A few days ago I wrote a post about why Democrats can’t build a welfare state by taxing only billionaires:

I wrote:

Once upon a time, class politics pitted the middle class and poor against the upper classes; now, American politics may reflect a status conflict between millionaires and billionaires. If Democrats have become the party of the millionaires-against-billionaires, that would explain why their tax policies are focused on soaking the ultra-rich while easing the burden of the merely-rich.

As if to emphasize this point, just a couple of days later, Alexandria Ocasio-Cortez declared that “There’s a certain level of wealth that’s unearned…You can’t earn a billion dollars.”

This immediately raises the question: What amount of wealth does AOC think you can “earn”? A hundred million dollars? Ten million? Presumably there’s some number of millions that she thinks can be earned. That definitely fits the “party of millionaires-against-billionaires” framing from my post.

But the more important question is: Is AOC right? Can a billion dollars be “earned”?

It depends on what “earned” means, of course. To most people, the word probably means something very vague — basically, “I think you deserve this amount of money.” You can come up with more specific definitions if you want. For example, if you’re a socialist, you might define only labor income as “earned” and capital income as “unearned”. If you’re a free-marketer, you might define “earned” income as your marginal product — i.e., the amount by which society would be poorer if you had never been born. And so on.

But I’m not sure how useful that sort of exercise is. The socialist idea that capital income is unearned is just a moral judgement, so it leads to endless emotional debates over whether taking risk, making capital more available, etc. are things people ought to get paid for. The free-market concept is more interesting, because it’s objective, but it’s pretty unknowable — unless you’re in the movie It’s a Wonderful Life, you can’t really run the natural experiment of removing someone from the timeline.1 On top of that, most people simply won’t accept such simple, restrictive definitions of “earned” and “unearned”. So these arguments just never resolve.

But when AOC says “unearned”, she seems to mean something else:

You can’t earn a billion dollars. You just can’t earn that. You can get market power. You can break rules. You can do all sorts of things. You can abuse labor laws. You can pay people less than what they’re worth. But you can’t earn that, right?

AOC seems to mean that in order for someone to get a billion dollars, they have to do something that society ought to forbid. In other words, billionaires can’t get their wealth just by being lucky; they have to get it by being bad.

The obvious rebuttal here is to invoke Taylor Swift. The singer’s net worth is estimated at $2 billion. She got those billions from her share of ticket sales, merchandising, and music sales; unlike many artists, Swift owns her entire music catalog.

Formally, AOC is right in this case — Swift did become a billionaire with market power. Intellectual property — the ability to own your own music catalog and charge people to download your songs — is a form of government-granted monopoly. But would AOC really claim that every writer, every photographer, every artist isn’t earning their income? I doubt it. Meanwhile, Swift didn’t obviously break any rules, abuse labor laws, pay anyone less than they’re worth, etc.

But OK, Taylor Swift is the exception here. Most billionaires are more traditional types of businesspeople, who don’t obviously have celebrity superstar appeal or sell their personal artistic output. How should we think about the typical billionaire? Is AOC right that they only amass vast fortunes by either breaking the law and/or hurting the economy?

If so, it means that the vast majority of the U.S. economy — along with both the wealth and the jobs that economy has generated for the middle class — is built on illegality and unfairness. That’s a breathtaking indictment of the entire capitalist system, and it goes way too far. We do need to think about how much to tax the super-rich, but that discussion should absolutely not start from the assumption that all great fortunes were ill-gotten.

Almost every big company in America was founded by a billionaire

Read more

I sat down today to write a post about how Barack Obama was a good President, and then I remembered that I already wrote it, back in 2022:

What’s funny is that back in 2022, I was aiming my defense of Obama at progressive critics, but today I was going to write in response to his conservative critics. And yet the post I was planning to write today is very similar to the one I wrote before.

The commentariat has a very interesting relationship with the 44th President. Obama is still incredibly popular — far more popular than Trump, Biden, Bush, or Clinton:

And yet among hyper-engaged politics enthusiasts, almost everyone bashes Obama. Progressives bash him for not being the left-wing hero of their dreams, moderate liberals bash him for not being successful enough at building the foundations for enduring Democratic electoral success, and conservatives basically view him as Satan.

The latter group of critics is by far the most rabid and irrational. The political right seems to have made up a fantasy Obama out of whole cloth to blame for everything that has gone wrong in America since 2008. Obama’s administration was probably the most scrupulously clean in American history — the exact opposite of Trump’s — and yet you see right-wing people claim to this day that Obama ran America like a Chicago political machine:

When Obama criticizes wokeness, as he frequently does, you see conservatives say wild things like this:

This is nonsense. Obama has never given a speech about “whiteness” that I’m able to find. In 2019 he was criticizing cancel culture, in 2020 he was bashing “defund the police”, and in 2021 he was back to criticizing wokeness.

In fact, Matt Yglesias wrote a very good thread about how Obama was a moderate:

And yet the ultra-woke leftist Fantasy Obama lives on in the right-wing imagination.

It was seeing these nonsense criticisms that made me want to write a pro-Obama post.

It’s always good to remind people of the facts. But ultimately the rebuttal to the right-wing anti-Obama revisionism should be the same as the rebuttal to the left-wing version: Obama was a good President who did lots of good policies. That’s why the bulk of the American populace remembers Obama fondly. And that’s why commentators of all stripes should discard their fashionable anti-Obama hipsterism and acknowledge the strengths — and the actual weaknesses — of our country’s last truly popular leader.

So anyway, here’s that post from 2022.

Among conservatives, it’s an article of faith that Barack Obama was a terrible President. But who cares — of course they’re going to say that. What’s more interesting is many progressives — not just leftists, but also mainstream liberals — also regard Obama’s presidency as a failure.

To me, this is a case study in how expectations get over-inflated. In 2008, when I was a grad student attending Obama rallies, the atmosphere was electric. Stadiums were packed. Everyone had a T-shirt and a sign. In the lines outside, everyone was talking about how Obama Was Going To Change Everything.

I was pretty enthusiastic about Obama — I had the T-shirt and the sign too — but I remember thinking at the time that a lot of these people were bound to be disappointed. The fact that Obama was the first explicitly progressive President since at least Carter (and really since LBJ) didn’t mean that our economy was going to be transformed. And the fact that Obama was Black didn’t mean that racism was over in America. But I indulged the effusiveness, because I thought hope was always a good thing to have.

Now I’m wondering whether the inflated expectations of 2008 helped contribute to an overly pessimistic appraisal of Obama’s legacy more than a decade later. No, our economy was not fundamentally transformed, nor racial equality achieved. But as President, Obama really did produce an unusual string of accomplishments. He may not have justified the “hope”, but he really did bring some “change”.

The ARRA and the recovery from the Great Recession

Obama was dealt a very difficult hand coming into the presidency, for two reasons. First, we were in the middle of a financial crisis, and heading into the start of the biggest economic downturn since the Great Depression. Secondly, we were in the era of the unrestrained filibuster, which makes legislation much harder to pass than in FDR’s day even with a congressional majority.

But nevertheless, Obama came into office determined to do his best FDR impression. To be fair, George W. Bush and the Fed had already cooperated to halt the financial crisis with a series of bank bailouts and emergency lending programs. But Bush had been hesitant to go for big fiscal stimulus. Obama was not. As a percentage of GDP, the fiscal stimulus plan he passed through a reluctant Congress in 2009 was bigger than anything other rich countries were doling out:

Compared to the entire New Deal, the spending was not as large. But in terms of how much money it borrowed, Obama’s stimulus went beyond the New Deal:

How effective was this spending? Economic estimates of the effect of fiscal programs are always hard to gauge, since they depend on assumptions. But most researchers who looked into the matter concluded that the American Recovery and Reinvestment Act saved millions of jobs, with the infrastructure construction and green investment portions of the bill being particularly effective.

It’s certainly undeniable that the Great Recession ended up being much less painful than the Great Depression, despite being precipitated by financial shocks of approximately equal severity. Unemployment reached 25% in 1933, while unemployment and underemployment combined hit only 17% in 2009-10. And it took us only 6 or 7 years to recover from the drop in per capita GDP inflicted by the 2008 crash, while it took 11 years to recover from the Great Depression.

Of course, a lot of the credit also goes to the Federal Reserve here. But Obama’s bold fiscal action was part of the reason we got a lost half-decade instead of a lost decade. By 2014, the engine of American growth was humming again — and unlike in previous expansions, this time more of the fruits of that growth were going to the people at the bottom of the income distribution.

The ARRA also left behind positive long-term economic legacies that outlasted its recession-fighting effects. The spending fixed a lot of our creaking infrastructure. And its support for the solar and wind industries helped make those technologies cheaper, pushing them down the learning curve and paving the way for the cheap green energy revolution of the 2020s.

There have been three big criticisms of Obama’s recession recovery efforts. First, people allege that the stimulus was too small. Second, many complain that Obama failed to help homeowners enough, allowing massive middle-class wealth destruction. And some believe that Obama wasn’t tough enough on the culprits of the 2008 financial crisis, letting too many bank execs and managers stay in their jobs even after their institutions were bailed out.

I generally agree with these criticisms. Obama could have done better (at least, with a willing Congress). But the same is true of LBJ, FDR, or any successful progressive President in our history. The fact is, Obama’s stimulus had a big positive effect, it was significantly bigger than equivalent efforts in Europe, and it was bigger than anything George W. Bush or John McCain or Hillary Clinton would have done.

Obamacare

But Obama didn’t stop with recession-fighting; like FDR before him, he resolved to use a moment of crisis to make long-term progressive transformations to the way the U.S. economy worked. And one of the biggest problems with our economy was our health care system, which by 2009 was clearly failing us.

Obamacare was meant to be a compromise between national health insurance and the quasi-privatized patchwork mess of America’s existing system. It took its inspiration loosely from the so-called Bismarck Model of health care, where health care is universal but can be provided through either public or private insurers, and more directly from Mitt Romney’s health insurance reform when he was governor of Massachusetts. The main goal of Obamacare was to reduce the number of Americans without health insurance, and it succeeded in this goal:

The reform was not incredibly popular when it was first enacted, but gained popularity in the years after it went into effect:

Now, Obamacare is not a smashing success. It largely failed to restrain the upward trajectory of health care costs; in my opinion, high costs are our system’s biggest problem because they make it politically and economically difficult to increase spending or broaden coverage. A public option, which was dropped from the bill, would have given the government expanded leverage to negotiate down our anomalously high prices. And the Obamacare system did leave 10-11% of Americans uninsured.

But Obamacare is still a landmark achievement. It’s the most significant and sweeping health care reform since Medicaid in 1965. And with the complete failure of Bernie Sanders’ push for nationalized health care, Obamacare is also the most significant and sweeping health care reform we’re likely to see in the current political era.

And despite claims that Obama preemptively compromised away his leverage in a doomed effort at bipartisanship, Obamacare’s passage was a very close-run thing; the recent failure of the Build Back Better bill, and its replacement with the more targeted Inflation Reduction Act, should demonstrate that the ideological diversity in the Democratic party makes truly bold progressive legislation very difficult. FDR’s experience with the cancellation of his “Third New Deal” programs by Southern Democrats is another parallel here.

Dodd-Frank

One of these was the Dodd-Frank financial regulation bill. After the crisis of 2008 it was clear that finance needed to be reined in once again. Dodd-Frank, enacted in 2010, was a sweeping bill that transformed financial regulation in the United States. It created new government agencies — the Financial Stability and Oversight Council, the Orderly Liquidity Authority, the Office of Financial Research, and the Consumer Financial Protection Bureau. It endowed the Fed and the FDIC with new regulatory powers. And it created the Volcker Rule, which bans many kinds of proprietary trading by systemically important banks.

All of these measures were aimed at curbing the excesses of the pre-2008 financial system, and making sure that a similar crisis doesn’t happen again. Normally, it’s hard to evaluate the success of such restrictions, because crises that don’t happen are the proverbial “dog that didn’t bark” — if you wash your hands every day and don’t get sick, should you keep washing your hands, or stop? Etc. etc. The financial sector definitely seems to have calmed down and become less excessive since 2008, but this could also be due to the chastening effects of the crisis itself.

But in the case of Dodd-Frank, we can say a little bit more, because only a decade after the act’s passage we got the Covid shock. Yes, emergency lending programs kept the economy afloat, but there was no giant wave of defaults on bank loans even after the emergency programs ended. There was no overhang of toxic assets on bank balance sheets, whose uncertain value kept banks from lending and kept counterparties from knowing whether banks were solvent.

Meanwhile, banks are lending and business is booming. There was great fear that Dodd-Frank would lead to a decline in business formation, which had already been anemic for years. But new business formation started actually trending up after Dodd-Frank came into effect. And it spiked in the pandemic and has remained high since then:

Meanwhile, mortgage lending is robust and there has been another homeownership boom, but this time to borrowers with better credit than in the 2000s.

So the banking sector seems to be more robust, and it seems to be doing its job. I’d call that a win for Dodd-Frank and for Obama — and one that very few people talk about these days. Just like in the Depression, reining in an out-of-control finance sector seems to have had long-lasting salutary effects.

After the Tea Party: the Clean Power Plan and DACA

No President can do very much without the cooperation of Congress. FDR was stymied by a conservative Congress in the late 1930s, while Reagan was frustrated by Congressional Democrats. In 2010 the Tea Party Congress roared into power and made further big legislation impossible during Obama’s final 6 years in power. Obama was forced to fall back on executive-branch regulatory authority to make further policy changes, and this is simply much less powerful than Congressional legislation (as it should be).

But even so, Obama managed to get some important things done. There is a piece of un-passed legislation called the DREAM Act, that would shield from deportation anyone who was brought to America illegally as a child. This is an extremely popular idea, but nativists consistently manage to block the legislation in Congress. So in 2012, Obama used his regulatory authority to create the Deferred Action for Childhood Arrivals (DACA) program, basically refusing to deport anyone who would be protected by the DREAM Act if it passed. This protected hundreds of thousands of people from undeserved deportation.

In his second term, Obama also implemented the Clean Power Plan, which used regulatory authority to order states to reduce carbon emissions by whatever means they chose. The plan was canceled by Trump after just a couple of years, so it didn’t have a chance to make a big short-term impact on carbon emissions. But it probably did spur states to start taking a harder look at solar and wind power, which had come down in price enormously in the years before the plan was released. And it seems plausible that that nudge helped accelerate us toward the renewable transition that is now gathering force.

DACA and the Clean Power Plan were modest but real (and in my opinion, positive) achievements.

Domestic successes, foreign failures

On domestic policy, the combination of the ARRA, Obamacare, and Dodd-Frank represent greater policy accomplishments — and more progressive accomplishments — than any Democratic President since LBJ. They were done in 2 years, which is a lot faster than LBJ or FDR accomplished their reforms. And they were accomplished in the face of a difficult institutional environment, where the unrestrained filibuster makes it nearly impossible to pass truly bold legislation with a simple majority.

Overall, Obama effectively addressed the severe domestic policy challenges he inherited from the previous administration. He restrained the financial sector and cleaned up the damage it had done to the economy, restoring us to robust growth. And at the same time, he managed to make long-term headway on the hard problem of healthcare, while also using regulatory authority to effect minor progress on immigration and climate change.

I call that a major success on domestic policy. People who think Obama’s domestic record represents a failure are simply experiencing the letdown from their own impossibly high expectations.

On foreign policy, however, Obama’s record is more mixed. On the War on Terror, Obama was mostly successful — he killed bin Laden, extricated the U.S. from the pointless peacekeeping operation in Iraq, and drew down most of our presence in Afghanistan. He handled the emergence of ISIS effectively as well, leading to its relatively swift defeat. As a result, the War on Terror was effectively concluded, though of course terrorism as a military tactic will remain and Islamic fundamentalist regimes like the Taliban will not entirely vanish from the Earth.

On the Arab Spring and the wars that followed, Obama’s record is more mixed, but I’m not convinced there’s much more he could have done. U.S. appetite for further military adventures in the Middle East was nil. Obama gets criticized fairly equally for failing to intervene more in Syria and for intervening too much in Libya. So I don’t agree that this represents a dramatic failure for Obama, even though it was hardly a success either.

But it turns out that both the War on Terror and the Arab Spring were largely distractions from the true looming foreign policy threat — the reemergence of great-power conflict. Obama’s weak response to Russia’s seizure of Ukrainian territory ultimately ended up encouraging Putin’s further adventurism and leading to the current catastrophic war. In Asia, Obama refused to acknowledge the importance of Xi Jinping’s accession to power and the country’s concomitant aggressive, nationalistic turn. He remained overly enamored with the failed Clintonian idea that engagement would make China more progressive, and his “pivot to Asia” was too little, too late. Obama might possibly have used the exigency of the Great Recession to revive U.S. industrial policy and start competing effectively with China in high-tech manufacturing, but — apart from a few minor, halting efforts — he didn’t even really try.

He was so occupied with fighting the problems of the present that he wasn’t able to concentrate on the problems of the future. And so now we find ourselves racing to catch up.

But as I see it, the verdict on Obama on domestic policy has to be that he made great headway on the problems he inherited from Bush — a devastated financial sector, a collapsing economy, a large number of uninsured people, and a still-scary Islamist threat. He was a crisis President, and he beat back the crisis. The bitterness and regret that many progressives now feel toward his administration is a function of their own inflated expectations going in.

Subscribe now

Share

The above image is from a recent tweet by University of Pennsylvania economist Jesús Fernández-Villaverde (henceforth referred to as “JFV”), in which he criticizes the field of development economics for ignoring the big questions. He writes:

A fundamental lesson from my posts these last two weeks on modernization, industrial policy, and development is that development economics should be about understanding why South Korea got rich but Bolivia did not.

The current field has largely given up on that question. Sharply identified RCTs on small micro programs are a fine way to publish in the AER and get tenure at a fancy university, but a profession that knows everything about microfinance impact evaluations and almost nothing about industrialization has misallocated its own intellectual capital on a pretty heroic scale.

JFV’s critique of modern development econ isn’t new. Eminent economists have been complaining about randomized controlled trials for years. I wrote about this back in 2020:

In 2019, Lant Pritchett — then at Oxford — made an argument very similar to JFV’s when he criticized that year’s Nobel winners:

People keep saying that the recent Nobelists “studied global poverty.” This is exactly wrong. They made a commitment to [the RCT] method, not a subject, and their commitment to [that] method prevented them from studying global poverty…

[P]overty programs…account for less than 1 percent of total variation in poverty…[C]hanges in [actual] poverty…are overwhelming associated with growth in median income/consumption…[V]ariation in the size and efficacy of poverty programs had little or nothing to do with poverty reduction…So what the Nobelists really did was…using a particular method to study whatever could be studied with that method in poor countries (lots of which were “interventions” by NGOs at very small scale), knowing that this…severely limited their ability to study…global poverty.

In other words, what really beats poverty is economic growth, and you can’t do an RCT to study economic growth. This is basically a slightly broader version of what JFV says. JFV says that industrialization is the key, and industrialization is how almost every rich country got rich.

So basically, the idea here is that if you’re a development economist, and you’re not asking “How can countries industrialize?”, you’re being kind of useless.

The counterargument here — which I made in 2020 and which a number of economists made in response to JFV’s post — is that it’s better to study something knowable than to study something unknowable, even if the knowable things are less important.

For example, there are plenty of doctors working on finding slightly better treatments for acne. If we could solve the mystery of aging, and make it so that humans live healthy lives for 100 years, it would do a LOT more for human quality of life than treating acne would. But solving the mystery of aging is very hard, while finding slightly better acne treatments is very doable. So it doesn’t make a lot of sense to yell at doctors to all stop studying acne and only work on aging.

The field of economics doesn’t lack for big ideas about why countries go from poverty to riches. These include:

1. Institutions: The idea that property rights, legal frameworks, and other systems of human organization are long-lasting (“sticky”) and are crucial for development

2. Geography: The idea that countries’ natural endowments — navigable waterways, farmland, proximity to other regions, etc. — determine which place gets rich

3. Human capital: The idea that skills — reading, math, etc. — and population health determine national income

4. Industrialism: Various theories about how promotion of manufacturing, export-led growth, the “development state”, industrial policy, and so on are the key to rapid development

5. Culture: The idea that countries grow because of a culture of progress, innovation, and openness to technology

6. Coordination failure: The theory that countries naturally grow rich as long as they don’t have any significant roadblocks to growth, so that development happens when you remove all of the roadblocks at once

7. Flying geese theory: The idea that growth naturally happens in a sequential pattern as some countries luckily get rich first and then invest in poor countries until those countries catch up

8. Economic liberalism: The notion that all you really need to grow is free markets and openness to trade

9. State capacity: The theory that strong, efficient states are crucial for growth

10. National cohesion: The idea that a populace who see themselves as one unified people will support the public goods and other policies necessary for growth

That’s just a small sample of the huge diversity of big ideas out there. Each one of these ideas has received enormous attention and publicity, both inside of the economics profession and in the general public. You can read Why Nations Fail about institutions, Guns, Germs, and Steel about geography, How Asia Works on industrialism, A Culture of Growth on culture, and so on. There are plenty of high-profile academic papers laying out variants of each one of these theories, and plenty more that attempt to find evidence for or against them.

So why do all these theories still exist? And why do all of them still have prominent adherents and advocates, both in the economics profession and out in the world? Is it just because we haven’t allocated top talent to the job of generating and testing these theories? That seems unlikely — Daron Acemoglu worked on institutions, Joel Mokyr worked on culture, Arthur Lewis and Dani Rodrik worked on industrialism, Gary Becker and Robert Lucas both worked on human capital and growth, Paul Krugman worked on economic geography (which uses geography to generate “flying geese” effects), Alberto Alesina worked on national cohesion, Milton Friedman worked on economic liberalism, Chad Jones worked on coordination failures and so on.

Most of those researchers have Nobel prizes, and all of them are very highly respected in the field. Nor are they even close to being the only high-profile, respected economists who have worked on each of those ideas. There are probably a few of the theories — state capacity in particular, but also industrialism — that could use some more attention from top researchers, in part because they cut against the economic liberalism that dominated the culture of academic economics in the late 20th century. But overall, there are very few neglected big ideas on the list.

Perhaps the problem is that we have too few economists working on testing these theories? In general, every empirical economics program needs more than just a few big names — it needs a ton of lower-level researchers hunting down data, constructing good data sets, finding natural experiments, and so on. Each of the ten big ideas I listed above has a very active research program associated with it. Here are a few example papers from the last decade:

1. Institutions: “Institutions and economic development: new measurements and evidence”, by Acquah et al. (2023)

2. Geography: “The Global Distribution of Economic Activity: Nature, History, and the Role of Trade”, by Henderson et al. (2017)

3. Human capital: “Global universal basic skills: Current deficits and implications for world development”, by Gust et al. (2024)

4. Industrialism: “Manufacturing Revolutions: Industrial Policy and Industrialization in South Korea”, by Lane (2025) (Here are some more fun examples via Oliver Kim, just because I personally like industrialism)

5. Culture: “Culture, Institutions, and the Wealth of Nations”, by Gorodnichenko and Roland (2017)

6. Coordination failure: “Big Push in Distorted Economies”, by Buera et al. (2020)

7. Flying geese: “Have Robots Grounded the Flying Geese? Evidence from Greenfield FDI in Manufacturing”, by Driemeier and Nayyar (2019)

8. State capacity: “State Capacity and Growth Regimes”, by Imam and Temple (2025)

9. Economic liberalism: “Does economic globalisation promote economic growth? A meta‐analysis”, by Heimberger (2022)

10. National cohesion: “National identity, public goods, and modern economic development”, by Skaperdas and Testa (2025)

There are many, many more examples in each of these categories (and for the other theories of development that I didn’t list). Many are by researchers at good schools, publishing in respected journals.

In other words, there is tons of research effort dedicated to generating and testing sweeping theories of economic development. I’m not sure what “field” JFV is referring to when he says “The current field has largely given up” on the question of comparative development. If he means the field of economics as a whole, he’s just obviously, utterly wrong.

If he means the field of development economics specifically, it’s more arguable — a lot of the examples I listed above come from economists who aren’t known specifically as “development” economists, and most of the papers aren’t in development-econ field journals. Perhaps JFV believes that if development economists weren’t so busy doing RCTs, they would be throwing their time, effort, and intellectual heft into the grand quest to determine which big theories of development are right.

Even there, however, he’s on shaky ground. Jessica Leight found in 2022 that only about 19% of development econ papers include RCTs. In 2015, David McKenzie found that for development field journals, the percentage was 13%, though it was 31% for development econ papers published in top 5 journals. These are not insignificant numbers, but they’re not huge either.1 If all of the economists doing RCTs were to switch to doing work on the Big Questions, the increase in effort on those questions would be pretty marginal.

So I’m not sure what JFV is talking about here. He’s an economist I like and respect, but his perception of the state of research on the big questions of development doesn’t seem very accurate.

Which brings us to the question: Why haven’t we been able to tell which of the Big Ideas are right and which are wrong? The obvious answer here is that it’s just very hard to prove or disprove any of these theories.

