe we won’t continue thriving as a nation in terms of our own strength if we don’t get better at manufacturing, construction, deployment, innovation, and cyclical experimental policy. There’s something for us to learn and compete with there.

On the narrower level, there’s a view that has taken hold in Washington that some version of decoupling is the way forward. One place where I’m uncertain — not certain I disagree, but the conventional view is so dominant that I’m more interested in the counter-argument — is Tom’s argument from the Huawei campus and his other experiences. He suggests we should do with China in the 2020s what we did with Japan in the 1980s and 1990s when they were outcompeting us on cars: create joint ventures in America where we develop their technological and manufacturing processes and embed them in our own companies. China did this with us too.

In Washington, this is considered virtually unsayable. I’d like to hear a better argument against it than I’ve heard because it’s not obvious that our current approach will accelerate the sophistication of our manufacturing chains.

My view is similar to Dan’s — I’d like us to have more precise conversations about means and ends. But that’s difficult in the current political atmosphere where you have to out-compete others to be symbolically tough or hawkish. (...)

Regarding what we need to do to accelerate our manufacturing and innovative ecosystems, the question of whether we should be decoupling or trying to couple and do tech transfer, engaging in more direct competition with products like Chinese EVs while heavily subsidizing our own industries with clear goals — that doesn’t seem completely crazy to me.

by Jordan Schneider, Ezra Klein, Derek Thompson and Dan Wang, ChinaTalk |  Read more:

Image: YouTube/Zhong Zaiben (钟在本)
[ed. Nice to read such a thoughtful discussion of US/China policy. More informed than most. I've exchanged a couple emails with Dan Wang and am very much looking forward to his upcoming book Breakthrough. If you're unfamiliar with Dan and his annual essays on everything China, I highly recommend you check out his: 2023 letter and 2022 letter.]

Don McCullin, Taking Gifts to the Sea Gods (Bali, 1982)

via:

Krasnov Theory

In February 2025, a rumor circulated online that U.S. President Donald Trump was recruited as an "asset" by Russian intelligence in the late 1980s and given the codename "Krasnov," following allegations from a former Soviet and Kazakh security official, Alnur Mussayev. 

The claim spread on TikTok, Facebook and X, where one account published a thread in response to the rumor, purporting to tie together evidence to support it. (...)

That user wrote: "Now that it's been reveals that Trump has been a Russian asset for 40 years named Krasnov by the FSB, I will write a simple thread of various pieces of information that solidifies the truth of everything I've written." At the time of publishing this article, the thread had been viewed more than 10 million times.

• In February 2025, Alnur Mussayev, a former Soviet and Kazakh security official, claimed in a Facebook post that U.S. President Donald Trump was recruited in 1987 by the KGB, the intelligence agency of the Soviet Union, and assigned the code name "Krasnov."

• Mussayev's post didn't state whether he personally recruited Trump or simply knew about the recruitment, nor did it state whether Trump actively participated in espionage or was just a potential asset.

• Trump did visit Moscow in 1987, but there is no clear evidence suggesting he was actively recruited by the KGB during that trip or at any other time.

• Mussayev's allegations that Trump was recruited by the KGB at that time don't line up with Mussayev's documented career path. Several biographies of him on Russian-language websites suggest that at the time Trump was supposedly recruited, Mussayev was working in the Soviet Union's Ministry of Internal Affairs, not the KGB.

• Trump's pro-Russia stance (compared with other U.S. presidents) has fed into past allegations that he is a Russian asset — for instance, the 2021 book "American Kompromat" featured an interview with a former KGB spy who also claimed the agency recruited Trump as an asset. Again, however, there is no clear evidence supporting this claim.

The claim gained traction when the news website The Daily Beast published a now-deleted story (archived) titled "Former Intelligence Officer Claims KGB Recruited Trump," using only Mussayev's Facebook post as a source. The article described Mussayev's allegations as "unfounded." We contacted The Daily Beast to ask why the story was deleted and will update this story if we receive a response.

We also reached out to Mussayev for comment on the story and will update if he responds.

Meanwhile, Snopes readers wrote in and asked us whether the rumor that Trump was recruited to be a Russian asset was true. Here's what to know:

The allegations don't line up with official records

The allegations originated from a Facebook post that Mussayev published on Feb. 20, 2025 (archived). The post alleged that in 1987, the KGB recruited a "40-year-old businessman from the USA, Donald Trump, nicknamed 'Krasnov.'" Mussayev claimed he was serving in the KGB's Moscow-based Sixth Directorate at the time, and it was "the most important direction" of the department's work to recruit businessmen from "capitalist countries."

Mussayev's post didn't specify whether Trump participated in any spying, only that he was recruited. In an earlier post (archived) from July 18, 2018, he described Trump's relationship with Russian President Vladimir Putin as follows:

Based on my experience of operational work at the KGB-KNB, I can say for sure that Trump belongs to the category of perfectly recruited people. I have no doubt that Russia has a compromise on the President of the United States, that for many years the Kremlin promoted Trump to the position of President of the main world power.

