I have a simple rule: people with no electoral experience should not run a democratic country. The rule isn’t absolute: two of the five US Presidents who had never held elected office did an excellent job (though both had won major wars).1 Moreover, if I had to violate my rule and pick someone with no electoral experience to lead a nation, I’d avoid bankers because they tend to focus narrowly on financial markets. I’d take a chance on an entrepreneur, a community leader, or a local deacon.

Canadian Prime Minister Mark Carney violates all of my rules. Unlike every head of state in Canadian history, he had never run for any office. He had never served in cabinet or as a Member of Parliament. When he entered the Liberal leadership race, he had no seat in the House of Commons, which is just plain weird in a parliamentary system. And as a former Goldman Sachs Managing Director and governor of both the Bank of Canada and the Bank of England, he is the consummate global banker.

Who is Mark Carney?

Carney was born in 1965 near the Arctic Circle, in Canada’s Northwest Territories. He grew up in Edmonton and attended Harvard on a scholarship, played ice hockey, and earned a BA in economics. He joined Goldman Sachs in 1988, took a 3-year leave in 1991 to earn a doctorate in economics at Oxford, and returned to work in Goldman offices in London, New York, and Toronto.

In 2003, Dr. Carney left Wall Street to join the Bank of Canada first as deputy governor. In 2008, Conservative prime minister Stephen Harper named Carney governor of the Bank of Canada, where he helped steer the country through the global financial crisis. He chaired the G20 Financial Stability Board, which put him at the center of international financial reform. In 2013, the British government named him governor of the Bank of England. Carney became the first non-Brit to hold the job in the bank’s three-century history, and he found himself navigating the Brexit upheaval.

Canada’s 2025 Election

In early 2025, Carney entered the race to lead the Liberal Party. First, he had to elbow his way past an obviously qualified woman. Chrystia Freeland had deep roots in the Liberal Party as Trudeau’s talented deputy prime minister and finance minister. Trudeau had positioned her to succeed him. But Trudeauisme had run its course, so Carney won the Liberal leadership nomination in a landslide, taking nearly 86% of the vote.

Getting the nomination was much easier than winning the election. By January 2025, Trudeau’s Liberals trailed the Conservatives by 20-25 points. Conservative Pierre Poilievre capitalized on Trudeau’s pretty-boy, nepo-baby unpopularity and ran on cost of living, crime, and housing. Every poll showed Poilievre as headed for a historic win.

Donald Trump rescued Mark Carney. Trump launched a trade war with Canada, imposing steep tariffs on Canadian imports and threatening a 25% tariff on Canadian-made cars. Then he kept announcing that Canada should become the 51st US state and threatened to use “economic force” to make it happen. Trump kept up his anti-Canadian shitposting until Canada’s election day, and he absolutely torpedoed the Conservatives.

Canadians rightly concluded that Trump was insulting their sovereignty. Carney adroitly surfed the resulting wave of patriotism. He ran an “elbows-up” ad with comedian Mike Myers that went viral globally and reshaped Carney’s image from a sober banker to a resolute Canadian patriot. Worth watching.

Conservative Pierre Poilievre’s “Trump light” populist style worked only as long as Trump remained a political abstraction. Once Trump threatened Canada, any resemblance to him immediately became a liability. Instead of a referendum on Trudeau, the election was suddenly about who could best confront a rogue US President. Voters concluded that a central banker who had steered economies through crises like Brexit was better suited than a Trump syncopant to lead a trade war with the US. Voters shunned Poilievre so hard that he lost his seat in Parliament.

Carney won 43.5% of the vote and 168 seats, just 4 short of the 172 seats needed for a parliamentary majority. Without seeming to care, Trump had destroyed yet another erstwhile ally. (One political scientist suggested that Trump had helped the Liberals so much that “they ought to pay him”).

Carney’s Davos Moment: At the Table, Not On the Menu

On January 20 of this year, Carney spoke at the Davos conclave of the world’s power brokers and thought leaders. Carney began his address in his serviceable, non-native French, asserting that the world is in “a rupture, not a transition.”

Carney declared that the US-anchored, rules-based order was gone. The great powers now used economic integration as a weapon. They imposed tariffs to gain leverage and exploited their control of global financial infrastructure to coerce weaker nations. They gleefully weaponized their supply chains for national gain.

Carney urged the world’s middle powers to act together, saying, “If we’re not at the table, we’re on the menu.” He argued that shared investment in resilience is cheaper than each nation building its own fortress, and that middle powers should build “variable geometry” coalitions that could shift with shared interests. Attendees gave his remarks a standing ovation. The following day, Trump used his time at Davos to demand that the world show more gratitude for US military protection. Once again, Trump magnified Carney’s popularity.

Since Davos, Carney has produced a preliminary but landmark agreement with China, a multi-billion-dollar deal with India, and addressed parliaments in Australia and Japan. He signed trade and security deals with the UAE and Ecuador. His efforts to deepen Canadian-EU ties have led to (probably unrealistic) talk of Canada as an “associate member” of the EU.

Last month, Carney, the presidents of Brazil and Kenya, and the European Council president announced Partners for Multilateralism (P4M), which aims to link middle powers and developing countries on international law reform, security, supply chains, climate and AI. Although details are thin, Mexico’s President Claudia Sheinbaum called his analysis “in tune with the current times,” and Australia’s Anthony Albanese said it matched Australia’s view.

A Liberalism With More Spine Than Spin

Let’s pull back and grant two counterclaims to this narrative. First, without Donald Trump’s ham-fisted intervention, Mark Carney could well have joined Justin Trudeau and a long list of poll-tested liberals who lose elections to national populists.2 But Carney broke the pattern by arguing for liberalism on moral grounds, not just technocratic ones. He has repeatedly told voters that a strong economy is necessary not simply for prosperity, but as a prerequisite of a just society. He defends, in moral terms, Canadians’ freedom to follow their own religious and ethical lights. At a time when most liberals reach for spreadsheets, Carney reminds the world of first principles.

Second, Carney is in an unfair fight. Canada is one-tenth America’s size. Seventy percent of Canadian exports go to the US, but only 12-15% of US exports go to Canada. As a result, Canada’s core economic infrastructure (especially its massive oil and gas pipelines and rail networks) runs north-south, so Canada cannot easily redirect its primary exports to other global markets. Finally, as Carney knows well, the Bank of Canada is heavily constrained by the US Federal Reserve. When rates diverge too far, Canada pays for it in the exchange rate and in capital flows (although, to be clear, the BoC is independent and has gone its own way many times, including this week).

Nonetheless, Carney is reshaping 21st-century liberalism. Third-way and abundance liberals embrace slashing regulations to promote growth. Some would cut taxes to fight inflation. Carney has done both while preserving Canada’s welfare programs. He suffers neither weakness nor wokeness – talking about gender far less than Trudeau did and rejecting conventional liberal group-think. He slashed immigration and wound back criminal justice reforms that had softened law enforcement.

Carney has broken with liberal orthodoxy so often that populists struggle to caricature him. With approval ratings over 70%, he is the most popular Prime Minister in Canadian history. This lets him do what leaders of other mid-sized democracies, like Macron in France, Burnham in the UK, and Merz in Germany, cannot: commit his nation, knowing that his people will follow.

At Davos, Carney challenged Thucydides’ famous observation that “the strong do what they can and the weak suffer what they must”, arguing that the rules of the liberal order never bound the great powers.3 As an ex-banker, Carney understood that this order rests on a dense web of trade, investment, and cultural ties that can be mobilized to counter global hegemons. He is very clear that Canada seeks resilience to prevent anybody from controlling its markets or threatening its sovereignty, territorial integrity, democracy, or rule of law. In his telling, international cooperation is not a code for great-power domination. It lets different nations that share fundamental values pursue their own preferences.

Moral Leadership

Along with Pope Leo, Carney is emerging as one of the free world's moral leaders. The two men share more than a faith. Carney is a practicing Catholic who knelt for the Eucharist at Leo’s inaugural Mass in May 2025, then met the new pope in the basilica with his wife and daughter. This past May, the two spoke by phone after the publication of Leo’s encyclical, Magnifica humanitas, on the encyclical’s themes and on peace in the Middle East and Ukraine. Leo has offered to host peace talks between Russia and Ukraine at the Vatican. Carney has spent the year trying to build a coalition of middle powers that say no to bullies.

At Davos, Carney told a story about a greengrocer from Václav Havel’s 1978 essay “The Power of the Powerless.” The grocer put a sign in his window saying “Workers of the world, unite!” He did not believe the slogan, nor does anyone who walks past it believe it. But like all grocers, he hung the sign each morning because, as Havel tells it, “one must do these things to get along in life”. The sign is a ritual that signals compliance.

Carney argued that systems built on pretense like this hold together only as long as people play along. The day even a single greengrocer removes his sign, the illusion starts to crack. Carney urged companies and countries to “take down their signs” and “live the truth” by recognizing that America was abandoning the security order they all relied on.

Pope Leo and Mark Carney both know the power of moral authority. They know the story of French foreign minister Pierre Laval, who in 1935 urged Joseph Stalin to ease up on his suppression of Catholics. Like Trump trying to humiliate Ukrainian President Volodymyr Zelensky, Stalin responded, “How many divisions does the Pope have?”

Joseph Stalin did not live to see John Paul II elected pope and return to his native Poland, where millions gathered to hear his message. Stalin never saw Solidarity rise from the shipyards of Gdansk the following year, or the Berlin Wall fall a decade later. Pope John Paul II and Solidarity helped bring down communist rule in Eastern Europe without commanding a single division. They commanded conviction, and the confidence it gave millions of ordinary people to stop obeying.

As two emerging leaders of the free world, Mark Carney and Pope Leo are making a similar bet. Divisions can win military battles, but moral authority decides whether the greengrocer puts up his sign. Tyrants who mock moral authority often end up buried by it.

Subscribe now

Share

America’s two leading AI frontier model builders, Anthropic and OpenAI, have both filed to go public. Anthropic is widely expected to release confidential financials and start its road show next month, while OpenAI will wait until 2027. These will be two of the largest public offerings in US history, with Anthropic’s opening market value potentially exceeding SpaceX’s $1.77 trillion record.

Two powerful forces are squeezing these companies’ valuations. First, lower-cost, often open-weight, models are putting pressure on prices. Second, as if the extraordinary cost of data centers, mounting political opposition, and the growing risk of drone attacks weren’t enough, both companies face growing product liability concerns.

AI companies are not easy to value. On the one hand, AI is by far the fastest-growing technology in human history. ChatGPT reached 100 million users within two months, much faster than Instagram, Facebook, Twitter, or Netflix. According to the Federal Reserve Bank of St. Louis, 54.6% of US adults aged 18–64 adopted generative AI within three years of public introduction. At the same three-year post-launch mark, personal computers reached 20% adoption, and the commercial internet reached 30%.

AI also improves shockingly fast. METR measures how long a task an AI can complete, based on how long it takes a human to do it. Two years ago, it could complete a task that takes a human four minutes. Earlier this year, the best AI could complete a task that takes a human 17 hours. METR is discontinuing the metric as some models now do work that takes a human 17 weeks. All we really know is that the AI you’re using now is the worst you will ever use.

But unlike software, AI demands extraordinary levels of capital investment. AI pessimists often argue that infrastructure booms become asset bubbles that eventually burst, triggering a stock market crash. They cite 19th-century canals, railroads, 1920s electric utilities, and telecom carriers laying optic fiber in the 1990s. But they ignore the 1950s Interstate Highway System, rural electrification, the Bell telephone network, and municipal water, sewer, and bridge systems. None of these build-outs led to a market crash.

Why Bubbles?

Why do some infrastructure booms lead to stock market collapses and others don’t? Four things need to line up to create a bubble:

1. A new technology with high but uncertain demand. Canals, railroads, electrification, and most utilities qualify. So, obviously, does AI.

2. Market competition that guarantees redundant capacity (government projects like freeways, monopolies like the Bell system, and municipal water and sewer systems do not produce investment bubbles). AI competition is obviously fierce and highly personal.

3. Private capital markets that let people bet on the future before it arrived (nobody traded shares in Hoover Dam or early rural electrification schemes).

4. Debt lets builders invest far more than the cash they generate. This is an especially potent factor. Debt inflated the canal, railroad, utility, and telco bubbles. In contrast, the dot-com bubble hurt investors more than lenders. The Nasdaq fell 78 percent and led to a very mild recession. Venture investors lost a lot of money, but it didn’t trigger a chain of defaults that spread into the banking system. Some economic historians call debt-driven collapses “Leveraged Bubbles”, which tend to produce deeper recessions and slower recoveries than stock-only bubbles. If AI is a bubble, it is increasingly a leveraged one.

AI checks all four boxes: a technology with high but uncertain demand, financed by competing private companies with a mix of private capital and growing amounts of borrowed money. The numbers are famously massive. UBS analysts forecast investment by hyperscalers and the builders around them at $1 trillion this year, $1.4 trillion next year, and $1.6 trillion in 2028. This exceeds the cash they generate, so analysts expect all six of the largest hyperscalers to post negative operating cash flow in 2026 and 2027. They borrow money to fill the gap.1

Historically, not all infrastructure bubbles have been bad. When the bubble burst, investors lost their shirts, companies went bust, and working families went hungry, but the country itself was richer. Once someone bought them out of bankruptcy, they put the canals, railroads, power grid, and fiber networks to work. This may be less true of AI because GPUs and specialized accelerators make up half of the cost of data centers. With specialized chips, IT hardware accounts for 65% to 75% of total initial facility spend. The chips inside a data center are far more valuable than the building or generators that make it run.

Unlike canals, railroads, or fiber, however, chips become obsolete very quickly. Every 12-18 months, a new generation of AI chips arrives, typically delivering 2-4 times more computation per watt. Because electrical power is the bottleneck limiting AI growth, running older, less efficient GPUs becomes economically irrational within 2-3 years.2

Do Falling Prices Predict Asset Bubbles?

Falling prices can mean two very different things.

• A lower price can simply reflect falling production costs. This happened when steel rails and bigger locomotives improved efficiency and drove rail prices down, or when multiplexing technology let fiber carry much more information. When unit costs fall, but volume grows enough to keep revenue rising, a price decline is great news.3

• Falling prices can also reflect investment that grows faster than demand. In that case, revenue per unit of capacity falls, and high-cost producers struggle. Railroad rates fell by more than half between the early 1870s and the 1890s. Similarly, long-haul fiber rates fell by up to 90 percent in the early 2000s.

A price change alone cannot tell you which one you are looking at. You need to understand utilization and revenue per unit of capacity. But what is a “unit of capacity” in AI?

• It is not task complexity. As noted earlier, METR has given up trying to measure how complex a human task an AI can complete.

• Not parameter count. Nor does a model’s parameter count measure AI capacity.4 A large model may have up to 3 trillion total parameters, but during inference it activates only a fraction of those weights (35 to 80 billion) per token. For this reason, efficient small language models might use 70 billion parameters and run on an Apple Mac. These local models require no data centers, but Stanford researchers concluded that they “can successfully answer 88.7% of single-turn chat and reasoning queries.”5 Qualcomm, Google, and Apple are all producing mobile phone chips that can run AI models. At the margin, they reduce demand for data centers.

• Not total compute required. Efficient models require much less compute. Reflection AI, an Nvidia-backed startup, says its Beam model matches China’s GLM-5.2 on advanced reasoning with 3 to 4 times less inference compute. Although independent tests haven’t confirmed this claim, many observers expect these efficiencies to grow rapidly. Even if parameters grow and overall token demand rises, these improvements could trigger a cascade of halts in data center construction and debt defaults.

• Not token use. Token use is also a poor measure of capacity. Famously, token prices are already falling. According to OpenRouter, OpenAI and Anthropic charge $1-2 per million input tokens, while Chinese models like GLM-5.2 list at $0.76 to $0.98 per million input tokens. Other models approach OpenAI or Anthropic: GLM-5.3 lists at $1.40 and Qwen3.8 Max at $2.00. But token prices can also be misleading. AI frontier models use tokens differently, so lower token prices often mask a higher cost per completed task.

The ideal operational metric is cost per correct resolution, not price per token. If this falls as usage grows, we are in the 1920s electricity case, where demand growth offsets declining costs. If price per correct resolution falls because supply outruns demand, it’s more like fiber. Unfortunately, we have no industry-standard measure of cost per correct resolution (or even a definition of correct resolution).

That said, open-weight models are clearly closing in on frontier-model quality at a fraction of the price. Overall corporate spending on tokens is falling as a result. This compresses margins for Anthropic and OpenAI and threatens anyone whose capex investments were built on premium pricing. But whether demand for AI will keep growing as prices fall remains an open question.

The Second Squeeze: Hidden Liabilities

Suddenly, liability for damage that their advanced models can cause looms large over Anthropic and OpenAI’s plans to go public. Neither company has fully released their most advanced models because of safety and behavioral concerns. Anthropic restricts Claude Mythos Preview to a small group of organizations through Project Glasswing. OpenAI says it will not release GPT-6.1 Astra.

Scholars and legislators debate whether to impose strict liability on AI developers. At present, no US state or federal authority requires AI developers or deployers to maintain liability insurance. Nonetheless, dozens of states are considering requiring AI companies to carry insurance. (OpenAI is backing an Illinois bill that would limit liability for catastrophic harm.)