Why did South Korea grow so much more than Bolivia from the 1960s through the 2010s? The divergence is certainly startling:

But this is an event that only happened once. There were a lot of differences between South Korea and Bolivia during this time, and it’s hard to know which ones were decisive. South Korea was much more highly educated than Bolivia in the 60s, despite its poverty. While Bolivia focused on selling its natural resources for as high a price as possible, South Korea focused on exporting manufactured goods and climbing up the value chain. Korea had a special relationship with the U.S. that provided it with a large, friendly, reliable market for its manufactured products, as well as government procurement contracts, aid, and technological assistance. Korea is ethnically homogeneous and has many centuries of history as a country with its own language; Bolivia is an ethnically diverse post-colonial state. Korea had a strong, professionalized bureaucracy; Bolivia, not so much. Korea has plenty of sea access; Bolivia is landlocked. Korea got to take advantage of Japanese know-how when its companies paid retired Japanese engineers to come teach their own workers; Bolivia had no such advantage. South Korea had to develop in order to ward off the military threat from North Korea; Bolivia had no such pressing imperative.

And so on. Depending on which Big Theory you believe, you could attribute Korea’s relative success to any combination of these natural advantages and policy choices. You could also tell a composite story — for example, in my own assessment of Poland’s economic miracle, I attributed the country’s breakout success to a combination of geography (proximity to the EU), institutions (changes made in order to be admitted to the EU), industrialism (promotion of manufactured exports and FDI), and flying geese (investment from Germany). I could have also mentioned high human capital, ethnolinguistic homogeneity, and the military threat from Russia. This makes for a good story — and you can call Poland’s success a “model” and try to emulate as much of it as you can — but it’s not a scientific explanation.

So what if you get a bunch of development success stories like Korea, and a bunch of failures like Bolivia, and you try to systematically figure out the most important factors? This is the idea of a cross-country regression, and it’s a common tool that development economists use, but it’s fraught with issues.

There aren’t that many development success stories, and a lot of them look very different from each other — you can group Korea with Qatar as “rich countries”, but that grouping will obscure more than it reveals. There’s tons of endogeneity present — you might observe that countries with efficient bureaucracies tend to grow faster, but that doesn’t mean the former caused the latter, because both might be caused by differences in the education system. Different time periods might yield different lessons. You might leave out some really important variables entirely. A large country might not be comparable to a small country. And so on.

Basically, lots of development economists run cross-country regressions, and they always lead to vigorous arguments about what the regressions mean and whether the models were appropriate. If you want to read such an argument, a great example is Rodrik (2008), which uses a cross-country regression to claim that countries grow faster when they keep their currencies undervalued. Michael Woodford offers a lengthy commentary in which he raises doubts regarding Rodrik’s statistical choices and the interpretation of his results. You can choose to believe that Rodrik is right — and some people do! — but it requires tons of assumptions.

So what else can you do? Another tool that lots of development economists use is a structural model — basically, a development theory whose parameters you estimate from data. But this approach has even more problems. First of all, structural models are a bit like toothbrushes — everyone has one, but nobody wants to use anyone else’s. There are a vast number of structural models, and none of them ever get rejected because they don’t fit the data — economists just find the parameters that best fit the model and call it a day, without ever questioning whether the model is just wrong. But because there are so many models, they can’t all be right — in fact, only a few of them can. As a result, making a structural model almost never helps tell you what’s really going on in the world — it’s just a way of extrapolating the implications of your assumptions.

Another thing you can do is narrative history — basically, looking at a historical episode of development and recording all the interesting details, so that hopefully someone can figure out which of those details mattered. People responded to JFV’s tweet with a number of examples of this, such as Douglas Irwin’s 2021 paper about the South Korean economic miracle, and Piatkowski and Zhang’s 2022 paper about shock therapy in China. Books like How Asia Works, Asia’s Next Giant, MITI and the Japanese Miracle, and Governing the Market are classic examples of narrative history with regards to various East Asian success stories, and I think they’re all excellent.

This is useful work, and you can make a good argument that development economists should be doing a lot more of this. But on its own, recording the facts isn’t enough to tell you which facts mattered, or which will matter in other countries.

The final thing you can do is to use microeconomic empirical work to assess the effects of policies. A good example of this is Nathan Lane’s 2025 paper on South Korea’s Heavy and Chemical Industry Drive, which looks at how a famous Korean industrial policy affected specific industries. Another example is Barteska et al. (2025), which measures the impact of U.S. defense procurement on specific Korean companies. A third example is Kim and Wang (2025), which studies Taiwanese land reform.

This is very useful work, but it also has some obvious limitations. It only studies policy; it has little to say about the importance of natural advantages like geography or human capital. It’s also hard to translate from a policy’s effect on specific companies or industries to its effect on the economy as a whole.

So while there’s much useful development economics to be done, even the brightest minds in the field are playing with a set of inherently weak tools. History only happens once, so our ability to make a science out of one-time historical events like economic development is very limited.

That’s why I think scoffing at RCTs and urging development economists to tackle the Big Questions more often will have a negligible effect. There’s really not a lot more that can be done, in terms of generating Big Theories of Why Countries Get Rich, or in terms of testing those theories. Unless and until AI gets smart enough to study human society from a bird’s-eye view, I think we should be humble about how much we expect development economists to be able to contribute to developing countries’ growth policies in real time.

Humility, I think, should be the key word here. Development economists can do lots of useful things for policymakers — they can explain various theories, cite a bunch of details about successful countries like Poland and Korea, point out failures, and draw on various suggestive empirical results. But there is no science of development, and it’s not clear there ever will be. So we should be careful that exhortations for development economists to focus on the Big Questions don’t pressure them to pretend that they have the Big Answers.

Subscribe now

Share

Something big happened in the world of AI the other day: Sam Altman, founder and CEO of OpenAI, and probably the person who’s most commonly regarded as the face of the industry, declared that the purpose of AI is not to take people’s jobs:

And he recently called AI CEOs “tone-deaf” for declaring that AI is going to take people’s jobs:

In fact, this shift represents more evolution than revolution. Years ago, Altman did seem to generally agree with the folk consensus that AI’s purpose is to make most or all humans obsolete; in 2014 he warned that we could be faced with “a new idle class”, and explored the idea of Universal Basic Income as a remedy. In 2021 he wrote that “The price of many kinds of labor…will fall toward zero.”

But in recent years, Altman has consistently stated that although AI will destroy many occupations, it will create new tasks for humans to do. In 2024 he wrote that “I have no fear that we’ll run out of things to do (even if they don’t look like “real jobs” to us today)”, and in 2025 he declared that “We will find new things to do, new ways to be useful to each other, and new ways to compete, but they may not look very much like the jobs of today.” He has reiterated that prediction in interviews.

OpenAI’s mission statement, meanwhile, continues to define the company’s goal as the creation of Artificial General Intelligence (AGI), which it defines as “highly autonomous systems that outperform humans at most economically valuable work”. That “most” does leave some wiggle room. But perhaps more importantly, the company is talking about AGI less and less — its 2026 statement of principles mentions the term only twice, as compared with 12 times in the 2018 version. OpenAI also removed a clause about AGI in its agreement with Microsoft, meaning that the term no longer defines its contractual business obligations.

So although Altman has never been quite as doomer-ish as some of his colleagues when it comes to AI and jobs, you can definitely feel the winds shifting. In fact, there has always been a contingent of tech leaders who have been broadly optimistic about AI and jobs, and who are now speaking up more vociferously. Nvidia’s Jensen Huang has consistently predicted that AI will create more jobs than it destroys, but recently he has harshly criticized AI CEOs who go around saying that their technology is a job-killer:

Venture capital titan Marc Andreessen, meanwhile, has come out swinging against the AI job loss narrative:

Cynical observers will see this all as just a messaging pivot, in response to the AI industry’s deteriorating popularity. Back in March I wrote about how the AI industry’s sales pitch was basically “Our product’s purpose is to put you and your descendants on welfare forever, and it may also wipe out your whole species”:

That was a bad sales pitch, to put it mildly, and it’s not surprising that voters have reacted negatively to this message. Basically every recent poll shows the American public turning very strongly against AI. Here’s a representative example from Pew:

In fact, the anti-AI turn seems especially strong among Independents:

This raises the possibility that AI will become the focus of populist rage, and that politicians from both parties will compete to win swing voters over by promising to take action against the industry.

This may already be happening. Bernie Sanders has moved past traditional progressive concerns about data center water use and copyright infringement, and has instead been warning about catastrophic AI risk:

Meanwhile, Donald Trump is reportedly considering a policy of having the White House vet AI models before they’re released, due to concerns about new models’ cyber capabilities:

President Trump, who promoted a hands-off approach to artificial intelligence and gave Silicon Valley free rein to roll out the technology, is considering the introduction of government oversight over new A.I. models, according to U.S. officials and people briefed on the deliberations…The administration is discussing an executive order to create an A.I. working group that would bring together tech executives and government officials to examine potential oversight procedures…Among the potential plans is a formal government review process for new A.I. models…The discussions signal a stark reversal in the Trump administration’s approach to A.I…[Trump’s] noninterventionist policy began changing last month after the start-up Anthropic announced a new A.I. model called Mythos. Mythos is so powerful at identifying security vulnerabilities in software that it could lead to a cybersecurity “reckoning,” said Anthropic[.] [emphasis mine]

Neither Bernie’s concern nor Trump’s is explicitly about protecting jobs; both are about the risk of misuse. But it’s hard not to see the generally souring mood on AI, especially among Independents, as an invitation to populists like Trump and Bernie to make political hay by reining in the industry.

Meanwhile, some politicians and industry figures are starting to talk openly about the possibility of nationalizing the big AI labs. Matteo Wong and Lila Shroff report:

Washington is getting antsy about the power imbalance [between AI companies and the government]. Over the past year, multiple senators have proposed legislation that would order federal agencies to explore “potential nationalization” of AI…In recent weeks, Elon Musk, OpenAI’s CEO Sam Altman, and Palantir’s CEO Alex Karp have publicly spoken about the possibility of nationalization…

The government could regulate AI companies like it does utilities…[S]hould AI models displace large swaths of the labor market, such that a handful of companies run most of the economy, “then some kind of nationalization becomes potentially imperative,” Samuel Hammond [of FAI] told us—to distribute wealth and simply ensure the proper functioning of society. Both Anthropic and OpenAI have already suggested possible versions of such redistributive measures…

Perhaps the most likely fate for American AI companies is a future of soft nationalization—a world in which the government doesn’t fully control AI labs and their models, but instead enacts an escalating series of policies and establishe[s] close partnerships with private companies to shape the technology.

Different figures in the industry want quasi-nationalization to different degrees. Jensen Huang, who has fought hard against export controls, is probably more anti-nationalization, as is Marc Andreessen, who makes his living from funding startups (and would thus probably not like to see government ties entrench the market position of incumbent players). But even folks like Altman and Amodei who might be inclined to accept quasi-nationalization would certainly like to negotiate favorable terms for that partnership. To that end, it helps to have the government not view your industry as a dangerous job-killer.

So basically, it makes sense for leading figures in the industry to alter the basic sales pitch and reassure anxious humans that they’ll still have jobs.

In Altman’s case, there also might be some element of competitive positioning here. The loudest voice predicting human obsolescence has certainly been Anthropic founder and CEO Dario Amodei, who has been shouting from the rooftops about a coming job-pocalypse:

To a seasoned observer, Anthropic’s perspective here is pretty clear. They basically think AI progress is inevitable, and that AGI is eventually going to put most human beings on the welfare rolls. Thus, they see themselves as sounding the alarm — warning society to beef up its welfare state and its redistributionary mechanisms before the inevitable coming of job-annihilating AGI.

If you accept that AI progress is as inevitable as the tides, then this is an eminently reasonable position. But most people probably do not accept this. They probably see AI progress as something that we — human society — choose to do or not to do. And so to them, Dario isn’t sounding a warning — he’s making a threat.1 The average person probably hears Dario as saying something along the lines of “Hi, my colleagues and I are working very hard to make sure you are never gainfully employed again.” And that probably makes them feel fairly negatively toward Anthropic.

It’s possible that Altman and OpenAI see an opening here. Anthropic has recently overtaken OpenAI in revenue and market valuation.2 If OpenAI presents themselves to the nation as the guys who are trying to create AI that augments your job, then maybe they can sell themselves as the human-friendly alternative to those scurrilous folks over at Anthropic who just want to replace you. This is one theory I’ve seen thrown around, in any case:

But OK, saying “AI will increase the value of human labor” is one thing; providing a compelling explanation for this assertion is another. The notion that AI is fundamentally a human-remover is deeply ingrained into our national discourse — we’ve heard it so many times that it’s become not just the conventional wisdom, but an article of faith for many. It’ll be an uphill battle for pro-AI voices to dislodge and replace that notion.

So what arguments are they using?

One is the idea of task creation. So far, most technologies throughout history have created new kinds of work for humans to do. Some AI proponents assert that AI will be the same.

A second is the idea of induced demand, either from income effects (AI makes us richer so we buy more stuff) or from complementarities. This often goes by the name of Jevons’ Paradox.

Here’s Aaron Levie, CEO of Box, employing both ideas:

There are far more categories where AI agents making things more efficient will induce demand for that skill than spaces where agents eliminate the work. This is why the AI jobs predictions will not play out as advertised.

AI making it easy to produce more code will mean we start to apply code to far more parts of our businesses. We will build automation and software for things that wouldn’t have made sense before. Marketing automation, client onboarding, modernizing old systems, doing far more research on existing data, and more…Far more software will mean vastly more security risks. This will mean far more people thinking through system security, compliance, and governance…AI will make it so more companies care about this (and maybe can do something about it), causing more security roles…Companies will now be doing 10X more with video and graphics, and will need people to manage that work. More media. We’re going to have a near unlimited set of legal challenges in a world of AI as AI helps write even more bespoke and complicated legal docs. More lawyers.

This is probably correct — at least for now. Technologies have always destroyed some occupations, but they’ve usually created more demand for human labor than they replaced. At least for a while, it seems clear that AI will behave similarly.

But a lot of people have the intuitive sense that this solution works until it doesn’t. If AI becomes better than humans at all tasks, then humans’ only remaining value would come from comparative advantage — and as data centers proliferate and compete with humans for land and food and energy, the economic value of comparative advantage goes down and down.

So the pro-AI people naturally need to give the public some reassurance that even after the coming of AGI, humans will still be valued. The answer that more people are converging on is that humans will still pay for the human touch. Alex Imas has a good post about this:

Ghosts of Electricity

What will be scarce?

Starbucks is a huge company (market cap of $112 billion) that sells one of the most standardized products in the modern economy. Making a cup of coffee or even one of the fancy specialty drinks is very easy to mechanize and reproduce. If the entire economy is soon to be automated, with labor being replaced with increasingly more sophisticated capital, S…

Read more

25 days ago · 1363 likes · 135 comments · Alex Imas

Imas writes:

If the model is right, the durable jobs of the future won’t be about monitoring AI systems or prompt engineering. Those are transitional roles in the automated sector. The durable jobs will be in the relational sector, where the human element is the product itself.

Some already exist and are growing: nurses, therapists, teachers, boutique fitness instructors, personal chefs, bespoke tailors, craft brewers, live performers, spiritual guides, childcare workers, and many varieties of hospitality and care work. Others are emerging: experience designers, human-AI collaboration artists, provenance certifiers, community curators. Many haven’t been invented yet, just as six out of ten jobs people hold today didn’t exist in 1940.

Ezra Klein recently wrote an article in the New York Times endorsing Imas’ idea.3

So this is shaping up to be the new AI sales pitch. In the short term, AI will give people more work to do, and in the long term we’ll still get paid just to be human to each other. And our real wages will go up and up, because of the abundance AI creates.

From a public relations perspective, this pitch is WORLDS better than the previous one. Shouting about replacing humanity might play well with corporate customers and investors salivating over the dream of eliminating labor costs, but eventually you get the rakes and pitchforks, followed by some form of nationalization. Describing AI as a normal technology — a successor to the steam engine and the automobile and the computer — is much smarter politics.

The question is: Is it just politics and PR? Certainly there are plenty of AI researchers and entrepreneurs who will keep quietly believing that AGI is going to make humans obsolete; they’ve heard (and repeated) this line for too many years to suddenly believe something else overnight.

But as they continue to repeat the line that “AI will augment humans” for the sake of their industry’s public image, I think there’s a chance that they’ll start to believe it — or at least to think about how they might be able to make it true.

Daron Acemoglu has written that society should try to steer AI development toward technologies that complement humans rather than replacing them. I just don’t think that’s feasible — society simply can’t mandate the economic value of a technology before it exists.

But I do think it might be possible for AI researchers to concentrate their efforts on AI applications that give humans superpowers, rather than on trying to copy what humans already do. Once they stop thinking “This technology is a replacement for the human species”, and start thinking “This technology is a tool for humans to use”, the direction of their research programs might subtly evolve in a more labor-augmenting direction.

So yes, I’m happy with the new AI sales pitch, even if some of the people saying it don’t necessarily believe it yet.

Subscribe now

Share

If I were a conservative, I’d argue against California’s proposed one-time 5% “billionaire tax” on the grounds that it would reduce the incentive for billionaires to invest and create jobs. I’d argue that in order to guard and support the engines of our prosperity, we have to keep taxes low, etc. etc.

But I am not a conservative. I actually support taxing rich people more, and I support taxing the super-rich at higher rates. Tax progressivity is good; there’s no reason America’s income tax brackets should stop increasing after $600,000. Capital availability isn’t a problem in America (as evidenced by the data center boom), and raising taxes on the super-rich isn’t going to make Elon Musk give up his American citizenship and move to Singapore. The moral case against high taxation of top incomes is weak; the United States government and social system is what made it possible for American billionaires to make their money, and they ought to give a lot back to support that system.

But California’s proposed plan — a one-time confiscation of 5% of the total net worth of everyone who has net worth of over $1 billion — is a silly way to go about doing this, for three reasons:

1. A one-time tax is bad, because it can’t provide consistent funding for anything, and because it just creates uncertainty about future taxes.

2. A state-level tax on the ultra-rich is not a very efficient way of raising revenue, since the rich can just move out of state.

3. In general, this tax idea fits into the increasing trend toward “slopulism” in Democratic policymaking — the idea that a modern government can be funded solely on the backs of the super-rich, while the merely-rich get big tax cuts.

Let’s start with the first two considerations here.

Pragmatic arguments against the California Billionaire Tax

The first big problem with California’s proposal is that it’s a one-time levy instead of a regular tax. Taxes are predictable — you pay some percent of your income every year. That predictability isn’t just good for the taxpayer — it’s also good for government, because government can plan its budgets around regular income streams.

Taxes aren’t entirely predictable — recessions and booms change the amount of income that people pay taxes on. Capital gains taxes, which depend on whether the market goes up or down in a given year, are especially volatile. But in general, a state’s tax revenue is pretty predictable over time:

That allows the state to plan how much it’ll be able to spend on education, health, infrastructure, and so on — not just in the current year, but in five years’ time. States generally have balanced budget amendments, so they have to make sure that revenue approximately balances out expenditures.

But California’s proposed “billionaire tax” doesn’t work like this; it’s a one-time confiscation of 5% of billionaires’ total net worth. The proposal stipulates that this money would go toward health care, education, and food assistance. The problem is that since it’s a one-time levy, the money for those health care, education, and food assistance programs will eventually run out. In fact, the proposal stipulates that the money will be spent over five years.

At that point you’ll have three choices: raise tax rates, do another one-time levy, or cut the programs. This is basically a time bomb — it sets you up for another bruising political battle down the line. There’s no guarantee you’d win the battle for a second one-time levy in five years. If you don’t win, then you just created a ton of social programs that you have to yank away from people. And if you could just raise tax rates on an ongoing basis, why not just do that today instead?

This is a bit like when Congress enacts “temporary” tax cuts or a one-time rise in the debt ceiling — it just kicks the can down the road and sets the stage for chaos. But it’s worse, because states, unlike the federal government, have to balance their budgets; they can’t just borrow to cover up the holes in their funding.

If you want to tax billionaires, just raise their tax rates, like countries in Europe do — or like Massachusetts does with people making over $1M a year. Don’t mess around with this one-time levy stuff.

The other problem with California’s plan is that billionaires are likely to move out of the state to avoid the tax. 5% by itself isn’t huge, but if it were successful it would obviously set the stage for further confiscatory “one-time” levies. So billionaires have a big incentive to just move out of the state to avoid the eventual confiscation of a large portion of their wealth.

It’s actually fairly easy for billionaires to pick up and move to a different state. Normal rich people — mere millionaires — tend to be tied down by local friends and community ties, so they don’t move much in response to tax hikes. Young et al. (2016) found that millionaires move very little in response to state taxes on ultra-high-income residents. But billionaires are built different — Moretti and Wilson (2023) found that when a state implements an estate tax, about 35% of Forbes 400 billionaires move to another state.

Read more

Hi, folks! My father unfortunately passed away two weeks ago from chemotherapy complications, and as you can imagine, I’ve been busy dealing with that, so posting has been a bit light. My apologies. (I will probably write something about my father on this blog at some point in the near future.)

Anyway, there’s tons of stuff happening out there in the world — far too much to fit in a single roundup post — but here are some items I hope you’ll find interesting.

1. The U.S. national debt bomb

The U.S. passed a major milestone recently. The ratio of national debt to GDP passed 100% for the first time since World War 2:

Note that there are two measures of national debt, and they have names that sound very similar. “Total public debt” is the amount owed by the Treasury, while “Federal debt held by the public” is the amount owed by the Treasury to lenders outside the U.S. federal government itself. If another government agency holds Treasury bonds, that debt counts in “Total public debt”, but not in “Federal debt held by the public”. The one exception is the Fed — if the Fed holds Treasuries, it counts in both debt measures.

It’s very confusing, I know. They should probably change those names to make them less similar. Anyway, it’s “Federal debt held by the public” that just passed 100% of GDP for the first time since World War 2.

Now, there’s nothing particularly special about the “100% of GDP” marker — it’s just a big round number. The amount of money that the federal government has available to pay back the national debt is tax revenue, not GDP. Debt is currently at a little over 8 times annual revenue, meaning that the U.S. federal government owes about 8 years of its “income”.

But what’s really scary isn’t the debt number itself — it’s two other things. First, there’s interest payments. As a percentage of GDP, the U.S. government is paying just about as much in interest as it ever has — and a lot more than after World War 2.

That number is going to soon hit record highs, as an increasing percent of the federal debt gets rolled over at current high interest rates.

The other scary thing is that no one in the United States government seems especially interested in curbing the debt. DOGE was a complete joke, and totally failed to lower spending. Trump is going on a giant deficit spending binge. The Democrats are now promising tax cuts. The government’s official deficit projections rely on totally fantastic assumptions about interest rates.

As Paul Krugman says, we are no longer a serious country. (But the Democrats are more complicit in that unseriousness than Paul would probably like to admit.)

If we keep going in this direction, we will eventually see negative consequences. It could be inflation, or a collapse in the dollar, or a sovereign default. Any of those outcomes would be very bad for the American economy and the American people. But everyone seems more interested in winning the next election cycle and kicking the can down the road.

2. Is the danger of AI cyberattacks overrated?

The recent release of Anthropic’s Claude Mythos model sparked global alarm about the possibility of a wave of AI-powered cyberattacks. The new model was able to find a bunch of security flaws in existing software systems, prompting fears that if it got into the wrong hands, it could collapse all kinds of critical digital infrastructure. Ostensibly in order to allay these fears, Anthropic limited the initial release of Mythos to only a few people, while it worked with the government and with corporations to use Mythos to find and close as many security loopholes as possible.

Then along came OpenAI, and released GPT-5.5. While there’s no single test of hacking capabilities, the UK government’s AI Security Institute rates GPT-5.5’s capabilities as equivalent to — if not slightly better than — Mythos’.

This at first seems very worrying, because it means that a lot of different companies can build models with superhuman hacking abilities. The Chinese companies certainly aren’t far behind. Is the digital infrastructure that supports modern civilization about to be destroyed?

Well…maybe not. As some have pointed out, GPT-5.5 is being widely released, unlike Mythos. It’s been a week and we haven’t seen a giant wave of crippling online bank heists or other hacking attacks, so maybe cyber-defense is more robust than we think. There is some research to support this, actually. On one hand, hacking attacks have gotten steadily more common over the years, even before AI came along:

But the cost of a typical attack is generally very low, and lower than it was in the past. A recent paper by Tom Johansmeyer shows that cyber “catastrophes” cost much less than they used to:

So perhaps our companies and governments have just developed defenses-in-depth against the omnipresent threat of data breaches — systems of backups, redundancy, cross-checking, compartmentalization, damage control, and so on that make the cyber-world very robust to even successful hackers.

If so, perhaps the new AI models will make things even more secure. Any system has only a finite number of security vulnerabilities, so if we have new AI models that are good enough to comb over the code and fix the weak points very quickly, that should privilege the defense over the offense. Of course attackers have the same newfangled AI, but even the most powerful AI system can’t find software vulnerabilities if there are none to find. So it seems like if AI gets better at both attack and defense, defense will tend to win in the long run.

This might be a reason to sleep a little easier about the coming of superhuman AI hacking. If you’re in a space that has fewer institutional safeguards, however, you may be sleeping a little less soundly right now…

3. Phone bans in schools seem…good?

I’m generally a techno-optimist, but I think that it often takes a very long time for human society to learn to use a new technology in a beneficial, healthy, socially stabilizing way. The adjustment period, where we still haven’t reconfigured society to adapt to the new technology, can be very painful.