Trump did visit Moscow in 1987, reportedly to look at possible locations for luxury hotels. However, several Russian-language websites (of unknown trustworthiness) with short biographies of Mussayev revealed a discrepancy: While Mussayev claimed he worked in the Sixth Directorate of the KGB in 1987, those online biographies, including one from the Moscow State Institute of International Relations, placed him in Kazakh KGB counterintelligence from 1979 until 1986, when he moved to the Soviet Union's Ministry of Internal Affairs.

It is absolutely possible that the public timeline of Mussayev's work history was established by the KGB as a cover for more covert activities. At face value, however, information on Mussayev's background does not completely align with what he claims.

Other sources corroborated that the Sixth Directorate's main focus was not foreign intelligence. The journalist and author W. Thomas Smith Jr.'s book "Encyclopedia of the Central Intelligence Agency" states that the directorate was responsible for "enforcing financial and trade laws, as well as guarding against economic espionage," in line with the counterintelligence descriptions present in the online biographies. Meanwhile, the First Chief Directorate was the KGB's main espionage arm.

by Amelia Clarke and Jack Izzo, Snopes |  Read more:
Image: Twitter

[ed. True? Does it matter? Results are the same.]

AI, Cartoons and Animation

[ed. See also: What if Making Cartoons Becomes 90% Cheaper? (NYT). Better cat videos?]
via: YouTube/X

No part of the entertainment business has more to lose — and gain — from A.I. than animation.

The $420 billion global animation industry (movies, television cartoons, games, anime) has long been dominated by computer-generated imagery; Walt Disney Animation hasn’t released a hand-drawn film since 2011. In other words, unlike much of Hollywood, animation companies are not technophobic.

Even with computers, however, the process of making an animated movie (or even a cartoon) remains extraordinarily expensive, requiring squadrons of artists, animators, graphic designers, 3-D modelers and other craftspeople. Studios have a big incentive to find a more efficient way, and A.I can already do many of those things far faster, with far fewer people.

Jeffrey Katzenberg, a former chairman of Walt Disney Studios and a co-founder of DreamWorks Animation, has predicted that by next year, it will take only about 50 people to make a major animated movie, down from 500 a decade ago. If he were founding DreamWorks today, Mr. Katzenberg said of A.I. on a recent episode of the podcast “The Speed of Culture,” he would be “jumping into it hook, line and sinker.” (NYT)

via:

Cool Tips

Cool tips
via:

[ed. Probably won't remember any of them (except maybe the broken key).]

Kazuo Ishiguro Reflects on Never Let Me Go, 20 Years Later

On the Decades-Long Creative Process Behind His Most Successful Novel

While I’d been busy writing my fourth and fifth novels, my study had mysteriously transformed itself around me into a kind of miniature indoor jungle. Everywhere were dusty mountains of scribbled-on pages and precarious towers of folders.

In the spring of 2001, however, I began work on my new novel with renewed energy, having just had the room entirely refurbished to my own exacting specifications. I now had well-ordered shelves up to the ceiling and—something I’d wanted for years—two writing surfaces that met in a right angle. My study felt, if anything, even smaller than before (I’ve always preferred to write in small rooms, my back to any view), but I was immensely pleased with it. I’d tell anyone interested how it was like being ensconced in the sleeper compartment of a period luxury train: all I had to do was revolve my chair and reach out a hand to get whatever it was I needed.

One such item now readily accessible was a box file on the shelf to my left marked “Students Novel.” It contained handwritten notes, spidery diagrams, and some typed pages deriving from two separate attempts I’d made—in 1990, then in 1995—to write the novel that was to become Never Let Me Go. On each occasion I’d abandoned the project and gone on to write a completely unrelated novel.

Not that I needed to bring down the file very often: I was quite familiar with its contents. My “students” had no university anywhere near them, nor resembled at all the sort of characters encountered in, say, The Secret History or the “campus novels” of Malcolm Bradbury and David Lodge. Most importantly, I knew they were to share a strange destiny, one that would drastically shorten their lives, yet make them feel special, even superior.

But what was this “strange destiny”—the dimension I hoped would give my novel its unique character?

The answer had continued to elude me throughout the previous decade. I’d toyed with scenarios involving a virus, or exposure to nuclear materials. I even dreamt up once a surreal sequence in which a young hitchhiker, late at night on a foggy motorway, thumbs down a convoy of vehicles and is given a lift in a lorry hauling nuclear missiles across the English countryside.

Despite such flourishes, I’d remained dissatisfied. Every conceit I came up with felt too “tragic,” too melodramatic, or simply ludicrous. Nothing I could conjure would come close to matching the needs of the novel I felt I could see dimly before me in the mists of my imagination.