Insurance companies themselves appear to have little interest in underwriting AI risk. Several carriers, including Berkshire Hathaway, Chubb and Travelers, have won state approval to exclude AI damage from their standard commercial liability policies. A few specialty insurers offer tailored AI error or agent-risk policies, but they cover far less than the multi-billion potential liabilities facing AI frontier platforms. None cover the risk that models could facilitate cybersecurity attacks or manufacture bioweapons. Investors have noticed, and credit default swap prices on AI companies have surged. (These derivatives serve as insurance against default for any reason)

Can platforms be held liable for harm caused by their algorithms? Probably. In August, Meta agreed to pay up to $17.1 billion to settle claims from dozens of states that Facebook and Instagram were designed to addict children. Although the deal admits no liability, a New Mexico judge separately ordered Meta to pay nearly $1 billion, a ruling that comes closer to proving platforms can be held liable. Florida is already suing OpenAI, claiming its products posed risks to children and that the company failed to warn the public about them.

Earlier this month, I argued that we should subject AI to basic disclosure, liability, and monitoring. A concrete first step is to require frontier model companies to carry enough insurance to protect consumers from harm. Requiring them to insure against the risk they take would force frontier and open-source companies to rigorously evaluate their products before releasing them.

Modern Times

Aiming for the Stomach: A Pragmatic Blueprint for AI Regulation

In 1904, American socialist Upton Sinclair went underground in the Chicago stockyards to expose how badly meatpackers were treating their immigrant workers. After working for seven weeks in a slaughterhouse, he published his famous exposé, The Jungle…

Read more

a month ago · 3 likes · 2 comments · Marty Manley

The problem, of course, is that AI’s potential harm is substantial, even if its potential benefits are too. The world likely doesn’t have enough insurance capacity to cover this level of tail risk. Edward Best, a partner at the law firm Willkie Farr & Gallagher, told the New York Times that “Maybe they could get a billion dollars of insurance, but is that going to help a company that large? I mean, they need the capacity of the entire insurance industry.”

Conclusion

The AI buildout and associated IPOs are squeezed between two forces. Open-weight models and efficiency gains are pushing prices down, while potential liabilities are pushing costs up. Equity investors can absorb both risks, but lenders cannot. But hyperscalers now borrow against revenue projections based on premium prices and manageable levels of legal risk. To go public, both Anthropic and OpenAI will need to forecast prices realistically and put a number on poorly understood, hard-to-insure liabilities. This is unwelcome news as they try to dress up pretty for their coming-out party.

Pay attention to three signals.

1. Is AI revenue growing as fast as investment in data centers? Right now, the answer is plainly no. AI revenues are growing rapidly from a modest base, but they still lag far behind data center capex. Historically, many companies have required $1.50 to $2 of revenue per dollar of capital invested to cover operating overhead, power, and operating profits. This means that the market would need over $1 trillion in annual enterprise software spend just to break even. Today, pure-play foundation lab revenues total between $130 billion and $160 billion annually.

2. What share of announced data centers actually break ground? If hyperscalers cannot get utilities to build the generating capacity and transmission lines they need, then hyperscaling isn’t possible.

3. How large is the first big AI liability verdict? If an agent swarm originating from an Anthropic or OpenAI model hacks a bank or breaks into a large company and causes damage, serious lawsuits will quickly follow. Even if legislators do not mandate liability insurance, investors will insist on it. Investors can move their money elsewhere in seconds, and they will.

If these metrics go badly, keep Hemingway in mind. Things go south gradually when projects slip or a jury rules against you. But when banks or private-market lenders call a loan, things can collapse suddenly.

Subscribe now

Share

Before Trump launched his ill-considered Iran War, many of us thought that Democrats would take the House in the November midterms, but were unlikely to take the Senate. That still could prove true, but polls now show that Democrats have a solid chance of taking full control of Congress. If Democrats win Senate and House majorities on November 3 and are sworn in on January 3, they will be the dog that caught the car. Their big question will be: now what?

Three voices will compete to steer Congress: the 2028 left-versus-moderate campaign agenda, the strong urge to hold Trump accountable, and improving the affordability of everyday life in America. Agenda, accountability, and affordability all matter, but they have vastly different political consequences. To grow as a party, Democrats should prioritize affordability, then accountability, and worry least about their 2028 agenda.

• Agenda 2028. After the midterms, many leading Democrats will immediately surface a manifesto for the 2028 presidential election. The left will claim that its leftism did the trick, and moderates will do the same. Manifestos will be everywhere. This is understandable. After all, Donald will be the lamest of ducks, and the 2028 presidential campaign will begin as soon as the midterms end. Building on midterm momentum will be top of mind for many Democratic leaders.

  Many groups will compete for electoral attention. Advocates for climate, labor, AI, women, immigrants, housing, LGBT, education, energy, and civil rights policies will clamor to be heard. Their issues matter, as does the 2028 election, but not yet. First, Democrats need to focus on measures to improve everyday affordability and, to a lesser extent than they’d prefer, to increase executive accountability.

• Accountability. Many Dems want redress, if not revenge. There will be massive pressure to hold Trump and his cronies accountable for their corruption, chaos, illegal orders, and other violations of their pledge to protect and defend the Constitution. Senators will salivate at the chance to haul Trump’s clown-car cabinet before Congress. They badly want to grill Pete Hegseth about his purge of high-quality military leaders, ammo shortages, and security leaks; Pam Bondi on her suppression of the Epstein files, weaponized prosecutions, and termination of DOJ ethics officials; RFK, Jr. for compromising CDC and FDA independence on behalf of an anti-vax agenda; Howard Lutnick for his shambolic tariff implementation and epic financial conflicts. The list is endless – as is the appetite to expose, humiliate, and possibly impeach the wrongdoers.

  This work is important. But it matters less than showing working Americans that Democrats take their standard of living seriously. If Democrats want to earn the vote of everyday working families, they need to put their concerns first.

• Affordability. Democrats should remember why working families went for Trump to start with: Dems had become a party of professionals that ignored inflation and a chaotic southern border. A Democratic Congress should focus on raising pay and reducing living costs, especially in health care, housing, and energy. In each area, legislators should begin long-term reforms while giving Americans immediate income increases or expense reductions. And Democrats should embrace immigration reform as part of this affordability agenda.

Trump will obviously be tempted to veto these bills. But a president who vetoes a series of measures designed to raise wages and benefits, reduce the cost of health care, housing, and energy, and address immigration intelligently will leave his successor with a great deal to answer for. By championing higher living standards, Democrats can make it very difficult for Republicans to claim to represent the interests of working-class voters.

Increase Incomes

Affordability is more than rising prices—it also includes stagnating wages. RAND found that if worker incomes had grown with the economy the way they did between the end of World War II and 1975, the median full-time worker would have earned between $92,000 and $102,000 per year in 2018. Instead, that worker earned about $50,000. If you apply RAND’s findings to today’s median full-time paycheck of about $65,000, a median worker would earn $120,000—$133,000 today.1

That’s not the only reason wage growth has slowed in the US, but in 1975 the private sector was 25% unionized. Today, it is less than 6%.

Democrats should push hard to correct this. In the long term, Congress should experiment with sectoral bargains backed by wage subsidies for low-wage industries. But Congress can do at least three things in the near term.

• Index the minimum wage. We effectively have no federal minimum wage. It has been stuck at $7.25 since 2009, so that today it affects only 1% of all workers. Even a high minimum wage of $17 or $18 will erode the way $7.25 did. So set the federal minimum wage at 60% of each state’s full-time median wage. The range would run from about $18/hour in high-wage states like Massachusetts, Washington, and Alaska to $12 in lower-wage Mississippi, Arkansas, or West Virginia. Instead of setting a one-size-fits-all minimum wage, this approach respects regional differences. More importantly, it resets the wage floor each year.

• Ban noncompete agreements. About 30 million workers are bound by noncompete clauses. Banning them would raise pay by almost $300 billion a year, according to the FTC (the Chamber of Commerce disputes the math). Oregon banned noncompetes for low-wage workers, raising their pay by 2% to 3%. It costs taxpayers nothing, and Trump’s own FTC chair calls noncompetes “pernicious.”

• Make work pay. Start by expanding the Earned Income Tax Credit. Today, workers without kids only get about $600 in credit. The 2021 Biden budget temporarily tripled it. Reinstating a higher credit would cost roughly $140 billion over ten years. Democrats should also reverse Trump’s SNAP and Medicaid cuts and restore the 2021 Child Tax Credit, which cut child poverty to a record low of 5.2%. These are the most market-friendly, worker-friendly, anti-poverty measures we have.

Cheaper Health Care

Today, a family of four on an employer plan pays about $27,000 a year in health care insurance premiums. Workers pay for all of this: $6,850 as a co-share of the premium taken out of their paycheck, and the rest through wage increases they never see. Viewed this way, health care now accounts for a quarter of average household spending.

We plainly need health care reforms that reduce costs. We need to train more physicians, retool fee-for-service models, and restructure hospital markets. This requires years of policymaking, whether payors are private or public. Some of that work falls to state lawmakers, governors, attorneys general, and insurance commissioners. Medicare for All polls well, with 66% favoring a national plan, but it will take years to pass and may not address the underlying cost problems. Congress can cut costs now by restoring income-based premiums and targeting the biggest health care line items.

• Limit hospital charges. Hospitals account for almost 40% of health spending, in part because they charge employer plans 254% of what they charge Medicare for the same services. The Urban Institute estimates that capping payments at 125% of Medicare rates in 2022 would have saved Medicare $107 billion. When Oregon tried it, outpatient prices fell by 25%. Congress should phase in a price ceiling indexed to Medicare reimbursement rates, knowing that private insurance will follow Medicare’s lead. Won’t hospitals fight it? Of course they will. Democrats should welcome the fight.

• Require site-neutral payments. Medicare pays much more for an X-ray in a hospital clinic than for one you get in your doctor’s office. Same with dialysis. The government has begun closing this gap for drug administration and may do the same for imaging. The Congressional Budget Office says that simply making all payments site-neutral would save Medicare $157 billion over ten years. It would save commercial health care plans even more.

• Restore the Obamacare premium subsidies. In 2025, Trump allowed enhanced health care credits to lapse to fund tax cuts for the wealthiest Americans. No surprise, average deductibles suddenly jumped 37%. Democrats should restore these credits and make more people eligible. It would cost $335 billion over ten years, assuming Democrats cap premiums at 8.5% of income to keep coverage affordable for middle-income workers.

• Negotiate prices for more drugs. More than 80% of US adults want the federal government to negotiate lower drug prices. Today, the feds negotiate for only ten meds, although this number is scheduled to rise steadily.2 Democrats can and should expand this to hundreds of currently excluded medications, including GLP drugs, cancer and heart disease meds, and the specialty biologics used to treat autoimmune conditions. Lawmakers should also extend the $2,000 annual prescription spending cap and expand the $35 monthly insulin cap to all private employer plans. Doing this would reduce health care costs for 160 million Americans.3

Cheaper Housing

Households spend a third of their money on housing, though spending varies widely by region. Making housing affordable over the long term requires more construction, especially in blue cities. It can take time for Congress to create laws or incentives to reform zoning and permitting. In the short term, Congress should:

• Create a large carrot to reward cities and states that reform zoning. The new budget offers cities a $200 million annual grant to reform zoning to permit more housing. That is a baby-cut carrot — far too small to matter. Congress should tie existing large transportation grants and HUD block grants to reward states and cities that eliminate single-family-only zoning, parking minimums, and barriers to accessory dwelling units. California now does something like this at a state level.

• Ban algorithmic rent-fixing. Congress should pass the End Rent Fixing Act to outlaw software like RealPage or YieldStar that landlords use to fix prices. This not only protects renters, but it helps preserve starter homes for actual families.

• Put builders back to work. We need to rebuild our building trades with a mix of training, continued wage increases, and smart immigration reform. 92% of construction firms report difficulty finding qualified talent. Even with wages averaging $40/hour, well above the private-sector average, the industry believes labor shortages cost it $11 billion annually.

  Demographics are not helping. More than 20% of the construction workforce is over 55 and nearing retirement, while fewer than 3% of young people enter the building trades. Immigrants make up more than 25% of the construction workforce.

Cheaper Energy

US households spend about 11% of their income on energy. This is split roughly evenly between home energy and gasoline. The single biggest step Congress could take to reduce energy prices would be to open the Straight of Hormuz by ending Trump’s war on Iran. This will require distasteful concessions to a horrible regime, so Democrats will be deeply tempted to let Trump swing in the breeze on this issue. Electricity prices are the second front.

Congress can design electricity markets for affordability by requiring grid upgrades, ensuring that data centers reduce electric bills, and making it easy for consumers to invest in technologies that lower energy costs.

• Make data centers pay their way. Twenty-three states have already approved large-load tariffs to help achieve this. Congress should require data centers to cover the complete incremental capital costs of new power generation, high-voltage transmission lines, and local distribution upgrades required to bring them online. Even if demand for data centers is overstated, many are coming, and local communities can and will demand labor and community agreements.4

• Require utilities to upgrade. Replacing old cable with advanced conductors could save $85 billion by 2035. FERC already requires planners to consider it. Congress should make it mandatory before asking ratepayers to fund new lines.

• Cap utility profits. Regulators grant monopoly utilities higher profits than private investors demand. In 2024, regulators authorized an average return on equity of 9.7%, but 34 major investment firms forecast an average long-term equity return of 6.7% for the broad US market. A UC Berkeley study estimates that consumers pay $7 billion a year to cover these excess costs. The Lowering Utility Bills Act would set returns at the low end of the allowed range nationwide.

• Preserve incentives and rebates for households that want to invest in energy efficiency. The IRS offered investment and production tax credits that reduced the cost of generating new clean energy. It also offered consumers rebates for high-efficiency heat pumps, panel upgrades, and weatherization. This provides instant savings of $8,000-$14,000 for household energy retrofits that can cut monthly heating and cooling bills by 20-40%. Better yet, conservation and local generation place downward pressure on power prices.

Will Trump veto all of these initiatives if Democrats pass them? Let’s find out and give Democrats the chance to promise they'll sign these measures if elected in 2028.

Make Immigration an Affordability Policy

Democrats need to follow the successful leads of Bill Clinton and Barack Obama and embrace increased border security, workplace enforcement, and restricting public benefits for undocumented immigrants. They should also support skill-based immigration policies that help create high-quality jobs and raise incomes for everyday Americans. In other words, make immigration an affordability issue.

• Enforce borders. Every nation needs secure borders. Without borders, no nation can reliably improve living conditions for its citizens. This requires enforcement not just by border agents, but by employers, banks, and government service offices like the DMV. And it requires that we deport those who arrive or remain here illegally. Clinton and Obama could say this without gloating or advocating cruelty. Every Democrat should do likewise.

  Democrats should insist that nobody ever set foot in the United States illegally.

• Legalize workers who are already here. Documented workers earn 19% to 22% more than undocumented workers with the same education and age. A Center for American Progress study also projects that increased legalization boosts wages slightly for other American workers. Congress needs to legislate a path to citizenship for those who have lived here for several years, paid taxes, and obeyed the law. These folks should pay a fine, demonstrate English proficiency, pass a citizenship test, and swear an oath of allegiance. Democrats should insist that nobody ever set foot in the United States illegally.

• Bias immigration towards workers with skills we need. At the high end, technically skilled immigrants create higher-paying jobs for Americans already here. They are a positive-sum force in the economy, which is why we should staple a green card to diplomas earned by overseas PhD students. We should admit other immigrants when high wages alone don’t solve labor shortages. Some of these jobs are essential but less skilled, like farm work, home care, or child care. Others are skilled construction trades jobs. For example, immigrants fill about 28% of direct care jobs, even as the hourly wage for a home aide rose from $23 to $34 between 2019 and 2025. The Careworker Visa Act creates a visa for childcare and eldercare workers with wage standards and anti-retaliation protections. 5

  The point is to admit immigrants based on what they contribute to American working life, as Canada does. Democrats should be clear: like citizens of every other country, Americans get to decide who we want to invite to live here based on who we think will contribute most.

A relentless focus on affordability will not come naturally to many Democrats. Many understandably want to focus on the White House and on pointing out that the current occupant is ethically, temperamentally, and cognitively unfit for the job.

But it is dangerous to confuse Trump’s crushing unpopularity with deep political support for the Democratic party. The Democratic brand remains deeply unpopular. It is so toxic that three red state candidates for Senate, Todd Achilles in Idaho, Brian Bengs in South Dakota, and Dan Osborn in Nebraska, are running without any party affiliation. Bill Hill is running for Congress from Alaska the same way. Even a majority of Democrats describe their party as “elitist”.

Many Dems sensibly want to use Congress to embarrass the president and tie his hands. But the single most important task of a Democratic Congress will be to show the American people that it understands their economic pain and is prepared to respond to it urgently. Doing this builds the credibility Democrats need to hold Trump accountable and defeat his successor in 2028.

Subscribe now

Share

The half-life of conventional wisdom may be shrinking. In early 2026, Ukrainian defense against a Russian invasion showed tangible progress; Iran was withstanding US attacks and choking off global oil supplies; and Donald Trump and Benjamin Netanyahu were having their way with the world. Today, none of these are certainties.

Darkness Over Ukraine

Although Ukraine is holding the ground war for now, the drone war has shifted fundamentally. Ukraine cannot consistently intercept Russia’s new low-cost “Jet Shaheds” from the ground. Unlike classic Iranian loitering-munition drones cobbled together out of Styrofoam and lawn-mower engines, the new jet-powered Shahed drones function more like cheap cruise missiles. They fly much higher and faster and can outrun the anti-aircraft guns that Ukraine relied on to knock down over 90% of incoming propeller drones. Ukrainian interception rates have fallen to about half; Russia now regularly hits Ukrainian population centers, so the civilian exodus from the country is accelerating.