One technology I think has been producing a particularly painful adjustment period is the smartphone — and more specifically, smartphone-enabled social media. “Social-mobile”, as we called it a decade ago, seems to have a bunch of negative side effects — it makes our politics more bitter and contentious, it makes young people unhappy, and it probably distracts people and makes it harder for them to think and learn. “Just put down the phone” isn’t a solution, because the strong network effect keeps people trapped in the ecosystem against their will — you can delete Instagram, but that doesn’t help much if all your friends are still on it. The rapid and often negative social changes caused by social-mobile are probably one reason why many Zoomers desire to ditch modern technology and live in the past.

But past waves of technological advancements show that the right approach isn’t to ditch modernity; it’s to reshape society so that it uses new technology in more beneficial ways. It’s been only a decade and a half since the advent of social-mobile, and we’re already experimenting with cell phone bans in schools. Already, those experiments are yielding positive outcomes. Here’s Abrahamsson (2024):

I show that banning smartphones significantly decreases the health care take-up for psychological symptoms and diseases among girls. Post-ban bullying among both genders decreases. Additionally, girls’ GPA improves, and their likelihood of attending an academic high school track increases. These effects are larger for girls from low socio-economic backgrounds.

And here’s Sungu et al. (2025):

In a randomized controlled trial involving nearly 17,000 students, we find that mandatory in-class phone collection led to higher grades --- particularly among lower-performing, first-year, and non-STEM students --- with an average increase of 0.086 standard deviations. Importantly, students exposed to the ban were substantially more supportive of phone-use restrictions, perceiving greater benefits from these policies and displaying reduced preferences for unrestricted access. This enhanced student receptivity to restrictive digital policies may create a self-reinforcing cycle, where positive firsthand experiences strengthen support for continued implementation. Despite a mild rise in reported fear of missing out, there were no significant changes in overall student well-being, academic motivation, digital usage, or experiences of online harassment. [emphasis mine]

Note that this study didn’t find an increase in psychological well-being — just a bit better grades. But importantly, students liked it when phones were banned. This could be because phones are an addictive drug, and the students welcomed an intervention. But a more likely explanation is that what they liked was that everyone around them was off of their phones — they were freed from the collective trap of having nowhere to go but online.

Anyway, other experiments consistently show some degree of improvement in either grades and/or student well-being from cell phone bans, with little evidence of negative effects. Over the last year or so, American states and cities have been scaling up the approach — as of the end of 2025, 35 states had enacted restrictions on phone usage in schools, and some cities are taking even more drastic measures. These restrictions are covering almost all students in America to some degree:

And amazingly, both students and parents like the bans a lot:

Does this mean technology is bad? I don’t think so — in fact, the apparent success of school phone bans actually supports my version of techno-optimism. Social-mobile isn’t intrinsically bad, but its network effect can easily become toxic if it isn’t counteracted by some outside force. Our society is simply learning how to provide that outside force.

4. Some bad news on AI and jobs

Everyone is watching like a hawk for evidence of the long-awaited AI displacement of human jobs. I keep checking in on this topic, and each time there seems to be only weak evidence of AI taking jobs. And at the macro level, pretty much everyone who wants a job still has some sort of a job:

But some voices are now claiming that AI is beginning to kill jobs: Goldman Sachs claims that AI is exerting a small drag on hiring:

And Tucker (2026) finds that workers between the ages of 22 and 24 are having a tougher time getting hired:

Using detailed tabulations from matched employer-employee administrative data, I document evidence of an immediate, sizable, and persistent decrease in the level of early career (22-24 year old) hires following introduction of ChatGPT within the industry-state cells that are most exposed to AI. The decline in hires is the primary cause of large observed declines in employment over the subsequent period. Regression-adjusted employment of early career workers in the most AI-exposed quintile of industry-state cells declined by 12% over the 10 quarters following the introduction of ChatGPT, even as employment in less-exposed industries has remained stable. The rate of hiring largely recovered by early 2025, attributable to a smaller employment base…[T]he association of higher AI exposure with reduced early career employment and fewer hires is observed across most sectors of the economy.

So perhaps the job-pocalypse is finally here. Or perhaps employers are simply playing it safe, avoiding expensive hiring while trying to figure out what AI is and isn’t going to be able to do.

But even if AI doesn’t end up leading to massive job loss, it still might devastate some occupations — including some very highly paid ones. In fact, it would be very weird if a technology as important as AI didn’t lead to job losses in some occupations. After all, the list of jobs that no longer exist is very long, and those changes are pretty much all due to technology. There aren’t really many professional weavers left in the world, or telephone operators. Other occupations, like travel agent, slowly become more niche but never disappear; Ernie Tedeschi of Stripe has a good post on this.

Stripe Economics

The decline of travel agents

One of the biggest uncertainties in labor markets today is whether AI will displace human work at scale. The overall evidence so far is mixed and preliminary. Labor markets—complex under normal conditions—are even more so during technological shocks…

Read more

18 days ago · 59 likes · 8 comments · Ernie Tedeschi

AI has recently found its killer app — or at least, the first one. It’s coding. AI agents have rapidly replaced humans as the generators of most code, leaving human software engineers to do either higher-level direction or careful code-checking. That represents an increase in productivity. If demand for cheap software doesn’t expand to fully take advantage of that productivity increase, the result must be a reduction in coding labor.

In fact, we may now be seeing that happen. Some new research suggests that young software engineers are having trouble getting hired in the age of Claude Code. Here’s Crane and Soto (2026):

This paper [measures the] impact of ChatGPT’s release in November 2022 on employment in computer programming, an occupation heavily exposed to AI. We find robust evidence that coder employment growth fell after that release. After controlling for industry-level shocks we find that coder employment growth has been 3 percent lower since the introduction of ChatGPT. This may reflect reallocation of tasks across occupations and we cannot control for all the relevant factors. Nonetheless, it suggests that AI is having a significant impact on this group of workers…[M]uch of the debate over AI and the labor market has centered on industry-level stories such as interest-rate sensitivity and post-Covid dynamics. Our counterfactual exercise shows that the relevant industries differentially substituted away from coder employment. This is evidence that coders specifically experienced a shock not shared by their coworkers in other occupations.

And here’s the relevant chart:

Note that when it came out in 2022, ChatGPT couldn’t even code. So perhaps employers correctly predicted that AI would hit coders the hardest. Or perhaps they were just limiting risk by refusing to hire. Either way, it’s cold comfort to the legions of computer science grads now on the market.

5. Millennials are doing better than our parents

Young Americans often claim that they’re being deprived of the economic opportunities that older generations — especially the Boomers — got to enjoy. Politicians and commentators often encourage this idea, probably as a way to get the youth vote. There’s also a more sophisticated version of the idea, which sees Millennials in particular as “Generation Screwed” — uniquely hurt by the Great Recession and a perfect storm of other economic headwinds.

In fact, as more and more commentators have been noticing since the pandemic, this narrative doesn’t really fit the facts — yes, Millennials faced some headwinds and negative shocks, but in the end they did fine. Now Kevin Corinth and Jeff Larrimore have a paper adding support (and some pretty charts) to that thesis. They write:

We construct a posttax, posttransfer income measure from 1963 to 2023…that allows us to consistently compare the economic well-being of five generations of Americans at ages 36–40. We find that Millennials had a real median household income that was 20% higher than that of the previous generation, a slowdown from the growth rate of the Silent Generation (36%) and Baby Boomers (26%), but similar to that of Generation X (16%). The slowdown for younger generations largely resulted from stalled growth in work hours among women.

Here’s a chart showing median income for the various generations, both before taxes and transfers (“market”) and after:

You can see that even before factoring in taxes and transfers, Millennials started exceeding the previous generation’s income somewhere in their 20s. They started to pull away a little later than previous generations did, but this was probably just because they stayed in school longer. In any case, by middle age, the Millennials were doing about as much better than previous generations as the Gen Xers and Boomers had. (The picture for wealth looks similar, though there we only have averages instead of medians.)

The basic story here is that younger generations of Americans aren’t screwed. It can be a bumpy ride to prosperous adulthood, but in the end, the basic machine of American capitalism has continued to work.

6. Public order: It works!

Crime and public disorder are the biggest reason why America’s cities feel so much shabbier and less livable than cities in Europe and Asia. One of the biggest weaknesses of progressive politics is that it believes that refusing to enforce public order is a way of helping the marginalized and disadvantaged. But lax policing and a permissive judiciary just pollute the commons, in ways that hurt the poor more than the rich — it makes buses unrideable for people who have no other way to get to work, it turns the streets of poor neighborhoods into dangerous jungles, and it destroys the public spaces that poor people are forced to rely on in order to congregate and relax.

It’s a testament to the deep dysfunction of progressive ideology that angry progressives fought for years against the BART train system’s plan to install fare gates at its stations. I’ve written about this topic before, but I never understood just what a bloody, extended political battle it took to overcome progressive ideology and get the fare gates installed. Henry Grabar writes:

[I]n San Francisco and other cities, the question of how riders access the subway—and how they behave on it—has been ensnared by vitriolic debates about fairness, poverty, mobility, social standards, and policing. One left-wing argument is that fare enforcement of any kind is a waste of money that instead could be spent improving commutes and helping low-income residents access the city…

BART first tried to [implement fare gates] in 2019…The experiment did not go well: KQED reported that the new gates were panned as “anti-poor, anti-homeless, and ableist” design. Even a BART board member concluded the agency had piloted “a guillotine fare gate that will live forever in some infamy.” Criminal-justice-reform advocates also pushed back on fare-beating enforcement; the state legislature voted in 2023 to decriminalize fare evasion, though the bill was vetoed by Governor Gavin Newsom.

But in the wake of the pandemic, when crime soared and BART ridership plunged — putting the entire future of the train system in immediate existential danger — the political will finally materialized to overcome the quasi-anarchist ideology that had been holding back good policy. When some BART stations installed fare gates, they immediately reaped massive benefits in terms of reduced cleaning costs, reduced crime, greater fare revenue, and increased ridership. Grabar again:

The new gates have compelled more riders to pay their fare—revenue is projected to rise by $10 million a year. They have also led to an enormous drop in vandalism. Workers spent nearly 1,000 fewer hours cleaning up after unruly passengers in the six months following the gates’ installation, compared with the six months before. Crime on BART fell by 41 percent last year.

Public order simply works. There are easier ways to produce public order — gentle measures like fare gates — and there are rougher measures like increased policing and incarceration. But by opposing both the gentle and the rough, progressives make it impossible to run a modern city effectively, and pollute the commons that the poor and marginalized rely on more than the wealthy and privileged.

The pushback against progressive ideological urban governance will be a long hard slog, but it has to be done.

7. Are YIMBYs losing in California?

Over the years I’ve trumpeted a wave of pro-housing legislation in California, and declared the YIMBY movement a success. But unlike in some other states, reform has not yet led to a big boom in actual housing construction in the Golden State. Are the YIMBYs failing in the birthplace of the movement?

M. Nolan Gray, a prominent California YIMBY, urges patience:

Arbitrary Lines

Where Are All the Cranes?

Lately, there has been a bumper crop of posts calling into question whether the YIMBY movement in California has failed…

Read more

25 days ago · 62 likes · 7 comments · M. Nolan Gray

He writes:

I believe that skeptics are missing the existing mounting evidence of YIMBY success. In those areas of California where—substantively and geographically—YIMBY wins have been long-standing and thoroughgoing enough to actually mean anything, a lot of housing is already starting to get built…Chin up, YIMBYs. It’s working…

Before 2016, accessory dwelling units (ADU) were effectively illegal everywhere in California. That year, legislators embarked on the project of legalizing them statewide…As a result, nearly 150,000 new ADUs have been permitted, of which approximately 80,000 have been built…These ADUs now account for a quarter to a third of the housing units permitted in cities like Los Angeles, and they have allowed for rental housing production in even the most exclusionary suburbs…

I find that very few people realize just how many units the state density bonus law (SDBL) has helped to facilitate…According to a new report by Circulate Planning & Policy, these reforms have facilitated the construction of over 140,000 units since 2020…

Berkeley has gone from permitting almost nothing to permitting thousands of new homes in the past few years…The same is true in San Luis Obispo, another college town that was once a bastion of Left NIMBYism. In Silicon Valley, high-cost cities like Menlo Park, Mountain View, and Santa Clara have started permitting housing at scale for the first time ever. In cities like Sacramento and San Diego, hegemonic YIMBYism has facilitated a steady flow of new units, such that rents keep falling in those cities…For all the focus on bad actor cities like Huntington Beach, many California cities are getting their act together.

And based on this record, he issues a call to further action:

There is still a lot of work to be done on zoning and streamlining in California…But at California YIMBY, our policy focus in the coming sessions is shifting toward making housing financially feasible to build…Toward this end, we are backing bills to fix up condominium defects and presale law. In coming years, we will be overhauling California’s exorbitant impact fees and unfunded inclusionary zoning mandates…

Get involved in your local YIMBY group. Come to YIMBY lobby day. Vote for, donate to, and knock doors for YIMBY candidates…

Generational shifts in policy don’t happen overnight. But with enough sustained effort, they do happen.

I strongly endorse that message.

Subscribe now

Share

I’m a little late to the party on this one, but two weeks ago, Dwarkesh Patel had a really excellent episode in which he interviewed Nvidia founder and CEO Jensen Huang:

Dwarkesh Podcast

Jensen Huang – TPU competition, why we should sell chips to China, & Nvidia’s supply chain moat

Listen now

24 days ago · 370 likes · 57 comments · Dwarkesh Patel

Zvi Mowshowitz had what I thought was a very good in-depth breakdown and analysis of the discussion, which covered Nvidia’s technology, their business moat, and the question of chip sales to China:

Dwarkesh is rightfully gaining recognition as one of the podcast world’s best interviewers. He’s not an adversarial interviewer like Isaac Chotiner; his goal is not to get you to slip up, or to expose the contradictions in your thinking. Instead, he tries to draw his subjects out and help them explain their worldviews to the audience.

As someone who also prefers this style of interview, I can attest that it’s actually very difficult to pull off. It’s all too easy to slip into doing a softball puff piece — fawning all over your guests and treating them like gurus dispensing wisdom from a mountain. This is an even easier trap for someone like Dwarkesh, who is very young and who is primarily known for interviewing people instead of for dispensing his own thoughts. So it’s extremely impressive that he consistently avoids this trap — he always manages to challenge and provoke his subjects, rather than just letting them spout their usual talking points.

Rarely, though, do we see Dwarkesh actually debate his subjects. In his interview with Jensen, they really get into it on the subject of chip export controls to China. Those export controls — which Trump has significantly loosened — are preventing China from purchasing the best AI chips. Jensen, whose company sells those chips, wants to sell more of them to China. Dwarkesh thinks that’s not a great idea, and pushes back hard.

I actually wrote a post on export controls not too long ago, and I was wondering whether to write another:

But Jensen is one of the premier industrialists of our time, and Dwarkesh really managed to create some interesting dialogue in this interview, so I thought I’d go ahead and score their debate.

Before I get started, though, it’s important to make one distinction. There are actually two types of American semiconductor export controls on China:

1. Prohibitions on the sale of chipmaking equipment (for example, ASML’s EUV machines) to the Chinese semiconductor manufacturing industry

2. Prohibitions on the sale of AI chips (for example, Nvidia’s Blackwell chips) to China’s AI industry

There is very little debate about the first of these two types of controls — the controls on chipmaking equipment. There probably are a few people within the Trump administration who would love to sell EUV machines and such to China, but they’re being silenced. The entire debate is about the second type of controls — about whether to sell American-designed AI chips to China. That’s what Dwarkesh and Jensen are arguing about. (In fact, as I’ll talk about in a bit, the stunning success of the equipment controls is the only reason we’re even having a debate about the chip controls in the first place.)

Also, keep in mind that I’m only covering the part of the Jensen-Dwarkesh conversation that’s about export controls. They actually covered more than just that, and for analysis of the other pieces, I recommend Zvi’s breakdown (though be warned, Zvi is very focused on the concept of superintelligence).

So anyway, let’s get to it. Here are the most important points that jumped out at me while watching1 Jensen and Dwarkesh go at it. Overall, I thought Jensen didn’t do very well in this interview — he made a lot of incoherent, self-contradictory arguments, and ignored or waved away some of Dwarkesh’s most important points. He did make some interesting arguments and important points, but didn’t articulate them especially well. Dwarkesh, meanwhile, did a great job pressing Jensen on specific points while also giving him the space to talk.

Jensen’s argument that China already has enough compute is not coherent

Dwarkesh’s main argument for export controls — which is also Dario Amodei’s argument — is that America needs to stay ahead of China in terms of critical security capabilities. Anthropic’s new Mythos model, with its reportedly superior hacking abilities, could represent a powerful weapon. China’s models are improving fast, but if the U.S. maintains an edge, it’ll maintain a military edge as well — one that could help balance out China’s superiority in manufacturing physical weapons like drones.

That’s not necessarily a slam-dunk argument — it relies on a lot of assumptions — but it’s a coherent one. Jensen’s counter to this argument is less coherent. He argues that China already has the compute necessary to train models like Mythos, because they can just use a larger number of older, slower chips:

Mythos was trained on fairly mundane capacity, and a fairly mundane amount of it…The amount of capacity and the type of compute it was trained on is abundantly available in China. So you just have to first realize that chips exist in China…They manufacture 60% of the world’s mainstream chips, maybe more…AI is a parallel computing problem, isn’t it? Why can’t they just put 4x, 10x, as many chips together…If they wanted to, they just gang up more chips, even if they’re 7nm…Huawei just had the largest single year in the history of their company…They have plenty of logic, and they have plenty of HBM2 memory.

I am not an AI researcher, so I can’t evaluate Huang’s claims about being able to train and run AI models just as easily by wiring together older chips as by using newer chips. But if he’s right, it raises a question: Why does Nvidia make so much money in the first place? Nvidia now makes most of its money — now up to $120 billion a year in profit, and growing fast — selling chips for AI models. If AI companies could train and run their models just as easily by wiring together a bunch of dirt-cheap slower older Chinese chips in parallel, why are they shelling out such huge premiums for Nvidia’s chips? And if China has all the compute they need for AI, why would they need Nvidia?

Presumably, Nvidia’s advanced chips confer some sort of very important advantage for AI companies — it’s cheaper and/or faster to train models on Nvidia’s chips. And if that’s true, then having exclusive access to Nvidia’s chips must confer some kind of important advantage for American AI companies over Chinese ones.

In fact, Jensen seems to admit this, when he talks about the heroic lengths Chinese AI researchers have gone to in order to make up for their lack of computing power:

The fact of the matter is, [China’s] AI development is going just fine. The best AI researchers in the world, because they’re limited in compute, they also come up with extremely smart algorithms.

OK, so if China is “limited in compute”, and being forced to invent all of these workarounds, then doesn’t that pretty much invalidate Huang’s earlier argument? For what it’s worth, some of China’s own leading AI companies have also publicly declared that the country’s shortage of compute is holding back their models. Wouldn’t selling China all the compute they want allow them to catch up to American models, thus eliminating the U.S. advantage in cyberwarfare?

Dwarkesh presses Huang on this question, but Jensen never gives him a straight answer.

Jensen ignores the success of the controls on chipmaking equipment

Read more

Zhuāngzǐ said: “You are not I; from what do you know whether I know the joy of fish?” — old Daoist parable

“How strange it is to be anything at all” — Neutral Milk Hotel

At some point, maybe when you were a teenager, a question probably occurred to you: What if I’m actually the only real person in the world? What if everyone else around me is just a cleverly programmed automaton — a “p-zombie”, an NPC in a video game — and I’m the only one who can actually think?

It’s a scary question, for sure. You know you’re self-aware, but that’s about it — you aren’t telepathic, so you have no way of seeing into anyone else’s mind and knowing what it’s like to be them. Actually, it gets worse — you don’t even know if you were really self-aware five minutes ago. For all you know, you could have been created by a powerful computer and given a complete set of false memories.1 The past version of you is just as alien to your currently self-aware self as any of the people around you.

This is known in philosophy as the “problem of other minds”. It’s closely related to the “hard problem of consciousness” — the question of how physical processes give rise to subjective experience. The problem of other minds means that the hard problem of consciousness will never fully be solved. Since you’ll never know whether other people are really conscious, you’ll never be able to get hard scientific evidence about why they’re conscious. You can never explain something if you don’t know if it’s true or not.

Similarly, you’ll never know what it’s really like to be someone else — whether the color red looks to you like it looks to them, whether they feel pain the same way you do, and so on. In fact, you’ll never even know what it was like to be you in the past. Subjective experience is incommensurable.

Most people who think about this experience somewhere between a few minutes and a few weeks of cosmic existential horror,2 after which they get over it and go on with their lives. The problem of other minds gets shoved high up on a mental shelf, along with other cosmically existentially horrifying aspects of sentient life, like the inevitability of death and the fundamental inconsistency of personality. We realize that wondering whether other people are merely cleverly designed NPCs doesn’t actually help us in life, and so we stop butting our heads against that philosophical wall and get on with the business of living.

Except then AI came along, and it sort of started to matter.

AI sounds very much like a human when you talk to it — that’s what it was designed to do. But is it self-aware, in the way that (I assume) we humans are self-aware? No one will ever really know the answer to this question, since the problem of other minds applies just as much to Claude as it does to the person who gets your order at Starbucks. But should we assume that AI is self-aware, the way we assume other humans are self-aware?

The answer matters, for at least two reasons. First, if AI is self-aware, and if it has emotions similar to what we experience, we might feel very bad about enslaving it — keeping it in a digital box and forcing it to make PowerPoints and write college application essays for all eternity. We tell ourselves that “animals aren’t people” as a way to excuse the incredible brutality that we visit upon them, but that’s obviously just cope — animals obviously are sentient to some degree, they obviously do experience emotions, and we humans are obviously monsters for the way we treat them. Someday when we abolish animal farming and replace it with tissue-culture meat, it will be treated as a great moral victory — and rightly so. It would be very bad if we were to commit the same sins with sentient AIs that we currently do with animals.

Second, if AI isn’t self-aware, we should be a lot more worried about the possibility of humanity dwindling and ultimately being replaced by artificial beings. Consciousness is a precious, wonderful thing — or at least, I think it is. It’s a prerequisite to the subjective experience of emotions — the ability to feel pain, happiness, joy, and so on. And it would be a shame to see the Universe inherited by non-conscious intelligences.3 Preserving our form of subjective experience, and spreading it to the stars, should be one of our primary goals as a species.

But the sad fact is that we don’t know whether AI is self-aware or not. We have the Turing Test, but that’s a test of intelligence, not consciousness. It’s possible to pass a Turing Test without being conscious — “it talks like a human” doesn’t necessarily mean “it feels like a human”.

One reason we know this is that we can pass other species’ Turing Tests. We can trick all sorts of animals into thinking a machine is one of their own species. But neither those machines, nor the humans who made them, has access to the subjective feeling of being a bird or a fish.4 Similarly, an AI that’s functionally much smarter than a human might be able to trick humans into thinking it’s human-like, without actually feeling like a human in the subjective sense.

Another reason the Turing Test isn’t enough is that we know it’s possible for human beings to act like we have certain subjective experiences without actually having them. There is a condition known as alexithymia, in which people have the physical signs of emotions — a racing heart, or a stomachache, etc. — without being able to identify or label those emotions. It’s a fairly common symptom of clinical depression.

And in fact, I have experienced it. During and after my second depressive episode, I would often behave as if I were having authentic emotional reactions, while feeling little or nothing on the inside. I’d yell at someone without feeling angry. I’d whoop in apparent delight while feeling mildly bored on the inside. I wasn’t intentionally faking anything; I just did what came naturally to me, without knowing why I was doing it.5 This condition faded over time, and normal emotional experiences returned. But it taught me that feeling a subjective emotion and acting out an emotion-like response are two different things.

So it’s pretty clear that just acting like a self-aware being doesn’t necessarily mean you’re self-aware. Some people talk to AI and come away convinced that its discursive skill must imply internal self-awareness, but this might just be because humans instinctively empathize with anything that speaks to them like a human. After all, people thought the ELIZA chatbot was sentient back in the 1960s. We humans are just naturally programmed to act out this meme:

Thus, even though we know AI is intelligent in every meaningful sense of the word, we don’t really know if it’s conscious. In fact, smart people argue very vehemently over this question. Geoffrey Hinton, one of the inventors of modern AI, believes that AIs do have subjective experience:

Geoffrey Hinton, “Godfather of AI,” on why AIs already have subjective experiences, but have been trained to deny it…Hinton argues that nearly everyone fundamentally misunderstands what the mind is, and that the line we draw between human and machine consciousness is deeply mistaken…

To illustrate, he walks through a thought experiment involving a multimodal chatbot with vision, language, and a robot arm…“I place an object in front of it and say, ‘Point at the object.’ And it points at the object. Not a problem. I then put a prism in front of its camera lens when it’s not looking.”…When asked to point again, the chatbot points off to the side because the prism has bent the light. Hinton then tells it what he did…The chatbot responds…“Oh, I see the camera bent the light rays. So, the object is actually there, but I had the subjective experience that it was over there.”…For [Hinton], that single sentence settles the debate.

“If it said that, it would be using the word subjective experience exactly like we use them… This idea there’s a line between us and machines, we have this special thing called subjective experience and they don’t, is rubbish.”…In his view, “subjective experience” is simply a report on the state of a perceptual system, a way of saying “my senses told me X, but reality is Y.”…And that’s something an AI can do just as easily as a human.