But now in 2001, as I returned to the project, I could feel something important had changed—and it was not just my study. (...)

***

There might have been other factors around at that time: Dolly the Sheep, history’s first cloned mammal, adorning the fronts of newspapers in 1997; the writing of my two previous novels (The Unconsoled, When We Were Orphans) making me feel more sure-footed about taking deviations from everyday “reality.” In any case, my third attempt at “the Students Novel” went differently to before.

I even had a kind of “eureka” moment—though I was in the shower, not a bath. I suddenly felt I could see before me the entire story. Images, compressed scenes, ran through my mind. Oddly I didn’t feel triumphant or even especially excited. What I recall today is a sense of relief that a missing piece had finally fallen into place, and along with it a kind of melancholy, mixed with something almost like queasiness.

I went about auditioning three different voices for my narrator, having each one narrate the same event over a couple of pages. When I showed the three samples to Lorna, my wife, she picked one without hesitation—a choice that concurred with my own.

After that I worked, by my standards, pretty rapidly in my refurbished study, completing a first draft (albeit in horribly chaotic prose) within nine months. I then worked on the novel for a further two years, throwing away around eighty pages from near the end, and going over and over certain passages.

***

In the twenty years since its publication in 2005, Never Let Me Go has become my most-read book. (In hard sales terms, it overtook quite quickly The Remains of the Day despite the latter’s sixteen years head start, Booker Prize win, and the acclaimed James Ivory film.) The novel has been widely studied in schools and universities, and translated into over fifty languages. It has been adapted into a movie (with Carey Mulligan, Keira Knightley, and Andrew Garfield as Kathy, Ruth, and Tommy—and a superb screenplay, appropriately, by Alex Garland); a Japanese stage play directed by the great Yukio Ninagawa; a ten-part Japanese TV series starring Haruka Ayase; and most recently a British stage play written by Suzanne Heathcote.

This has meant that over the years I’ve been asked many questions about the novel, not just from a range of readers, but from writers, directors, and actors wrestling with the task of transferring this story into a new medium. Reflecting on these questions today, it occurs to me that the great majority of them can be gathered into two broad categories.

The first might be summarized by this question: “Given the awful fate that hangs over these young people, why don’t they run away, or at least show more signs of rebellion?”

The second group of FAQs is slightly harder to characterize, but essentially comes down to: “Is this a sad, bleak book or is it an uplifting, positive one?”

I’m not going to attempt here to answer either of the above, partly because I don’t wish to give spoilers in an introduction, but also because I feel quite content, even proud, that this novel should provoke such questions in readers’ minds. I will however make the following observation—which may possibly make greater sense after you’ve finished the book.

It seems to me that these most-asked questions about Never Let Me Go arise because of tensions concerning its metaphorical identity. Is this story a metaphor about evil man-made systems that already exist today—or are in imminent danger of existing—ushered in by uncontrolled innovations in science and technology? Or, alternatively, is the novel offering a metaphor for the fundamental human condition—the necessary limits of our natural lifespans; the inescapability of aging, sickness, and death; the various strategies we adopt to give our lives meaning and happiness in the time we have allotted to us.

It may be both a strength and a weakness of this novel that it often wishes to be both of the above at one and the same time, thereby setting certain elements of the story in conflict with one another.

by Kazuo Ishiguro, Lit Hub |  Read more:

Image: Never Let Me Go 

'News' in 2025: Eye of the Beholder

Our qualitative research (conducted with 57 Americans in August 2024) and a survey of 9,482 U.S. adults in March 2025 confirm the idea that what news is varies greatly from person to person. And each decides what news means to them and which sources they turn to based on a variety of factors — including their own identities and interests.

These findings build on decades of our own and others’ research on the changing dynamics of news consumption, illuminating key distinctions between what news was and what it is today.

What was news?

Before the rise of digital and social media, researchers had long approached the question of what news is from the journalist perspective. Ideas of news were generally tied to the institution of journalism as media “gatekeepers” determined what was newsworthy, producing and packaging information with a particular tone or set of values for a passive audience.

“The journey we’re on is 30 years ago, the platforms or places where you could be told something you don’t know aside from being personally told by a friend of yours was very limited,” said Nicholas Johnston, the publisher of Axios. “Now, it’s essentially infinite.”

What is news?

In the digital age, when people are exposed to more information from more sources than ever, researchers — including at Pew Research Center — increasingly study news from the audience perspective, as audiences themselves define what “news” is.

“That makes everyone like a wire service editor,” said ProPublica managing editor Tracy Weber. “That’s good. That’s also bad because they’re not trained to be the best wire service editor.”

The reality of exactly how audiences make these assessments is complicated. News is less central to most people’s experiences on digital platforms and social media than journalists often hope. Some work — including our own — finds the definitions people hold for news are not always consistent with their actual behaviors. There are also clashes between what people believe others think of as news and what “feels” like news to them.