Ukraine’s drone and defense sector punches even further above its weight than Israel’s does. The Pentagon could kill three birds with one stone (reduce costs, help an ally, and retool the US military) by outsourcing huge swaths of US military procurement to Ukraine. Kyiv, meanwhile, is developing new ground turrets, advanced electronic countermeasures, and a small missile called Nika and will deploy them soon.

This isn’t over, but without foreign help, the Ukrainian economy may collapse. Russia is systematically targeting the country’s economic engines. They have destroyed more than 70% of steelmaking, shut down enough grain shipping ports to threaten global food supplies, and targeted data centers and energy infrastructure as winter approaches. Ukraine has retaliated by knocking out most of Russia’s oil refining.

Meanwhile, Ukraine’s feckless allies have restricted Ukraine from using Western weapons inside Russian territory. They discourage strikes on oil assets because they prefer stable retail fuel prices. This makes it tough to take out a sprawling Russian oil refinery that requires a drone flight of several hours deep into the Urals or Western Siberia. Russia has become much better at electronic jamming and protecting its most sensitive refineries.

The stakes are very high. Without a massive increase in financial aid and European boots on the ground, Ukraine could run out of the resources it needs to sustain the war another year. If Ukraine were to fall, a colonized Ukraine would hand Russia a military-industrial powerhouse. Many experts expect Russia to target Moldova, the Baltic states, and Denmark next. This would completely fracture the European collective security framework, since EU nations are failing to build defensive capabilities to match the reality of modern drone warfare (they did approve a 90 billion Euro loan to Ukraine in April, but money is not mobilization).

As Europe slowly realizes that Russia’s attack on Ukraine is an amuse-bouche for Putin’s planned banquet, they are badly divided on how to respond. The pro–Putin AfD is ascendant in Germany; pacifists prevail in Ireland and Switzerland; and France is shut down because (checks notes) um…why exactly is France shut down?

Iran Tests Its Pain Threshold

Just because it was idiotic for Donald Trump and Benjamin Netanyahu to start a war with Iran does not automatically mean they will lose. During the last 60 days, Iran’s economy has experienced serious inflation, a currency collapse, and new waves of international financial pressure.

The Iranian rial has hit historic lows. A year ago, a dollar bought a million Iranian rials. Now it buys 2.7 million, meaning the currency has lost more than half its value. Iran’s own data show consumer price inflation approaching 90% this year, with grocery prices even higher. Everyday goods are becoming unaffordable for many families.

The US Treasury has aggressively targeted foreign financial institutions, regional banks in Turkey and the UAE, and the networks of shadow banks used by Iran to launder oil revenues and gain access to US dollars. They have cracked down on foreign financial institutions linked to covert networks used by the Islamic Revolutionary Guard (IRGC). This has choked the inflow of US dollars and UAE dirhams to Tehran, helped depreciate the rial, and fueled domestic inflation (something Trump is proving good at).

Meanwhile, the IMF projects that Iran’s real economy shrank 5-6% this year because of shipping restrictions, a shrinking oil sector, and severe infrastructure damage. This has fueled a massive internal budget deficit and severe shortages of critical imports like medicine.

The Iranian regime is obviously both ruthless and resilient. The formal economy may collapse within a year, but this hardly means that the regime will. They have a proven ability to transfer resources, including aid from China, to the IRGC and transfer pain to everyday people. They can create hawala currency exchange houses and front companies outside the reach of formal banking regulations. They can sneak oil to China’s teapot refiners. But the economic pressure on Iran and the Iranian people is crushing and has become much worse.

Donald and Bibi: Going Down?

During the past six months, Donald Trump and Bibi Netanyahu have both encountered several high-profile policy reversals and mounting scrutiny of their personal ethics. Both face looming electoral defeats.

• Trump’s war, tariffs, inflation, corruption, support for AI, and ICE raids have become incredibly unpopular. He has done an about-face on mail-in voting and now actively encourages Republican voters to make use of mail-in ballots. He may not like the results.

• Trump now resembles Joe Biden: old, stumbling, and out of touch with American living costs. He has no off-ramp to the war he started, and Netanyahu has no interest in helping him to find one. Prediction markets foretell a very tough 3-4 weeks for both men. Many expect Bibi to lose (don’t count on it), and PredictIt gives 60% odds that Trump will lose control of the US Senate. (If Trump loses the Senate, expect Samuel Alito to resign the Supreme Court within days so that lame-duck Senators can confirm his replacement.)

We are well past peak Trump. His speaking events have taken on the tone of an oldies band’s farewell tour. At the start of Trump’s term, most polls showed that most voters viewed Trump favorably. Today, every poll shows large majorities disapproving of his performance by large margins. Presidents usually become less popular before a midterm election, but Trump’s ratings are in the territory of Presidents Truman and Bush, both of whom were neutered by midterm voters.

In both politics and warfare, momentum can be treacherous. Leaders work hard to build momentum, and it often leads them to overreach. Tailwinds suddenly become headwinds, testing leaders’ endurance and resolve. Zelensky, the IGRC, Trump, and Netanyahu are all facing setbacks. Count none of them out.

Subscribe now

Share

In 1904, American socialist Upton Sinclair went underground in the Chicago stockyards to expose how badly meatpackers were treating their immigrant workers. After working for seven weeks in a slaughterhouse, he published his famous exposé, The Jungle.

Americans ignored the plight of meatpacking workers but panicked at Sinclair’s description of the spoiled, filthy, contaminated meat that was ending up on their plates. “I aimed at the public’s heart”, Sinclair later said, “and by accident I hit it in the stomach”. Within two years, Congress had put federal inspectors on the slaughterhouse floors.1

AI is now having its Jungle moment. Every podcast and editorial page is reporting fights over data centers, fears of AI-induced job losses, or warnings of an AI-triggered apocalypse. Washington, Sacramento, and Albany are trying to write rules without any agreement on what those rules should do.

Three debates are driving this fight: How risky is AI? Who sets the pace? And who enjoys the benefits?

Three Debates

How dangerous is AI? Is advanced AI a general-purpose technology that we will absorb over several decades, or an agent that could outrun human control?

This is a core disagreement. Princeton’s Arvind Narayanan and Sayash Kapoor see AI as a “normal technology”. They argue that institutional bottlenecks slowed the adoption of electricity over generations, and conclude that AI will do the same. They do not argue that AI is harmless – simply that its costs and benefits closely resemble earlier general-purpose technologies.

Yoshua Bengio, Geoffrey Hinton and many safety researchers in frontier AI labs disagree. They see AI models as goal-directed systems that can improve themselves. To them, AI looks less like software or even technology. It is more like an emergent form of life.

This question determines your sense of urgency about AI policy. But both camps should want the same first step: a way to measure what AI systems actually do and establish rules of accountability.

Who sets the pace? Would any American regulation automatically favor China, or does a race without regulation make winning dangerous?

Many observers see the race to self-improving AI as a race for global domination. From their perspective, any regulation will slow America down and hand China the lead from which the US may never recover. Opponents counter that Beijing already regulates its own models more tightly than the US does. They argue that nuclear arms-control treaties show that rivals can cooperate to reduce catastrophic risk.

A similar fight takes place at home. New York’s RAISE Act and California’s SB 53 already require frontier developers to publish safety frameworks and report critical incidents. These laws have not slowed major AI labs. Nonetheless, the AI industry wants Washington to preempt laws like these. Ideally, we would begin with regulations designed to reveal and contain actual risk, not simply slow the pace of development.

3. Who wins? Who pays? Will the gains from AI be shared broadly, or will they concentrate wealth and power in a few companies and pass the cost of failure onto the public?

A handful of firms now dominate AI data centers, models, and policy. Their spending on data centers is growing faster than the revenue they produce, even as public opposition to data centers grows. If the buildout becomes a debt-fueled bubble, pensioners and taxpayers will absorb the losses. If AI eats jobs faster than it creates them, workers will.

The framework I propose leaves this debate for another day. It starts by revealing underlying risks. It also forces those who want to capture AI’s upside to pay for its risks without relying on regulators to oversee every possible problem.

A Pragmatic Start

We have done this before. Instead of settling every debate in advance, we gather evidence, sharpen incentives, and limit the damage a new technology can cause. Every mature industry with products that can kill people uses three tools: disclosure, liability, and monitoring. These elements reinforce each other and, if well-executed, combine to increase public trust in AI regulation.

Today, even corporate frontrunners are campaigning for consistent disclosure and transparency standards. In June, OpenAI published the thoughtful Democratic Governance of Frontier AI: A Blueprint for a Federal Framework. Although it is a corporate document, the blueprint explicitly calls for harmonized legal frameworks to address cyber risks and outlines the need for transparency, accountability mechanisms, and independent oversight. It serves as a good benchmark for what the industry is currently willing to accept regarding mandatory disclosure.

I. Disclosure: measurements that matter

Disclosure comes first. Not today’s voluntary safety testing reports, which inevitably sound like marketing collateral. Instead, we need to set standard metrics that inform policy debates and force companies to confront the social consequences of the risks they take.

Which metrics? The National Institute of Standards and Technology (NIST) has developed an AI Risk Management Framework that creates a starting vocabulary for regulators. It suggests metrics that measure a model’s autonomy, potential for weaponization, and operational risk.

• Autonomous Capability. Regulators need metrics that measure how independently a model can operate. For example, how complex a task can a model complete on its own? Organizations like METR track how lengthy a task an AI model can complete. Task lengths have been doubling roughly every seven months. Can the model being tested acquire additional resources? Can it copy itself? Can it resist shutdown? How often does it try to escape its sandbox using tools like browsers or code interpreters? Like a Volkswagen diesel engine, does it behave differently when it thinks it is being tested?

• Weaponization Potential. These metrics quantify how much a new model creates opportunity for catastrophic harm, such as how much the model helps a novice build a chemical, biological, radiological, and nuclear (CBRN) weapon. Labs must report the measurable difference in time, accuracy, and viability when a non-expert attempts to design a biological pathogen or chemical weapon with the model versus without it. They would report how well a model can find and exploit new CBRN vulnerabilities. Can it convert these findings into a working attack? Tests report improvement in these abilities. A big enough improvement triggers public disclosure and deployment safeguards.

• Operational and Infrastructure Risk Reporting. Regulators need continuous visibility into the physical and procedural realities of model development. We already do this not just with slaughterhouses, but with major banks and airlines. As models grow bigger and more powerful, the disclosure requirements need to increase. There is usually less reason to regulate small models, although this might change. But a training run that crossed a standardized computational threshold would trigger public safety monitoring. California’s SB53 uses 10^26 FLOPs as a reporting threshold. These thresholds age as training gets cheaper, so they may need to be paired with capability triggers.

  AI labs should file near-miss and incident logs, just like pilots do. These reports would include internal testing incidents on models that never reach the public. Labs would formally log vulnerabilities like hacking capabilities that they discover, even if they never release the model under review to the public.

  Labs should also report their security benchmarks. Labs should publicly report their security benchmarks so that, as their models grow, the infrastructure protecting them from theft by state adversaries grows with them. These benchmarks would cover software, like cybersecurity or hardened compliance layers that AI models badly need, and the physical hardening of the data centers that house model weights.

II. Liability: unlearn the lessons of social media

In 1996, long before the rise of social media, Congress passed Section 230 of the Communications Decency Act. It shielded websites from liability for user-posted content. This helped to nurture the early Internet; I founded a used book company that benefited from it directly.

Section 230 created an enduring moral hazard. It let social media platforms capture the upside of engagement while assuming no responsibility for algorithmic harms. We let social media companies pollute our commons and have been cleaning up ever since.

AI is different because its output is the company’s own product. Congress should not grant foundation models safe harbor protection – they should face ordinary negligence and product-liability law. Legal scholars are already considering how this would work. In the Yale Journal on Regulation, Ketan Ramakrishnan argues convincingly that courts and legislatures must abandon safe harbors for AI developers and shift from a negligence standard to strict product liability.

Cyberattacks should receive special attention. If an autonomous agent hacks into a system without authorization, the party deploying the agent would pay a fine. AI labs would make very sure that their foundation models could not pick digital locks. Application companies would make a special effort to not inadvertently modify foundation models to enable hacking. Companies whose agents accessed unauthorized systems would be liable, whether or not the agent steals anything. A $10 million minimum penalty per breach would force every lab and software developer to prevent their agents from hacking.

III. Monitoring: Bring back resident inspectors

Upton Sinclair published The Jungle as a magazine serial in 1905 and as a book in 1906. It was a national sensation that caught President Theodore Roosevelt’s attention. The president sent investigators to Chicago. Their report led to the Meat Inspection Act, which Roosevelt signed within months. The new law put federal inspectors, now part of the USDA, on the slaughterhouse floor.

On-site inspection worked, so regulators began to try it in other industries. The Federal Reserve and the Comptroller of the Currency still embed examiners inside the largest banks to monitor risks in real time.2 The Nuclear Regulatory Commission places inspectors on-site at every nuclear power plant. They oversee systems that are complex, high-risk, and mathematically dense – and they have unrestricted access to the plant. They rarely interfere with operations.

Likewise, both open- and closed-source frontier labs also need public safety monitors with access to training runs, test results, and model weights. As with the NRC inspectors, they need the authority to halt a dangerous deployment before it can harm the public.

The Objections

Aviation and pharmaceuticals are both heavily inspected and highly innovative. Their experience demonstrates that liability, disclosure, and monitoring policies can work without damaging businesses. But they are not risk-free. At least five smart objections deserve answers.

• It doesn’t address the real risks. Product liability is a preposterously insufficient way to address the risk of an AI-designed pathogen that could kill 100 million or more people. Health care technologists like Bill Gates, intelligent observers like Noah Smith, and random Substackers like me take this risk very seriously. The lack of a hardened compliance architecture in today’s open- and closed-source models makes this risk especially worrisome. Unfortunately, we still lack a scientific or political consensus on the nature of AI risk. We can implement disclosure, liability, and monitoring regulations without one.

• Regulation can inadvertently entrench incumbents. Massive frontier labs can afford to house federal regulators and absorb lawsuits. Startups and open-source developers cannot. Critics will argue the three-part policy outlined here will inadvertently kill open-source AI and preserve the oligarchy of a few gigantic frontier labs.

  This is a serious concern. Legislation needs to set high compute and revenue thresholds for regulation. This is why California’s SB53 reserves its heaviest duties for developers with more than $500 million in revenue. We do not want a university lab that is fine-tuning an open model to have to deal with a federal inspector.

• Causation is messy. Tort law allows you to hold a company liable only for harms clearly caused by its product. But AI is developed across multiple companies: a lab trains a base model, a startup fine-tunes it for a specific application, and a user abuses it with a malicious prompt. Who should be held responsible for the harm?

  This is not a new problem. It appears in businesses from auto parts companies that supply Big Three carmakers to drugmakers whose products physicians can misprescribe. Tort law has sorted out chains like this for more than a century. It can do it again.

• Exposing trade secrets. Some critics worry that embedded regulators pose a massive security vulnerability because weights and training algorithms are the jealously guarded crown jewels of frontier AI labs. A regulator who is modestly paid compared with the engineers they oversee, yet enjoys unfettered system access, is a natural target for foreign espionage. But bank examiners and nuclear inspectors already handle sensitive secrets. They hold clearances, work on site, do not take materials home, and go to prison if they leak. AI monitors should work the same way.

• Government is too slow. USDA inspectors and bank examiners monitor processes that are well-defined and relatively stable. Meat and bank ledgers don’t change much. But AI capabilities change monthly, and federal pay cannot match Anthropic’s. Critics will argue that federal monitors will lack the technical talent and operational agility to audit frontier models in real time.

  So borrow talent like the US Digital Service and the Defense Digital Service already do. Embrace secondments – giving technically talented engineers an opportunity to serve a 1-2 year tour of public duty with their equity still vesting before returning to their labs. Recruit heavily from the platform safety teams at frontier AI labs – many of whom want exactly this job. Strict cooling-off rules and selecting candidates from several competing companies can keep the revolving door of seconded engineers from descending into regulatory capture.

Aim for the Stomach

This triad of policies will not decide whether AI is a normal technology or a new species. It will not determine whether China should govern the pace of AI development or whether AI investments are a bubble. It is not designed to do these things. But it can establish the institutional muscle required to act if a genuine disaster were to arrive.

Upton Sinclair accidentally reformed an entire industry by enabling the public to smell the stench of contaminated meat on their dinner plates. AI has not yet found its stomach—that visceral, undeniable crisis that turns abstract anxieties into immediate political demands. When that moment arrives, we do not want to debate philosophy. We want metrics and rules in place and inspectors at the ready.

Subscribe now

Share

I remember learning that all of the profits from video rentals came from late fees. This means that Blockbuster was never in the business of helping you watch movies. They were in the business of hoping you would forget to return your VHS cartridge. It was a predatory model, and I was not sorry to see them go.

Many companies earn money in ways that are invisible to their users. Often a visible business builds a trusted brand and attracts customers, while a hidden one generates profits. In many cases, the hidden business earns so much more than the visible one that it really is the company.

Hidden businesses take many forms, but most of them are disguised ad agencies, banks, or real estate businesses.

Hidden Ad Agencies

A business that knows you want to buy something has a choice: it can sell you the item you want, or it can charge someone else for the privilege of showing it to you first.

• Social media platforms pioneered hidden ad models. Meta opens its annual report by declaring its mission to build the future of human connection. They offer users a place to make new friends, keep up with existing ones, and see media curated to your interests. Financially, however, Meta users’ attention and data are inventory. As the saying goes, if you are not the customer, you are the product. Advertising produces more than all of Meta’s profits. Google’s ad share is falling as cloud services grow, but in 2025 it was almost $300 billion of the company’s $400 billion in revenue. The same goes for Reddit (94% ad revenue) and Snap (89%). At their core, these are ad agencies.