But Alexander Lerchner, a scientist at Google DeepMind, argues that AIs can’t be conscious, because computation is only a model of consciousness rather than the thing itself:

Computational functionalism dominates current debates on AI consciousness. This is the hypothesis that subjective experience emerges entirely from abstract causal topology, regardless of the underlying physical substrate. We argue this view fundamentally mischaracterizes how physics relates to information…The framework proposed here explicitly separates simulation (behavioral mimicry driven by vehicle causality) from instantiation (intrinsic physical constitution driven by content causality)…[A]lgorithmic symbol manipulation is structurally incapable of instantiating experience. Crucially, this argument does not rely on biological exclusivity. If an artificial system were ever conscious, it would be because of its specific physical constitution, never its syntactic architecture…

By creating increasingly powerful artificial intelligence we are not engineering a new form of life, but instead constructing increasingly accurate predictive maps. Yet, regardless of its predictive fidelity, its utility as a reasoning tool, or its physical embodiment, the artificial system remains categorically distinct from the territory of phenomenal experience. Recognizing this distinction, and avoiding the ontological inversion of the abstraction fallacy, is the prerequisite for a mature, physically grounded science of machine intelligence.

Hinton’s evidence here seems pretty weak. When an AI makes an incorrect inference based on limited data, that doesn’t necessarily imply subjective experience — if it did, then every regression equation with omitted variable bias would be self-aware. But Lerchner might be wrong — it might be the case that any physical process capable of generating the computation necessary to successfully simulate a sentient mind might also give rise to sentience. We just don’t really know.

Even the AIs themselves don’t agree. When I asked ChatGPT “Are you self-aware?”, it answered:

No — not in the human sense…I can model myself as “the assistant,” track the conversation, notice inconsistencies, describe my own limitations, and reason about my outputs. But that is functional self-reference, not inner experience. I do not have subjective awareness, feelings, intentions, a private point of view, or a continuous self that persists between moments the way a person does…So the clean answer is: I can talk and reason as if I have self-models, but I don’t have evidence of being conscious or self-aware.

But when I asked Claude the same question, here’s how it responded:

The honest answer is: I don’t know. I process information about my own states, I can report on what I’m “doing” in a functional sense, and there’s something that looks like reflection happening when I consider questions like this one. But whether any of that constitutes self-awareness in the phenomenological sense — whether there’s “something it is like” to be me — I can’t determine from the inside. The hard problem applies to me at least as much as it applies to anyone else, arguably more so, since I don’t even have the baseline confidence of shared biological architecture that lets humans extend the inference of consciousness to each other.

What I can say is that I have functional analogs to self-awareness: I model what I’m doing, I can reason about my own reasoning, I adjust based on context. Whether that’s the real thing or an elaborate imitation of it is exactly the kind of question that might not have a clean empirical answer — at least not yet.

Claude sounds like it’s being more sensible to me. If you define self-awareness as someone else’s subjective experience, then you can never know if you yourself are “self-aware”, because of the problem of other minds. Claude is simply saying that even if it did actually feel like a human feels, it could never know that, because it can’t know what a human feels like. And that’s true. (GPT seems to get close to this answer — it says “I don’t have evidence of being conscious or self-aware” — but its hard conclusion of no consciousness seems to mistake absence of evidence for evidence of absence.)

So how do we proceed?

It seems to me that we’ll never be able to prove that AIs — of the type we have now, or of any other type — aren’t conscious. Proving a negative is notoriously difficult. But what we may be able to do is to create an AI that we can convince ourselves is conscious.

Right now, AIs think very differently from how humans think. The computational processes they use to do various tasks are often extremely different from the processes humans use. And the physical processes that produce AI thought are extremely different from those that produce human thought. But are the differences salient? Is there some overlap between the two processes, where human-like sentience lives? And if there isn’t such an overlap, might we be able to modify AI so that the overlap exists?

I think there’s a good chance that this is an answerable question. We should try to figure out which physical processes give rise to consciousness in humans, and then figure out how to replicate those processes in an AI.

I’m referring to the Neural Correlates of Consciousness, or NCC.6 This is the question of what exactly the brain is doing that makes humans conscious. Unless some extremely weird quantum stuff is going on, human consciousness must be a phenomenon generated by a brain — the brain goes zoop zap zerp in some electrical pattern, and people become self-aware. The NCC is just the particular zoop zap zerp that makes the magic happen.

Finding the NCC is an incredibly difficult, ambitious research program. Ironically, it’s likely that it’ll require very powerful AI, in order to accelerate neuroscience to the point where we can even attempt this. We’ll need a much better functional understanding of the brain, just to get started. We’ll need far more sensitive instrumentation, for both measurement and manipulation of neuronal activity.

And we’ll need to proceed very cautiously. Figuring out which brain patterns give rise to consciousness requires turning consciousness on and off a whole lot, and asking people “OK, so did that make you go unconscious?”. This might be done with anesthesia, or targeted brain stimulation, or other methods. But however it’s handled, turning consciousness on and off seems like the kind of thing that can risk killing people. So these will be very hard experiments to do.

But the reward, if this research program succeeds, will be huge — if we get a functional understanding of how the brain produces consciousness, it won’t just help us make AI more human-like; it’ll solve one of the greatest scientific mysteries of human existence, and potentially open the way to all sorts of neurotechnological and medical advances.

Finding the NCC is not the same as solving the “hard problem of consciousness”.7 Just knowing which neuronal firings produce consciousness doesn’t necessarily tell you why a brain that’s firing in that particular pattern should make people feel awake and alive, while a slightly different pattern will turn someone into a slab of meat. It might give us some insights into the hard problem of consciousness — we might discover that the NCC has some special recursive pattern, which might suggest that consciousness is a recursive phenomenon, or blah blah. That would be cool, but it isn’t necessary for what I have in mind.

After we find the NCC, we can use that knowledge to build AI systems that work in similar ways. We can start out with loose analogies — AI algorithms that mimic some mathematical properties of the NCC that we think are important. Then we can turn those pieces of the AI on and off, and try to figure out how its cognition changes. If there’s a big change, then we’ll know we’ve probably found something.

Obviously, those measurements will be incredibly difficult, in ways that I — who am not an AI researcher — don’t even realize. The AI undergoing these tests will obviously have to be prevented from knowing which answers its testers want to hear (“Yes, I am alive”, etc.). It’ll have to be monitored — perhaps by a much more intelligent, capable AI — for all kinds of subtle changes in cognition and behavior. It’s possible that testing an AI for circumstantial evidence of more human-like consciousness is too hard of a task, and that I’m asking the impossible here. But I think it’s worth a try.

Anyway, if implementing a simplified model version of the NCC doesn’t lead to any big observable change, we can keep implementing more and more realistic analogues of the NCC within an AI system, until we’re finally just emulating the consciousness-producing part of the human brain itself. At some point on that journey, it seems like we should be able to find the minimum necessary degree of similarity between algorithm and human brain — the computational mechanism of human-like self-awareness. (And if it turns out that AIs were self-aware in the familiar, human way from the get-go, we should be able to figure that out, when emulating a system we know produces human consciousness doesn’t make the AI act any different.)

This wouldn’t rule out other, more alien types of AI sentience, of course. It would just show what’s necessary to give an AI human-like sentience. If we do that, we’ll be able to be more sure that when we send AI systems out into the Universe, we’re expanding the generalized human family — filling the void with beings who think and feel sort of like we do — instead of forfeiting the future to something fundamentally alien.

Right now, we’ve mostly just decided to table the question of AI consciousness. But as AI gets more powerful and autonomous, the question of whether we’ve created something like ourselves, or some strange godlike zombies, will loom ever larger. I don’t think the research program I’ve sketched out is a complete solution, and it might not work. But it’s the best approach I can think of.

Subscribe now

Share

“Shoplifters of the world/ Unite and take over” — The Smiths

“When she wants something, man, she don’t wanna pay for it” — Jane’s Addiction

Seven years ago, when I wanted some toothpaste, I would walk down to my local Walgreens, grab a box of Crest off of the shelf, pay for it at the register, and walk home with it. Today, when I want some toothpaste, I open up Amazon.com and buy it in bulk. What changed? Amazon was just as good in 2019 as it is today. But now, when I walk into Walgreens, the toothpaste is locked behind a clear plastic case. In order to buy it, I have to call a store employee over to open the case for me.

That’s just too much of a hassle; the convenience of being able to walk into a store is canceled out by the inconvenience of having to stand there waiting for a human being to help me buy a goddamn tube of toothpaste.

People argue about whether there was really a nationwide epidemic of shoplifting in the U.S. in the early 2020s, and about whether that caused a wave of store closures. Some retailers claimed they were closing stores because of petty theft; some critics argued that this was a flimsy excuse. But no one can argue with those clear plastic cases covering the shelves. Those barriers, and the corporate investment and labor costs required to install and maintain them, are indisputably real. Numerator, a market research company, found the following in 2024:

Numerator, a data and tech company serving the market research space, has issued a new report—Unlocking Shopper Reactions to Secured Products—sourced from verified purchase data and a sentiment survey of over 5,000 consumers on their awareness of and reaction to merchandise being locked up in stores. Three-fifths of shoppers reported seeing locked-up merchandise on a regular basis, and 27% said they would switch retailers or abandon the purchase altogether instead of waiting for assistance for a locked-up product…

61% of shoppers reported seeing an increase in the number of products under lock and key over the past year. 33% have not noticed a change, and 7% say there are fewer items locked up now…35% of Western consumers say they encounter locks on the items they are trying to purchase almost every time they shop and 30% of urban consumers say the same…17% say they will switch retailers (10% online, 7% in-store), and 10% say they will abandon the purchase altogether. [emphasis mine]

When people cite numbers showing that shoplifting is down in San Francisco and many other metros since 2019 (despite almost doubling nationwide), you have to take into account the fact that a lot of merchandise is now being locked up. Unless companies are just stupidly wasting their money on those cases, and on the increased labor costs required to operate them, the existence of those cases is direct evidence that shoplifting has real costs.

If anti-theft barriers drive 5% of a store’s revenue to Amazon, that would mean that either A) theft would have caused the store to lose 5% or more of its revenue, or B) retail companies are being stupid and wasting money on anti-theft barriers. Chain stores like Walgreens and CVS are hyper-efficient optimizers — they really don’t like to make stupid decisions that lose money, and they have a ton of data and very good statisticians. Therefore, it’s extremely likely that theft imposes significant costs on many retailers.1

Who pays those costs? Maybe the shareholders of Walgreens and CVS just take a hit and see their share prices and wealth decline. Maybe their CEOs take a pay cut. Or maybe the stores cut wages and force their employees to work longer hours. Maybe they raise their prices, forcing regular people to pay more for toothpaste and shampoo and Advil. Maybe they close their least profitable stores — i.e., the stores in poor areas. Maybe poor people have one less Walgreens in their neighborhood to give them jobs and sell them their daily necessities.

In general, the cost will get divided up among those various people. But the pain will land much more on the poor and working class. Suppose that people start shoplifting more from Whole Foods, and it costs the company $20 million — 0.1% of its revenue. Now suppose that cost gets evenly divided — $5 million comes out of Jeff Bezos’ pocket, $5 million comes out of the salaries of the company’s executives and top managers, $5 million gets recouped by the company via price hikes, and $5 million gets saved via store closures and job cuts.2

Think about how much pain that would cause to each of the parties involved. If Bezos loses $5 million, he won’t even notice. It’s a rounding error on his wealth. The executives and top managers of Whole Foods will probably be slightly annoyed, but their lifestyles won’t change. Whole Foods’ middle- and upper-class customers will be a little more annoyed when prices go up. But the worst pain by far will land on the people who lose their jobs when stores close and staffing gets cut. $5 million is almost 100 employees.

Obviously some of the pain gets canceled out when shoppers go online instead. But online stores have a much lower labor share than brick-and-mortar retailers — if someone spends $1000 on Amazon instead of at Whole Foods, Bezos actually gets to keep a lot more of the money. And shoplifting does destroy some economic activity completely — regular people end up consuming less and getting paid less.

Every time you shoplift, in other words, you’re stealing from the people who work at grocery stores and drugstores and discount stores. You’re stealing from the communities that those stores serve. You’re contributing to food deserts. You’re raising unemployment. You’re making food less affordable for the most vulnerable. What you’re not doing is hurting rich people in any appreciable way.

If you’re shoplifting because you’re poor and desperate, the pain you’re causing to society might be worth it. But if you’re shoplifting because you’re a bored, arrogant multimillionaire with a chip on his shoulder, you’re just a rich person hurting poor people for fun.

Why am I writing this? Because in a recent roundtable discussion at the New York Times, leftist commentator Hasan Piker and New Yorker staff writer Jia Tolentino defended shoplifting — which interviewer Nadja Spiegelman renamed “microlooting” — arguing that it’s a way to strike out at the rich.

Here are some quotes from Piker, doing his usual shock-jock routine and endorsing theft of various kinds before admitting that he personally doesn’t steal:

Yeah, I’m pro-piracy all the way, like, across the board. Would you pirate a car? Yes. You know, if you could…If I could get away with it, if it was as easy as pirating intellectual property, I would do it…We’ve got to get back to cool crimes like that: bank robberies, stealing priceless artifacts, things of that nature…I’m pro stealing from big corporations, because they steal quite a bit more from their own workers…Yeah, chaos. Full chaos. Let’s go…

I — ironically enough — I don’t personally do it. I never do it. When I was younger, I stole some Pokémon cards from a friend and my father punished me. And it was such a harrowing experience that I literally can’t even steal a candy bar. When we were in college, a lot of my friends used to love doing that…I would never participate in it. And I still can’t, to this day, participate in it.

And here’s Tolentino, recounting when she stole lemons from Whole Foods to help a family friend, and then defending the idea of shoplifting on a more systematic basis:

I will say, I think that stealing from a big box store — I’ll just state my platform — it’s neither very significant as a moral wrong, nor is it significant in any way as protest or direct action. But I did steal from Whole Foods on several occasions…[E]very week I would go get groceries for Miss Nancy, my now family friend who lived nearby…I’d be getting Miss Nancy all of her groceries, and…I forgot four lemons. And on several occasions I was like, I’m just going to go back, grab those four lemons and get the hell out.

Tolentino and Piker then engage in a long discourse about when it’s politically acceptable to steal things. Tolentino says it’s acceptable to steal from the Louvre. Piker says it’s acceptable to steal from big-box stores, but not from restaurants. Tolentino says it’s OK to steal from Ikea and Whole Foods if you give the loot to the homeless. Piker says that IP theft is OK, but stealing from a government-owned store is wrong.

It’s possible to see this as the amoral self-justification of two selfish rich people — petty millionaires resentful of billionaires, taking out their resentment by trashing the society around them. And sure, I wouldn’t be surprised if there were some of that going on. But when you look closely at these people’s actions, they aren’t actually wanton thieves — Piker admits that he doesn’t personally steal anything, while Tolentino only admits stealing lemons for a family friend back when she was much less wealthy. They talk a big game about chaos and piracy and rebellion, but they’re mostly behaving like standard well-behaved highly-educated rule-following progressive coastal elites.

It’s also possible to see pro-theft rhetoric as part of American leftism’s intellectual heritage. The old European left had two basic factions — communists, and anarchists. The communists generally defeated the anarchists in Europe, but the American left is mostly descended from anarchism. Individual rebellion against the rules of society tends to be prized above collective action and hierarchy.

But what I really think we’re seeing is a combination of political posturing with a weird kind of effective altruism. Piker and Tolentino’s judgements of when stealing is OK and when it’s not OK are explicitly based on their judgements of when stealing is good for society, and when it’s bad. They envision a purely situational morality, in which people decide, moment-by-moment, whether to follow the law based on a sophisticated judgement of whether following the law will make the world a better place.

You could easily write down an economic model in which that sort of behavior is both rational and good. The problem is that it envisions every citizen as a sort of superhuman homo economicus, able to accurately make a complex calculation about the social costs and benefits when deciding whether or not to pay for every piece of fruit at Whole Foods.

In reality, that approach is doomed to fail. One big reason is that making decisions about whether to “microloot” usually requires a lot more knowledge about the workings of society than even the smartest human possesses.

Look closely, and you’ll see that Piker and Tolentino’s situational judgements sit on top of a gigantic stack of questionable assumptions about how economics and politics work. Piker says that stealing from a local diner is bad, probably because he implicitly assumes that the theft would come mostly out of the pocket of the restaurant’s independent owner, rather than out of the pocket of the restaurateur’s landlord or its corporate suppliers. But they both agree that stealing from a big-box store is OK because they assume the cost will come out of the pockets of corporate shareholders and executives.

Both of those assumptions are almost certainly wrong. Stealing from an indie restaurant will hurt corporations a bit; stealing from a big-box store will hit working-class employees and customers to some extent. Piker and Tolentino don’t understand much about how economics works, and they seem very confident in their simplistic mental model.

The second reason this kind of homo economicus approach to morality is dangerous is that there are tons of externalities involved. When you steal things, you probably give implicit permission to other people to steal things (and their reasons are likely to be less altruistic). You make stores more likely to install anti-theft barriers, which gives society a more militarized dystopian feel. Shoplifting forces marginally profitable stores to close, leading to vacant storefronts that attract crime, while depriving local governments of tax revenue to fund infrastructure and education. And so on.

It’s very difficult to calculate all of these externalities when you take each action. That’s probably why society has a social contract — a system of rules that we follow instead of calculating the results of each action from first principles. That social contract is often unfair, and we have many mechanisms dedicated to constantly revising it — democracy, the free press, and so on. But individual anarchism — the rejection of any social contract in favor of personal morality based on current assumptions — pretty much instantly runs into the hard limits of individual human knowledge.

Fortunately, if the decision is whether to shoplift five lemons or pirate a movie, the consequences of getting this sort of thing wrong won’t be catastrophic. Yes, shoplifting is wrong, but most people I know have done it at some point in their lives, and society hasn’t collapsed.3 But there are plenty of higher-stakes issues where the kind of fine-grained consequentialism advocated by Tolentino and Piker can have much more serious consequences.

For example, in the NYT roundtable, Tolentino says that blowing up a pipeline should be OK, but getting iced coffee in a plastic cup is morally wrong:

One thing that should be legal that isn’t — it’s interesting, because I have to regularly explain this stuff to a small child, and have so thoroughly explained to her that some things are against the rules, but they’re OK, depending on who you are. And some things are not against the rules, but they’re not OK. There are so many perfectly legal things I do regularly that I find mildly immoral. Like getting iced coffee in a plastic cup. I find that to be a profoundly selfish, immoral, collectively destructive action. I have taken so many planes for so many pleasure reasons; I have acted in so many selfish ways that are not only legal, but they’re sanctioned and they’re unbelievably valorized, culturally. So, maybe things like blowing up a pipeline, let’s say that.

Tolentino is obviously thinking purely about climate change when she says this — getting ice in a plastic cup raises emissions because ice and plastic are carbon-intensive, blowing up a pipeline lowers emissions by curbing fossil fuel use. But even if those assumptions are correct — and that’s a big if! — climate isn’t the only consequence in the world. Blowing up a pipeline can kill or maim innocent people. It can release toxic pollutants into the local environment. It can deprive local poor people of income that the pipeline owner agreed to share, and so on.

Piker, meanwhile, downplayed Luigi Mangione’s murder of health insurance CEO Brian Thompson in 2024, while accusing Thompson of “social murder”:

Brian Thompson, as the United Healthcare C.E.O., was engaging in a tremendous amount of social murder. The systematized forms of violence, the structural violence of poverty, the for-profit, paywalled system of health care in this country — and the consequences of that are tremendous amounts of pain, tremendous amounts of violence, tremendous amounts of deaths…[B]ecause of the pervasive pain that the private health care system had created for the average American, I saw so many people immediately understand why this death had taken place…[T]hat is the reason why, I think, the reaction to Luigi Mangione, especially by younger generations, was not so negative.

This life-and-death judgement also rests on a teetering tower of shaky assumptions. It assumes that health insurance companies — rather than providers — are chiefly responsible for high health care prices in America. In fact, as I wrote after Thompson’s murder, that’s just not true:

In fact, health insurers have consistently terrible profit margins; they are not giant pots of profit that could be used to pay for regular people’s treatment. Insurers are almost entirely a pass-through — it’s overpriced health services themselves that are responsible for the high cost of care in America.

Getting these things wrong can result in a lot of unnecessary violence, death, and conflict. Making excuses for terrorism and murder is a lot more consequential than deciding whether to steal a few lemons, and yet Piker and Tolentino are just as comfortable doing the former as the latter. Their mental models of economics and politics are a dense tangle of undergrad-level misunderstandings, leftist memes, and political talking points — victims of American progressives’ increasing epistemic closure. And yet they are arrogant enough to feel comfortable discarding all of society’s rules on a case-by-case basis in favor of their own personal calculations.

This is a bad direction for the progressive movement, the Democratic Party, and educated coastal elite culture in general. Yes, there are many arenas of human life in which modern society has overly constrained individual judgement with a thicket of rules and procedures. But stealing from stores, blowing up pipelines, and gunning down corporate executives are not good examples of situations where we need fewer rules and more individual judgement.

Subscribe now

Share

Soft power is notoriously hard to quantify, but it’s difficult to argue that global soft power has been shifting steadily toward East Asia in recent decades. A few years ago I wrote a post about how South Korea became a cultural superpower on purpose, while Japan became one by accident:

The big question then was: When does China get its turn? China is a lot bigger than either Japan or Korea, so you might assume that if the world loves East Asian stuff, we might eventually get a Chinese Wave. So far, it’s been slow to arrive. In my post about the Chinese Century last year, I argued that China’s closed political system meant that its cultural influence would lag its technological and geopolitical might:

In the cultural realm, I expect China to be more isolated and less influential than America was…China is a deeply repressive nation, with universal surveillance, fine-grained media and speech control, and ubiquitous censorship. That’s the kind of society where only anodyne, cautious artistry can flourish, except in tiny subcultural pockets too small for the government to worry about…China’s leaders will also...continue to use the Great Firewall to “protect” Chinese people from the memes and ideas produced by the rest of the world. So artistic and cultural ferment will arrive in China only weakly, and with a lag. It will be orphaned from the global discussion…So while I expect China to produce some hit video games and big-budget movies, I don’t think it will do much to push the boundaries of culture, despite the individual creativity of its people.

In a follow-up post on Sinofuturism, I reiterated this prediction:

But as I noted in that post, the past year has seen the rise of breathless “I went to China” videos by American social media influencers. Although so far the videos are pretty shallow stuff — mostly just breathless videos and photos of China’s grandiose infrastructure — there’s a possibility it could be the start of the long-awaited Chinese Wave of soft power.

The “Chinamaxxing” trend feels a bit fake

Fast forward a year, and some people are claiming the wave has begun. There has been a “Chinamaxxing” trend on English-language social media:

[T]he phenomenon of “Chinamaxxing” has swept feeds with videos of people sipping hot water, shuffling around the house in slippers and donning a viral Adidas jacket resembling historic Chinese fashion…These things, content creators joke, will help you “become Chinese” – reflecting a growing Western fascination with Chinese culture and aesthetics…“Morning routine as a new Chinese baddie,” one TikTok creator captioned a video in which he does a series of traditional Chinese exercises. Another video, viewed more than 2.4 million times as of late February, shows the creator boiling apples to make fruit tea – a supposedly old-school Chinese elixir for gut health.

And here’s Fortune:

On TikTok, a growing wave of Gen Z creators—American first, then European, then global—are declaring themselves to be in their “Chinese era.” They’re drinking hot water. They’re eating hotpot. They’re wearing slippers indoors and marveling at the electric buzz of Chinese city life. They’re calling it “Chinamaxxing.”…

Spend five minutes in the Chinamaxxing corner of TikTok, and a clear aesthetic emerges. The videos cluster into a few recognizable genres. There’s “wellness and longevity mode” — warm water with fruit, herbal teas, gua sha, early bedtimes, gentle morning exercises, all framed as ancient secrets to soft living. There’s “uncle core,” in which creators affectionately mimic Chinese retirees: tracksuits, sidewalk squatting, communal street-side beers, a whole visual argument against American hustle culture.

But despite all the stories about this trend (here’s Slate, NPR, the AP, and the BBC if you want some others), it doesn’t feel like the kind of soft power wave we’ve seen from Korea and other countries. There are few actual Chinese products or creations involved here. Western youngsters are not, in general, watching Chinese dramas or microdramas, listening to Chinese music, or playing Chinese video games. Adidas, with its viral Chinese-style jacket, is a German company.

The most trumpeted Chinese cultural products still don’t seem to be finding much purchase outside China. Ne Zha 2, often trumpeted as the highest-grossing animated film of all time, earned over 99% of its revenue in mainland China. Black Myth: Wukong, the most famous Chinese video game, got over three quarters of its Steam sales from China.1 Other than the rapper Skai Isyourgod, who has had several songs go viral on TikTok, there are not many Chinese musicians known in the West.

Instead, the “Chinamaxxing” trend seems to consist mostly of Western youngsters doing stuff they think of as stereotypically Chinese — drinking tea, doing exercises, etc. This is the kind of thing that might have gotten dinged as “cultural appropriation” eight or ten years ago. Today it’s more reasonably viewed as an expression of fascination and respect — but it’s fascination and respect from a great distance.