Our study finds that people generally say something is more likely to be news if it is factual, up to date, important to society and unbiased. But more than half of Americans also say it’s at least somewhat important for their news sources to have political views similar to their own.

Opinions on whether something is “news” or “not news” also aren’t black and white. Research has found that people view “news” as more of a continuum than a simple yes-or-no question. People classify content as more or less “news-like,” and this varies across platforms and sources, as well as from one person to the next.

Platforms and sources: 

The rise of digital and social media platforms has changed how people experience news. Where people encounter information can influence what they experience as news, and their standards often vary depending on how and where they find it.

For instance, when our qualitative research participants scrolled through their own social media feeds, they drew from cues related to both the topic and its source — including that source’s political orientation — to make case-by-case evaluations about whether the information was news or not.

It is a reflection of the changing nature of how news is distributed that audiences can, and regularly do, make distinctions about what kinds of content they see online. Part of the scrolling experience is making snap judgments about whether content is news and whether to engage with it.

Individuals: Each person brings their own mindset and approach to navigating today’s information environment. For instance, many of our participants acknowledged that their personal identities — including their age, race, ethnicity, gender, religion and especially their political leaning — influence how they consume and think about news.

It also influences how they feel about it. Many of the most common emotions people associate with the news they get are negative: angry, sad, scared, confused.

These feelings can vary depending on people’s political opinions and the political context. Our March 2025 survey found that Democrats are more likely than Republicans to say the news they get makes them feel angry, sad and scared.

But the vast majority of Americans also say news makes them feel informed at least sometimes. And despite all the complications and challenges of the current news environment, many of our participants also view news as an essential part of their life.

Kirsten Eddy, Neiman Lab |  Read more:

Image: via

[ed. See also: What is news, anyway? (NL).]

A pair of slip shade wall light fixtures from the late 20’s and early 30’s. Manufactured by the Lincoln Company.
via:

Art Deco house, San Francisco
via:

via:

Farhad Moshiri (Iranian, 1963-2024), Control Room, 2004. Embroidery on black velvet, 50 x 68 cm

via:

Iggy Pop

“Iggy Pop, lead singer of The Stooges, a late 1960s/early 1970s rock band influential in the development of the nascent hard rock and punk rock genres. In this interview he talks about the method behind the madness.” [ed. here's the song they played that night - Five Foot One]

Wiktor Jackowski (Polish, 1987), suspended river, 2025, oil on canvas, 90 x 120 cm
via:

May 18, 2025: Big Bad Billionaire Bill

AKA: Medicaid Death Watch

Tonight, late on a Sunday night, the House Budget Committee passed what Republicans are calling their “Big, Beautiful Bill” to enact Trump’s agenda although it had failed on Friday when far-right Republicans voted against it, complaining it did not make deep enough cuts to social programs.

The vote tonight was a strict party line vote, with 16 Democrats voting against the measure, 17 Republicans voting for it, and 4 far right Republicans voting “present.” House speaker Mike Johnson (R-LA) said there would be “minor modifications” to the measure; Representative Chip Roy (R-TX) wrote on X that those changes include new work requirements for Medicaid and cuts to green energy subsidies.

And so the bill moves forward.

In The Bulwark today, Jonathan Cohn noted that Republicans are in a tearing hurry to push that Big, Beautiful Bill through Congress before most of us can get a handle on what’s in it. Just a week ago, Cohn notes, there was still no specific language in the measure. Republican leaders didn’t release the piece of the massive bill that would cut Medicaid until last Sunday night and then announced the Committee on Energy and Commerce would take it up not even a full two days later, on Tuesday, before the nonpartisan Congressional Budget Office could produce a detailed analysis of the cost of the proposals. The committee markup happened in a 26-hour marathon in which the parts about Medicaid happened in the middle of the night. And now, the bill moves forward in an unusual meeting late on a Sunday night. (...)

Cohn explains that Medicaid cuts are extremely unpopular, and the Republicans hope to jam those cuts through by claiming they are cutting “waste, fraud, and abuse” without leaving enough time for scrutiny. Cohn points out that if they are truly interested in savings, they could turn instead to the privatized part of Medicare, Medicare Advantage. The Congressional Budget Office estimates that cutting overpayments to Medicare Advantage when private insurers “upcode” care to place patients in a higher risk bracket, could save more than $1 trillion over the next decade.

Instead of saving money, the Big, Beautiful Bill actually blows the budget deficit wide open by extending the 2017 tax cuts for the wealthy and corporations. The Congressional Budget Office estimates that those extensions would cost at least $4.6 trillion over the next ten years. And while the tax cuts would go into effect immediately, the cuts to Medicaid are currently scheduled not to hit until 2029, enabling the Republicans to avoid voter fury over them in the midterms and the 2028 election. [ed. emphasis added]

The prospect of that debt explosion led Moody’s on Friday to downgrade U.S. credit for the first time since 1917, following Fitch, which downgraded the U.S. rating in 2023, and Standard & Poor’s, which did so back in 2011. “If the 2017 Tax Cuts and Jobs Act is extended, which is our base case,” Moody’s explained, “it will add around $4 trillion to the federal fiscal primary (excluding interest payments) deficit over the next decade. As a result, we expect federal deficits to widen, reaching nearly 9% of GDP by 2035, up from 6.4% in 2024, driven mainly by increased interest payments on debt, rising entitlement spending and relatively low revenue generation.” (...)