• Retailers have copied this model. Social platforms sell your attention; retailers sell your purchasing intentions. We think of Amazon as a retailer, but selling stuff is a notoriously low-margin business. Advertising is not. Amazon’s ad revenue jumped to $68.6 billion in 2025. Although the company does not disclose profits from this revenue, they almost certainly exceed the $34 billion Amazon made in retail. This helps explain why ads now enshitify your Amazon search results.1

• Home delivery companies also operate hidden ad agencies. Grocery delivery is a demanding, low-margin business. But customers often become loyal and can be monetized with ads. In 2025, Instacart’s ad revenue was $1 billion, which effectively accounted for its entire cash flow. Uber and DoorDash may look like logistics companies, but like retailers and social media companies, they are ad agencies at their core.

Hidden Banks

Like ad agencies, retailers and delivery apps convert your attention into advertising dollars. But airlines, hotels, and coffee shops convert your payments into an unregulated bank. They monetize your payments and credit and issue currency that they control.

• Airlines are the most glaring example. American Express paid Delta $8.2 billion in 2025 for originating Amex travel cards—a figure that dwarfed the airline’s $5.8 billion operating profit. When analysts stripped loyalty revenue out of airlines’ 2024 results, the operating margins for Delta, United, American, and Southwest all turned negative. European carriers, hindered by EU caps on credit card interchange fees, actually have to make money flying airplanes. American carriers do not. They mint currency (miles), sell it wholesale to Wall Street, and let merchant fees subsidize the unprofitable business of aviation.2

• Hotels and coffee shops want in on this game. For example, Starbucks operates a massive unregulated bank. Customers pre-loaded $4.3 billion onto Starbucks cards in a single recent quarter. They effectively loaned the company $2 billion interest-free. Better yet, customers lose their gift cards (a term the industry touchingly calls “breakage”). Add the interest on the float to money gained from lost gift cards and Starbucks quietly pockets roughly $300 million a year for doing nothing. It’s a bank with no FDIC insurance costs that makes money when a customer defaults by losing their gift card.

  Marriott’s co-branded card fees rose over 8% to $716 million in 2025. It expects these fees to grow about 35% this year by charging card issuers more. Marriott franchisees want in: 51 owners of 990 hotels wrote to corporate leadership in March demanding a share. They know that renting rooms is hard work. Collecting fees from affiliate cards is much easier.

  Even a fading retailer like Macy’s makes more money from branded credit cards than from selling clothes. Their credit card revenue ($669 million) eclipses their net income ($642 million) – meaning that Macy’s is no longer just a retailer. It is a hidden bank.3

Hidden Real Estate Companies

You don’t need digital data or a branded credit card to run a hidden business. Sometimes, the most lucrative asset is the ground the customer is standing on.

• Fast Food. In the 1950s, McDonald’s CFO openly admitted the company was in the real estate, not the hamburger business. It remains true today. McDonald’s owns the land under most of its restaurants and collects more in rent than in hamburger royalties from its franchisees. Because the lease and franchise come as a package, McDonald’s locks in captive tenants who cannot switch landlords. Jack in the Box and Wendy’s follow the same model.

• Casinos. Similarly, when Caesars was forced to split its physical properties from its casino operations, the market revealed that the buildings themselves were far more valuable than the gambling business.

The Tail Wagging the Dog

It is tempting to argue that these hidden businesses are a net positive for consumers. After all, Amazon’s ad revenue helps subsidize Prime shipping, and airline credit cards keep flight routes open that might otherwise close.

But hidden businesses operate largely without the regulations that govern traditional industries. Starbucks takes deposits without deposit insurance. Retail media networks operate without the disclosure norms of traditional advertising. Airlines issue currency and devalue it whenever they like (credit Pete Buttigieg for using the DOT to investigate this).

More importantly, a fast-growing hidden business inevitably degrades the visible one. Delta will adjust its cabins and routes not to serve passengers, but to sell more Amex cards. Amazon will worsen its search experience and enshittify its site to cram in more sponsored listings. Eventually, the hidden business steers the visible one. And much like Blockbuster hoping you forget to return your VCR tape, today’s hidden businesses aren’t designed to help; they are designed to exploit.

Share

Subscribe now

As the Great Financial Crisis descended in 2009, Peter Thiel gave up on democracy. The PayPal multimillionaire and first investor in Facebook declared democracy “incompatible with freedom”. By freedom, Thiel meant freedom from government rules and taxes. His notorious essay blamed his libertarian woes on the extension of voting rights to women and welfare beneficiaries.

In his own imperious way, Thiel was highlighting a real dilemma. In most economies, a small group of highly paid people earns much more than folks in the middle. In the US, those who earn $450,000 or more per year (the top 1%) take home 15% of all pay. This share has doubled since 1979. The top 10% ($200,000 and up) receive 40% of all pay. Economists say that income skews right – and wealth famously skews even further right than income.

As a result, median (midpoint) incomes are lower than mean (average) incomes. A handful of professional athletes, rock stars, or senior law partners making $25 million/year raised the average 2024 pay for a full-time US worker to $84,603, but this had little effect on the median income of $65,000. Obviously, the gap between mean and median incomes widens as high earners capture more of the income pool.

How much does this matter? One view says it’s irrelevant because high earners are not a club with a permanent membership. There is a lot of churn, especially at the very top. The vast majority of people who reach the top 1% or the top 10% of income earners do not stay there for long. Research by Thomas Hirschl and Mark Rank found that more than 53% of American workers spend at least one year in the top 10% of income earners over their working lives, but fewer than .6% of Americans manage to stay in the top 10% for ten consecutive years.

On the other hand, the top 20% churns at a much lower rate. Economists Gerald Auten and Geoffrey Gee found that two-thirds of taxpayers who reached the top 1% were out within a decade. But they did not fall far. Nearly 70% of those who reached the top 1% remained in the top 10% two decades later, and about 70% of those in the top quintile remained there a decade later. The Federal Reserve found similar results using more recent data. So although membership in the very top income bracket is often fleeting, the US manages to produce a class of affluent, professional-income households that is durable enough to form stable political interests.

This group now controls the Democratic party.

Does Democracy Hurt High-Income Voters?

In a democracy, median voters decide elections. This means that voters in a country with a large gap between median and mean incomes have an incentive to elect politicians who will tax the rich and use public services, infrastructure, or direct transfers to redistribute wealth downward. This is less true if income is shaped like a bell curve, in which median and mean incomes converge.

Peter Thiel is not the only high-earner who assumes that democracy shifts political power towards poorer voters, leading to higher taxation of income, if not wealth. Scholars have long theorized that for this reason, economic elites have a rational incentive to suppress democracy to protect their capital.1

But is it true? Does democracy actually expose economic elites to the risk of redistribution via higher taxes, weaker property protections, or more competition? If it were true, then assets held by high earners should lose value as a country becomes more democratic or as democratization becomes more likely. Markets will price in the risk of future redistribution, so researchers should be able to use asset prices to test Acemoglu and Robinson’s highly influential theory that elites use democracy to avert revolution by shifting legal power to median voters.

Recent research by Max Miller at Harvard Business School tested this. He found that democracy really is more expensive for high earners. He matched stock market data on equity returns from 90 countries from 1817 to 2018 against a database of regime transformations to identify “democratic episodes”. He exploited the Vatican II endorsement of democracy as a shock to Catholic autocracies, and used an instrumental-variables approach with “waves” of democratization to build a theoretical model to interpret the magnitudes.2

Miller found that markets treat democratization as equivalent to a financial crisis – meaning investors demand a much higher risk premium to compensate for the risk of future redistribution. Dividend yields typically rose about 20% in the five years before a country democratizes. Neither coups nor political crisis produced as big a reaction as the threat of democratization. Catholic-majority autocracies saw excess stock returns fall by roughly 6–11 percentage points after papal pronouncements during Vatican II simply raised the odds of democratization.

Were markets rational or just gripped by fear? Maybe both. Miller shows that democratization led to real redistribution: the government's share of GDP rose by about 4.8 points, the Gini coefficient fell by about 2.3 points, and labor’s income share rose by about 6.7 points. This suggests that investors’ fears were rational, not just sentimental. Miller’s analysis suggests that most (~58%) of the risk premium was due to taxes and redistribution. The rest was due to increased competition that eroded government-protected rents.

This is not surprising, but why does it matter? Because the world’s oldest democracy has come to rest on two parties, one increasingly authoritarian and the other built to serve the interests of high-income voters. This is unsustainable, and it matters enormously.

When Did Democrats Become the Party of High-Income Professionals?

Before the 1980s, the Democratic coalition remained grounded in the New Deal and was recognizably working- and middle-class. Its backbone was non-college-educated white voters, who made up two-thirds of Democrats as recently as 2016. The big tent held unionized factory workers, urban ethnic minorities, Southern Dixiecrats, and Prairie populists. Money came from private sector union PACs, individuals, and traditional liberal donors.

During the 1980s, deindustrialization and globalization shrank the unionized manufacturing workforce that anchored the Democratic coalition (and progressive coalitions in Europe and Canada as well). At the same time, the expansion of higher education created many more college-educated professionals.

This educated population was not homogeneous. One group of grads, consisting of teachers, journalists, nonprofit workers, and government workers, was more hostile to the private sector, more female, and more unionized. They moved steadily left. College-educated business and technical professionals remained more moderate as a group. Over time, this branch of the party became anti-business and placed increasing emphasis on cultural and social issues. This made the party more attractive to educated professionals and donors – and less visibly working-class in its economic and cultural identity.

Like a Hemingway character going bankrupt, this transition was gradual, then sudden. In 2008, 36% of Republican white voters were college-educated, compared with 32% of Democratic voters. By 2020, 52% of white Democrats were college-educated – a 20-point swing in twelve years. By then, college-educated whites outnumbered non-college whites among Democrats for the first time (remember, only 40% of Americans have a four-year college degree). According to Pew, white non-college voters now make up 57% of the Republican coalition, while Democrats increasingly rely on college grads. The gap is especially pronounced among women.

Ruy Teixeira, whose 2002 book with John Judis, “The Emerging Democratic Majority”, predicted growing Democratic strength, has become one of the most prominent critics of this trend. He points out that the education gap between college and non-college Democrats hit 27 points in 2024, more than double its 2016 level when Trump beat Hillary Clinton. French scholar Thomas Piketty coined the term “Brahmin Left” to describe progressive parties around the world that college graduates dominate.

Political fundraising reinforced this trend. As union density fell, labor PAC money dried up, especially from unions representing private-sector workers. Wealthy donors from Wall St., Silicon Valley, and Hollywood – and the law firms they spawned – took their place. The Supreme Court’s Citizens United decision removed limits on most PAC spending, which accelerated this trend. About 20 million people contribute money to campaigns, but 12 megadonors accounted for 7.5% of all federal political contributions between 2009 and 2020. Roughly 20% originated from less than 1% of the population living in the top 100 highest-giving U.S. ZIP codes.

Even the party’s small-dollar donor base, which Democrats often present as a counterweight to megadonors, skews “wealthier, more educated, older, and whiter than the average American”, according to Brookings. This ideologically energized small donor class has pulled Democratic legislators much further left on racial, immigration, transgender, and climate issues and further disconnected them from working-class and non-white voters.

Compare this with how well non-college voters believe that Democrats speak to their actual priorities.

Democrats Have Drifted Away from Redistribution

A party reveals its priorities not in its slogans but in its spending and actions. Democrats still favor redistribution, but the party’s attention has drifted elsewhere. Political scientists tracking Democratic platform language since 2012 find that references to “growth” have dropped 32%, while “climate,” “LGBT/LGBTQI+,” “equity,” and “environmental justice” each rose by triple-digit percentages.

Modern Democrats have a different relationship to redistribution than their predecessors. The wealth of many college-educated professionals lie in credentials, skills, experiences, and networks of professional relationships. These assets are largely insulated from the threat of income redistribution because it is portable across state lines, industries, and national borders. No state can easily seize, tax, or redistribute what is inside a software engineer or an attorney’s head.

One telling indication of this is the growing number of US college grads who have moved overseas. Although a small share of the total population (5.5 million Americans were living abroad as of late 2024), this number is growing quickly.3

• In 2025, 2.2 million people left the US, and 180,000 of them were US citizens – the first time America experienced a net loss of citizens since the Depression.

• A 2026 survey indicates that serious intent to relocate internationally has quadrupled year-over -year. Nearly every state in the European Union reports record numbers of Americans arriving to live and work.

• Over 90% of Americans living abroad have at least a bachelor’s degree. They are not mostly retirees: mid-career professionals aged 35 to 54 account for 41% of overseas Americans and are the fastest-growing group of emigres.

• Renunciations of citizenship have skyrocketed. The US Treasury reports 30,000 people are now in line for renunciation appointments. Gallup reports that serious intent to relocate quadrupled from 2025 to 2026, especially among women.

In contrast, the conservative “Merchant Right” is much more likely to attract owners of auto dealerships, regional contractors and developers, and businesses that depend on resource extraction. These companies rely on fixed physical assets and local regulatory capture. They are highly vulnerable to taxation and redistribution because they cannot move to Portugal.

The Case For Predistribution

So-called “predistribution” policies have very different economics from traditional redistribution. The difference is when and how the state intervenes to address inequality. Traditional redistribution accepts the existing division of profits between capital and labor and taxes economic winners to compensate losers. Predistribution directly challenges the labor/capital division by seeking to increase labor’s bargaining power so workers capture a larger share of corporate revenues upfront.

Redistribution and predistribution rely on very different policy mechanisms.

• Predistribution policies include steps like raising minimum wages, strengthening collective bargaining rights, reforming corporate governance (e.g., codetermination or tying executive pay to workers’ pay). It also includes antitrust enforcement and investments in public education and skills training.

• Redistribution policies rely on fiscal policy. Its tools include progressive income and wealth taxes, universal basic income (UBI), food assistance, housing subsidies, and state-funded healthcare.

Both redistribution and predistribution policies aim to create a more equitable society, but they rely on fundamentally different economic mechanisms and carry very different political implications.

• Ex post vs. ex ante. Redistribution is a two-step process. It allows the market to allocate income and wealth according to its own logic then uses taxes and transfers to correct inequalities ex post, or after the fact. Predistribution intervenes in market rules to ensure a fair distribution of wealth before taxes and transfers. It aims to fix the market ex ante so that it naturally produces more equitable outcomes.

• Economic visibility. Redistribution is highly visible. Taxpayers can see money leave their paychecks – and it often generates political resentment. But predistributive policies are embedded in the “rules of the game.” When a worker earns a higher wage due to strong collective bargaining or minimum wage laws, they view it as earned income rather than a government handout. It is much less susceptible to the political stigma associated with welfare.

• Agency. Many voters see redistribution as paternalistic. They resent their taxes being used to “pay losers”. In contrast, predistribution is typically seen in terms of power and dignity, as when laws strengthen unions or democratize ownership.

• Political difficulty. Predistribution is often a heavier lift because it is less flexible. A government can tweak a tax bracket or expand a tax credit in a single bill. Still, predistribution often requires taking on entrenched financial interests to reform labor laws or break up powerful monopolies.

Acemoglu and Robinson asserted that elites suppress democracy to protect their capital. Miller showed that historically they have had good reason to do so. But the modern Democratic Party has found a way to avoid the challenge of redistribution completely: it simply changes the subject from class conflict to cultural progressivism. This enables the high-earning coalition to protect its economic privilege, neutralize the threat of redistribution, and maintain the moral high ground.

Not surprisingly, the Brahmin left has given up entirely on predistribution, even though, unlike the asset-bound “Merchant Right”, they have little to fear from the redistribution they nominally still promise. By changing the subject, Democrats have not just insulated their affluent base from the threat of redistribution; they have built a coalition willing to leave working-class voters to fend for themselves.

Subscribe now

Share

China’s flooding of global markets with electric vehicles has created a cascade so severe that The Wall Street Journal reports there are not enough ships to get the 10- 12 million cars it wants to export out of the country.

We have all heard praise for the far-sighted industrial policy that made this happen. Beijing technocrats sagely targeted batteries and EVs, spun up state-owned car companies, showered them with subsidies, dictated the terms of joint ventures with foreign carmakers, and executed a twenty-year plan with a discipline and patience that democracies cannot muster. I have hinted at this myself.

New Research Tells a Different Story

A new paper in The China Journal tears this story apart. Fengming Lu spent three years interviewing the officials, engineers, and entrepreneurs who built China’s EV industry. He and his co-author were trying to solve a puzzle. They knew that in 2014, China’s ten largest automakers were all joint ventures with foreign firms. All but one of the Chinese companies were massive state-owned enterprises (SOEs), not private firms.

The research question was why the story had completely changed ten years later. Take a look at the 2024 table.

The private firms like BYD, Chery, Geely, and Great Wall that took over the market did so despite provincial governments investing heavily in their state-owned competitors. Together with Beijing, they awarded nearly every early EV research grant to state firms. The flagship pilot programs were built around them. But the companies that won were the ones that Beijing had initially ignored.

The researchers wanted to figure out why. They concluded that the EV boom grew not from central industrial planning, but from three intersecting forces. Central rules that favored a few large incumbents had actually locked most Chinese cities out of the car business. As a result, dozens of local governments that were desperate for tax revenue wanted in. They teamed up with private investors to create the companies that actually won. Except, of course, that winning means flooding overseas markets, which imposes massive costs.