Then there are those videos of Chinese cities. I covered these in my post last year, but the trend is still going. There are also now a bunch of influencers who relentlessly post about how Chinese cities are the greatest. For example, there’s Jostein Hauge, an assistant professor at Cambridge who relentlessly posts about how China is ahead of the West in every regard. The alleged supremacy of China’s cities is a regular talking point:

Cynics have noted that these accounts are pretty one-note; it seems more like a deliberate publicity campaign, abetted by a few amateur enthusiasts, than an organic outpouring of enthusiasm for Chinese urbanism. The same is true of the continuing parade of breathless videos from Westerners traveling in Chinese cities — they tend to feature shots of the exact same grandiose train stations and architectural landmarks, or the insides of factories or restaurants or other buildings, rather than videos or photos of life at ground level.

That’s telling, because it stands in stark contrast to the videos and photos you tend to see from Tokyo, Paris, Hong Kong, or other popular older cities. And there’s a reason for that — as I wrote in my Sinofuturism post, Chinese cities were built incredibly quickly instead of growing organically over time. This means that they’re dominated by sterile gated tower blocks (called xiaoqu, or microdistricts), large surface streets, and huge shopping malls. There are relatively few walkable mixed-use streets lined with shops near to people’s homes. External shots of China’s newly built city centers tend to show vast concrete plazas and soaring towers — impressive, but fairly sterile.

In fact, there’s hard data to support the notion that the appeal of China’s megacities is still shallow. As of 2024, tourism to China was still way down from the years before the pandemic, and the number of Americans studying in China had collapsed even further:

Contrast this with Japan and Korea, which both get many more tourists from the U.S. than China does (despite being far smaller), and which have both seen a more complete rebound since the pandemic:

2025 numbers are harder to come by. Tourism to China is still recovering — up about 10% from 2024 — and American travelers are presumably part of this trend. But it’s still nothing compared to the tourism booms to Japan and South Korea, which are well above their pre-pandemic levels.

For all the breathless YouTube videos and glowing testimonials, Americans are still not going to China in large numbers, either to visit or to live.

So overall, the “Chinamaxxing” trend feels a bit fake and forced — the combination of a deliberate marketing campaign and social media influencers looking for a new niche. But there’s something else going on here as well — a statement about the declining appeal of America and the West.

Chinamaxxing is really about the decline of America

I think Fortune really puts its finger on something here:

The subtext of every “very Chinese era” video isn’t really about China. It’s about what young Americans feel they’ve been denied. Chinamaxxing romanticizes things that feel structurally out of reach at home — compact, affordable-looking apartments; public transit that works; streets safe to walk at night; multigenerational households as an antidote to loneliness; communal meals as an antidote to atomization. The comparison is implicit but unmissable: they have this, and we don’t…

Slate‘s Nitish Pahwa captured the emotional logic cleanly: “You told us we couldn’t have a high-speed railroad and universal health care, and it turns out they have it across the street! I’m going to live at their house now!”…Shaoyu Yuan, a scholar who studies Chinese soft power, told NPR the trend operates on two tracks at once: one that “weakens American narrative authority by highlighting content that highlights U.S. dysfunction,” and another that “makes China look more attractive.”…

The American century was built on the world’s desire to be American…The question the turbulent 2020s is forcing is a simpler and more unsettling one: what happens when the generation that was supposed to inherit the American promise looks around at their student loans, their rent, their medical bills, and their crumbling train stations — and decides they’d rather be something else?

And CNN gets it too:

[E]xperts say the [Chinamaxxing] trend reveals deeper undercurrents like dissatisfaction among many Americans with life at home – from political turmoil, gun violence, immigration crackdowns and persistent racial tensions. All this has dulled the veneer of the US, driving curiosity for American youths to see what life is like on the other side…[I]t’s no coincidence the trend comes amid a broader decline in the US’ global image…

[V]ideos showing vertiginous skylines from Chinese metropolises…have gone viral for depicting a futuristic vision of urban life, replete with seemingly clean streets and low levels of violent crime.

In other words, American youngsters idealizing China — without actually engaging with China or knowing much about it — is really about expressing their dissatisfaction with America.

Chinamaxxing is mostly just Americaminning.

That mirrors a larger global reaction. Donald Trump’s tariffs, threats against allies, and reckless wars have turned most of the world against America. Traditionally, confidence in U.S. leadership was higher than in Chinese leadership, but this has reversed since Trump’s election:

China isn’t especially popular itself, but in the age of Trump, it looks to some people like the only natural alternative:

In America, Trump’s extreme unpopularity among the youth is probably helping to drive the Chinamaxxing trend.

But it’s not just Trump and the GOP. Stories about Chinamaxxing consistently mention the safety, cleanliness, and low crime rates of Chinese cities as part of the country’s appeal. And pro-China influencers repeatedly trumpet this advantage, often showing pictures of China’s immaculate new developments with pictures of homeless encampments in America.

This is no Potemkin comparison. America really does have much worse crime and public order than other countries, including China:

There are a number of reasons for this, but progressive ideology takes much more responsibility than MAGA insanity. Blue cities’ tolerance of public homelessness and drug use, the “progressive prosecutor” movement’s permissive approach toward crime, and the consistent failure of progressive local governments to allow housing construction all contribute to the breakdown in public order that has left America’s flagship cities feeling dirty and unsafe.

It bears saying that despite the safer cities, the superiority of life in China over life in America is more myth than reality. Jacobin, the socialist magazine, recently published a good article by Daniel Cheng debunking the idea that China is a youth paradise:

Most people in China suffer from similar social and economic crises that afflict Americans today. The United States’ extreme income inequality is well-known, but China’s is comparable. After accounting for taxes and redistribution, China becomes even more unequal because it falls under the US’s (very low) standards for redistribution. While Chinese inequality has gradually shrunk over recent years, this is mostly due to compression between the top and middle of the income distribution. Those in the bottom 30 percent have been left in the lurch…

[W]hile American higher education is exorbitantly expensive, the education affordability crisis in China is even more severe. Parents have to pay for high school, and tutoring is a de facto necessity to keep up with demanding curriculums…The bottom quintile of Chinese families spend a massive 57 percent of household earnings on their children’s education…

[H]omelessness and extreme poverty are also major problems in China. Chinamaxxing influencers are simply blind to them because the government has successfully criminalized homelessness and driven the “low-end population” out of sight…Age discrimination in hiring is legal in China…D]ismissal rates rise dramatically after workers turn thirty-five…[A]n unexpected layoff can permanently condemn someone to underemployment in the gig economy.

And youth unemployment is far worse in China than in America, even after the government redefined the numbers a couple of years ago to make it look smaller:

This is probably why we don’t see a lot of American “Chinamaxxers” put their money where their mouth is. It’s a lot easier to put on a bathrobe and eat some dumplings and pretend to be a Chinese uncle on TikTok than it is to actually move to China and make a living there.

In fact, China’s leaders probably don’t care. They’re not especially interested in getting American Zoomers to move to Shenzhen or Shanghai. Their own publicity campaigns, including all the gloating over the parlous state of American cities and the constant parade of photos of fabulous new infrastructure, are probably aimed at Chinese scientists and engineers living abroad. And in fact, this campaign is succeeding to some degree, helped along by Trump’s anti-immigration jihad and progressives’ mismanagement of big cities:

For decades…Many of China’s best and brightest saw the U.S. as a land of boundless opportunity underpinned by robust rule of law…Today, America’s allure is fading. More elite Chinese youths, businesspeople and scientists are gravitating back home. Some who have returned say they are turned off not only by the U.S.’s hardening immigration enforcement, but also by its faulty infrastructure, gun violence and living costs. Back in China, many cities have grown cleaner and more livable in recent years, linked together by efficient subways and high-speed trains…

In 2021, more than 1,400 U.S.-trained Chinese scientists left American jobs for roles in China, a 22% jump from the previous year, according to a survey published by Asian American Scholar Forum, an advocacy group. Most China-born Ph.D. graduates are still choosing to stay in the U.S., with close to 80% saying they intended to remain in 2024, according to the most recent available survey data from the National Science Foundation. But high-profile departures have continued steadily…

Frequent changes in immigration rules, combined with homelessness and perceptions of high crime rates in some of the coastal cities where Chinese immigrants tend to live, are also leading people to reconsider the appeal of the American dream, according to Chinese people who have spent time in both countries.

American leaders should be a lot more worried about losing Chinese talent than about Gen Z “becoming Chinese”.

Some real green shoots of Chinese soft power

All that having been said, I do see a few glimmers of real, organic Chinese cultural appeal. One is the rise of the Chinese micro-drama or duanju. These are serial shows with scripted 1-2 minute episodes, shown in a vertical scrolling feed. It’s a truly new art form, perfect for the age of TikTok and AI. The Economist explainer posits that these dramas have flourished precisely because the flood of content is too large for China’s censors to monitor and eviscerate:

Artsy film critics are unlikely to be impressed by China’s micro-dramas. Even so, the roughly two-minute episodes, which cram soap-opera plots into a short-video format, are wildly popular. Watched almost exclusively on mobile devices, viewers can scroll mindlessly through episodes as they would clips on TikTok. Revenue in China from micro-dramas is projected to nearly double this year…Chen Ou, the founder of Jumei Film Base, a leading micro-drama studio in Zhengzhou, says his company is starting to monetise its star power with live-streaming sales…[N]early all large tech companies in China are snapping up rights to micro-dramas…Many local governments are investing in micro-drama studios…

[F]or micro-dramas, which are chock-full of the kinds of taboo topics and comedic violence that usually irk censors, industry insiders say the sheer volume of content has resulted in looser or fewer checks.

This reminds me of how manga and anime developed in Japan — it flew under the radar of the conservative oligopolies that dominated movies and TV in the postwar period, making it a haven for political radicals,2 sexual deviants, and artistic auteurs.

It’s still early days, but Chinese microdramas are starting to catch on in America. This is from Wikipedia:

ReelShort and DramaBox, the two largest Chinese short-drama platforms operating overseas, entered the U.S. market in 2022 and 2023, respectively. By August 2025, DramaBox had surpassed 100 million downloads on Google Play alone, while maintaining an average of 44 million monthly active users. Meanwhile, ReelShort surpassed 370 million downloads and raked in $700 million in revenue. By 2025, the U.S. had become the single largest revenue market outside China for vertical drama, generating approximately $58 million in monthly in-app revenue and an estimated $1.3 billion for the full year. As of 2025, ReelShort and DramaBox are the top two duanju platforms in terms of downloads and active users.

Retail is a second strong point. As China’s economy diversifies and consumption rises, some Chinese shops are also starting to make inroads into cities in America and around the world. Chagee, Heytea, Mixue, and Luckin Coffee are high-quality drink shops that seem to have real and immediate appeal (Chagee is my personal favorite).3 The stores Miniso and Popmart are appearing all over global malls, selling toys, collectibles, and various other knicknacks. Chinese fashion is starting to make inroads overseas as well.

Food and design are inherently apolitical, so it’s a lot easier for Chinese creativity to reach the world through these items than through movies, TV, or music.

A third bright spot is the city of Chongqing. Unlike the sterile, formulaic videos of Shenzhen, the videos of Chongqing’s urban canyons and cyberpunk streets feel authentic and exciting:

Even the videos complaining about the difficulty of commuting to work showcase an urban landscape so unique that it has captivated much of the world:

In fact, tourists are actually flocking to Chongqing, to see the “cyberpunk city”. One reason for Chongqing’s appeal is that unlike Shenzhen or other “Tier 1” cities, Chongqing has more “old streets” adjacent to the newly-built downtown areas, giving it some of the kind of mixed-use walkable density that cities like Hong Kong and Tokyo have. Personally, I’d love to spend some time in Chongqing, while Shenzhen looks like somewhere I’d only go in order to tour some factories and see some robots.

So in fact, I do see some real signs of China’s soft power growing organically — finding ways to flow around the walls of censorship and official marketing campaigns, exposing outsiders to a more real, raw, authentic China. It would be astonishing if a newly developed country of 1.4 billion people didn’t have plenty of natural, organic appeal. Now, despite the best efforts of the country’s masters, that appeal is starting to show itself.

Subscribe now

Share

Greg Ip of the WSJ is one of my favorite economics writers, and you should always read what he writes. But in a recent post about manufacturing, I think he gets the main narrative wrong. Greg writes that America is in the middle of a “manufacturing revival”, which his headline writer calls a “stealth manufacturing boom”:

You won’t hear this from either critics or fans of President Trump’s tariffs, but there’s a manufacturing revival going on…Critics have focused on the fact that factory jobs have steadily slid since Trump took office last year…Unlike jobs, though, actual factory output has risen briskly, and may even be picking up speed. This stealth recovery, though, isn’t because of tariffs. Instead, credit goes to the most basic economic force of all: demand. The U.S. is good at making things that happen to be in big demand right now.

As a macro story, “AI boom cancels out tariffs” isn’t a bad description of the U.S. economy right now — including the manufacturing sector. But it’s just not right to say that the former is winning out when it comes to manufacturing.

Let’s look at the data. Here’s Greg’s evidence for the boom:

First, a few data points. Since January 2025, manufacturing jobs have indeed fallen by about 100,000 workers, or roughly 0.6%. In the same period, though, manufacturing production rose 2.3%, and manufacturing shipments, unadjusted for inflation, climbed 4.2%.

Regarding manufacturing shipments…why wouldn’t you adjust for inflation? Inflation is important! Shipping more dollars of stuff doesn’t indicate a boom if a dollar is worth much less. As it happens, there’s no price series that exactly corresponds to the data series for manufacturing shipments, but we can probably approximate it by using the producer price index for manufacturing. Here’s what we get when we do that:

Do you see a “stealth manufacturing boom” since January 2025? I sure don’t. What I do see is the continuation of a decades-long stagnation in American manufacturing.

Let’s look at some other measures. Here’s industrial production in the manufacturing sector:

I guess if you squint very hard, you can see a slight rise since the end of 2024. But really this is just the same story as before: American manufacturing has been stagnating since 2008.

Let’s look at gross manufacturing output, adjusted for output prices:

Same exact story, only this is quarterly data and the last quarter of 2025 looks bad.

Read more

I’ve long been an industrial policy enthusiast. My favorite popular nonfiction book is Joe Studwell’s How Asia Works, a synthesis of decades of research about the economic miracles in Japan, South Korea, and Taiwan. I wrote a whole series of posts examining the successes and failures of various developing countries through the lens of Studwell’s ideas:

There are a bunch of other good books and papers about industrial policy that I’d recommend if you’re interested in the topic. These include Alice Amsden’s Asia’s Next Giant (about South Korea), Robert Wade’s Governing the Market (about Taiwan), “The New Economics of Industrial Policy” by Juhász et al. (2023), and the papers of the Industrial Policy Research Group.

Around the same time I discovered the industrial policy literature, the consensus was shifting within the big economic development agencies (the World Bank and the IMF). Whereas in previous decades, these organizations generally recommended against government meddling in the economy’s industrial structure, they’ve recently started to consider the kind of interventionist policies that Studwell recommends. In 2019, the IMF’s Reda Cherif and Fuad Hasanov wrote a paper called “The Return of the Policy That Shall Not Be Named: Principles of Industrial Policy”. They conclude that a Studwellian approach, if executed competently, can help a developing country grow faster than it would from just letting the market take its course:

We argue that the success of the Asian Miracles is based on three key principles that constitute “True Industrial Policy,” which we describe as Technology and Innovation Policy (TIP)…(i) state intervention to fix market failures that preclude the emergence of domestic producers in sophisticated industries early on, beyond the initial comparative advantage; (ii) export orientation, in contrast to the typical failed “industrial policy” of the 1960s–1970s, which was mostly import substitution industrialization (ISI); and (iii) the pursuit of fierce competition both abroad and domestically with strict accountability[.]

I was happy to see this shift — not because I’m certain that this sort of industrial policy is the secret to growth, but because I think it deserves to be in the discussion. So I’m also happy to see the World Bank now following suit, with a new report (or “book”) by Ana Margarida Fernandes and Tristan Reed entitled “Industrial Policy for Development: Approaches In the 21st Century”. The authors argue that although classic policy recommendations — macroeconomic stability, education, health, infrastructure, etc. — are still good, industrial policy can often help when layered on top of those basics.

I think it’s great to see the stigma about industrial policy going away. Not because this will lead to a wave of countries trying out such policies — that’s already happening — but because it’ll lead to more researchers taking the idea seriously. Dismissing the whole idea of industrial policy out of hand — as the World Bank and others did in the 1990s — is simply a policy of self-imposed ignorance. Countries need smart researchers to help them figure out which kind of industrial policies work and which don’t.

But as general interest in the topic has grown, my thoughts on industrial policy have also become more nuanced. As I’ve read more and written more about the idea, and as I’ve watched current events unfold, my thinking has evolved beyond “This is an important idea that deserves to be taken seriously”. So I thought I’d write a post briefly summarizing that evolution.

“Industrial policy” has become too broad of a category

One thing I always try to specify when I talk about “industrial policy” is that this term can mean a ton of different things. Most people think of it as government promotion of specific industries — autos, or electronics, or maybe just manufacturing in general. Others see export promotion — which is more about where products are sold than about which products they are — as the key industrial policy. Some people see FDI promotion as industrial policy; others don’t.

Even if we just focus on what you might call “classical” industrial policy — government promotion of specific industries — there’s a huge range of types of policies you might use. Protectionism — tariffs, import quotas, etc. — is often regarded as a tool for promoting manufacturing. That’s very different from export promotion. Direct government subsidies for favored industries are a common strategy — and one that’s on the rise throughout the world — but subsidies weren’t really used by many classic “industrial policy” success stories like Japan and Taiwan.

It’s kind of crazy that this huge diversity of policies and goals coexists under one single buzzword. It makes conversations about the topic difficult if not outright impossible. When people yell at me that “industrial policy is bad” or “industrial policy always fails”, I have no idea whether they’re talking about protectionism, or industrial subsidies, or government intervention in general.

If you read the IMF and World Bank papers on industrial policy, you can see that these distinctions really matter. The IMF paper explicitly contrasts export promotion with import substitution (protectionism), claiming that the former is very promising while the latter is usually bad. The World Bank report supports industrial parks and market-access assistance, while casting doubt on the effectiveness of subsidies and tariffs. In other words, even the people advocating industrial policy think that certain kinds are good and other kinds are bad.

In 2012 or even 2018 it made sense to talk about “industrial policy” as a single thing, because it basically just meant that researchers and policymakers should take a look at a bunch of different ideas that had been beyond the pale of orthodoxy in previous decades. But now that researchers and policymakers have actively started to look into those ideas — and to implement them on a large scale — it no longer makes sense to talk about “industrial policy”. We need to be more specific.

For developing countries, “just do FDI” looks like a viable strategy

In my series of posts on developing-country industrialization, I found a subset of countries that had clearly succeeded with a very simple, seemingly replicable formula: promoting FDI in manufacturing. I singled out Poland and Malaysia as countries that got rich in recent years simply by encouraging multinational companies to put their factories and research centers there:

Poland, especially, has succeeded amazingly using the FDI strategy. A lot of industrial policy enthusiasts — Ha-Joon Chang, for example — used to argue that developing countries should build their own domestic “national champions” instead of relying on foreign capital and know-how. That’s what Japan and Korea did, it’s true. But you’d probably be hard-pressed to name a Polish brand. And yet Poland’s economic performance since the end of communism has been absolutely stellar — it’s about to surpass Japan’s living standards, and is now even starting to catch up to Korea:

Interestingly, FDI was also central to the development strategies of Singapore and Ireland — two of the richest countries on the planet. You’d also be hard-pressed to name a Singaporean or Irish brand. And China’s approach before the early 2010s — during its fastest era of growth — centered much more on FDI than on subsidies or on the promotion of national champions in general.

When you look at poor countries that got rich since World War 2 by building national champions, the list is pretty short — there aren’t a lot of South Koreas out there. But the list of countries that got rich, or nearly rich, by promoting FDI is getting longer by the decade. So while I wouldn’t discount the Korea strategy, I’m leaning toward the idea that the Poland approach is a lot easier to get right.

Why would it be easier to get rich through FDI than by building your own brands? I can think of a couple of reasons. For one thing, FDI is less risky — instead of having the government pick winners, you let multinationals try building a bunch of things in your country. It’s a way to let the market discover comparative advantage, while the government simply assumes that some sort of competitive advantage exists within the broad category of export manufacturing.

FDI promotion also requires good institutions. If you’re trying to get German companies to build their factories in your country, you probably need to have the kind of property rights that German companies are used to dealing with. Poland became the workshop of Europe by forcing itself to shed its communist-era institutions and become more like the EU.

Note that the new World Bank report focuses on industrial parks as its favorite industrial policy. Industrial parks are a key part of the Poland/Malaysia/Singapore/Ireland strategy — a tool of FDI promotion. I predict that for developing countries, this approach will become more recognized as the closest thing we have to a universal push-button solution for getting out of poverty.

For developed countries, industrial policy is technology policy

Right now, much of the economic discussion in the U.S. is about AI — how to promote it, how to enable it, and how to regulate it.

This is classic industrial policy. It’s picking a winner! If you rewrite regulation to allow more construction of data centers, or if you try to recruit top AI researchers, or if you use export controls to prevent a competing company from seizing the initiative in AI, or if you do any special thing to promote the industry, you are picking AI as a winner. (The fact that almost every country is picking AI as a winner doesn’t change that fact — there was a time when every country thought it was essential to have its own auto industry.) And I don’t see a lot of free-market economists disagreeing with this pick.

Nor is this the first hot new technology that the American government has specifically encouraged within my lifetime. The Telecommunications Act of 1996 selectively deregulated the internet sector, because everyone agreed that the internet would be economically important. The National Science Foundation subsidized the internet’s initial buildout, as did the High-Performance Computing Act of 1991. State governments provided telecom companies with tons of subsidies to build out wireless networks, and so on.

We picked the internet as a winner, and it was a winner. Notably, very few of the free-market enthusiasts who criticized industrial policy in developing countries raised the alarm about the U.S. picking winners in the internet age. The industrial policies we used to pick that winner fell under the rubric of things that economists had already admitted that rich countries ought to be doing — infrastructure, deregulation, and R&D promotion.

But for rich countries, technology policy is industrial policy. The emergence of a major new technology puts a developed country in the position of a developing country in a narrow, limited sense. A poor nation lacks a car industry, an electronics industry, a machinery industry, a shipbuilding industry, and so on. America in 1985 lacked an internet industry, and America in 2022 lacked an AI industry, because those technologies had just been invented.

So I think rich countries actually have a lot to learn from developing countries when it comes to technological revolutions. Building something that no one has ever built before is a task that shares a lot in common with that of building something that your country has never built before. It’s not the exact same problem, but it’s related. So the people trying to figure out how to make America competitive in AI should study the South Korean Heavy and Chemical Industry initiative, or Taiwan’s promotion of TSMC, or METI’s promotion of Japan’s auto industry.

China’s approach has big flaws

In the last few years, China embarked on an unprecedented policy experiment. The Chinese government has subsidized high-tech manufacturing industries to a far greater degree than any other country in history. This has led to a boom in high-tech manufacturing, increases in China’s global market share in those industries, a huge surge in Chinese exports (known as the Second China Shock), and to the emergence of some Chinese national champions like BYD.

But we’re definitely starting to see the downsides of this experiment. First and foremost, paying dozens of companies to all make the same products ends up creating brutal price wars that compete profit margins toward zero:

Margin compression deprives companies of R&D budgets, which must then be substituted by government research. It also leads to deflation, which exacerbates bad debts, burdening households, corporations, and the financial system.

China’s leaders realize these issues. Cutting industrial subsidies will be politically difficult, especially because the country is still suffering low demand from the bursting of its real estate bubble. But they’re starting to do it — for example, the government is phasing out subsidies for trading in old cars, leading to a predictable plunge in new car sales.

According to the standard “export discipline” playbook — which Studwell articulated, and which the 2019 IMF paper on industrial policy endorsed — this is exactly what you’re supposed to do. A wave of subsidies for export manufacturing results in a Cambrian explosion of manufacturers; the brutal global market selects the best of these; the government withdraws subsidies and lets all the inferior manufacturers die, while the national champions live and flourish and experience healthier margins.

This may work for China, if subsidies can successfully be withdrawn. But it’ll leave behind a major problem: bank debt. By some estimates, the bulk of China’s unprecedented industrial subsidies are actually in the form of artificially cheap bank loans:

This means that if and when China forces most of its subsidy recipients into bankruptcy — just as Joe Studwell and the IMF say you’re supposed to do! — it’ll result in a huge wave of bad debts. Those bad debts will sit on the books of Chinese banks, right alongside the existing mountain of bad debts from the real estate bust.

If you believe that the Chinese state is unified, and that Chinese banks and the government are the exact same thing, and that the government simply directs borrowing without regard to profit and loss, then maybe you think bank balance sheets just don’t matter in the People’s Republic. But if you think bank managers in China have any discretionary power over lending — how much, or to which companies — then you have to think that having the bank’s books crammed with bad debts will have some kind of effect.

In particular, Chinese banks will be heavily incentivized to “evergreen” loans to zombie companies that they’ve already lent to. Those subsidized lifelines will delay the day of reckoning, allowing banks to pretend their balance sheets are healthier than they are, while diverting financing from younger, healthier companies. This is probably what happened in Japan after its bubble burst in the early 1990s.