The continuing Republican insistence that spending is out of control does not reflect reality. In fact, discretionary spending has fallen more than 40% in the past 50 years as a percentage of gross domestic product, from 11% to 6.3%. What has driven rising deficits are the George W. Bush and Donald Trump tax cuts, which had added $8 trillion and $1.7 trillion, respectively, to the debt by the end of the 2023 fiscal year.

But rather than permit those tax cuts to expire— or even to roll them back— the Republicans continue to insist Americans are overtaxed. In fact, the U.S. is far below the average of the 37 other nations in the Organization for Economic Cooperation and Development, an intergovernmental forum of democracies with market economies, in its tax levies. According to a report by the Center for American Progress in 2023, if the U.S. taxed at the average OECD level, over ten years it would have an additional $26 trillion in revenue. If the U.S. taxed at the average of European Union nations, it would have an additional $36 trillion. (...)

So with the current Big, Beautiful Bill, we are looking at a massive transfer of wealth from ordinary Americans to those at the top of American society. The Democratic Women’s Caucus has dubbed the measure the “Big Bad Billionaire Bill.” (...)

Speaker Johnson hopes to pass the bill through the House of Representatives by this Friday, before Memorial Day weekend.

by Heather Cox Richardson, Letters from an American |  Read more:

Image: Speaker of the House Mike Johnson (R-La.). Bill Clark/CQ-Roll Call, Inc via Getty Images

[ed. See also: How to Take Medicaid from Millions of Americans, in Less Than 72 Hours (The Bulwark); Who Changed and America Has Become Trump's Battered Wife (Slack Tide); and, State-by-State Data on Health Insurance Losses from Medicaid, ACA Cuts (Joint Economic Committee).

The Internet is for Extremism

Everything seems insane on the internet.

It’s 2023, and Donald Trump still dominates American political discussion. The internet is filled with wild MAGA nonsense, and if you follow politics online you’ve probably also learned what a Tankie is against your will. But it’s not just politics - everything seems insane. Influencers are doing crazier and crazier stunts to go viral. Pop culture fights happen more often and with more venom. Niche communities seem to fall into deranged niche drama more easily than ever.

To understand how Donald Trump used the internet to take over American politics - and why everything else is also going insane - we first need to understand MrBeast.

You have to go bigger

YouTuber MrBeast could fairly claim to be the biggest online content creator in the world. He’s the most-subscribed individual creator on Youtube. He has more than 290 million followers across his YouTube channels and his videos have collected more than 45 billion views. And it’s possible that no one in the world has thought as deeply about how to go viral as he has.

MrBeast has talked at length about his obsession with YouTube, producing content, and going viral. He often talks about how he’s been uploading videos since he was 11 years old, how he’s probably spent 40,000 hours discussing content creation tactics for the YouTube platform. He faked going to community college to live at home with his parents and make content 15 hours a day. He’s the kind of guy who has extremely detailed (and evidence backed) opinions about the facial expressions that go on video thumbnails, how often a video should jump cut, and what types of videos will get views. So it’s worthwhile to think about some of his earliest viral videos, what he’s making now, and what it says about the nature of virality. (...)

MrBeast was 19 and a small-time YouTuber, nowhere near a household name. He was offered the biggest sponsorship he’d ever been offered to date - 5,000 dollars - and his immediate reaction was ‘Double that and let me give it away to a random homeless person’. He ended up being right, and the video went insanely viral. He knew that the bigger the number (especially if it could break into five digits and be 10,000 dollars) the better the video would do.

The instinct to go bigger has informed virtually everything MrBeast has done since then. He soon had a new video giving away $20,000 to homeless people, then $100,000, then an actual house. He is always pushing the limits, doing bigger and wilder and more, and not just when it comes to giving away money. He’s driven through the same drive-through 1000 times straight. He spent four million dollars to enact a real life Squid Game and bought a train so he could run it off a cliff. He spent two days buried alive in a coffin and a week stranded on a raft in the middle of the ocean. He’s given away a private island a cured 1000 blind and deaf people.

The biggest and probably most knowledgeable content creator on the planet has one philosophy - if you want people to watch, push things to the extreme. And this rule doesn’t just govern YouTube videos. It governs everything we do online.