“Go To The Mountains”

As is so often true in China, the story begins with Mao Zedong. Not the wily revolutionary, who spent his middle years fighting the Japanese from the mountains of Shaanxi province in northwestern China, but the aging and delusional leader trying to govern the world’s most populous country.

Mao was a strong guerrilla leader, and remained one even when trying to govern. His 1958 “Great Leap Forward” was a campaign to rapidly transform China into a rich, modern industrial society. Mao forced peasants into communes and required them to build primitive metal furnaces to make steel. He urged the wholesale slaughter of sparrows, which he thought ate too much grain. This led to massive swarms of insects that ruined crops and left people starving. The resulting chaos and famine killed between 15 and 45 million people – one of the worst disasters in human history.

By 1964, seeing that the Great Leap had taken China backward and not yet inspired to launch the mass psychosis known later as the Cultural Revolution, Mao fixated on the risk of nuclear war. He declared that China needed a Third Front – a defense area safe from Russian or American atomic weapons. China must “prepare to go to the mountains”, meaning relocate entire industries into the wilds of western China.

China gave it a try. They moved nearly 4 million workers and family members and diverted 40% of all capital construction to the construction of more than 1,100 large- and medium-sized industrial and mining enterprises. They built more than 8,000 kilometers of railroad track in the mountains of western China, which Mao deemed safe from atomic bombs.

The result was a massively fragmented auto and steel industry. Instead of expanding existing coastal steel mills, Mao ordered the construction of an enormous steel complex in Panzhihua, a remote canyon in Sichuan province, solely because it was buried deep in the mountains and hard to bomb.

The “Second Auto Works” (now Dongfeng Motor) was built in the remote Shiyan mountains. The factory’s assembly lines were deliberately scattered across dozens of different valleys so that if one workshop was bombed, the others could keep producing trucks for the army.

The Third Front (and, to a lesser extent, the Great Leap Forward) created a large population of geographically stranded, vertically integrated, locally protected producers that were individually too small to be efficient and too politically costly to kill. Eventually, the surviving companies moved to larger population centers: Dongfeng to Wuhan, Shaanqi to Xi’an, Hongyan to Chongqing, all abandoning the mountain sites the Third Front strategy had put them in. But the seeds were planted: even the smallest and most remote provinces in China were sure they could build a car company.

Provincial and Municipal Ventures

Wuhu is a middling city in Anhui, a middling province. It wanted to get in on car production, but in the 1990s, central policy reserved passenger car production for six anointed state enterprises and their foreign partners. Until the 2010s, Chinese law protected these incumbents by requiring new carmakers to invest at least 1.5 billion yuan. This kept out the riff-raff.

Wuhu began by convincing First Auto Works (FAW) to open a chassis plant there. When the city sold its cement works in 1996, it used the proceeds to buy a used engine line from Ford of Britain, poached FAW’s engineers, and started its own carmaker. It named the firm Chery and appointed the deputy mayor as chairman. Although the carmaker received grants from Beijing and remained a regional SOE through the 2000s, it eventually raised private capital and became China's second-largest automaker. Anhui provincial and Wuhu municipal governments still own a quarter or so of Chery.

This pattern – a Chinese city that could not get state champions to return their phone calls deciding to back local entrepreneurs to start a new car company happened repeatedly over the next thirty years. Li Shufu struggled to build a company that made refrigerator parts, then decorative building materials, then motorcycles and scooters. By 2002, the company, now called Geely, began assembling sedans. In 2010, they bought Volvo Cars from Ford. When most state banks refused to finance the deal, Geely convinced Goldman Sachs and four city governments to invest. Each city got a Volvo plant or a headquarters building for its trouble.

Rules loosened in 2015, spawning hundreds of startups, many of them EVs. In 2019, only 12 firms actually held government licenses to build EVs. But there was a loophole designed to protect incumbents: a company authorized to build gas cars could convert to EVs. The rule was a golden ticket to every failing regional automaker and led to more than 100 new EV companies.

These companies survive in part because city or provincial governments will rescue struggling carmakers. The city of Hefei invested 7 billion yuan to rescue NIO at the bottom of the market. When its investment doubled within months, Chinese media proclaimed it the Hefei Model, and soon every mayor in China wanted to imitate them.

As large state enterprises began losing money, they also lost support from their hometowns. Beijing’s BAIC was China’s top EV seller from 2013 to 2019 and outsold Tesla globally in 2017. It built the electric taxis that enabled Beijing to hit its air-quality targets, but in doing so they abandoned hybrids and made fewer cars that private buyers wanted. Instead of backing them, Beijing shifted its bets to Li Auto and Xiaomi. Likewise, Guangdong funneled billions into XPeng because local regulators doubted that hometown favorite GAC would successfully manage the transition to EVs. In short, Chinese carmakers competed not only for customers but also for municipal and private investors.

Lessons

The researchers highlight three important lessons from Chinese carmaking.

• Industrial policies benefit from competition. For the industrial policy debate now consuming Washington, it suggests the power of competing investments. China’s real advantage was dozens of governments competing to build winners with their own money and their own consequences. Its advantage was not an enlightened center, which in the end was too rigid to stop them.

• State-owned enterprises rarely work well long-term. The pattern of private-sector collaboration with local governments to outperform more established SOEs is evident in many other sectors. Pharmaceuticals, steel manufacturing, and frontier AI models all fit this pattern.

• Nobody wants the hometown company to fail. The approach carries the risk of massive wasted capacity, price wars, and municipal debt. China currently has 130–150 active automobile companies competing in its hyper-competitive domestic market. The paper mentions that Nantong, a city in Jiangsu, sank 6.6 billion yuan into EV sports car maker Sailin and built few, if any, cars.

Even though dozens of cities are determined not to let their company fail, a massive industry shakeout is underway because Chinese customers have stopped buying. Domestic car sales are down a stunning 20% over last year. Most analysts predict that by 2030, only 5-10 car companies will be profitable. Last year, Beijing warned local governments to stop launching EV projects. Next year’s message is likely to be even tougher.

Subscribe now

Share

No organization in human history has ever spent as much money as the United States Federal government. The feds spend about $7.5 trillion per year – almost a quarter of all spending in our economy. Raising taxes to support this level of spending is unpopular, so Washington borrows 26 cents of every dollar it spends. The result is a federal debt of almost $40 trillion. That’s almost $120,000 per person and well over one year of US economic output.

Forty Months

In 2030, which is 40 months away, a mix of debt and demography will make today’s debates about “affordability” look quaint.

• The debt crunch. In 2030, the US federal debt as a share of the economy will break the record we set during World War II. But during World War II, 85% of federal spending was for warfighting. Today, three-quarters of federal spending goes to Social Security, Medicare, other mandatory programs, and interest payments on the national credit card. Congress only controls a quarter of our spending.

  By 2030, US debt will amount to $300,000 per American household, but no magic alarm goes off on the day we cross this threshold. All that happens is that creditors notice that their risk is steadily growing, so they raise interest rates. This week, US borrowing costs hit a 25-year high as investors grew nervous about Trump’s mounting profligacy. On Thursday, a $25 billion Treasury auction of 30-year bonds fetched yields as high as 5.22 percent. The last time yields were this high, in 2001, the government was running surpluses, so it simply shut down long-term borrowing. The Treasury Department stopped auctioning 30-year bonds for almost five years. Today, the rising rates on long bonds represent a stealth tax increase because they drive up the price of every mortgage and auto loan. Worse, the interest bill grows – and it already costs every US household about $7,600 annually.

• The retiree crunch. In 2030, every retired senior will depend on just 2.5 workers. In 2030, the last of the boomers will finally retire, and for the first time, seniors will make up 20% of the population. For contrast, it was about 12% in 2008. Old people like me not only increase Social Security and Medicare spending, but thanks to declining immigration and birth rates, we push these expenses onto a smaller working population. When I was born, six working people paid taxes to support every retiree. By 2011, there were just four. Today, there are 2.7.

  This isn’t sustainable financially or politically. The US designed social services based on twentieth-century demographics. We assumed that a growing population would continue to drive economic growth and pay these bills. We worried about overpopulation, not underpopulation.

• The demographic crunch. 2030 is also when the CBO projects deaths will exceed births for the first time in US history. A year ago, the agency put that milestone in 2033. It keeps moving closer. From 2030 on, the only source of US population growth will be immigration. Social Security, which has run annual deficits since 2010, is projected to become insolvent in 2032, triggering a 22% across-the-board cut in benefits.

  Medicare is a bigger budget buster than Social Security. Of the projected $138 trillion budget shortfall over the next 30 years, $109 trillion is from Medicare. Rising health care costs, combined with an aging population receiving far more in benefits than it ever paid in taxes, isn’t sustainable. The Medicare Hospital Insurance trust fund is expected to become insolvent in 2033, forcing an 11% cut in hospital payments.

  These baseline forecasts assume no surprises. If the US needs to borrow in response to an unexpected recession, pandemic, or protracted war, the situation worsens faster.

There are only three ways for a country to reduce a deficit this large: grow faster, raise taxes, or suffer inflation. There is no fourth door. The rest of this post looks at this looming trilemma.

Can AI Help the US Outgrow Its Debt?

Can growth help the US manage $40 trillion in debt? Only if the economy can grow more than 4.3% per year – roughly double the recent trend. Is there any technology that could double our growth rate? Probably not, although if there is, it is artificial intelligence.

By 2030, the US will have invested $3 trillion in AI infrastructure, give or take. The US is betting $500 billion to $1 trillion per year on AI, or roughly 2% of GDP.1 We have made massive bets like this before. Measured as a share of GDP, the largest ones have been:

Today’s AI bet is more like war spending than investments in long-lasting railroad tracks, highways, and electrical grids. About half of a data center’s cost is specialized chips that need to be replaced every 3 to 5 years. And unlike past software or internet booms, the physical bottleneck for AI is the capacity of our electrical grid. This forces tech giants to invest directly in new power generation and nuclear partnerships.

We have no idea whether this massive bet will pay off or fail utterly. Forecasts range from apocalyptic doom to growth rates that would render federal borrowing far less problematic. We do not know whether open-source models from China or elsewhere will turn “frontier model” companies into merchants of commodity tokens.

What we know for sure is that AI will be a vastly more powerful and dominant force in 40 months because almost every metric we use to measure AI is growing exponentially.

• Raw compute. The raw compute required to train frontier AI systems doubles every six months or so – far faster than Moore’s Law.

• Task length. Evaluations by safety and research organizations show that the length of autonomous tasks (such as coding or multi-step reasoning) that AI can successfully complete doubles every 7 months.

• Token consumption. Google went from 9.7 trillion tokens in May 2024 to 480 trillion in May 2025 to 3.2 quadrillion in May 2026. That is 330x in two years.

Forecasters keep underestimating AI growth because of three compounding and confounding forces. Cost per token continues to fall faster than expected. Most agents now run continuously instead of waiting for a human to type a request. And tokens per query keep rising as complex tasks generate longer prompts and longer responses.

Fundamentally, however, anything that doubles every six months becomes one hundred times bigger in 40 months. Will 2030 AI be 100 times stronger? And if it is, will economic growth exceed 4% per year?

Not likely. There is a massive difference between computational scaling in a frontier lab and technology dissemination in the wild. While compute inputs may double rapidly, translating scale into breakthroughs in reasoning requires architectural advances. Physical limits such as data center power availability, grid capacity, and chip manufacturing constraints mean that exponential trends will run up against real-world frictions.

One of these frictions is political. Americans are mobilizing against data centers. Economic growth is popular if it increases opportunities and pay — but AI, big tech, and data centers that drive today’s growth are not.2

If we cannot outgrow our debt, the United States will either raise taxes or accept higher inflation, which will show up without a vote and hit low-wage families hardest. Taxes are never popular, but they can be democratic. So which taxes should we favor?

A Framework for Evaluating Taxes

Tax debates are eye-rolling because they are so predictable. First, someone proposes a tax. Then someone else says it will kill jobs, let the rich off easy, or never raise the money that sponsors claim. Often, all three objections are correct – and irrelevant because they critique three different goals. So let’s start with the goals. Warning: tax goals are contradictory.

All taxes are a trade-off between yield, justice, and feasibility.

Yield: can this tax fund the government we want? This comes down to a tax's revenue potential, breadth, and stability during a recession.

• Size. A tax that raises a percent or two of federal revenue is a token gesture. It may be defensible, but the political capital burned to enact it is wasted. For example, inheritance taxes in most rich countries typically produce one to two percent of revenue. You may favor high inheritance taxes (I do, to inhibit dynastic wealth, not to generate revenue), but the fiscal juice is rarely worth the squeeze.

• Breadth. Most of this is what tax nerds call “breadth of incidence”, meaning how many people pay the tax. A broad-based tax generates more revenue at lower rates. These taxes are easier to enforce because there are few or no exceptions. And they are politically easier because there is no small affected group with a concentrated interest fighting a public that will never care as much as the affected group.

• Stability. Finally, we should prefer taxes that are stable across the business cycle. As California has discovered, taxes on corporate profits or capital gains swing violently with the economy. We built our state revenue system on top-bracket capital gains, so every downturn forces spending cuts at exactly the time the state needs to spend more, not less.3 In contrast, regressive VAT and payroll taxes barely move in a recession.

Justice: who should pay, and does this tax make them do so? Justice is a matter of progressivity and horizontal equity, or evenhandedness.

• Progressivity rests on what economists call declining marginal utility. A thousand dollars matters more to someone earning thirty thousand than to someone earning three million. Progressive taxes help counter the tendency for income growth to concentrate at the top, which under a flat tax system yields less revenue over time. And of course, some progressivity is just basic Willie Sutton, who famously robbed banks because “that’s where the money is”.

• Evenhandedness. This just means that two people in identical circumstances should pay the same amount of tax. This sounds trivial, but it matters because far too much of the tax system treats identical economic outcomes differently. A worker who earns wages and a partner who earns distributions owe wildly different sums for the same income. Likewise in California, where Prop 13 limits reappraisals, on a street of houses with identical market values, one neighbor paid property taxes that were one-tenth of mine and another paid 30% more.

Feasibility. Feasibility asks whether a tax is collectible, how it might distort behavior in unintended ways, whether it is durable, and whether the costs of implementing it are acceptable.

• Collectibility. A tax that cannot be collected won’t work. European Value Added Taxes (VAT) are nearly self-enforcing because every business requires documentation from its suppliers to claim its invoice credit. Wealth taxes, on the other hand, can be hard to collect if they rely on valuations of assets without an established market price, such as private company equity, art, or closely held real estate. Every valuation becomes an argument won by the side that can afford better lawyers.

• Distortions. Some taxes change behavior in unintended ways. Payroll taxes are easy to collect, but they raise the price of hiring – something we want more of, not less. If you hire a worker, your taxes go up. But if you buy a machine, you can depreciate it and deduct interest costs, which reduces your taxes. Lawmakers did not intend to subsidize automation relative to hiring – but they did.

• Jurisdictional fit. For a tax to be feasible, it needs to fit its jurisdiction. Land cannot move, so a wealth tax focused on property works well, even at the county level. Income is leaky at the edges, so taxing it works nationally, but imperfectly. Taxing corporate profits barely works at all when a large company can simply shift profits to its Dublin subsidiary.

• Durability. To work, a tax needs to endure. Payroll taxes are distasteful, but they persist because workers and employers believe that they are buying future Security. European VATs are hard to repeal because they are invisible. Estate taxes are the opposite. They are highly visible, targeted at a small number of very affluent people, and often portrayed as threatening family farms or businesses. Legislatures repeal them all the time.

• Transition costs can kill excellent ideas. Since Henry George first proposed it, many economists have favored the land value tax because it applies to all land, making it broad-based. It is minimally distortionary, since land is hard to move, increase, or decrease. It is also hard to evade. The problem is getting it started: it reduces the value of land that current owners bought at full price. A new VAT does the same thing: it produces a one-time jump in prices. These costs are real, concentrated, and charged to people who did nothing wrong. These transition costs are often the real barrier to enacting these taxes. Debates about the steady state are arguments about the wrong thing.

Obviously, yield, justice, and feasibility can conflict. A tax that everyone pays hits the poor. A payroll tax is excellent on breadth and horrible on progressivity. Should progressives always oppose regressive taxes on principle?

Demand Progressive Outcomes, Not Progressive Taxes

No. Americans should judge the progressivity of a tax system, not the progressivity of each tax. Many countries fund progressive governments with regressive taxes. Compared with US progressive income taxes, Nordic VATs are reactionary. But measured by outcomes, these countries are dramatically more egalitarian because a broad, efficient base pays for transfers so large that they swamp the regressivity of the taxes themselves.

American progressive impulses demand that taxes themselves be lighter on the poor than the rich — only to discover that progressive taxes built on a narrow base raise too little revenue and mobilize opposition quickly. Americans end up with a system that looks progressive on paper, but actually collects less than it needs and delivers a thinner welfare state than in countries that impose a 25% tax on groceries.

Let’s apply the yield, justice, and feasibility framework to specific taxes:

• Land value taxes score highly on neutrality and progressivity but fall short on revenue potential and transition costs. It is a genuinely good tax that cannot serve as a primary federal tax. Moreover, enacting it in places without property taxes is very tough because it requires a one-time wealth confiscation that few politicians can justify.

• Payroll taxes bring in a lot of money, are terribly unfair, but are brilliant on feasibility and durability. This is why payroll taxes are a revenue workhorse that we cannot get rid of. We should fix the distortion by equalizing the treatment of labor and capital rather than by shrinking the base.4

• Income Taxes. Income taxes are moderate in breadth because, in a typical year, about 40 percent of filers owe no federal income tax. The base narrows even further in recessions. But the revenue potential is high — it delivers roughly half of federal receipts. Stability is mediocre because it depends heavily on capital gains paid by top-bracket earners, so it swings with asset prices.