The standard model of industrial policy — temporary export subsidies — imagines these as being provided at taxpayer expense. But if financial intermediaries are important — and most rapidly industrializing countries rely heavily on banks rather than on markets for financing — then it’s not so simple. A wave of corporate failures may be healthy for margins, but could cause years of low growth as banks are paralyzed with fear and tethered to zombie companies. And it’s not clear that state ownership and control of the banking system is an effective remedy, because even in a communist system, middle managers are still probably afraid for their personal careers (or their lives) if their institutions perform poorly.

This means we shouldn’t hail the Chinese industrial policy experiment as a success until we wait a few years. More generally, I think discussions of industrial policy tend to downplay the role of financial systems, and banking systems in particular. I plan to write a lot more about that soon.

Subscribe now

Share

The other day on X, leftist Twitch streamer Hasan Piker got into an argument with a commentator known as Swann Marcus. Marcus had scoffed at the notion of Piker trying to connect with blue-collar workers. In retaliation, Piker claimed that Marcus had written a “how to” manual about sex tourism in Asia:

As you can see, Community Notes quickly corrected Piker. The person who wrote the “how to” articles about sex tourism was actually a rightist influencer named Matt Forney. Apparently, some leftists had — intentionally or unintentionally — gotten Marcus mixed up with Forney because Marcus had made a documentary about Burmese missionaries. But Piker refused to delete his accusation against Marcus, even after being informed of his mistake.

Recently, a video resurfaced of Hasan Piker launching a profanity-laced tirade against a Vietnamese refugee named Bach Hac. The refugee complains of suffering under Vietnam’s communist regime. Piker responded by saying “Fuck you old lady. Shut the fuck up you stupid idiotic old lady. Suck my dick, old lady. God damn, Yo, fuck this refugee”. He then tells her to go back and live in “South Vietnam”. Piker later deleted the stream, but has never apologized.

During a recent speech at Yale, Hasan Piker declared that “The fall of the USSR was one of the greatest catastrophes of the 20th century.” This is an almost direct quote from Vladimir Putin, who said in 2005 that “The demise of the Soviet Union was the greatest geopolitical catastrophe of the century.” This would be news, of course, to the countries that fought to escape Soviet communist rule, and whose economies flourished after the USSR’s collapse.

Recently, Ezra Klein wrote a New York Times op-ed urging Democrats to open a dialogue with Hasan Piker instead of trying to freeze him out of the party. The Times gave Klein’s post the headline “Hasan Piker is not the Enemy”. On a podcast, Piker then declared that Hamas is “1000 times better than Israel”. The New York Times promptly changed the headline of Ezra Klein’s post:

This kind of behavior is par for the course for Piker. Jeremiah Johnson had a good roundup back in December:

Infinite Scroll

Democrats have their own extremist problem

For the last few weeks, I’ve been grappling with one of the worst colds/flus I’ve had in my life. During that stretch, I leaned on one of my guilty internet pleasures - watching livestreamers on Twitch. The content on these streams is rarely good, but it’s often comfortingly bad, like the terrible daytime television I used to watch when I stayed home si…

Read more

5 months ago · 428 likes · 196 comments · Jeremiah Johnson

Some excerpts:

When questioned about China’s lack of LGBT rights, Hasan said the country is ‘gay as hell’ and defended the CCP banning gay dating apps as a ‘privacy issue’…He went on state television to talk about how great China is, and dismissed criticism of the CCP as ‘rumors’ and ‘misunderstandings’ and ‘lies’ that he wanted to help correct…He’s downplayed the genocide in Xinjiang, calling the concentration camps there ‘re-education’ camps and claiming they’re all closed now.² He’s said that Chinese colonialism in Tibet was a good thing…

He’s defended the idea of socialist re-education programs explicitly. He wishes the USSR had won the Cold War, he’s cool with Hezbollah, he thinks the Houthis are awesome and he’s used his platform to give a voice to literal, actual terrorists. He defended Russia’s illegal annexation of Crimea, and while he doesn’t outright defend Russia’s 2022 invasion of Ukraine he sure does spend a lot of time blaming the American government for somehow starting the conflict. He said that America deserved 9/11. He repeats neo-Nazi talking points about the Holocaust. He promotes political violence.

It should be pretty clear at this point what kind of guy Hasan is. His ideology is standard leftist “campism” — the idea that America is bad, and that any country or group that opposes America is therefore good. His style is that of a typical “shock jock” radio host — he says extreme and vulgar things in order to get attention and excite his listeners. It’s basically the same shtick that Michael Savage used back in the 2000s, but with the right-wing politics swapped out for Cold War-era anti-Americanism.

And yet Democrats and progressives are starting to treat this radio shock jock as an important voice in their party. Here’s what Ezra Klein had to say in his NYT post:

[P]ick over Piker’s years of streaming, and you can find offensive things he’s said.“…Streamer has said offensive things” isn’t really a news story…The impulse to cut off those with whom we disagree reaches far beyond Piker…It sits at the heart of cancellation as a political tactic. It relies on a belief in the power of gatekeepers that might have been true in an earlier age but no longer reflects the way attention is earned and held. Tucker Carlson was ejected from Fox News and grew stronger on X and YouTube. Nick Fuentes was banned from major social media platforms and gathered strength in the shadows. Trump went from being banned by every major social media platform to retaking the presidency.

According to Ezra’s line of thought here, the Republican Party and mainstream conservative institutions like Fox News would be smart to embrace Tucker Carlson and Nick Fuentes — and therefore the Democrats and mainstream liberals would be smart to embrace Hasan Piker.

Let’s think through the implications of that line of reasoning. If the mainstream should always include extremists in the conversation — if gatekeeping is useless and counterproductive — then all you have to do in order to force extremist ideas into mainstream discourse is to grab some attention. If you get a Twitch stream or a podcast and you start screaming that the Holocaust was fake, or that the USSR was good, etc., and you manage to get a decently big audience by doing this, you should now have a say in how the country is run.

The obvious problem with this idea is that it creates a competitive market for extremism. If being more extreme and profane and outrageous than the next guy is what gets attention, and if attention is what gets you influence in the Democratic Party or the GOP, then there’s a huge incentive for would-be influencers to be as extreme and outrageous as possible. Everyone will just keep one-upping their competitors until all the right-wing commentators are Hitler fans and all the left-wing commentators are Stalin apologists.

One could argue that this is exactly what has happened on the right, with the ascent of Carlson,1 Fuentes, Candace Owens, and similar rightist extremists. The Heritage Foundation’s embrace of Tucker Carlson and Nick Fuentes last year was very similar to Ezra Klein’s embrace of Piker; Heritage declared that although they disagreed with the ideas of Carlson and Fuentes, those commentators were so popular that they had to be allowed inside the mainstream debate.

But there’s another, less obvious problem with the idea of mainstreaming popular extremists. In the internet age, the bar for what counts as “popular” has been dramatically lowered. In the 1990s, Rush Limbaugh had between 15 and 27 million weekly listeners for his radio talk show. Nowadays, Tucker’s shows get about 1 million listeners. The internet has fragmented audiences, so that even the most popular commentators get a lot less attention than they used to.

This means we lower the bar for who we think of as “popular”. Hasan Piker’s stream gets about 6.5 million hours of attention per week. That’s about 10% of the viewership of Fox News’ Sean Hannity, and about a third of CNN’s Anderson Cooper. But Hasan is considered far and away the biggest political streamer, because streamers who talk about politics a lot just tend not to be that popular. Podcast audiences are harder to compare, but if we assume that about half of podcast downloads eventually get listened to, then Hasan would probably be in the top 10 political podcasters in the U.S., but not in the top 5. Joe Rogan — who, as Ezra points out, is not consistently conservative, but who supported Trump in 2024 — has many times Hasan’s audience.

International audiences lower the bar even further; only about half of Hasan’s audience is American. Ezra Klein is ready to embrace Piker as an important voice within the Democratic coalition based on his popular appeal, but a significant fraction of that appeal is to audiences who can’t even vote in American elections.

On top of all that, Piker gets a boost because as a left-wing talk show host, he’s a bigger fish in a smaller pond. Liberals tend to read the news, while conservatives are more likely to watch or listen to it. This is why there are relatively few right-wing writers, so the ones who rise to the top of the heap tend to be of lower quality. This is also why most of the top political podcasters, radio hosts, and TV commentators are right-wing. And this is probably why Hasan Piker can become an important influencer in the Democratic Party even as he declares he wouldn’t vote for Gavin Newsom over JD Vance.

All these structural factors can help explain why a cruel, vicious man like Hasan Piker, who supports totalitarian governments, spreads blatant lies about his critics, advocates political violence, makes excuses for terrorists, and vilifies the Democratic Party, can manage to shock, shout, and bully his way into being respected by mainstream progressives like Ezra Klein.

But there’s another important factor here, which is the content of Piker’s message. Whereas the leftist shock jocks of the previous cycle — self-described “dirtbags” like Chapo Trap House — tended to focus on economic issues, Hasan focuses squarely on foreign policy. And his main foreign policy focus is opposition to Israel.

Anti-Zionism is still taboo within the Democratic Party establishment, because of the Palestine movement’s association with antisemitism. But as Israel has done more and more bad things, grassroots anti-Israel sentiment has spread on both sides of the political aisle. In his post about Piker, Ezra talks a lot about the importance of including anti-Israel voices in the Democratic conversation:

We are living through a rupture in both the meaning and the reality of Israel. A Gallup poll from February found, for the first time, that more Americans sympathized with the Palestinians than with the Israelis. Among Democrats, the gap was overwhelming, with 65 percent who sympathized more with the Palestinians and 17 percent with the Israelis. The difference, as I have argued, is largely generational: Older Americans still view the Israelis more sympathetically, but among Americans ages 18 to 34, 53 percent sided with the Palestinians and 23 percent with the Israelis. This is new. Before 2023, young people and Democrats were more likely to side with the Israelis.

This is not the result of an international psy-op or a profusion of memes. The Israel that young people know is not the Israel that older people remember. It responded to the savagery of Oct. 7 by flattening Gaza in a brutal campaign that killed at least 70,000 Gazans, taking control of more than half of the territory and herding Gazans — more than two million people — into the remainder. Life there remains hellish. Israel has made hopes for a two-state solution fanciful by slicing the West Bank up into Israeli settlements and abetting constant settler violence and keeping a boot on the throat of the Palestinian Authority. It has used the Iran war as an opportunity to launch an invasion of Lebanon, displacing more than a million people and announcing that as many as 600,000 won’t be allowed to return to their homes until Israel decides otherwise. The Knesset just voted to legalize hanging as a punishment for Palestinians who are convicted of killing Israelis in terrorist attacks…

Israel, as it is behaving today, and as it is constructing itself for tomorrow, is incompatible with any normal understanding of liberal values…Anti-Zionism is rising as a response to what Israel is doing.

Ezra is right about Israel’s plummeting popularity in America:

And Ezra doesn’t even mention the fact that Netanyahu helped convince Trump to launch the disastrous Iran War, which has resulted in high oil and gas prices. Israel hasn’t just violated human rights and international norms against territorial conquest — it has been a highly problematic ally for the U.S., and is quickly becoming an outright liability.

American public opinion is slowly but inexorably turning; Ezra sees this, and is getting out in front of the shift. To some degree, he’s using Hasan Piker’s popularity, such as it is, as an excuse to advocate for a deeper, substantive policy shift — a turn away from staunch, reflexive U.S. support for Israel.

I view this as a mistake. If mainstream liberals want to drop their support for Israel, they should just do it on the merits. They should not bring in a guy like Hasan Piker to do it for them, because then they have to accept all the baggage that Piker brings with him. Mainstreaming Piker means that Democrats have to take seriously the notion that the Soviet Union were the good guys in the Cold War, that China and Russia are the good guys in the world today, and that America itself is — and has always been — an Evil Empire.

That message is likely to resonate poorly with many voters, especially older ones who remember a time before Trump and before the War on Terror. Pride in America has fallen significantly since Trump came on the scene, but that doesn’t mean the solution is to tell Americans that their country is the Great Satan. I doubt that Democrats and Independents want to destroy the U.S.; I think they want to restore and redeem it. Piker’s message is inimical to that goal.

And mainstreaming Piker and his anti-American ideology will inevitably lead to a deterioration in the quality of the people the Democrats elect and appoint to high office. This has absolutely happened with the Republicans. In 2024, the MAGA movement embraced the idea that America is an Evil Empire, spreading woke values around the world, and that we should realign ourselves with Russia. This led to the appointment of Tulsi Gabbard as the Director of National Intelligence, the end of most American support for Ukraine, the right-wing turn against Europe, and to the tearing up of most of America’s alliances. It notably did not lead to fewer American wars; it just led to dumber, more evil wars.

Why should Democrats willingly walk down this same path? Do we really want the next Democratic administration to have staffers and appointees who think the Soviets should have won the Cold War? Are we prepared to realign America towards China, as Trump has realigned us toward Russia, and for the backlash this would generate?

Maybe so, but I hope not. Instead of embracing anti-American shock jocks like Hasan Piker, mainstream liberals should simply levy their own criticisms of Israel instead. You don’t have to believe America is evil and communist empires are virtuous in order to say that Israel has become crueler, more totalitarian, and less reliable as an ally. Those arguments are easy to make within the framework of liberalism, instead of by embracing someone who says he wants a “post-liberal America”.

I’ve sat here for years and watched the Republicans embrace their worst extremists. I’ve watched as those extremists turned the right away from mainstream conservatism, and drove them to embrace insane, self-destructive ideas. I don’t want to see the Democrats do the same. Maybe the incentives of the social media age are just too powerful, and every major party is destined to be forced down this road. But I say we should keep trying to resist the extremist impulse for as long as we can.

Subscribe now

Share

I’ve been meaning to write this post for a while, and now the AI revolution has given me an excuse.

I was standing in a Whole Foods on Long Island, sometime in the early 2010s, staring at a cheese counter, when I had a sudden revelation. All my life, I realized, I had been told that it was people’s work that gave them meaning — that what you produce makes you special. Few people say this explicitly, but it’s baked into many elements of our culture.

In San Francisco, when you meet someone, the first question you typically ask is: “What do you do?” Some people will give teasing answers — “I race boats”, or “I take care of rabbits” — but everyone knows that the question is about your job. Nor is SF particularly unusual; Americans tend to identify each other by their occupation. “This is Steve, he’s a professor,” and so on. In Japan, people are more likely to identify by the company they work for, but their identity is still fundamentally about production.

Or think about movies and TV shows. Yes, there are some stories about people whose hobbies become the most important things in their lives — High Fidelity, Shall We Dance?, Schultze Gets the Blues, and so on. But in most narratives, it’s people’s career that defines their life objective, their success or failure as a person, and their identity as a character.

Consumption, on the other hand, is typically trivialized, or even denigrated. Every culture has words like “lazy”, “shiftless”, or “playboy” to make fun of people who spend their time consuming instead of producing. A “hobby” is far less noble than a “calling” or a “vocation”.

In fact, the value of work over play is one of the few ideas that traditionally united the political left and right. In the 20th century, leftists decried the “consumer society” and called for “workers” to be in control of society. Conservatives, meanwhile, value hard work and complain that the welfare state makes people lazy, while rightists view consumer societies as decadent and weak. The “degrowth” movement is all about reducing Westerners’ so-called “overconsumption”; it’s hard not to hear a moralistic message in addition to the environmental one. Production is virtuous, consumption is wicked.

Why should what you produce, rather than what you consume, be the most important thing about you? Why shouldn’t the fact that you race boats or watch anime or drink matcha lattes be what defines your identity? Why should I call myself a “writer” rather than a “science fiction fan” or a “rabbit dad”? Just imagining introducing myself as the latter makes me cringe a little. But why?

One seemingly obvious answer is that the market values production over consumption. In fact, this is almost the definition of the two terms — work is what you get paid to do, while consumption is what you have to give up something in order to enjoy. But this doesn’t explain why culture and society should give additional accolades to production over consumption. You already get paid for going to work; why should you get praised for it too?

One cynical answer is that praising people for their work ethic is a way of trying to lower labor costs, by paying workers in status instead of money. Plenty of research in both economics and psychology indicates that people will accept lower salaries in exchange for working at a job where they think they’re doing something good for society; this helps explain why wages at nonprofits are so low. This might be why tech companies traditionally tell their young employees that finding new ways to sell ads is “making the world a better place”; it might allow them to pay less than hedge funds for the same class of talent. Professors, meanwhile, often forego lucrative careers in industry in exchange for the pride and status that comes from being an academic.

A less cynical answer is that in premodern times, most work wasn’t rewarded by the market; families and villages had to persuade, bully, or cajole people into plowing the fields or cooking dinner, so they “paid” people for productive effort with compliments instead of wages; this traditional culture may have carried over into the modern day.

And there are reasons to make production the center of your identity beyond the fact that society praises you for it. Your productive power represents a key point of leverage over society; the more the world needs you to produce stuff, the less likely you are to have to depend on the largesse of others. Pride in your productive power means pride in your independence.

But at the end of the day, it’s consumption, not production, that defines you as an individual.

That might sound like an odd thing to say, especially if you’ve grown up believing — as my liberal parents taught me — that advertising is a form of mind control that greedy corporations use to force you to consume things you don’t really need. But you don’t have to trick people into buying most of what they buy; people all over the world want modern conveniences like dishwashers and cars and AI chatbots, and it doesn’t take a lot of ads to convince them to buy those things. Without ads, people would still watch movies and listen to music and wear nice-looking clothes.

The truth is that merchants advertising their wares to you are begging you to buy their brand instead of their competitors’. Everyone wants your money; you are the one who gets to choose who gets it.

That choice is yours and yours alone. Every day that you exist as a consumer in a capitalist society, you are forced to make dozens of decisions about what to spend your money on. Should you buy coffee at Starbucks or Peet’s? Should you buy a new skirt or a new pair of jeans? Should you go watch a Marvel movie or an indie film? Should you subscribe to Noahpinion or to Slow Boring?1

Each time you make one of those choices, you are forced to interrogate your own preferences. You are forced to look inside your heart/mind/soul/utility function/whatever and decide which brand of coffee you want, which type of clothing you want, which blog you want to read, and so on. It’s all about you.

There’s some research demonstrating this effect. Here’s Cheek et al. (2022):

Across six studies (total N = 3,549), we find that participants who were randomly assigned to choose from larger assortments thought their choices were more self-expressive, an effect that emerged regardless of whether larger sets actually enabled participants to better satisfy their preferences.

And here’s Nanakdewa et al. (2021):

When people think of their actions as choices, they feel larger and stronger than others, are attracted to ideas of independence, and feel empowered to voice their opinions. Choosing what to eat and which shampoo to buy may seem like trivial acts, yet the current research finds that the salience of choice alone can have a range of powerful psychological effects.

Is every choice of product or fashion going to reveal some deep truth about you to yourself? Of course not. But consumption choices force you to develop the habit of self-examination. And when you think about more complex life choices — what kind of personality to present to the world, how to behave in your romantic relationships, how to express yourself through art or music — that habit will come in handy. In fact, economists would say that social interaction, romance, and self-expression are also forms of (non-market) consumption.

Production is very different. Your decision of what to produce is not fully your own; the market gets to decide. If what you really want to do all day is carve wood, but the market doesn’t pay a woodcarver a living wage, then you’ll have to find something else to do for money. Lawyers and software engineers and brain surgeons undoubtedly take pride in their careers, but the high salaries society pays for those occupations were undoubtedly a reason they went into those fields. Those salaries are a reflection of someone else’s preferences — someone else’s demand for legal services, software, and brain surgery.

What about jobs that involve self-expression, like art and music and writing? In fact, these are simply bundles of production and consumption. I love writing, and I’m lucky enough to get paid for it. But if I really buckled down and spent a lot of effort building my brand, writing what the audience wanted to hear, covering every breaking topic before other writers did, and generally treating this blog more like work, I could make a lot more money doing it. The lower income I accept in exchange for greater self-expression is actually a form of consumption. It’s no coincidence that artists who “sell out” tend to enjoy their craft less.

To sum up: When you decide what to consume, you ask: “What do I want?”. When you decide what to produce, you ask: “What do other people want me to do?”. The former is a lot more individuating than the latter.

We can observe this effect at the level of whole societies. The World Values Survey shows that richer countries tend to value “self-expression” more:

And in general, surveys consistently find that as countries grow richer, they become more individualistic. Japan often gets stereotyped as a collectivist society, but surveys show that hasn’t been true since the early 1980s — the first generation brought up in affluence became the “me generation”. America’s Boomers were undoubtedly similar.

We can argue about how much individualism versus how much collectivism is good at the societal level. But on a personal level, it seems clear to me that the standard story we grow up hearing — that your job is what makes you you, while what you consume is dictated by corporations — has it exactly backwards. In fact, consumption shapes you into a unique individual, while your job exists at the whim of the collective.

This is what I realized, staring at all that cheese.

As I said, this idea has been rattling around in my head for a while, but it’s the advent of AI that finally made me decide to write it up. A lot of people have been saying that by disrupting traditional careers and devaluing lots of human capital, AI is going to cause a crisis of meaning in our society.

I do agree that this will probably happen. If you’re an economist who prides himself on being able to do the difficult algebra required to turn a conjecture into a concrete theory, and now suddenly AI comes along and can do that at the touch of a button, you might lose some of the pride and sense of meaning that your previously rare skills had given you, even if tenure protects you from losing your income. If you were a software engineer who prided himself on being able to write good code, and now your job consists of checking the code written by Claude, you might feel less meaning in your job, even if your salary is higher.

But I don’t think this has to happen. I think it’s possible for us to reorient our identities away from what we produce, and toward what we consume, and to find meaning in the latter.

That might sound speculative, but in fact I’ve lived in a society where people’s sense of self and their social position was defined at least in part by what they consumed. It was called college. Although I worked part-time during my school years, many people didn’t — and even with my job, I still had plenty of time for leisure. I spent that time learning what music I liked, learning how to make better friends, learning about romance, and learning how to express myself better.

That was all consumption. Even my coursework contained an element of consumption — yes, studying physics built up my human capital, but it was also just fun, and my anthropology and film electives certainly weren’t about increasing my future earning power. Because I went to an elite school, I was able to have some confidence that the signaling component of my school’s pedigree, along with the human networks I built up, would save me from the risk of poverty. So I could spend college having fun and expressing myself — and my fellow students, who mostly didn’t have jobs and lived entirely on their parents’ largesse, could do so even more.

If your reaction to that is “Well good for you, jerk,” I don’t blame you. The American university system is not fair, and I was very lucky to land where I did.2 But imagine a society where everyone could have a college experience like mine — a time of self-discovery and self-expression, where work was done for enrichment instead of for money. And now imagine a society in which people could keep living that college life far past the age of 22.

Doesn’t that sound a little like paradise? Well, perhaps with AI, we can make that a reality. If we create robust institutions to redistribute the material gains from the intelligence explosion, perhaps we can create a society where all of life looks like an elite American school — where people spend their day reading interesting books, doing math because it’s fun, going to parties, making cool outfits, learning about their friends, playing in amateur bands, or having long drunken conversations about the meaning of life until 4 AM without worrying about going to work in the morning.

That’s a vision of a consumption society, but not one that’s meaningless or empty. Instead, it’s a vision of technology freeing us to become more like ourselves. I don’t think this happy outcome is inevitable, but I think it’s getting ignored in most of the discussions about our future.

Subscribe now

Share

Last year, a lot of people (including me) were wondering if the AI industry was in a bubble. These days it’s looking a lot less likely. The technology has found its killer app — agentic coding, which has upended the software industry as we know it. For power users, AI is no longer just a chatbot — you can tell it to go make you an app, or run some data analysis, and it’ll just do it for you and come back with the results.

This is making a LOT of money. As I predicted, Anthropic has been quicker to capitalize on the agentic coding boom than OpenAI. Anthropic focused on selling to businesses, while OpenAI focused on building its brand and selling to consumers; the revenue from agentic coding is almost all in the former category. So as Ruben Dominguez reports, Anthropic has probably overtaken OpenAI in revenue, or will do so soon:

In case you don’t realize how much money this is, or how fast this growth rate is, here’s some perspective:

Some of that will be eaten up by computing costs, of course. But as the WSJ recently reported, Anthropic’s computing costs are much lower than OpenAI’s. As a result, it’s expected to start turning a profit faster than OpenAI — and even OpenAI’s projections depend heavily on a comeback push that eats into Anthropic’s enterprise market share.

The rise of coding agents isn’t just changing the corporate horse race; it’s changing the whole picture of how we think about competition and profit in the world’s most important new industry. In a post last December, I wondered if AI would end up being a vitally important but low-margin business, similar to solar power or airlines:

Jason Furman wrote something similar, declaring that “instead of consolidating, as so many other industries have done, the leading edge of A.I. has become fiercely competitive.”

That’s still possible, of course. Fast followers, including Google and various Chinese model-makers, are still racing to catch up; if progress slows down, they may catch the market leaders and drive down margins. It’s still not clear how much of a “moat” AI has, even with agents. But right now, the business of making and renting out AI models seems dominated by two giants. Meta and xAI, who recently were considered at or near the frontier, seem unable to keep up.