The MrBeastification of Everything

Think of a relatively normal and uncontroversial thing to be a fan of. Let’s pick hot sauce - imagine yourself as a hot sauce aficionado. If you were living 30 years ago before social media, your options were pretty limited. Maybe you’d know a couple restaurants nearby with pretty spicy food. Maybe you read an extremely niche hot sauce magazine that would publish twice a year for a tiny audience. Maybe you knew of a mail order company that sold some really hot stuff, hotter than you could get at the supermarket. But fundamentally, even as an obsessive fan, your options were pretty limited.

Today, your options are not limited. There are dedicated hot sauce forums online. There are mountains of social content analyzing hot sauce, discussing hot sauce, watching celebrities eat incredibly hot chicken wings. There’s a hot sauce subreddit with hundreds of thousands of subscribers where people have incredibly strong opinions about this topic.

We can even measure this empirically - the hottest pepper in the world today is up to 10x as hot as the world’s hottest pepper in the 1990s. And it’s also far more accessible. You can buy hotter sauces than ever before, easier than ever before, and it’s all thanks to the power of the internet. Hot sauce is undergoing MrBeastification - always pushing for hotter.

There are times where it seems like this extremism is happening to everything online. Celebrity fandoms are more extreme than ever - in fact, it’s no longer enough to be a wildly deranged stan, you must also engage in the anti-fandoms that are now common. Sorority rush has gone from a relatively understated affair to a giant social media production requiring intense planning. Our financial scams have gone from straightforward ponzi schemes to meme stocks and cryptocurrencies that border on being actual cults. Fringe beliefs in every field - economics, vaccinology, history - are flourishing. The more conspiratorial and extreme the view, the better it tends to do on the internet. Everything is being pushed to be the most extreme version of itself.

The Incentives We’re Chasing

What’s causing this? It’s the social web. There are a couple of structural ways in which the internet empowers and incentivizes extremism. (...)

Fundamentally, the social dynamics of the internet turbo-charge this extremism. When the amount of content available online is near-infinite, why wouldn’t you gravitate towards the content that is the most of whatever it is you’re searching for? Why watch a video where a cook bakes a 10 pound cake when you could watch a 50 pound cake? Why watch someone give away a thousand dollars when you could watch someone else give away a hundred thousand dollars? MrBeast recognized this early on and he’s correct - this is how things work with social content. People always want bigger and crazier and more extreme.

by Jeremiah Johnson, Infinite Scroll |  Read more:

Image: MrBeast

[ed. Reveal parties, celebrity outfits, weed, $400 million gift planes... everything.]

Tundra tires
via:

‘Aquamosh’: Plastilina Mosh’s Weido Pop Masterpiece

During the commercial heyday of Mexican rock, bands made their mark by fusing different genres. Artists such as Caifanes mixed post-punk and arena rock with pre-Hispanic music, while others like Café Tacvba and Maldita Vecindad borrowed elements from ska, punk, son cubano, cumbia, disco, and more. But few artists deftly combined as much as Monterrey’s Plastilina Mosh.

The duo, composed of Juan José “Jonás” Gonzalez and Alejandro Rosso, made wildly exploratory music using both traditional instruments and state-of-the-art tools like computers and samplers. They made their mark right from the get-go with their 1998 full-length debut, Aquamosh. It was an amazingly creative and fun mishmash in which everything from lounge to industrial coalesced into a nearly flawless record. It helped establish Plastilina Mosh as auteurs of experimental hook-laden music that still sounds fresh decades after its release.

Plastilina Mosh started in Monterrey, Nuevo Leon in 1997. Jonaz had played in a metal band called Koervoz De Malta and Rosso, a classically-trained musician, played keyboards in the prog-leaning outfit Acarnienses. Both had interest in a wide range of music, from acid jazz to punk.

Around this time, Monterrey was becoming a mecca for music in Mexico. The Mexican rock boom – which started with bands like Caifanes and Botellita de Jerez in the late 1980s – had its epicenter in Mexico City. But as the 90s progressed, attention began to shift to Monterrey, with the G-funk-inspired Control Machete, the power pop-meets-rap rock of Zurdok, the Britpop-leaning Jumbo, the Latin rhythms of El Gran Silencio, and many more. The press dubbed this generation of bands La Avanzada Regia. Loosely translated, it means “The Regal Avant-Garde.” (“Regio” is a nickname for people from Monterrey.) (...)

Today, their status as elder statesmen in the Mexican scene is secure. They paved the way for more Mexican music fusionists like Nortec and 3BallMTY, groups that put together genres like norteño, cumbia with electronics, and hip-hop. Much like the Beastie Boys and Beck in the United States, the group’s music predicted a generation that’s grown up on short-form video and eclectic playlists, where hip-hop, corridos, and rock mix together without a second thought.

by Marcos Hassan, Udiscovermusic |  Read more:

[ed. Glad to see the boys finally getting their due. Here are a couple videos from their post-Aquamosh period:]
Plastilina Mosh returned with Juan Manuel in 2000, abandoning the adrenaline-inducing punk attitude of Aquamosh to delve into dance music, disco and trip hop, all with their fun-loving anarchic spirit in place. Later, they leaned toward melodic experiments with songs like “Peligroso Pop” and “Perverted Pop Song,” showcasing their ability to make picture perfect power pop without sacrificing their experimental instincts.