  But it is the easiest tax to make progressive. A graduated rate schedule can be tweaked to become more progressive or regressive without touching anything else. Unfortunately, income taxes fail the fairness (“horizontal equity”) test. Two people at identical economic income owe wildly different amounts depending on whether the income arrives as wages, carried interest, qualified dividends, or S corporation distributions. That gap has nothing to do with ability to pay.

  Administrability is excellent for wage income and poor everywhere else. Tax withholding from paychecks means that W-2 pay is almost perfectly compliant (credit in part a very young Milton Friedman, who helped design withholding taxes as a Treasury intern during World War II).5 Compliance with pass-through and proprietor income is far worse and represents perhaps the largest single source of unfairness in our tax system.

• Wealth taxes score high on progressivity and low on administrability, which drags down everything else. Most OECD countries that have tried wealth taxes have repealed them for this reason. The moral argument for wealth taxes is compelling: since 1990, in nominal terms, paychecks have grown by 18%, homes by 66%, and stocks by 1,500%. But no amount of moral argument overcomes the feasibility problems that plague wealth taxes.

  That said, treating borrowings against stocks as realization events subject to income taxes would be easy to administer and fair, since these loans provide current income to wealthy shareholders.

• Inheritance taxes are a very tempting solution for those of us who oppose the accumulation of dynastic wealth. Note, however, that the top estate rate was 77 percent from 1941 to 1976, yet the tax produced only 1 to 2 percent of federal tax revenues. Very high statutory inheritance rates have never produced large tax collections in the United States.

• Federal taxes on data centers. Data centers are immobile once built and score well on incidence, which is why local property taxes on them work well. A steep federal tax would prevent jurisdiction shopping and could be used to fund infrastructure build-out. Data centers employ relatively few workers once built, so taxing them does not threaten jobs. This tax would be quite popular, but fiscally irrelevant. An energy-based excise tax, restrictions on asset depreciation, and a gross receipts tax combined would generate perhaps $50 billion today.

  Even if the whole package were tripled by 2030, it would cover less than 5% of the projected federal deficit — serious money, but not a fundamental change.

I plotted the relative progressivity, incidence, efficiency, and revenue potential of different taxes on a grid, using Claude as a research assistant. It’s not perfect, but it helps to illustrate the trade-offs.

The Taxes Nobody Votes For: Tariffs and Inflation

So far we have evaluated taxes that Congress might pass. But two taxes shown above need no Congressional vote and never appear in the tax statistics. Tellingly, tariffs and inflation are the two taxes America is actually raising right now.

Tariffs. Tariffs affect everyone – they carry a broad economic incidence. They hit roughly 330,000 importers of record, who pass the costs through to consumers. Add the exporters hit by retaliation and downstream manufacturers buying intermediate inputs, and tariffs affect a lot of people. They are very efficient and collectible, which is why tariffs were the first federal tax. Collection happens at the border, prepaid, on a bonded party, with no filing season. For most of the nineteenth century, they accounted for about 90% of federal revenue for exactly this reason.

But as a tax, tariffs are self-defeating. Like tobacco taxes that fund anti-tobacco campaigns, if tariffs work, they disappear. Tariffs that successfully promote reshoring or steer consumers away from high-duty imports go away.

Tariffs are highly regressive. The Tax Foundation puts the average tax increase at about $900 per household in 2026 – and goods are a larger share of the budget for households at the bottom.

And as revenue generators, tariffs have disappointed. Calendar 2025 customs duties totaled $264 billion, up from $79 billion in 2024, with the average effective rate reaching its highest level since 1947. That is five or six percent of federal revenue.

But tariffs have not proven durable. In February, the Supreme Court held that IEEPA does not authorize the President to impose tariffs. The court concluded that the Constitution vests the taxing power in Congress, not the President. Since the decision, the Court of International Trade has ordered customs officials to refund roughly $165 billion through a new system built for the purpose. CBP has already refunded $100 billion in tariff payments. Trump, meanwhile, is appealing to limit who gets their tariff payments back.

Trump has learned the hard way that courts can repeal taxes that were never enacted by Congress. As a result, the federal government finds itself refunding revenue it has already spent. Unlike any other tax on this list, tariffs can be clawed back even after they are collected.

Inflation. Inflation affects everyone, so tax policy wonks would declare it to have the broadest incidence of any tax. There is no exemption, no threshold, no filing status. It is also very efficient. Nobody avoids it, and it has zero administrative costs. There are no returns, no audits, no enforcement. Because it favors people who hold real assets, such as land, rather than financial assets like cash and securities, inflation can hurt the wealthy slightly less.

Looked at another way, however, inflation is the least neutral tax of all. Every other tax distorts some price at the margin, but inflation degrades the price signal itself. If inflation is unexpected, it rewrites every contract in the economy simultaneously and without consent.

Most economists treat inflation as regressive, and if you measure flows like wages, it plainly is. After all, food and energy account for a larger share of household budgets at the bottom. But if you measure inflation’s impact on balance sheets, it’s quite progressive. Inflation weakens creditors, who tend to be older and wealthier, and rewards debtors, who tend to be younger and middle class. The largest single debtor in the country is the federal government, so the inflation tax is a transfer from bondholders to taxpayers.

What about the revenue potential of inflation? It is enormous and mostly invisible. Between 2021 and 2023, inflation did more to lower the US debt-to-GDP ratio than any fiscal legislation in decades, and no one had to introduce a bill.

Put inflation and tariffs side by side, and their similarities are hard to miss. Both score well on collectability and terribly on neutrality. Both are regressive when measured in terms of flows. Both raise serious amounts of money. And both are imposed without a legislative vote: one by delegated executive authority and the other by monetary policy.

But a tax nobody votes on is a tax nobody consents to. That should trouble us more than the regressivity, because consent to taxation is a cornerstone of democracy.

Taxes and Democracy: No Representation Without Taxation

In an eye-opening book, The Price of Democracy: The Revolutionary Power of Taxation in American History, Brookings senior fellow Vanessa Williamson argues that America’s fights over taxes have always been proxies for deeper conflicts over who counts as “We the People.” Poorer Americans have repeatedly built movements to tax everyone, themselves included, to fund a more equal nation. Wealthy Americans have responded by constraining the power to tax and by shrinking democracy itself, through voting restrictions, gerrymandering, and violence. In her telling, anti-tax politics has always been an anti-democratic project.

The pattern runs deeper than American history. Comparisons across centuries and countries tell the same story: a ruler who seeks to tax becomes dependent on taxpayers. Taxes build a connection between the ruler and the ruled that gives power to taxpayers. As representation develops, taxes often go up, because taxpayers gain a say in what the government does.

This is why many high-tax countries are also among the freest and most democratic. The Nordic states convert the broad bases described above into universal healthcare, free higher education, and extensive family leave, which reduce financial precarity and stress for citizens. Theirs is a view of freedom as a positive enabler of liberty: the financial Security to change careers, start businesses, or live without fear of medical bankruptcy.

The American Trilemma

Only three things can close a fiscal gap the size America faces: faster growth, higher taxes, or higher inflation.

Growth is always the preferred path, even if it is never painless. At the moment, we are making a high-risk bet that AI can help. AI’s exponential growth metrics are real and steep. But a technology whose impact and financial returns nobody can forecast is a hope, not a fiscal strategy.

So we arrive at taxes and inflation, which America has already started choosing. Tariffs took $264 billion out of American households last year without a Congressional roll call. When the US debt-to -GDP ratio fell between 2021 and 2023, it was not because Congress acted bravely to reduce borrowing. It was because of inflation.

That should worry us more than the fiscal math. Williamson demonstrates that the power to tax and the power to govern are tightly linked — and those who want to shrink American democracy always begin by attacking the tax base.

A government must answer to the citizens that pay taxes to fund it. A government funded by tariffs and inflation answers to no one — as Donald Trump knows all too well.

Subscribe now

Share

Between 1990 and 2011, China joined the global trading system and quickly became the world’s factory. What economists now call the “China Shock” refers to the destruction of 2.4 million American jobs in low-wage textile, apparel, furniture, and electronics factories during this period.

This need not have been a shock. The US economy turns over more than 5 million jobs each month and usually creates even more. This means that the China Shock was less than a month of normal job losses spread over twelve years. But the China Shock was concentrated geographically among families with few robust alternatives.1 The US did appallingly little to help affected workers, so by 2016 many of them turned to Donald Trump.

Trump likes to condemn all trade with China as a ripoff for the US. It isn’t. American living standards are higher thanks to low-cost Chinese goods. America exports high-tech software, electronics know-how, aerospace, and finance to China. We have lost factory jobs, but most high-skill, high-margin industries are (or were) complementary rather than directly competing with China. Whether the US helping China to move up the value chain faster than they would have otherwise (as Apple famously did) was generous, strategic, or stupid depends heavily on your view of whether China is a US customer, supplier, or rival. Obviously, China is all three – whether our President can get his head around it or not.2

Today we are in the middle of a second China Shock. It is nothing like the first, in part because China is nothing like what it was three decades ago. Adjusted for local purchasing power and cost of living, China became the world’s largest economy twelve years ago. Its $44 trillion economic engine is now one-third larger than the US ($32 trillion) or the EU ($31 trillion, a trading bloc, not a nation).3

Because China is exporting to technically advanced, high-skill, high-wage manufacturing sectors such as automobiles, industrial machinery, chemicals, and precision tools, the second China Shock is hitting Europe harder than the US. Germany, Europe’s largest economy, is especially vulnerable. Germany bet its economy, its national identity, and its legendary social contract on dominating the market for high-end gas cars. China bet on electric vehicles built rapidly, cheaply, and at “good enough” quality. China’s bet is paying off — and the result is China Shock 2.

More China Shocks are coming, and in recent weeks, the outline of China Shock 3 has come into view. The US is now placing the largest bet in world economic history on leading-edge, closed-source frontier AI models and the chips and data centers required to support them. Large tech companies are borrowing real money to finance this bet. But open-source Chinese models now offer “good enough” automated intelligence at a fraction of the prices offered by OpenAI, Anthropic, or Google.

China has already commoditized mobile phones, solar panels, batteries, and much more. If it does the same to AI, our too-big-to-fail hyperscalers will generate too little revenue to cover their debt. Americans who despise big tech will get the bill.

This post describes the Second China Shock and suggests ways to measure whether or not a Third is on the way.

The Second China Shock

China’s explosive growth pulled 800 million people out of poverty – one of the most impressive economic achievements in history. But in the process, the Chinese economy came to rely on a vast, state-subsidized manufacturing ecosystem that, in part, depends on weak domestic demand. Rather than raise wages and expand its domestic market, China directs its surplus capacity outward through exports and foreign investment. It’s a novel form of mercantilism that bears little resemblance to the colonial mercantilists of 17th-century Britain, France, or Spain.

China is now increasing its share of EV sales in both Europe and overseas markets. EU tariffs and trade rules can temporarily protect domestic markets from Chinese EVs, but they cannot protect exports, which make up most of the revenue for European car companies. A worker whose plant is closed cares little whether it is due to the loss of foreign or domestic sales.

The macro environment is also different this time. America’s China Shock came during a period of robust (~3%) underlying GDP growth, but the Eurozone is facing China Shock 2 amidst structural stagnation (GDP growth <1% and falling). This increases the economic and political pain of trade losses.4

China Shock 2 threatens not only the sectors that underpin Europe’s economy but also its much-envied social contract. In one of the great ironies of the age, Chinese communists are refusing to raise the living standards of the proletariat they profess to honor, choosing instead to flood European markets and weaken its struggling social democracy.

Nowhere is China Shock 2 more obvious than the German car industry. Cars account for about 5% of German GDP and ~773,000 direct jobs, plus several million indirect ones. Cars represent 37% of all German business R&D – and the next sector is not close.

Germany depends heavily on exporting cars to the rest of the world. More than three-quarters of German automotive revenue comes from exports – especially to China. In 2025, however, German auto exports to China fell 33%. Thanks to Trump’s tariffs, exports to the US fell 18%.

China is also squeezing German car sales in Europe. Despite the tariffs, Chinese brands took 10.9% of the European market in June and outsold Japanese brands for the first time in May. Much of that volume is sales of plug-in hybrids, which is how Chinese firms drove around tariffs written for all-battery EVs. Chinese cars are 20-30% cheaper to buy, cost less to operate, and are high quality, even at entry levels.

• Chinese cars are rapidly gaining market share. BYD and Chery hit 3-4% share in the UK in a fraction of the time Toyota and Hyundai needed. On a recent trip to the UK, most of the taxis or Ubers we took were hybrids or EVs made by BYD or Chery. Every driver not only praised the quality of the car, but had calculated precisely how much their new car saved them on gas each month. The result: local carmakers lost three points of market share in the first six months of this year alone.

• German families are now paying the price for their car industry’s stubborn attachment to internal combustion engines. Volkswagen forecasts up to 100,000 layoffs and has sought unsuccessfully to close four of the company’s ten plants. Porsche has declared that 20% of its workforce is at risk. BMW just announced 8,000 layoffs, and parts suppliers Bosch, Continental, and ZF announced another 33,000.

The German car industry is reacting to low-cost Chinese EVs the way the Swiss reacted to Japanese quartz watches: by reshaping their products from daily tools to luxury goods. Mercedes, Audi, BMW, and Porsche can retreat to upmarket niches, but Volkswagen is uniquely vulnerable (notwithstanding its Porsche and Audi holdings).

More than just cars are at stake. The German car industry is the load-bearing wall of the postwar European social contract. In Germany, codetermination laws give workers half the seats on corporate supervisory boards. The law turned fifty on July 1, and Volkswagen made the anniversary grim. In December 2024, IG Metall and the VW works council agreed to trade 35,000 jobs, 734,000 units of annual capacity, and years of forgone raises for a promise to keep all ten German VW plants open. Eighteen months later, management came back asking to close four plants. McKinsey had advised closing eight. The board, where the state of Lower Saxony votes 20% of the shares, said no. Only the original wage-and-hour concessions now stand. Investors took note, however, and VW’s stock has fallen more than 25% since February.

Codetermination finds itself in a corner. It gives workers dignity, information, and a negotiated pace of decline, which beats the unmitigated chaos that American workers face. German workers can also try to veto corporate responses to competitive pressure, but rejecting a restructuring plan without offering an alternative is a veto that may ultimately cost more jobs than it saves.

Worker influence is declining in Germany. Half of the country’s largest firms have found ways around supervisory boards. Union election turnout is at a record low, and the right-wing populist AfD is building support among workers at plants VW has threatened to close.

Are the Chinese Cheating?

Labor unions reflexively accuse Chinese carmakers of selling EVs at below-cost prices. This is one of the main justifications for tariffs against China (national security is the other, often stronger one).

To be sure, China provides all sorts of support to EV companies and to car buyers. Beijing stokes domestic EV demand through tax waivers, local fleet mandates, and consumer rebates—though, as in the US, many of these have lapsed. Some provinces offer financial incentives to retire gas cars and make it much easier to register an EV than a combustion vehicle. But subsidies to boost demand inside China don't explain why Chinese cars are crushing competitors abroad.

The Chinese government also subsidizes EV production. These subsidies include government financing for automotive R&D, state aid for land purchases, and low-cost loans. (Companies like Tesla that operate plants in China also receive these subsidies).

How big are these? Rhodium Group decomposed BYD’s $ 4,700-per-vehicle cost advantage over a Tesla built in China and found that direct government grants account for about 6 percent of it, roughly $292 per car. Vertical integration and lower overhead account for at least three-quarters of the advantage. Add plants running at very high utilization, suppliers paid on stretched terms, and development cycles able to ship a new model in eighteen months, and the Chinese have a structural cost advantage that tariffs can delay but never close.

Chinese cars became good, cheap, and ubiquitous in part because EV competition in China is a bloodbath. Most carmakers (56%) lost money in 2025. 227 models cut prices in 2024, bringing the industry profit margin to a record low 4.1%. Beijing has responded with an official “anti-involution” (反内卷) campaign and has summoned executives to justify below-cost pricing.

Chinese automotive overcapacity is stunning. Goldman estimates that the 126 car brands emerging from China could meet 100% of global EV demand. Once again, China has overbuilt an industry (it did the same in housing and solar panels). This time, however, it is exporting the wreckage. This past June, China became the first country to export over a million vehicles in a single month (partly to avoid tariffs that kicked in on July 1). Importantly, it did so as Chinese domestic car sales fell 21.1% in the first half of the year.

But China’s structural advantages gained through scale, vertical integration, and manufacturing know-how are formidable. For this reason, the US and Germany should combine tariffs with an invitation to China to open local assembly plants that guarantee workers the ability to bargain collectively and to have a voice, but not a veto, in workplace decisions. The US needs to force China to build an automotive supplier ecosystem in North America, just as China forced Apple to train the technicians that soon figured out how to build EVs.

Tariffs are a tactic in this fight; they are not a strategy. They only work on things that arrive in containers. This matters, because the Third China Shock is arriving over the internet.

The Third China Shock

The third China Shock is coming in AI, and it will hit the US far harder than the EU. America has built roughly 44% of the world’s installed data center capacity. The EU holds about a tenth.

AI is a treacherous industry to analyze because so much is happening so quickly and because US tech companies are taking four simultaneous risks: investment risk, revenue risk, political risk, and retreat risk.