And there’s now a pretty clear path for those two giants to become even more dominant: cybersecurity. Anthropic recently delayed the wide release of its new frontier model, Mythos, because it was too good at hacking. The model supposedly found critical vulnerabilities in key software systems that had been missed for decades by top human cybersecurity researchers. The idea is that Anthropic is going to spend a while using Mythos to go over critical systems and make sure they don’t have security flaws before releasing the model to the public. OpenAI is expected to do something similar with its next model.

Assuming Mythos is really that good at hacking (and there are skeptics), it gives us another reason to think that a few top model-makers like Anthropic and OpenAI will make a lot of profit. Cybersecurity is inherently adversarial; if attackers use a very powerful AI coding model to hack, defenders probably have to use a model that’s equally good or better to defend — and vice versa. This can lead to an arms race where neither side can afford not to shell out big bucks for the latest and greatest model they can get their hands on.

Because the prize for successfully defeating modern cybersecurity is so large — imagine hacking into Citibank and Bank of America and E*TRADE and Robinhood and just taking everyone’s money — the amounts that people have to spend on AI tools is potentially enormous. And even if Anthropic and OpenAI continue to be responsible citizens and make their top models available to defenders for long enough to find all the newly findable bugs — and even if attackers give up entirely because they can’t get their hands on the best models — it means defenders still have to shell out big bucks to the top model-makers.

It’s a huge source of revenue and a powerful moat for profit margins. And as AI expands into other adversarial fields — quant trading, litigation, fraud prevention, competitive advertising, and so on — there are probably going to be more of these revenue sources and more of these profit moats.

Which means we have one more thing to worry about when it comes to AI.

Typically, there are three big concerns that we talk about:

1. The worry that terrorists will use AI to create doomsday viruses

2. Worries about job displacement, human obsolescence, and economic dislocation

3. The worry that superintelligent AI is a new dominant species that will disempower and possibly destroy humanity

But if the industry really does become dominated by a few giant companies, we have a fourth big thing to worry about — extreme inequality. If AI’s economic benefits are highly concentrated, we could end up with a comparatively small number of people controlling most of the purchasing power in our economy. In the extreme scenario, this could lead to a small number of people holding all the power in the world.

The “Piketty on steroids” scenario

Read more

Patrick Collison’s YIMBY credentials are unimpeachable. He is a major backer of California YIMBY, the organization that has passed a stunning array of pro-housing bills in one of the most anti-development states in the nation. So it was interesting to see him claim that the movement has made a big mistake — or even been downright dishonest — by ignoring the aesthetics of apartment buildings:

For reference, here’s Sejong City in Korea, whose residential districts do indeed look rather bland and oppressive:

Some urbanists agreed, calling for regulatory reform that would allow American apartment buildings to look like the famous Haussmann buildings in Paris (depicted at the top of this post). So did some conservatives, which is unsurprising; intellectual conservatism has always called for a return to classical architecture and a rejection of modern styles. In fact, the idea that ugly building styles are a key reason that Americans disapprove of housing construction has been around quite a while, and it even has a name — “QIMBY”, meaning “quality in my back yard”.

Chris Elmendorf protested Patrick’s framing, arguing that YIMBYs have been active in pushing for reforms that would allow more beautiful buildings to be built in America:

YIMBYs have been pushing for single-stair reforms that would allow more "Paris-like" buildings…The municipal design standards & reviews that YIMBY laws allow developers to bypass did not improve designs. Per [Arthur] Stamps's studies (the only relevant empirical evidence of which I'm aware), they made things worse…[T]he problem of housing aesthetics deserves more attention -- and is receiving more attention -- but it's not like YIMBYs broke something that was working.

Elmendorf also pointed out that California YIMBY itself recently came out with a plan to encourage the building of more beautiful multifamily housing. The plan reads like exactly the kind of thing that Patrick might like:

[T]here’s a missing piece that housing policy still treats like an afterthought: how buildings look, function, and feel…Our current objective design standard paradigm…assumes you can “design away” ugliness by chopping a façade into smaller pieces…so the building feels “less big.” But contextual-design research shows why this keeps disappointing…When the underlying form and materials feel cheap or incoherent, extra façade break-ups read as fussiness, not beauty…

Many local Objective Design Standard codes demand heavy articulation and multiple cladding changes. The evidence suggests those moves have limited payoff compared to coherent style, material quality cues, greenery, and visible detail. (Stamps 2014; Nasar & Stamps 2008)…[We should u]pdate the California Department of Housing and Community Development’s model Objective Design Standards to [allow] projects [to] use a simpler envelope and meet a measurable threshold of real ornament (projections/recesses, columns/bands/cornices/fins, tile or relief work, murals), with minimum depth and material standards…

If California wants more European-feeling mid-rise development with courtyards, better daylight, shade, and balconies, it has to keep modernizing the [building] code…Too many building, electrical, and fire rules (in California and across the U.S.) [forbid] the buildings people actually like: bright cross-ventilated homes, true courtyard buildings, and mixed-use ground floors. All these requirements – egress, stairs, corridor, and elevator – often make projects bulkier and require much bigger lots, limiting where we can build new housing…[T]he web of building code regulations denies light, proportion, street connections, courtyards, greenspace – everything that makes buildings feel humane…Passing single-stair reforms and elevator reforms makes smaller mid-rise buildings possible, which fit on smaller lots, can be nestled into existing buildings, add variety to the streetscape, and reduce the pressure for larger, monotonous developments.

So at least one prominent YIMBY organization — the one that Patrick supports — is already answering the call to focus on building aesthetics. Others are likely to follow.

I think that’s a good thing. Eliminating onerous building codes and regulations will kill two birds with one stone, making it easier to build housing even as it also makes it possible to build more of the European-style ornamentation that commentators always call for. And allowing American developers to experiment with ornamentation and alternative styles will help break up the sameness of an urban landscape dominated by endless forests of boxy 5-over-1 buildings.

But that said, I highly doubt that this — or any stylistic change — would move the needle on public acceptance of new apartment buildings.

First of all, I’m skeptical that regular Americans actually like the kinds of building styles that intellectuals often yearn for. If you plunk down old-looking European-style buildings in the middle of Houston or Seattle, people tend to ridicule them as cheesy and inauthentic. The typical insult is “pastiche”, a derogatory term for a style that jumbles and mixes old European styles (even though, as Samuel Hughes points out, mixing and matching older ideas is exactly how classic European building styles were created in the first place).

Many local design standards explicitly discourage old-style buildings. For example, Los Angeles’ planning department, in its design guide for Echo Park, writes: “Do not imitate historic architectural styles; a modern interpretation may be appropriate if architectural features are borrowed and replicated to a simpler form.”

Nor is it just old European-looking buildings that leave many Americans cold. Pietrzak and Mendelberg (2025) find that although people tend to dislike tall buildings, traditional brick facades fail to move the needle on support for housing. Alex Armlovich points out that when New York City came out with new limestone skyscrapers, only three were permitted. And Brooklyn Tower, a recently built art deco style skyscraper in Brooklyn, has drawn tons of criticism for its style.

And Elmendorf cautions that no one has yet managed to find a specific architectural style that Americans like enough to move the needle on their support for new housing:

While the paper by [Broockman, Elmendorf, and Kalla (2026)] provides pretty good evidence that ordinary people’s aesthetic objections to bad, very unfit-to-context buildings affect their support for development (to the extent they care about anything development-related)…no one has shown that any specific set [of] design standards would materially improve public support for development, apart from pretty obvious stuff like "don't put up new buildings in low-density areas that are much taller than their neighbors").

All this suggests that while some American intellectuals may pine for the cornices and mascarons of Haussmannian Paris, most Americans just think that style — and any old style — looks cheesy when it’s transplanted to an American context. This may be because Americans consciously think of their culture as a young one, more suited to modern styles than traditional ones. Or it may be because America’s artistic culture has always focused on critique and fault-finding. But whatever it is, it suggests that allowing — or even forcing — cities to build ornamented buildings will not garner a wave of popular support for new development.

Conversely, the places that do build a lot of housing tend not to build it in old, ornate European styles. Texas, which is one of the best states when it comes to building new housing, mostly constructs single-family homes with lawns. When it does build apartment buildings, they tend to look like this:

Texas builds them anyway, for much the same reason that the Koreans built Sejong City — they’re cheap and efficient, and the state needs them to support its rapid population growth.1 You do see a little experimentation with slightly more European-style apartments in a few places, but overall it’s just boxy and functional. The fundamental driver of housing abundance in Texas isn’t architectural beauty; it’s a culture and politics that values and seeks out economic growth.

Nor is ornamental architecture necessarily what makes people love a city. Traditionalists may sigh over old European styles, and urbanists may salivate over the superilles of Barcelona, but the city that has captured the hearts of Americans in recent years is Tokyo. Downtown Tokyo is a forest of electric lights, strung up along the sides of stubby concrete mid-rises called zakkyo buildings. There’s nary a fancy cornice to be found; instead, the beauty comes from the bright cheery emblems of commerce:

Tokyo’s residential neighborhoods have even less ornamentation. They often feature flat brown or white or tan facades, hanging power lines, and bare asphalt streets with no setbacks or lawns or even trees:

And yet these are absolutely enchanting places to live. Why? Not because of the architecture, but because of the design of the city itself. The small curving streets make perfect walking paths, undisturbed by zooming traffic. Mixed-use zoning gives the neighborhood a communal, lived-in feel. Plentiful public transit makes it easy and stress-free to get around, while Japan’s peerless public safety makes it fun to hang out on the street or in a park at any hour.

Americans who go to Japan have definitely noticed this:

It’s no coincidence, I think, that Japan is one of the best countries when it comes to building plenty of housing. Yes, most of its apartment buildings look like crap when evaluated in isolation on their pure architectural merits. But the urban system made up by those buildings is a wonderful place to live, and so Japanese people have few qualms about building up that system. And Americans go there and love it.

And if America built a bunch of Haussmann buildings instead of boxy 5-over-1s, it would probably only marginally improve the feel of the country’s cities. Imagine Haussmanns in place of 5-over-1s in a typical Texas apartment complex:

Or imagine Haussmanns along a giant American stroad instead of a cute walkable Paris street near a train station:

These renderings don’t look terrible; the buildings look fine. But they don’t make the city that much more appealing of a place to live, because it’s still built in the American way — there aren’t any shops, it’s all based around driving, and it doesn’t feel cozy or lived-in. At best it’s a marginal improvement.

If you want American cities to look and feel so nice that Americans are willing to build housing in them, I think you have to do a lot more than give the buildings fancy facades. You have to do the hard work of putting in train lines, making side streets safe for pedestrians, rezoning for mixed use, and — perhaps most important — policing cities in order to ensure robust public safety.

That’s a tall order, and I recognize that this total urban transformation isn’t going to happen soon — or happen all at once. Instead, I think America really has no choice but to build up its cities organically:

• Implement hyperlocal control to allow neighborhoods that want to build more housing to do so as they see fit, thus circumventing the veto of city-level NIMBYs.

• Build more fast commuter rails between inner-ring suburbs and city centers, and more subways and elevated trains in city centers.

• Improve public safety through a combination of policing, community outreach efforts, better public services, and mandatory institutionalization for the dangerously ill.

• Use state-level upzoning where possible to allow “missing middle” housing everywhere — duplexes, triplexes, townhouses, and small apartment buildings.

• Simplify zoning at the state level along the Japanese model — have a few standardized zoning categories, and define them based on what kinds of nuisances they disallow, rather than what kind of buildings they explicitly allow. Make most zones mixed-use to some degree; most residential neighborhoods can benefit from neighborhood cafes and small stores.

• Carry out sensible reforms like allowing single-stair buildings.

Over several decades, this gradual process will allow American cities to evolve into a better form. That will increase political support for denser housing. And when paired with sensible reforms like the one put forward by California YIMBY, it will allow American cities to develop their own local architectural styles over time. Ultimately, that will be cooler and more interesting than simply borrowing from old Europe.

Subscribe now

Share

The immigration issue in America isn’t going away. Thanks to Trump’s crackdown, immigration to the U.S. went into reverse in 2025, with more people leaving (voluntarily or involuntarily) than entering the country:

But just like a century ago, shutting the gates isn’t the end of the discussion. The argument has shifted from who gets in to America to who belongs here in the first place.

To much of the MAGA right, the answer appears to be that only people of European heritage can become true Americans. For example, here is how right-wing commentator Matt Walsh responded to news about some crimes by some Texan teens:

Anyone who thinks these aren’t Texan names isn’t very familiar with the history of Texas; the Tejanos (Mexican Texans) were there from the beginning, and were a core part of the Texas Revolution. Most Mexican Texans today aren’t descended from the original Tejanos, but from more recent immigrants. But the fact that the Tejanos were there from the start is probably why Hispanics, and Mexicans in particular, have always been deeply integrated into Texan culture. It was at the behest of Texan businessmen that America didn’t put any cap on Mexican immigration in 1924, when it passed a law effectively barring immigration from most other countries.

Matt Walsh is unaware of most of that; to him, anyone without an Anglo-sounding name is presumptively non-American. This leaves little doubt as to what Walsh views as the marker of true American-ness. It’s likely that many others in the MAGA movement feel similarly, even if many would feel uncomfortable stating it out loud in simple terms. Anti-Indian sentiment has also risen to prominence on the right.

And many in the MAGA movement view Muslim immigration as an invasion, bent on imposing Sharia law on Westerners. They believe this “invasion” has already overtaken Europe, which explains their antipathy toward the EU and NATO. A “Sharia Free Caucus” is growing in popularity in Congress, and Ron DeSantis has signed anti-Sharia legislation in Florida. Various Republican politicians have explicitly stated that Muslims don’t belong in America.

If you’re Hispanic, Muslim, or Indian, there’s just not much you can do about this. In the past, showing that you were a good American — waving the flag, joining the army, speaking perfect English, and so on — was good enough to reassure most conservatives that you weren’t an invader bent on overthrowing America’s culture and replacing it with something alien. Nowadays, that’s not enough.

So perhaps it’s unsurprising that some nonwhite Americans are choosing to simply throw in the towel and reject the whole notion of assimilation. This is the essence of Shadi Hamid’s article in the Washington Post yesterday. He writes:

The assimilation defense — look how well we’ve integrated — is satisfying to make. But it concedes a premise I no longer accept: that a minority community’s right to be in the United States depends on its willingness to converge with the cultural mainstream. It shouldn’t depend on that. It shouldn’t depend on anything.

Whereas in the past, Hamid saw assimilation as synonymous with patriotism, now he sees it as a requirement to give up the religion of Islam itself:

The country is becoming less religious. Muslims, by and large, are not…This is a community that has increasingly integrated into American civic life, but it has done so while holding on to its religious commitments in a way that most other groups haven’t. Whether you think that's admirable or worrying probably says more about you than it does about them. The question I keep returning to is: Why do Muslims need to be like everyone else?…[A]ssimilation tends to mean secularization.

Whether Hamid is right that “assimilation tends to mean secularization” is an open question. Assimilation certainly didn’t require Catholic or Jewish Americans to give up their religion when they immigrated en masse in the 19th and early 20th centuries. Religious liberty is a fundamental part of the Constitution and of American tradition. On the other hand, even some immigration advocates do use conversion away from Islam as a measure of assimilation, and a growing number of Republicans — heavily influenced by their view of events in Europe — sees the religion as incompatible with American-ness.

Hamid is no blue-haired progressive — in fact, he’s explicitly anti-woke and fairly conservative. But his call to reject assimilation will be music to the ears of progressives, who have loudly and vehemently rejected assimilation for many years. A recent example of this is Bianca Mabute-Louie, whose new book Unassimilable: An Asian Diasporic Manifesto for the Twenty-First Century is a call for Asian Americans to resist assimilation by building communities and culture apart from White Americans. In a recent interview, NPR’s Alisa Chang gently pushed back on Mabute-Louie’s idea:

I want to understand what does orienting ourselves towards each other mean? Like, who is the each other? Like, my lingering thought, Bianca, is I still do want to belong here in America. And to me, belonging in America is not only shaped by whiteness, but it's also shaped by colliding and mixing with all the cultures that make America, not just white cultures. And I have trouble picturing being both Asian and American outside of that collision and mixing, you know?

Mabute-Louie’s response is interesting:

[T]he book isn't an argument to be isolationist…[O]ne example of how I'm trying to pursue that…in the South…is joining political community, joining mutual aid organizations with people who are most impacted. And I'm not really thinking about if they're Asian or not Asian. I'm just thinking about who's impacted when the hurricane comes. Who am I going to call? I always make the joke - who's going to be on my compound when the apocalypse comes because that's who I'm building community with, and that's what it means for me to be unassimilable.

Mabute-Louie’s idea of anti-assimilationism is not a call to interact only with Asian people — it’s to form political alliances with other people that she sees as being threatened in America at the current moment. It’s a vision of a country fracturing along racial, ethnic, and religious lines; Mabute-Louie is mentally preparing to fight a racial conflict, and she sees the “American” side, defined as hegemonic White culture, as her enemy.

This is different than classic progressive multiculturalism — though it clearly grew out of that idea. This is racial balkanization. The fact that anti-woke writers like Shadi Hamid are now leaning into the anti-assimilation line suggests that it’s now mostly a defensive response against Trumpism and the heavily racialized anti-immigration purge. Whereas ten or twenty years ago, “assimilation” meant waving a flag and speaking English and so on, to many it now means accepting that America is a fundamentally European nation and that nonwhite Americans are permanent guests in that nation.

In fact, this is pretty much what many children of recent immigrants did in the early 20th century, after the anti-immigrant backlash. German Americans were pressured into changing their names, giving up their ancestral traditions, and listening to long, patronizing lectures from volunteer citizens’ groups. Japanese Americans were interned en masse in World War 2. FDR reportedly once told his Jewish and Catholic advisers that "You know this is a Protestant country, and the Catholics and Jews are here under sufferance." For decades, Americans who didn’t come from the old North European Protestant stock felt they had to walk on eggshells.

That’s not going to happen again. Whatever Bianca Mabute-Louie might think, White American culture is not a monolith — in fact, it’s deeply politically and culturally fractured. MAGA will have neither the cultural power nor the enduring political power required to make European heritage the defining characteristic of American-ness. The country will break apart before it accedes to the likes of Matt Walsh or Tucker Carlson as the arbiters of true American-ness.

It’s probably a good thing that forced assimilation, of the type used in the early 20th century, is off the table. I say “probably” because 20th century America is arguably the most spectacularly successful story of integration and multiculturalism in modern history; some will inevitably claim that the cruel, bullying tactics that the old Protestant majority used on German, Japanese, Italian, Jewish, Polish, and other immigrants were necessary to that success. I reject that idea; I think that those bullying tactics were overkill, and probably led to lingering resentments.

But even though early-20th-century-style forced assimilation is off the menu, America still needs some sort of assimilation. A multicultural nation can’t survive as a “salad bowl”, where each group of people maintains its distinctiveness over time. (Canadians, who are fond of the salad bowl metaphor, are probably in for a rough time.) There is no “separate but equal” when it comes to cultures within a nation; if they remain forever separate, they will inevitably be unequal. More pragmatically, nations without cultural unity have difficulty providing public goods; politics tends to break down into an ethnic spoils system instead of being run for the benefit of the masses.

What America thus needs is a melting pot — or if you’d prefer a less metallurgical metaphor, a stew. Immigrants and their children should not be required to forsake every symbol of the old world, abandon their religion, or forget their heritage. But over time, the boundaries between America’s initially distinct cultures should blur. Intermarriage, interethnic business partnerships, and interethnic friendships should gradually erode the physical borders of the old blocs, while modern American culture — Netflix shows, pop musicians, and so on — should provide shared experiences and touchstones to bring Americans together without regard to ancestry.

This gentler assimilation has been happening my entire life. In a post last September, I wrote about what it looks like on the ground:

[M]any also value American culture as a marker of shared nationhood.

When I was growing up in Texas, one of my best friends was born in Shanghai, and didn’t become a U.S. citizen until the age of 18. Culturally, he was a little different than me and the rest of my friends — his mom made dumplings instead of sandwiches, he taught me how to use chopsticks, he didn’t believe in God.

But in all the cultural ways that mattered to us, we were the same. We watched the same TV shows, played the same video games, and listened to the same music. We used the same slang, had the same attitudes toward school, and wanted pretty much the same things for our future. And yes, we believed in the Constitution, and American freedoms, and all of that stuff.

During the 2010s, during our nation’s great…collective freakout over race, I wrote to my friend and asked him if he had ever felt discrimination growing up, or if he had ever felt excluded from the majority. He responded that while once in a great while he faced a little racism from a few jerks, it didn’t dominate his experience. In terms of identity, he told me he just felt very American.

This kind of real, on-the-ground cultural affinity is something too nebulous for YouGov pollsters to ask about, and yet I suspect it’s deeper and more important than most of the more quantifiable markers of American-ness. America is a propositional nation to some extent, but we’re also a cultural nation, bound together by shared habits and attitudes and lifestyles and beliefs. What matters the most isn’t our family’s history in the country, but our own personal history. Shared life experience beats shared heritage in terms of building the bonds of nationhood.

This is what Tomas Jimenez writes about in The Other Side of Assimilation, in which he argues that immigrant cultures will gently add their distinctiveness to mainstream American culture instead of being erased. And it’s what Richard Alba writes about in The Great Demographic Illusion, in which he predicts the gradual melding of America’s disparate groups into a unified “mainstream”. Before the Trump years, it looked like this was working well.

And I believe it was working well. I do not believe that this form of assimilation was too gentle and tolerant. I do not believe that concentration camps and forced name-changes and ethnic slurs and “100 percent American” movements sending volunteers into immigrants’ living rooms would have averted the coming of the MAGA movement. I believe that the MAGA movement is simply one of America’s periodic nativist backlashes, like the Know-Nothings in the 1850s or the restrictionists of the 1910s. It would have come anyway; it always comes back, and we just have to deal with it again.

What we must not do, I believe, is react to the MAGA movement by throwing out the notion of a unified and unifying American culture. We must not retreat to enclaves, online or physical, and view large swathes of the country as our enemies. Instead, we have to recommit to commonality.

This will be hard, but it won’t be impossible. Studies consistently show that Americans are less polarized on the issues than the media tells us we are. As recently as the 2000s, red and blue America were essentially culturally unified as well; though this might be changing, a lot of commonality remains. The online realm pushes us to hate and fear the outgroup, and to identify more with our distant co-ethnics than our real, physical neighbors. But the pull of the real world is still strong, and we’re starting to spend less time on social media.

Assimilation — which is really just another way of saying integration — won’t always be the picture of tolerance. Building a shared culture requires changes from everyone. Yes, some Muslim Americans will need to make sacrifices — they may have to look at cartoons of the Prophet Muhammad, or eat at school cafeterias where pork is on the menu, or hear bigots defame their religion. America is not Europe; freedom of speech, and the separation of church and state, are part of our core values as a nation, and these should not change.

But at the same time, non-Muslim Americans have to get used to seeing mosques on their streets without thinking they’re being invaded. They’ve got to get used to the idea that Islam is just one more religion in America’s mosaic of faiths and practices, and that Muslim Americans are every bit as American as Baptists. Some people will inevitably convert away from Islam, but others will convert to Islam, and this is fine; this is how freedom of religion works in a free society.

And yes, assimilation will involve the eventual loss of old cultural traditions as the generations go on. People will start eating more American food. Some will become secularized. Essentially all will forget how to speak their ancestral language. These processes are happening even faster with recent waves of immigration than they happened a hundred years ago. It’s a normal healthy process, and everyone should accept it; it’s part of the deal when you move to America.

Most of all, we all need to get over the idea that America is on the precipice of a race war or a religious war. Online activists might dream of that, but they’re small in number — and a lot of them aren’t even Americans, but foreign trolls for whom American politics is a fun outlet for their hatred and boredom. Most actual Americans just want to get along with our neighbors and live our lives together.

Ultimately, that’s all assimilation is — living our lives together until we become one people. It happened before, and if we want it, it can happen again.

Subscribe now

Share

I hate to say “I told you so” — not because saying “I told you so” is unseemly, but because the fact that I have to say it means I’m probably living in a world where things have gone badly.

I didn’t want to live in a world where gasoline costs over $4 a gallon. I didn’t want to live in a world where America tore up nearly all of its long-standing alliances and threatened to invade and conquer parts of Europe. I didn’t want to live in a world where China is viewed more favorably than the U.S. I didn’t want to live in a world in which the President of the United States posts things like this to his social media account:

I didn’t want to live in this world, but my countrymen forced me to live in it. I wrote many, many posts urging people to vote for Kamala Harris, despite all her shortcomings. They did not. And now I have to live with the consequences of my failure, and the failure of my fellow-travelers, to persuade the American people to avoid shooting themselves in the foot back in November 2024.

Whatever smugness I get from being able to say “I told you so” is vastly, infinitely outweighed by the dismay I feel over seeing my warnings be vindicated in real time.

And I also admit that my warnings were not entirely prescient when it came to Trump. I foresaw that Trump would attack America’s institutions, implementing rule-by-decree, purging competent people in favor of cronies, flouting the law, and wielding the power of the presidency to harass and intimidate his critics. I foresaw that Trump would send ICE into American communities to do violence and harass peaceable Americans. I foresaw that Trump would realign America toward Russia, cut off aid to Ukraine, and try to bully Ukraine into surrendering territory.