Castígame Diviértete Sé que gozas y me gusta 

[Punish me I know that I have behaved badly Have fun I know you enjoy and like]

Te quiero igual y, No sé ni como aguantar, Ni controlar mis deseos de morderte, En donde no te puedes mirar, Mientras busco distracción en el radio, o en tu conversación, o en una estúpida canción

[I love you the same and, I don't even know how to endure, Nor control my desire to bite you, Where you can't look, While I am looking for distraction on the radio, or in your conversation or in a stupid song]

y me detengo con tus ojos cristalinos, Como gotas de champány sin embargo no dejó de pensar en lo suave de tus labios cuando sueles besary el sabor de tu saliva cuando empiezas a amary entiende lo que digo esto es fácil solo sigue el manual no tiene tanto problema es cuestión de escuchar no busques una tangente es fácil como decir que yo te gusto como tú a mí

[And I stop with your crystalline eyes As shampán drops And yet he kept thinking In the soft lips when you usually kiss And the taste of your saliva when you start loving And understand what I say this is easy Just follow the manual You don't have so much problem It's a matter of listening Don't look for a tangent It's easy to say that I liked you like you like me like you like me]

[ed. Bob Dylan they ain't.]

Ping Pong Bot Returns Shots With High-Speed Precision

Ping pong bot returns shots with high-speed precision (MIT News).

Image: David Nguyen, Kendrick Cancio and Sangbae Kim (with video)

"Building robots to play ping pong is a challenge that researchers have taken up since the 1980s. The problem requires a unique combination of technologies, including high-speed machine vision, fast and nimble motors and actuators, precise manipulator control, and accurate, real-time prediction, as well as higher-level planning of game strategy."

“If you think of the spectrum of control problems in robotics, we have on one end manipulation, which is usually slow and very precise, such as picking up an object and making sure you’re grasping it well. On the other end, you have locomotion, which is about being dynamic and adapting to perturbations in your system,” Nguyen explains. “Ping pong sits in between those. You’re still doing manipulation, in that you have to be precise in hitting the ball, but you have to hit it within 300 milliseconds. So, it balances similar problems of dynamic locomotion and precise manipulation.”

Multi-Agent Risks from Advanced AI

Executive Summary 

The proliferation of increasingly advanced AI not only promises widespread benefits, but also presents new risks (Bengio et al., 2024; Chan et al., 2023). Today, AI systems are beginning to autonomously interact with one another and adapt their behaviour accordingly, forming multiagent systems. This change is due to the widespread adoption of sophisticated models that can interact via a range of modalities (including text, images, and audio), and the competitive advantages conferred by autonomous, adaptive agents (Anthropic, 2024a; Google DeepMind, 2024; OpenAI, 2025). 

While still relatively rare, groups of advanced AI agents are already responsible for tasks that range from trading million-dollar assets (AmplifyETFs, 2025; Ferreira et al., 2021; Sun et al., 2023a) to recommending actions to commanders in battle (Black et al., 2024; Manson, 2024; Palantir, 2025). In the near future, applications will include not only economic and military domains, but are likely to extend to energy management, transport networks, and other critical infrastructure (Camacho et al., 2024; Mayorkas, 2024). Large populations of AI agents will also feature in more familiar social settings as intelligent personal assistants or representatives, capable of being delegated increasingly complex and important tasks. 

While bringing new opportunities for scalable automation and more diffuse benefits to society, these advanced, multi-agent systems present novel risks that are distinct from those posed by single agents or less advanced technologies, and which have been systematically underappreciated and understudied. This lack of attention is partly because present-day multi-agent systems are rare (and those that do exist are often highly controlled, such as in automated warehouses), but also because even single agents present many unsolved problems (Amodei et al., 2016; Anwar et al., 2024; Hendrycks et al., 2021). Given the current rate of progress and adoption, however, we urgently need to evaluate (and prepare to mitigate) multi-agent risks from advanced AI. More concretely, we provide recommendations throughout the report that can largely be classified as follows. 

• Evaluation: Today’s AI systems are developed and tested in isolation, despite the fact that they will soon interact with each other. In order to understand how likely and severe multi-agent risks are, we need new methods of detecting how and when they might arise, such as: evaluating the cooperative capabilities, biases, and vulnerabilities of models; testing for new or improved dangerous capabilities in multi-agent settings (such as manipulation, collusion, or overriding safeguards); more open-ended simulations to study dynamics, selection pressures, and emergent behaviours; and studies of how well these tests and simulations match real-world deployments. 