1. Investment risk.

Alphabet, Meta, Microsoft, and Amazon, the four largest players in the data center race, have now committed nearly $2.4 trillion to leases, buildings, energy, and equipment for AI infrastructure. This is 5-10 times the investment these companies announced just one year ago. Back then, these companies generated more than $200 billion in free cash flow, and activists (foolishly) complained about stock buybacks. Today tech companies consume more cash than they produce – and stock buybacks are a distant memory.

Tech’s share of total US investment in chips and data centers now accounts for nearly all recent economic growth. These investments now account for almost 5% of GDP (chart) and for a record 40% of all US capital investment. AI-driven companies account for 45% of the stock market and about that share of market growth. Increasingly, AI investment represents a systemic risk that is too big to fail – and investors know it.

Worse than this moral hazard, AI investment is increasingly leveraged. A year ago, 10% of capex was funded by debt; today, it is 33%. Debt is now another constraint on data center buildout, in addition to chips, power, construction, and local political opposition.

2. Revenue risk.

The Volkswagen problem is repeating itself in AI software. Chinese models like the new Alibaba Qwen 3.8 max announced this week can do eighty percent of the work at one-fifth to one-tenth of the cost. Other Chinese models, like Moonshot’s Kimi K3, are approaching parity with Claude and are priced accordingly.

Evaluating AI revenue risk requires some caveats. To start, tokens are not tasks. To most companies, the cost per finished task matters more than cost per token, and Chinese models frequently require more tokens per task. Also, price gaps are rarely static. China may sell cheap tokens, take share, then move upmarket the way Toyota did when it launched Lexus. Complicating matters further, vendor financing overstates AI revenue and risk. 85% of Google’s net income in Q2 of this year was attributable to paper gains on its investments in Anthropic and SpaceX. For Amazon, it was two-thirds. If frontier lab valuations fall, these hyperscalers not only lose customer revenue, they lose investment gains that they have borrowed against as earnings.

Finally, it is very difficult to tariff AI. Cars ship in containers, but AI model weights ship over wires. Every protectionist instrument that Trump reached for on Liberation Day, or that Europe reached for in 2024, has no software equivalent.

With those caveats in mind, Artificial Analysis built an impressive cost-intelligence matrix of frontier models, showing the cheap-and-smart quadrant dominated by China. A similar chart prepared for automobiles in 2019 would have shown Chinese cars in the exact same position.

As with cars, low-cost Chinese models today can climb the value stack and become incredibly powerful tomorrow. Many industry analysts, including Stratechery’s formidable Ben Thompson, have argued that cheap Chinese models are good for everyone except OpenAI and Anthropic. Thompson argues that Chinese models enable chipmakers to sell more chips, data centers to run more models, and software firms to keep more margin.

But this is precisely what German automakers said about cheap Chinese batteries in 2019. They argued for commoditizing the inputs and retaining control of the value-added layer, which they took to be branding and mechanical engineering. It did not work because the party that commoditizes the input generates cash that enables them to walk up the stack. CATL did not remain a cell supplier; it evolved into a massive provider of energy technology, infrastructure, and supply chain ecosystems. BYD, now China’s largest automaker, also started as a battery company. Xiaomi, perhaps China’s fastest and most innovative carmaker, was a consumer electronics company known for cheap mobile handsets as recently as five years ago. Today it makes some of the finest cars in the world.

3. Political risk.

Until recently, political risk has been deeply underrated in Silicon Valley. No longer. At the exact moment that AI has provided America an indispensable tailwind to economic growth, opposition to it has become the most reliable applause line in American politics. Hatred of AI is one of the very few issues that bring Democrats and Republicans together.

At the moment anyway, Americans despise AI.

• They are 20 points more likely to say AI’s effect on society will be negative than positive. Seven out of ten oppose having a data center near where they live.

• Data Center Watch’s Q1 2026 report found at least 75 projects worth roughly $130 billion blocked or delayed in a single quarter, matching the whole of 2025 in three months.

• Active data center opposition groups more than doubled from 396 at the end of 2025 to 833 by March. More than 300 state data center bills were filed in the first six weeks of 2026. Statewide moratorium proposals appeared in 14 states.

Opposition to data centers has several sources. Some people oppose all industrial construction. Some are concerned about water and power use and pollution.5 Many resent the widespread use of non-disclosure agreements and the granting of large tax exemptions by local governments to fabulously wealthy tech companies, who do not always honor their promises.

But many people fight data centers because they find AI itself objectionable. As journalist Jasmine Sun reported after visiting several communities fighting data centers, AI is “...an avatar for a small group of Silicon Valley billionaires’ ability to impose their vision of the world onto everybody else without their consent.”

Ultimately, political uncertainty and volatility represent a tax on technology development. We saw the impact of an uncertainty tax on federal clean energy policy when it ping-ponged from Carter to Reagan to Biden to Trump. In July, New York Governor Kathy Hochul, a pro-growth Democrat, ordered a pause on all state environmental permits for projects at 50 MW and above for up to a year. Her order includes a plan to repeal sales tax exemptions for large data centers. It is the first statewide data center moratorium in the country, and unlikely to be the last.

The contrast with China is acute. China set long-term goals and pushed its companies to build global monopolies in solar, EVs, electronics, drones, rare earth minerals, and more. It is doing the same thing in AI, whether people like it or not. But it has drastically reduced pollution.

4. Retreat risk.

High-cost American frontier labs may retreat to serving a small number of high-paying customers, as German carmakers are doing. Frontier labs like Anthropic, Google, and OpenAI are likely to focus on cybersecurity, robotics, and biology and let cheaper open-weight models absorb most coding challenges and the 80-90% of humdrum queries. Under this scenario, criminals, spy agencies, and hostile states will fund the recursive self-improvement of the most advanced models. This will force a growing share of tech spending to fund AI dedicated to containing AI malware.

As he often does, Derek Thompson offered a pithy summary of these risks with a question: “Can the US government become China faster than Kimi can become Claude?”

The answer, unfortunately, is no. The experience of Germany illustrates why, and it should serve as a warning to the US and the industry it now depends on. Recall that the German auto industry is being swept away by Chinese imports for three reasons.

1. Germans invested in the wrong layer of the stack – in branding and mechanical engineering when value had shifted to batteries and software.

2. Germans faced a rival that got good faster than anyone expected, got cheap faster in the layer that matters, lived on 4.1% operating margins, and was willing to export the resulting wreckage.

3. Sluggish institutions. Germany’s cost structure is defended by unions built for a world where German cars were always globally competitive.

American AI firms face a similar challenge. They are investing in the wrong layer of the stack: expensive, closed, proprietary inference. But value is shifting rapidly to low-cost commodity systems, which are being adopted rapidly, especially in developing countries.

Like German automakers, AI firms face a rival that became both good enough and cheaper faster than anyone expected. US hyperscalers have spent more time worrying about “Are we investing enough?” than “Who will capture the value?”

We do not have the equivalent of Germany’s labor market institutions, which made its adjustment slow, negotiated, humane – albeit possibly fatal. Instead, adjustment in the US will be fast and brutal. Nobody in a data center town is getting eighteen months of works council meetings.

Instead, they are getting a loose confederation of groups raising hell and lawsuits to block data centers. This is what happens when a country has no institutional channel for industrial displacement. The hundreds of civic, environmental, and anti-AI groups blocking data centers are what codetermination looks like in a country with no codetermination.

Signs of a Third China Shock

Any analysis of AI should offer falsifiable indicators likely to reveal whether China is actually upending American AI and threatening our massive investments. Here are five measures to track:

1. Debt service. Do hyperscalers grow their cloud revenue fast enough to cover their data center debt and maintain a healthy global market share? If not, the AI caboose will try to outrace the cloud services engine – never a good idea.6

2. Prices. Does the price of open-weight models remain low or free — or do Chinese labs move up-market as K3 did? Put another way: how much of the inference market can China actually commodify?

3. Quality and adoption metrics. Do leading measures of AI dominance tip towards the US or China? Benchmarks like Stanford's 2026 AI Index report that as of March 2026, the top US model led the best Chinese model by 2.7% on their benchmark basket. Hugging Face tracks download volume & “likes” that reflect developer interest and open-source adoption.7 Other metrics include consumer web visits and monthly active users (MAUs) on major inference sites8, the launch of notable models, and patent volumes that indicate the future pipeline of AI dominance.9

4. Balanced growth. Does US GDP, unrelated to AI, recover and constitute at least two-thirds of US economic growth, vs. almost nothing today?

5. EU somnolence. Finally, does Europe actually build compute capacity, or do initiatives like InvestAI join the long list of “someday” mobilization pledges in Brussels? A $31 trillion EU economy that produces few frontier models and just 10% of global data center capacity is not in the US interest, even if Trump and some European leaders believe otherwise.

In 2000, the Clinton administration argued that bringing China into the World Trade Organization would accelerate its inevitable transition to liberal democracy. The administration was infatuated less with access to Chinese markets than with the belief that trade would lubricate the CCP’s “soft landing”. Today, this looks shockingly naive — part of America’s ongoing postwar habit of misreading foreign adversaries.

This habit plagues us still. Treating every interaction with Beijing as all-out competition is a form of wishful thinking, because it lets posture replace strategy. Tariffs are the clearest case. They are what a country reaches for when it has decided that something must be done and has not decided what. They will not save Wolfsburg or Detroit, and not even Donald Trump can tariff a model weight.

The first China Shock destroyed 2.4 million jobs, and we did almost nothing for the people who lost them. The bill arrived wearing a weird hairdo in 2016. The Second China Shock is being paid by German workers who have built institutions to negotiate the terms of their own decline. But codetermination is failing alongside the auto industry. The third China Shock will land on an economy in which AI accounts for 40% of capital investment and 45% of the stock market, and it will hit a nation with no industrial adjustment institutions at all. What we have instead is hundreds of citizen groups filing lawsuits to stop data centers, which is the only channel available to people with no other voice.

Building better labor market institutions is not charity toward displaced workers — it is the precondition for making large bets at all. A country that cannot reabsorb the losers produced by its own investments will eventually be prevented from making them. We got Donald Trump for ignoring 2.4 million people. Nobody knows what we will get for ignoring many more people this time.

Subscribe now

Share

Imagine Google Maps for navigating your life. An always-on AI agent would give you turn-by-turn personal and professional advice, advise you on tricky social and professional situations, and help you build new skills. Because it would remember everything you hear, read, say, and do, it could remind you of people you’ve forgotten, and track your personal and professional commitments. It would understand and shape your tastes in books, movies, music, food, sports, porn, investing, travel, and friends. It could spend or invest your money, organize your day, and monitor your eating, exercise, social life, and sleep. Some users will doubtless try to max everything, everywhere, all at once. Others will flee in horror.

Your always-on AI agent will be a relationship like no other because, on some dimensions anyway, it will seem to know you better than you know yourself. It could give you the skill, knowledge, judgment, and confidence to make better decisions more quickly. You would feel more intelligent, productive, and creative — perhaps with good reason.

Would you be more likely to turn such an agent off or launch a second one? Or ten more? Would you enjoy having a dozen agents competing to create the best possible you?

Intimate Devices

This is not science fiction, nor is it far off. Portable agents are arriving on devices you wear as glasses, carry as a phone or pendant, or — eventually — have implanted.

• The French-Italian eyewear brand EssilorLuxottica sold over 7 million Meta AI glasses last year, up from the 2 million in 2023 and 2024 combined. The company is now discussing expanding production to 20 million pairs a year.

• Meta is prototyping “super sensing” AI glasses that use cameras and audio recordings to capture a user’s every moment. It currently holds about 80 percent of the market for all-seeing, all-hearing devices, but Apple, Google, and Snap are all racing to launch rival products.

• OpenAI has hired former Apple designer Jony Ive to build a wearable, AI-enabled smart speaker with facial recognition designed to “learn information about who is using it and what's around them.”

• Apple, meanwhile, rebuilt Siri from the ground up — unveiled at WWDC in June, powered in part by Google’s Gemini models — as a full conversational companion that reads your screen, your calendar, and your mail, and acts across your apps.

• And Elon Musk’s Neuralink, which had implanted brain-computer interfaces in 21 patients as of early this year, says it plans to move to high-volume production and nearly automated surgery in 2026.

At first, you might not like an agent who knows so much about you — especially one prone to hallucinations that are different from your own. But your agent will adapt. And so will you.

“And so will you.” But what does “you” mean if a person is continually refined, directed, and improved by her secret agent? Where does a user’s mind end and an AI begin? Philosophers Andy Clark and David Chalmers argued back in 1998 that technology extends our minds beyond our skull — a notebook, faithfully consulted, is an inseparable part of your cognitive machinery. An AI, or a group of AI agents, that knows you better than most humans do, becomes part of who you are. Would shutting it off be liberating or terrifying?

Will an agent-filled world be weird? Maybe not. Hearing aids and contact lenses are strange at first — then you cannot live without them. Count the buds and headphones on the young people you pass each day. Computers have moved from room-sized to desk-sized to wrist-sized. Why not an AI-enabled watch, Fitbit, Oura ring, earbuds, or video glasses? In Shenzhen, China, I tested prototype glasses that translated Mandarin into English in real time as captions. It wasn’t ready for prime time, but it was nice to read Chinese easily.

Whose Agent Is It, Anyway?

In pondering what he calls our coming “exoselves,” Kevin Kelly imagines four kinds of relationships with agents that know us better on some dimensions than we know ourselves. He imagines a twin who thinks like you; a tutor-guardian whose judgment you defer to; a hired counselor who is intimate yet professionally bounded; or a hero — a better half who models the person you aspire to be.

Kelly’s taxonomy is about you and your agent. But how will my agent get along with everyone else’s? Does my wife’s or my kid’s agent have our family’s interests at heart? What do they think about me?

And Kelly’s agent-enhanced self is only half the story. How will your agent interact with mine, with governments, or companies? What can my agent know or surmise about you? Will it steer us towards or away from each other? Will people without agents be more human, or just more isolated? Will an agent with knowledge of the most intimate and mundane features of your life improve your choice of friends, business associates, and lovers? In what domains will you trust your agent to have better judgment, and which not?

And in this context, what does “privacy” even mean? We are getting an early answer. Meta’s glasses are already at the center of a mounting privacy fight. In many states, it is illegal to record audio of a third party without their consent. Privacy experts worry that always-on devices could violate data privacy or biometric data laws. These lawsuits concern agents that film and record you. Wait until they can reason and hallucinate about you in conjunction with other agents.

And who has real agency here? Can I direct an agent to help me become more thoughtful? More domineering? Greedier? More spiritual? Who can teach us how to use an agent wisely? Kelly’s answer is sobering: learning to live with an exoself will be one of the major lessons of a lifetime, and society will take years to work out best practices. We have had three decades to learn how to manage social media. How’s that working out?

This points to the deepest problem in any new social technology: trust. An agent this intimate requires users to trust not just the software but also the platform and the company behind it.

Do you trust any company that much? Do you trust Elon Musk with your brain implant? What happens if your agent is erased, captured, spied on, or subpoenaed? What if the company behind it goes bankrupt? Kelly predicts heart-wrenching stories of people losing their exoselves when a platform folds. Imagine grieving a business or government failure because it felt like an amputation, or hating a company that purchased access to your digital soul.

AI agents are just the first step. Thanks in part to AI, many technologies that will reshape individuals are accelerating rapidly. Neural implants, targeted proteins, customized pharmacology, and genetic engineering are all advancing rapidly. We will soon be able to customize our mind, body, and personality directly — not through habit and education, but through hardware and chemistry. And we can influence not only ourselves, but our children and their descendants.

The post-human age is within sight. What are the odds that we use these capabilities wisely?

Subscribe now

Share

I just spent a couple of weeks cycling around Ireland and the UK. The Republic of Ireland is outside the United Kingdom, so it was unaffected by Britain’s decision to leave the European Union (”Brexit”). As a result, economists often contrast the benefits of EU membership for a small, open, English-speaking Ireland with the cost of leaving for Great Britain.

Ireland always wins this comparison – although damage from Brexit is only part of the reason. Brexit didn’t make Ireland richer so much as it made Britain poorer. Since 2015, UK GDP per capita has flatlined, while Ireland’s tax engineering attracted foreign investment that partially offset its own trade losses. Apples-to-apples comparisons are tricky – but even a casual visitor sees the damage the UK inflicted on itself. The aging country voluntarily shrank its own talent pipeline for reasons that were more cultural than economic. As with many ruinous economic decisions, the cost of Brexit was slow, diffuse, and deniable, while the benefits were immediate and politically resonant.

The results are terrible. The UK now faces horrible fiscal arithmetic and severe labor shortages in labor-intensive sectors like construction, agriculture, and elder care. Research talent is fleeing to competing nations. And it’s a tough policy to undo, since the UK can only rejoin the EU by begging it for unfavorable terms.

While in Edinburgh, I visited the grave of Adam Smith, one of the earliest and most astute observers of capitalism. I recalled an exchange from 1777, when Smith’s young friend John Sinclair brought him news of General Burgoyne’s surrender to the American rebels at Saratoga and despaired that “if we go on at this rate, the nation must be ruined.”

Smith replied: “Be assured, my young friend, that there is a great deal of ruin in a nation.” Smith’s quote was both a warning and a timeless reminder of human resilience. He thought that successful societies were structurally robust and could absorb economic mismanagement, political blunders, and monumental crises without collapsing entirely. Smith was right about Britain in 1777. Whether he stays right depends on how much of its ruin the United Kingdom decides to spend.

America’s Brexit

The 250th anniversary of The Wealth of Nations and of our national founding is an occasion to reflect on one of America’s own stupid, Brexit-like decisions. A century ago, nativist forces in the United States built a strong public consensus to sharply reduce the number of people we would allow to move here.