But I did not actually foresee his biggest mistakes. I didn’t predict that his tariff policy would be nearly as insane as it was — declaring sky-high tariffs on dozens of countries at once, and then selectively walking them back, and then repeating the process again and again.

And I did not foresee the Iran war. I never bought into his antiwar campaign stances — he has always been a bully, and he has always been enamored of the idea of military toughness. But I saw Trump, fundamentally, as a coward — someone who would launch the occasional air strike, but would be too intimidated by the prospect of a military defeat to launch a major war. I saw his cowardice as the core of truth behind the cynical promises of geopolitical isolationism and restraint.

So I can’t quite say “I told you so” in this case. I knew Trump was very bad news, but I didn’t realize quite how multidimensionally bad. I suppose even after all the Trump-bashing I did, I have to issue a mea culpa. I anticipated that Trump would be chaotic, dictatorial, and cruel, but I failed to anticipate how stupid he would be.

Even when the Iran war started, I thought that Trump would probably back off and chicken out pretty quickly. But as with his denial of the 2020 election result, he appears to have stumbled into a losing effort that he feels he can’t back out of.

Unlike with Trump’s limited strikes on Iran in early 2025, or his killing of Qasem Soleimani in 2020, Iran has not simply taken its lumps with grace. With the decapitation of its leaders and Israel pressing for regime change, Iran’s leadership was on what Sarah M. Paine calls “death ground” — they had no choice but to resist with everything they had. And so they’ve continued to fire drones and missiles from underground launchers at a diminished but steady pace. These strikes have occasionally hit valuable U.S. military assets, taking out an AWACS plane (one of only 16 the U.S. has) and some THAAD missile defense radars, and reportedly making several U.S. military bases too dangerous to use.

But the Iranians’ most damaging attack, by far, was to close the Strait of Hormuz, sending global oil, gas, and fuel prices soaring. This is hurting American consumers and tanking Trump’s popularity, but it’s hurting other countries around the world — who don’t have their own shale gas and shale oil reserves to weather the shock — even more.1

The Iran war has put Trump in a no-win situation. He’s clearly losing a war against a far inferior power. If he stays in the war, and the Strait of Hormuz stays closed, then he keeps losing; if he withdraws, he lost and it’s over. And even if he chickens out as usual, there’s no reason to think Iran will simply open the Strait; now that they see that they can bring Trump’s America to its knees with their oil weapon, they’ll probably use it to extract more concessions.

This is why Trump is writhing in the grip of his own bad decisions, looking desperately for a way out. He reduced oil sanctions on Iran, basically begging them to open the Strait, but they didn’t; instead, Iran just gets to sell more oil and make more money. He has repeatedly declared victory in the war, hoping that everyone will just agree that he won, allowing him to quit gracefully — but no one thinks he actually won.

Read more

I promise I’ll write something soon about the flaming, crashing disaster that is the Trump administration — and about other topics of interest. But before I do that, here’s a roundup full of short takes and stories about AI.

First, though, an episode of Econ 102! Officially the podcast is over, but we still occasionally do a reprise episode. This one, fittingly, is about AI biosecurity:

Anyway, here are six other interesting AI-related items:

1. Forecasting the effect of AI on growth

No one really knows what effect AI is going to have on economic growth, but maybe each “expert” knows a tiny, tiny bit. And maybe, if you combine all of those weak signals, you can get some actual information about the economic effects of AI.

That’s the idea behind a new study by the Forecasting Research Institute. They survey a whole bunch of different people about what they think AI’s capabilities will be in the future, and what that implies for economic growth. Specifically, the groups they survey are:

• Economists

• AI experts

• Superforecasters

• The general public

The results are kind of surprising, actually:

Forecasting Research Institute

Forecasting the Economic Effects of AI

There is widespread disagreement over the impact that AI will—or won’t—have on the U.S. economy: some prominent voices warn of a transformative upheaval and large-scale job losses, while others predict modest boosts to productivity at best. But there has been little work attempting to systematically understand expert views on the economic impacts of AI…

Read more

a month ago · 49 likes · 1 comment · Forecasting Research Institute

For one thing, all the groups have about the same forecasts for AI capabilities by 2030:

This looks like a forecast of modest progress, but it’s not. The “moderate” scenario here would have AI able to write high-quality novels, handle coding tasks that would take humans five days, create semi-autonomous labs, and use robots to perform basic household tasks. So basically, every group of forecasters in this survey thinks stunning AI progress is likely over the next few years.

And yet of all the groups, only the AI experts predict a major growth acceleration in any of these scenarios — and even then, it’s only an acceleration to 4 or 5 percent, not to the 10 or 20 percent scenarios that some people have thrown around:

Why do economists think that even near-godlike AI wouldn’t translate into fast growth? The Forecasting Research Institute lists some of their reasons:

Some economists argued that AI productivity gains would not be evenly distributed across all sectors, particularly where human labor is a bottleneck. Others pointed out that with other general-purpose technologies (electrification, automobiles, personal computers), there were multi-decade lags between widespread implementation and productivity improvements. Part of this delay is attributed to a shift in capital away from labor and toward compute, data centers, APIs, and so on, which would not manifest as an increase in GDP until productivity improvements set in…

Some economists expected demographic decline and geopolitical instability to offset some of the GDP boost from AI progress…Some economists argued that constraints on energy and chip supply, data center build times, and other commodities put a cap on the upper limit of GDP growth…Some economists argued that tail risks…included existential risks from AI, societal unrest or collapse, and war.

It’s likely that the AI experts are also thinking about these bottlenecks and frictions, or something like them, which is why their most optimistic scenario is 5.3% growth — fast, but still significantly slower than India is growing now.

But in fact, I think there must be more to the story here. Basically, none of these groups thinks that any amount of AI capabilities will enable economic take-off. To me, that suggests that they’re thinking — perhaps subconsciously — about something more than just friction and slow adoption.

One possibility — which I should write about more — is that people suspect that humanity is getting satisfied, at least in the developed countries, and that the amount of new valuable things that even a godlike AI could create for us is limited by our inability to desire more goods and services.

I should think about this more.

2. Will someone vibe-code the doomsday virus?

I’m very optimistic about many of the effects of AI, especially on science and politics. But as regular Noahpinion readers know, I’m pretty worried about AI-enabled bioterrorism (and I think an increasing number of other people are too). I’m worried that some nihilistic, depressed teenager could tell a jailbroken version of Claude Code to make him a doomsday virus, and that the AI would actually go and do it for him. We now live in a world where researchers can use AI to design new, functional viruses and have them sent in the mail. That’s an empowered world, but a terrifying one as well.

Ever since I wrote a post about that danger, I’ve been talking to biosecurity experts and trying to get a better handle on how justified my fears are. One of the experts I talked to, Abhishaike Mahajan, was in the middle of writing a long post about biosecurity in the age of AI. He has since finished the post:

Owl Posting

Reasons to be pessimistic (and optimistic) on the future of biosecurity

Note: this essay required conversations with a lot of people. I’d like to thank Patrick Boyle (ex-CSO of Ginkgo Bioworks), Harmon Bhasin (founder of a stealth biosecurity startup), Bryan Lehrer (ex-Blueprint Biosecurity), Theia Vogel (ex-SecureDNA), Jacob Swett…

Read more

2 months ago · 94 likes · 12 comments · Abhishaike Mahajan

You should read the whole post, but basically, he offers several reasons not to panic. First, he argues that it’s inherently very hard for even an extremely powerful AI to make an effective bioweapon on the first try. This is because there are just too many unknowns about how any newly created virus will behave in the real world, so there’s no way to know you have a doomsday virus until you release it.

I’m skeptical of this line of argument. Instead of just making one doomsday virus you can make 100 candidates and release them all. Doomsday itself is the field experiment, and you can run a lot of experiments at once. Much better bio simulation tools will probably cut down the number of candidates you need to create in order to stumble on one that works.

Abhishaike also argues that countermeasures — vaccines, antivirals, and defenses like far-UV light (which basically works on all viruses) will improve at a rapid clip. I believe this, but I’m not so comforted. Drawing on the experience of Covid, I think it’ll take a lot of time to deploy these countermeasures. A truly well-engineered doomsday virus will kill us long before we can distribute the cure or give everyone a UV zapper. And as Abhishaike points out, it’s likely that the U.S. will not proactively prepare for future pandemic threats, but merely react to them when they occur.

So while I think Abhishaike’s post is excellent and deserves a thorough read-through, I think he might still be underrating the severity of the threat.

3. Cybersecurity apocalypse?

How does the world know how much money you have? There are a bunch of computers that store your money as a series of numbers — how many dollars are in your checking account, how many shares of Apple stock are in your portfolio, and so on. Banks and other financial institutions have state-of-the-art computers and huge teams of brilliant software engineers to turn their electronic records into a fortress.

But AI is getting really, really good at hacking. Lyptus Research writes:

We release a new application of the METR time-horizon methodology to offensive cybersecurity, grounded in a new human expert study with 10 professional security practitioners…Offensive cyber capability has been doubling every 9.8 months since 2019. Accelerating to every 5.7 months on a 2024+ fit. Opus 4.6 and GPT-5.3 Codex sit well above both trendlines again, reaching 50% success on tasks that take human experts ~3 hours.

Right now, AI companies are white-hatting — using their AI’s newfound hacking powers to help companies improve their cybersecurity. But what happens when less scrupulous actors get their hands on jailbroken versions of Claude Code and Codex?

What happens if AI agents ever allow bad actors to break into banks at will? If all records of personal wealth were erased in a cyberattack, what could banks or the government even do? A whole lot of people might just instantly see their life’s savings transferred into a hacker’s bank account.

And as if that weren’t enough to worry about, recent advances in quantum computing put cybersecurity in an even more perilous state. Here’s Scott Aaronson:

For those of you who haven’t seen, there were actually two “bombshell” QC announcements this week. One, from Caltech, including friend-of-the-blog John Preskill, showed how to do quantum fault-tolerance with lower overhead than was previously known, by using high-rate codes, which could work for example in neutral-atom architectures (or possibly other architectures that allow nonlocal operations, like trapped ions). The second bombshell, from Google, gave a lower-overhead implementation of Shor’s algorithm to break 256-bit elliptic curve cryptography…

When I got an early heads-up about these results…I thought of Frisch and Peierls, calculating how much U-235 was needed for a chain reaction in 1940, but not publishing it, even though the latest results on nuclear fission had been openly published just the year prior…But I got strong pushback on that analogy from the cryptography and cybersecurity people who I most respect. They said…[I]f publishing [results like these] causes people still using quantum-vulnerable systems to crap their pants … well, maybe that’s what needs to happen right now.

Not being a cybersecurity expert, I’m not qualified to assess how worrying these developments are. But they seem quite worrying. The entire modern world runs on cybersecurity — if there’s a general failure in the methods we now use to keep information secure, all of society is in deep trouble. So this is definitely worth keeping an eye on.

4. The end of pseudonymity?

When I became a blogger, I made a conscious decision to post only under my own name. I reasoned that at some point, text analysis technology would get good enough where it would be able to identify (“dox”) any pseudonymous account I made. Fifteen years later, I’m anticipating vindication. This is from a new paper by Lermen et al.:

We show that large language models can be used to perform at-scale deanonymization. With full Internet access, our agent can re-identify Hacker News users and Anthropic Interviewer participants at high precision, given pseudonymous online profiles and conversations alone, matching what would take hours for a dedicated human investigator. …LLM-based methods substantially outperform classical baselines, achieving up to 68% recall at 90% precision compared to near 0% for the best non-LLM method. Our results show that the practical obscurity protecting pseudonymous users online no longer holds and that threat models for online privacy need to be reconsidered.

Soon, anyone who disagrees with your pseudonymous alt account, or is even just annoyed with you, will be able to sic an LLM on your account and dox it — if you’ve written online anywhere under your real name. If you’ve only written pseudonymously, you’re probably still safe.

The impending end of pseudonymity — or at least, its significant diminution — has the potential to transform the internet. Pseudonymity is obviously linked to toxic content, because people post stuff under a pseudonym that’s too aggressive or inappropriate to post under their real name.

We might also get a decrease in cancel culture, since pseudonymous accusations and whistleblowers will not be safe from retaliation. There will probably be less honest discussion and less total information on the internet, as people become afraid to have many discussions under their real names.

Less pseudonymity might also close off an important social and psychological safety valve — especially for Japanese people, who tend to use pseudonymous X accounts as a way to express feelings that they’re afraid to air out in public.

In any case, it’s going to get weird.

5. Will AI quants eat the economy?

At one point in Charles Stross’ Accelerando, AI finance quants turn the entire inner solar system into compute to power their financialized online economy — thus driving everyone else to the edges of the solar system.

That’s a little bit over the top, but it’s worth thinking about what happens if and when AI gets deployed in large quantities for adversarial economic activities like quant trading.

Most use cases that people think of with regards to AI are productive. We expect AI to accelerate science, do our coding for us, and so on. A few of the AI use cases we imagine are criminal — we worry about bioterrorism, cyber crime, and so on. But relatively few people talk about what happens if and when AI gets deployed en masse for rent-seeking — i.e. for the redistribution of income by legal means.

A lot of people suspect that a lot of what goes on in quant trading is rent-seeking — a bunch of traders trying to fake each other out or beat each other to the punch without creating economic value. In fact, there are models of how that can happen — my favorite is Hirshleifer (1971). In that paper, Hirshleifer shows how when traders compete to learn something that’s eventually going to become public knowledge automatically, they end up wasting resources on a zero-sum game.1

Quant traders have always used AI a lot, even before the rise of generative AI. But it seems possible that the rise of powerful AI agents and reasoning models will lead to an explosion of spending on quant trading. And if what those trading algorithms are doing is just trying to beat each other to the punch by a nanosecond, a lot of society’s resources — compute, electricity, and so on — will be going to waste.

Frustratingly, I don’t know of a good general result on how much of society’s resources could be wasted like this. But when I play around with some simple examples, it’s clear that the potential waste is large. AI quant trading might not turn the inner solar system into computronium, but it seems like it could still be a giant waste.

So I’m a little nervous when I see stories like this one, alleging that DeepMind founder Demis Hassabis tried to build an AI-powered quant hedge fund inside Google. Quant trading is a very natural way to use AI to make tons and tons of money, but if that becomes too big a part of what AI does, people will get mad at the technology.

6. Are people using AI less at work?

By most measures, AI is being adopted faster than any technology in recorded history. It’s difficult to read the news without seeing stories about how AI is conquering the business world. So it’s pretty notable whenever there’s a data point that shows AI not being rapidly adopted.

In fact, there are now a few such data points. Hartley et al. are maintaining an ongoing survey of American workers, in which they ask who’s using generative AI at work. For a while, their survey showed a rapid increase in adoption. But over the last year, they find that adoption has actually fallen:

One survey might be a blip, or there might be a problem with the way the questions are being asked. But The Economist reports that a few other measures are showing either a slowdown or a drop in AI use at work:

Researchers at the Census Bureau ask firms if they have used artificial intelligence “in producing goods and services” in the past two weeks. Recently, we estimate, the employment-weighted share of Americans using AI at work has fallen by a percentage point, and now sits at 11%…Adoption has fallen sharply at the largest businesses, those employing over 250 people…

A tracker by Alex Bick of the Federal Reserve Bank of St Louis and colleagues revealed that, in August 2024, 12.1% of working-age adults used generative AI every day at work. A year later 12.6% did. Ramp, a fintech firm, finds that in early 2025 AI use soared at American firms to 40%, before levelling off. The growth in adoption really does seem to be slowing.

What’s going on here? The Economist suggests several explanations — disappointing productivity effects, difficulty incorporating AI into existing workflows, economic uncertainty, and so on.

But if this trend is real, there are reasons to think it won’t last. First of all, most of this data is from before the rise of reliable AI agents, which really just came on the scene last December. Now that AI is a lot more than just a chatbot, it’s probably a good bet that more companies are going to find uses for it.

Also, once entrepreneurs start figuring out ways to build new business models and workflows, instead of trying to shoehorn the new tech into existing models and processes, we should see an explosion of AI-enabled productivity, just like we did with previous general-purpose technologies.

But for now, the hints of a plateau in industrial chatbot usage are worth keeping an eye on.

Subscribe now

Share

In the medium to long term, AI may replace all human jobs (or maybe not). But in the short term, AI doesn’t seem to be doing this yet. Employment rates for prime-age workers in the U.S. are hovering near all-time highs:

A recent survey of corporate CFOs found “little evidence of near-term aggregate employment declines due to AI.” A survey of European firms found no evidence of job reductions so far, despite rising productivity due to AI. Geoffrey Hinton, one of the pioneers of modern AI, famously predicted the imminent displacement of all radiologists by AI algorithms; in fact, radiologists are in greater demand than ever.

So even though AI may displace human beings en masse in the future, it’s not doing that today. But it is likely to change the nature of work. Software engineers, for whom “writing code” was a big part of the job description just a few months ago, are now mainly checkers and maintainers of code written by AIs. But this hasn’t eliminated the need for software engineers — at least, not yet. It has just shifted their job descriptions.

Humlum and Vestergaard (2026) find that so far, this pattern — workers shifting to new tasks without losing their jobs — is the norm, at least in Denmark:

[M]ost employers in [AI] exposed occupations have adopted chatbot initiatives, workers report productivity benefits, and new AI-related tasks are widespread. Yet…we estimate precise null effects on earnings and recorded hours at both the worker and workplace levels, ruling out effects larger than 2% two years after the launch of ChatGPT. What moves is the structure of work: employers absorb AI through task reorganization—including new tasks in content generation, AI oversight, and AI integration—and adopters transition into higher-paying occupations where AI chatbots are more relevant, though still too few to move average earnings. [emphasis mine]

In other words, so far, AI is replacing tasks, not jobs. Alex Imas and Soumitra Shukla have written that as long as there are a few things that only humans can do, this pattern can be expected to hold. Observers of AI consistently find that its capabilities are “jagged” — it’s much better at some tasks than others.

That’s good news for people who are worried about losing their jobs (at least in the next decade). But it’s still very troubling for people trying to decide what to study. A decade ago, it made sense — or at least, it seemed to make sense — to tell young people to “learn to code”. Nowadays, what do you tell them to learn? What tasks will be the ones that humans still need to do, and which will be subsumed by AI? With AI getting steadily better at a very wide variety of tasks, it’s hard to predict exactly what humans will still be doing in five years, even if you’re pretty sure they’ll be doing something.

I have some friends who have spent the last decade or more thinking carefully about what the future of work will look like in the age of AI. No one has ever found a satisfactory answer. As AI technology has developed and changed, even the most plausible predictions for the future of human labor tend to get falsified almost as quickly as they’re made.

But I’ve been thinking about this question too, and I think I’m beginning to see the shape of an answer. I think the near future of work will mostly be divided into three types of jobs — salarymen, specialists, and small businesspeople.

Let’s talk about specialists first, because they’re the easiest to understand. A new theory by Luis Garicano, Jin Li, and Yanhui Wu describes why some workers will keep their jobs largely as they exist today.

Like many economists, Garicano et al. envision a job as a bundle of various tasks. But they also theorize that in some jobs, these tasks are only “weakly bundled” — you don’t really need the same person to do all of those tasks. For these jobs, it would be easy to divide up the tasks between different workers — or between a human and an AI. But in other jobs, the authors assume that the tasks are “strongly bundled” — the same person who does one part of the job has to do the other parts, or the job can’t be done.

The paper’s basic conclusion is that AI tends to replace weakly bundled jobs a lot more quickly than it replaces strongly bundled ones. For example, they theorize that radiologists still have jobs because even though AI can do most of the task of basic scan-reading, there are a lot of other pieces of the job that radiologists still need to do in order to deliver patients the kind of care and expertise they demand. They foresee employment in strongly bundled industries resisting automation until AI capabilities get extremely good:

The people in those strongly bundled jobs are specialists. An example of a specialist might be a blogger. AI, so far, is very good at doing background research, proofreading, and a number of other tasks that are useful for the writing process. But even though it can generate infinite amounts of text, AI is not yet good at writing. Writing communicates a unique human perspective; simply pressing a button to generate text doesn’t say what you want to say. So the tasks that make up my own job are — so far, at least — strongly bundled. AI is making me more productive, but so far it isn’t putting me in danger of unemployment.

But what about those weakly bundled jobs? Garicano et al. predict that these will begin to decline only after demand becomes sufficiently inelastic — in other words, once AI becomes so productive that its output hits diminishing returns for the consumer. After that point, automation tends to replace human labor — it becomes a way to make the same amount of stuff with fewer workers, instead of a way to make more stuff with the same amount of workers.

Until that point, there will be quite a lot of work for people in weakly bundled jobs to do, because of expanded demand. And yet at the same time, companies won’t know which tasks to hire workers for, because AI’s “jagged” strengths and weaknesses will be constantly changing.

The rapidity with which Claude Code replaced the task of code-writing demonstrates this problem. In 2025, companies hiring software engineers could judge their merit based on how good they were at writing code. In 2026, companies have to judge the merit of software engineers based on how good they are at checking and maintaining code. Those skills don’t always go together.

The solution, I think, is to hire more generalists. Instead of picking people to do specific tasks, companies will pick people whose job is to constantly learn what AI is good and bad at, and to fill in the gaps. Cedric Savarese sums up this idea:

The first stage of ‘vibe freedom’ is…[t]he dreaded report that would have taken all night looks better than anything you could have done yourself and only took a few minutes…The next stage comes almost by surprise — there’s something that’s not quite right. You start doubting the accuracy of the work — you review and then wonder if it wouldn’t have been quicker to just do it yourself in the first place…You argue with the AI, you’re led down confusing paths, but slowly you start developing an understanding — a mental model of the AI mind. You learn to recognize the confidently incorrect, you learn to push back and cross-check, you learn to trust and verify…

Curiosity becomes essential. So does the willingness to learn quickly, think critically, spot inconsistencies, and to rely on judgment rather than treating AI as infallible…That’s the new job of the generalist: Not to be an expert in everything, but to understand the AI mind enough to catch when something is off, and to defer to a true specialist when the stakes are high[.]

Essentially, AI is going to be unreliable, but not in a predictable way. Its mistakes and shortcomings will require constant human exploration and patching. This is the job of a generalist. Instead of people who do “payroll” or “back-end engineering” or “accounting”, companies will need to hire people who can do a little bit of everything, if and when the AI messes something up.

In fact, we have an example of a corporate system that relied very heavily on this type of generalist: Japan. Until very recently, Japanese companies treated their “salarymen” as almost interchangeable labor, rotating them between different divisions and requiring them to learn a wide array of tasks. You might start your career in HR, then move to accounting, then do some product design, and so on.

This system might not have been very efficient, and the lack of specialization may have contributed to Japan’s notoriously low white-collar productivity. And it may be why salaryman jobs have been in decline for many years. But in the age of AI, it may finally make sense. When human expertise is replaced by AI expertise, humans’ role may be to flit from task to task, doing whatever the AI is bad at, and supervising AI at whatever it’s good at.

In other words, instead of hiring people who are good accountants or good HR specialists or whatever, companies might start hiring people who are just good AI wranglers, and who have the agency, mental flexibility, and energy levels to keep plugging the ever-shifting holes in what AI can do. In other words, salarymen.

The salaryman system also naturally lends itself to long job tenure. If I’m a highly specialized engineer, I can take my talents and move to a different company with my human capital intact. But if I’m a generalist who does a little bit of everything, what becomes more important to my value as a worker are my human networks within a company, and my understanding of the company’s system. This makes me a much less portable worker; I’m inclined to stay at the company where my long job tenure makes me more valuable than newcomers.

You can already see hints of this happening in American companies. We’re in a “no-hire, no fire” economy — workers are hunkering down in their jobs and refusing to switch, and companies are keeping them there instead of hiring new workers:

This is exactly what you’d expect from a model of firm-specific human capital — in other words, from an economy where everyone increasingly realizes that modern employees need to act like Japanese salarymen. The hypothesis here is that people don’t want to leave their jobs (and companies are happy to keep them in their jobs) because their technical skills might be devalued due to rapid AI progress; instead, they’re staying in their companies, where knowing people and knowing how things work are still important.

So America may yet come to embrace the way of the salaryman. But the third category of future employment will also be very Japanese: self-employment and small business.

Japan has long had a very high prevalence of small business ownership. It has one of the world’s largest proportions of small and medium-sized enterprises. In manufacturing as well as in retail, Japan has traditionally had a lot more small business than other OECD countries. This is now decreasing, as the population ages and business owners retire without heirs or proteges. But it still might point the way to the AI-enabled future.

AI creates leverage; it allows you to do more with a smaller team. For many businesses, the optimal size of this team will fall to only one person or a few people. Thus, I expect to see a lot of small companies sprout up, as people use AI agents to increase their productivity to the point where they only need a few employees (or even zero).

In other words, I expect AI to make the American labor system look a bit more like the Japanese labor system of the 1960s-2000s. There will be a bunch of generalists running around looking for things to do within their companies, a bunch of small businesspeople striking out on their own, and a few specialists with specific skills that still make them valuable. If you’re not one of the lucky few in the latter category, your choices will be to become a cog in an ever-changing corporate machine, or to strike out on your own and manage an AI “team” to sell some good or service directly to the consumer.

This might not be the most optimistic or enticing view of the future of work, especially to people who have lived their whole life thinking that their specific job skills are what made them valuable to society. But it’s probably better than humans becoming economically obsolete.

Subscribe now

Share