• Mitigation: Evaluation is only the first step towards mitigating multi-agent risks, which will require new technical advances. While our understanding of these risks is still growing, there are promising directions that we can begin to explore now, such as: scaling peer incentivisation methods to state-of-the-art models; developing secure protocols for trusted agent interactions; leveraging information design and the potential transparency of AI agents; and stabilising dynamic multi-agent networks and ensuring they are robust to the presence of adversaries. 

• Collaboration: Multi-agent risks inherently involve many different actors and stakeholders, often in complex, dynamic environments. Greater progress can be made on these interdisciplinary problems by leveraging insights from other fields, such as: better understanding the causes of undesirable outcomes in complex adaptive systems and evolutionary settings; determining the moral responsibilities and legal liabilities for harms not caused by any single AI system; drawing lessons from existing efforts to regulate multi-agent systems in high-stakes contexts, such as financial markets; and determining the security vulnerabilities and affordances of multi-agent systems. 

To support these recommendations, we introduce a taxonomy of AI risks that are new, much more challenging, or qualitatively different in the multi-agent setting, together with a preliminary assessment of what can be done to mitigate them. We identify three high-level failure modes, which depend on the nature of the agents’ objectives and the intended behaviour of the system: miscoordination, conflict, and collusion. We then describe seven key risk factors that can lead to these failures: information asymmetries, network effects, selection pressures, destabilising dynamics, commitment and trust, emergent agency, and multi-agent security. For each problem we provide a definition, key instances of how and where it can arise, illustrative case studies, and promising directions for future work. We conclude by discussing the implications for existing work in AI safety, AI governance, and AI ethics. (...)

***

While coordinated human hacking teams or botnets already pose ‘multi-agent’ security risks, their speed and adaptability are limited by human coordination or static strategies. As AI agents become more autonomous and capable of learning and complex reasoning, however, they will be more easily able to dynamically strategize, collude, and decompose tasks to evade defences. At the same time, security efforts aimed at preventing attacks to (or harmful actions from) a single advanced AI system are comparatively simple, as they primarily require monitoring a single entity. The emergence of advanced multi-agent systems therefore raises new vulnerabilities that do not appear in single-agent or less advanced multiagent contexts. (...)

Undetectable Threats. Cooperation and trust in many multi-agent systems relies crucially on the ability to detect (and then avoid or sanction) adversarial actions taken by others (Ostrom, 1990; Schneier, 2012). Recent developments, however, have shown that AI agents are capable of both steganographic communication (Motwani et al., 2024; Schroeder de Witt et al., 2023b) and ‘illusory’ attacks (Franzmeyer et al., 2023), which are black-box undetectable and can even be hidden using white-box undetectable encrypted backdoors (Draguns et al., 2024). Similarly, in environments where agents learn from interactions with others, it is possible for agents to secretly poison the training data of others (Halawi et al., 2024; Wei et al., 2023). If left unchecked, these new attack methods could rapidly destabilise cooperation and coordination in multi-agent systems.

by Lewis Hammond, et. al, Cooperative AI Foundation |  Read more:

Image: via

[ed. Not looking good for humans. See also: The Future of AI: Collaborating Machines and the Rise of Multi-Agent Systems (Medium); and, Vision-language models can’t handle queries with negation words (MIT):]

***

In a new study, MIT researchers have found that vision-language models are extremely likely to make such a mistake in real-world situations because they don’t understand negation — words like “no” and “doesn’t” that specify what is false or absent.

“Those negation words can have a very significant impact, and if we are just using these models blindly, we may run into catastrophic consequences,” says Kumail Alhamoud, an MIT graduate student and lead author of this study. (...)

They hope their research alerts potential users to a previously unnoticed shortcoming that could have serious implications in high-stakes settings where these models are currently being used, from determining which patients receive certain treatments to identifying product defects in manufacturing plants.

“This is a technical paper, but there are bigger issues to consider. If something as fundamental as negation is broken, we shouldn’t be using large vision/language models in many of the ways we are using them now — without intensive evaluation,” says senior author Marzyeh Ghassemi, an associate professor in the Department of Electrical Engineering and Computer Science (EECS) and a member of the Institute of Medical Engineering Sciences and the Laboratory for Information and Decision Systems.

Vision-language models (VLM) are trained using huge collections of images and corresponding captions, which they learn to encode as sets of numbers, called vector representations. The models use these vectors to distinguish between different images.

A VLM utilizes two separate encoders, one for text and one for images, and the encoders learn to output similar vectors for an image and its corresponding text caption.

“The captions express what is in the images — they are a positive label. And that is actually the whole problem. No one looks at an image of a dog jumping over a fence and captions it by saying ‘a dog jumping over a fence, with no helicopters,’” Ghassemi says.

Because the image-caption datasets don’t contain examples of negation, VLMs never learn to identify it.

[ed. Great. Knowable unknowns vs. Unknowable unknowns.]