Two Republicans championed the cause: Representative Albert Johnson, a eugenics advocate who chaired the House immigration committee, and Senator David Reed, who explained the logic of banning Asian immigrants and severely restricting southern Europeans to the New York Times a month before passing his bill. Until the mid-1880s, Reed wrote, immigrants had come from “the same sources from which our country was originally colonized” and were easily assimilated. Then:

“Beginning about 1885, new types of people began to come. For the first time in our history, men began to come in large numbers from Italy, Greece, Poland, Turkey in Europe, the Balkan States, and from Russia.

“As these new sources of immigration began to pour out their masses of humanity upon our shores, the old sources in Northwestern Europe seemed to dry up, and whereas in 1890 the natives of Southern and Eastern Europe constituted about 8 percent of our foreign-born population, in 1910 they constituted 39 percent. This change brought new difficulties in the problem of assimilation. These new peoples spoke strange languages. It was not to be expected that they would readily fuse into the population that they found here.”

This view was widespread. The founder of the American Federation of Labor, Jewish immigrant Samuel Gompers, welcomed the Johnson-Reed Act because it reduced the supply of immigrant labor, which he believed threatened the wages of native-born workers. Adam Smith had a phrase for those who lobby the state to shut out their competitors: “the wretched spirit of monopoly.” Gompers wanted the AFL to monopolize the labor supply, and he died in December 1924, months after helping pass the law he believed would achieve that goal.

The Ku Klux Klan, which had resurged as an anti-immigrant organization, backed the bill, citing the benefits of racial purity. The Klan knew that the underlying motive for the law was ethnic, not economic, since it left Mexican immigrants untouched. After all, growers needed farmworkers.

Only six Senators and 71 House members voted against the bill, which Calvin Coolidge promptly signed. Among the dissidents was a freshman from Brooklyn named Emanuel Celler, who spent the next 41 years fighting the law and championing immigrant and civil rights.

The impact of the Johnson-Reed Act was swift: it cut annual immigration by 80 percent using national-origin quotas pegged to the 1890 census, which predated the waves of Italian, Polish, Jewish, Armenian, and Greek immigrants. The act categorically excluded all Asians, building on the Chinese Exclusion Act. Italy’s quota fell from over 40,000 to under 4,000. Almost as an afterthought, the law also created the US Border Patrol and the consular visa system, which is still the administrative foundation of American immigration enforcement.

Ruinous Legislation

Most historians and economists today judge the use of national and racial categories to restrict immigration in 1924 as a massive, Brexit-caliber error. It slammed the door just before the refugee crises of the 1930s, starved American science and industry of talent for a generation, and took 41 years to undo. As with Brexit and Trump’s anti-immigrant rants, the underlying passions behind the legislation were more cultural than economic.

The results endured for decades. The foreign-born share of the US population fell from roughly 13 percent in 1920 to under 5 percent in 1970. A century later, the legislation looks like an economic and humanitarian own goal.

• Economic effects. Adam Smith had seen this before. In The Wealth of Nations, he attacked England’s Law of Settlements, which barred poor laborers from moving between parishes in search of work, calling the removal of a man who had committed no crime “an evident violation of natural liberty and justice.” Johnson-Reed was the Settlement Acts scaled up to a hemisphere, and it failed the same way.

  Research by Ran Abramitzky, Leah Boustan, and others finds the law fell short even on its own terms. It promised that cutting immigrant labor would raise wages for native-born workers. It mostly didn’t. In areas that lost the most immigration, native workers saw little wage gain; instead, farmers mechanized, employers recruited Mexican and Canadian workers where they could, and some industries simply shrank. Marginal mines closed rather than paying more. Native-born workers actually shifted into lower-paying occupations in some affected areas because immigrant-fueled growth had been pulling them up the ladder.

  The act also cut off the flow of scientific and entrepreneurial talent for four decades. The counterfactual is unknowable, but the US won the twentieth century in physics substantially on refugee scientists who got in before or despite the quotas: Einstein, Fermi, Szilard, Bethe, von Neumann. Many more were kept out.

• Social and humanitarian effects. The quotas were still in force in the 1930s, and the US held to them, often under-filling them through consular obstruction, while European Jews sought to escape rising anti-semitism. In a dark chapter of US immigration history, the US refused to admit the 937 passengers of the MS St. Louis, who were fleeing Nazi Germany and sought refuge in Florida after Cuba revoked their landing permits. Worse, the German immigration quota went unfilled in most years of the Nazi era, even as hundreds of thousands waited. Anne Frank’s father applied for a US visa and never got one. The act ensured America’s doors were largely closed during the Holocaust, thanks to legislators who had been explicit in 1924 that keeping out Jews was among their goals.

  Johnson-Reed was openly and proudly built on racialist eugenics. Harry Laughlin of the Eugenics Record Office served as the House committee’s official “expert eugenics agent,” and the congressional record is full of racial pseudoscience that reads as simply grotesque today. Smith’s other book, the one his tombstone lists first, is about the human capacity to extend sympathy to strangers; the 1924 congressional record is a study in its refusal. Adolf Hitler’s praise for the quota system in Mein Kampf is not an endorsement that successful laws usually receive.

• Political effects. Within two generations, descendants of the groups excluded by Johnson-Reed became fully assimilated, prosperous, and politically powerful. This made the original premise (i.e., that Italians, Jews, and Slavs were racially unassimilable) look not just wrong but absurd. The alliance between organized labor and nativism lasted until the law’s economic realities and moral decrepitude forced the AFL-CIO to reverse its position in 2000. And it was Celler, by then the dean of the House, who co-authored the 1965 repeal.

  Unexpectedly, the law was so successful that it helped demobilize the Ku Klux Klan. Once the immigration issue was won, decisively and seemingly permanently, the Klan lost its most potent recruiting grievance outside the South (a lurid murder scandal involving its Indiana leader in 1925 did the rest). Membership cratered after 1925, and by 1930, the organization was a shell.

Were there any benefits? A minority of economists and historians argue that the immigration pause aided assimilation and compressed wages at the bottom, and that the Great Migration of Black southerners into northern industry was partly enabled by the cutoff of European labor. These effects are real but modest, and they cannot salvage the law’s reputation because it classified immigrants by race rather than using a more neutral Canadian-style points-based immigration selection scheme.

It had taken the United States four decades to admit that racial and national restrictions were a serious mistake. In six more decades, we would repeat the error.

Spending Our Ruin

The 1924 immigration law produced echoes a century later, when the share of foreign-born residents in the US once again approached 15 percent. Public revulsion at the chaos associated with Joe Biden’s neglect of border security contributed directly to the election of Donald Trump to a second term. The deportation machinery Trump inherited, from the Border Patrol to the consular visa system, was itself a Johnson-Reed legacy.

A report released this week by the Peterson Institute quantifies the costs of Trump’s executive orders that imitate the Johnson-Reed Act. By cutting out two-thirds of international student visas, for example, the US cuts off STEM workers with PhDs, 35% of whom are foreign-born. Peterson expects this will shrink the high-skill STEM workforce by about 6% overall and by 11.5% at the PhD level. That decline “would cut annual US growth by $240 billion to $481 billion within a decade – a loss the size of a mid-sized US state like Wisconsin or Utah.”

The public is starting to grasp these costs, and public opinion is swinging back as the costs of deportations become more visible and concrete. As with “Bregret” in the UK, Gallup found record-high shares of Americans calling immigration a good thing by mid-2025.

Standing at Smith’s grave, I took his sentiment as reassurance. But it’s also a warning. There is a great deal of ruin in a nation, but the quantity of ruin is finite. Johnson-Reed shows what a single act of self-harm can cost: forty-one years of closed doors, a generation of lost science, and the St. Louis turned away within sight of Miami. Britain is drawing down its stock of ruin on Brexit. America spent heavily in 1924 and is reaching for its wallet again. Smith assured young Sinclair that Britain would survive the loss of America. He never said ruin was free.

Subscribe now

Share

As a kid in the fifties and sixties, I loved wristwatches because they seemed alive and because every man wore one. But in the mid-nineties, this affection curdled into something strange. As I walked down Union Street in San Francisco with my infant son, I spotted an Oris wristwatch – an obsolete mechanical timekeeper. I bought it, loved it, and slowly, inexorably, pathologically, fell into watch collecting.

The hobby grew alongside my sons. As the new century arrived and Bush became Obama, Trump, Covid, January 6, and Trump II, watches helped order my world. Dials summoned me. Movements hypnotized me. The design, the social history, the absurd little feats of engineering required to keep a few hundred parts ticking on a wrist: all of it helped me cope with the cacophony. This post is about the politics, economics, and geography of these tiny machines, and why an inefficient, expensive object that nobody needs refuses to disappear.

The Politics of Timekeeping

People have always tracked life’s rhythms, but for most of history, humans needed calendars to track seasons, not clocks to track hours. By the Middle Ages, mechanical clocks signaled the rise of new forms of political power. As I wrote in an earlier post:

Time is a social construct, but timekeeping is political. When a public clock came to town, the priest, mayor, or business owner set it to local noon and began to coordinate worship, navigation, commerce, finance, and warfare. Eventually, pocket watches and railroads that required standard times eroded the power of church and municipal timekeepers. Once workers acquired pocket watches, bosses could no longer stretch the workday. Once soldiers got them, military attacks could be far more precise.

By 1400, at least twenty-five English towns had municipal clocks. By 1700, roughly half of England’s parish churches kept time for their parishioners. Accurate timekeeping underwrote the British Empire. Once clockmakers solved the longitude problem, sailors could fix their position at sea, and European powers could navigate, conquer, and govern across oceans. By mastering time, Europeans mastered the world.

From the start, a clock was a tool of discipline and direction. It structured the day, summoned people to work or prayer, and measured when they showed up. Soon, factory workers were required to “clock in”, so employers could quantify their work with unprecedented rigor.

Industrialization hardened this logic. Workers were hired not to finish a task but to surrender defined blocks of their day. The factory clock dominated work schedules because it enabled employers to measure punctuality and productivity with greater precision. The historian E. P. Thompson described this shift from task-based to time-based labor as a pillar of capitalist discipline. Much of life now unfolded on someone else’s schedule. The grim supervisor with a stopwatch became one of the founding caricatures of capitalism.

Railroads and telegraphs required standard time zones (until 1883, the US had 144 local times). Watches broke old information monopolies and built new ones, much as cellphones did a century later.

Pocket watches go back to the seventeenth century, but mechanical wristwatches did not become respectable among men until the First World War demonstrated their advantage in combat. They only lasted six decades. Nobody needed a mechanical wristwatch before 1910, and nobody needed one after quartz arrived in 1970. During that brief window, however, nearly every man and many women wore a small mechanical timekeeper on their wrist.

The Economics of Timekeeping

The Swiss watch industry has been declared obsolete twice. It survived both times by becoming something new.

In the late nineteenth century, America was well-positioned to destroy the Swiss watch industry. Watchmakers like Waltham and Elgin figured out how to use metalworking machines to mass-produce interchangeable parts and accurate watches at a scale and a cost that the shambolic Swiss cottage system could not match. When Waltham exhibited its manufacturing prowess at the 1876 Philadelphia Centennial, a shaken Swiss delegate, Jacques David, rushed home and published a famous report warning his compatriots to modernize or perish.

Over the next twenty years, most Swiss watch producers took his advice. After the First World War, they consolidated production into cartels that fixed prices and set production quotas. They used their control over the “Swiss Made” label to turn a national reputation for precision into a mechanism for price collusion.

The second near-death experience came from Japan. On Christmas Day 1969, Seiko launched the first production quartz watch. Quartz watches were cheaper, tougher, and much more accurate. They rendered mechanical movements obsolete and brought Swiss watchmaking to its knees. In 1970, Switzerland had more than 1,600 watch companies; by the mid-eighties, fewer than 600 survived. Watchmaking employment fell from roughly 90,000 to 33,000. Two-thirds of the industry simply vanished.

Those who survived pivoted from function to feeling. Houses like Rolex and Patek Philippe repositioned their watches as luxury goods. A watch became micro-engineered wearable art, an emotional purchase that nobody needed just to tell time anymore.

The pivot worked. According to the Swiss Watch Industry Federation, Swiss watch exports in 2025 totaled about 25.6 billion Swiss francs, or about $28 billion. The industry now ships fewer watches than it has in decades, about 14.6 million in 2025, a multidecade low, and only about 5 million of these were mechanical. The volume keeps shrinking even as revenues remain high, thanks to watches that cost tens of thousands of dollars.

Mechanical watch sales are astonishingly concentrated. Morgan Stanley and LuxeConsult, whose annual report is the closest thing this secretive industry has to an earnings call, estimate that out of roughly 450 active Swiss brands, just four (Rolex, Cartier, Audemars Piguet, and Omega) account for 55% of sales. Rolex alone accounts for roughly a third, with wholesale sales of nearly 11 billion francs and about 1.15 million watches a year. Shockingly to me, the 1.4% of watches priced above $60,000 account for 37% of export value and nearly all of the industry’s growth. The pivot to high-end luxury both devoured mechanical tool watches and saved them from extinction.

The Geography of Timekeeping

Today, mechanical watchmaking clusters in three places: the valleys north of Geneva, the region around Tokyo, and a small German village near Dresden. Each industry was shaped by its own history, and each developed a different idea of what a watch is for.

• Switzerland: Calvinist craftsmanship. Swiss watchmaking owes its existence to a religious crackdown that contained an important loophole. In the mid-sixteenth century, Geneva banned jewelry and ostentation. This rule would have ruined the city’s goldsmiths, except that Calvinism prized punctuality and the disciplined use of time, so, under its strict sumptuary laws, a pocket watch was designated a virtuous tool rather than a sinful decoration (four centuries later, the industry would quickly reverse this distinction). Because they could no longer ornament the exterior of watches, jewelers pivoted to refining clock and watch movements.

  Earlier Modern Times posts described the rise of Rolex in Switzerland and Seiko in Japan. Another outlined the eerie parallels between Rolex and Apple Computer.

  Today, large Swiss watchmakers are more stewards than innovators. This was not always true. During the seventies, the whimsical, artsy, and cheap quartz Swatch generated so much cash that its owners bought up a dozen brands damaged by the quartz crisis. Now, the Swatch Group owns Omega, Glashütte Original, Blancpain, Longines, and several other brands. Richemont, a conglomerate founded by a South African family whose fortune came from diamonds, gold, and tobacco, has been a surprisingly careful custodian of houses like Vacheron Constantin, IWC, Panerai, and A. Lange & Söhne, preserving their distinct identities rather than blending them into a single luxury mush. Adhering to the twisted logic of a true luxury retailer, Richemont destroyed hundreds of millions of dollars’ worth of unsold watches rather than discounting prices and cheapening its brands.

• Japan: weaponizing innovation. Swiss watchmaking grew from farmers seeking indoor work during long winters, but Japanese watchmaking grew directly from state policy. The Meiji Restoration prized manufacturing independence, so Japan adopted the Western fixed-hour clock, and Japanese firms quickly moved from importing clocks and watches to making their own.

  Seiko became the world’s most relentlessly innovative watch company (thanks in part to Citizen, a serious quartz competitor whose best watches are often overlooked). Seiko built Japan’s first wristwatch and produced a steady drumbeat of world firsts after the war. In 1969, it pioneered the quartz watches that took a wrecking ball to Switzerland’s mechanical watch industry. In 1999, it launched a movement that delivers quartz-grade accuracy from a mechanical caliber (Grand Seiko’s Spring Drive), a feat no Swiss maker has yet to match.

  There is a lovely irony here. Japan destroyed two-thirds of the Swiss industry using the very strategy, precise and affordable mass production, that had failed American industrialists a century earlier. The same weapon produced the opposite outcome because the Japanese kept innovating, whereas American watchmakers preferred to collude and try to corner the market.

• Germany: surviving catastrophe. German watchmaking lives almost entirely in Glashütte, a small Saxon town that has made watches and trained watchmakers since the 1840s. This picturesque village has a brutal history. The Allies firebombed it (along with civilians in neighboring Dresden) at the end of the Second World War. After the war, Glashütte fell inside communist East Germany, so the Soviets carted off their watchmaking machinery as reparations. They also merged the independent workshops into a single state combine. (Oddly, this protected Glashütte from the quartz crisis, since a planned economy saw little reason to chase cheap battery-powered watches.)

  After reunification, watchmaking reemerged quickly from Glashütte’s fertile soil. Walter Lange and the executive Günter Blümlein rebuilt A. Lange & Söhne from nothing. Within four years, watch enthusiasts spoke of Lange as a new Patek Philippe. Glashütte Original emerged from the old state-owned enterprise. Like Lange, it reflected a distinct German watchmaking philosophy: stunning hand-engraved bridges, oversized date mechanisms that snap precisely at midnight, and movements assembled twice so that horologists can test, take apart, clean, and reassemble them for delivery. The Germans take this approach for granted, which is the most German thing about it.

Beneath the industry giants sits a vast and lively world of independent watchmakers organized into three price tiers: insanely expensive, absurdly expensive, and merely expensive. The top tier features museum pieces that cost six figures and up. The fun lives in the more accessible tiers. H. Moser & Cie, a Swiss independent brand, trolls the establishment by making a fully mechanical watch that resembles an Apple Watch. Another Moser features a case made of real Swiss cheese to protest loose “Swiss Made” rules. Sinn, a German watchmaker, builds pilot’s watches proven on space missions. Oris, my first love, issued a watch that displays Kermit the Frog on the first day of each month. Here, instead of a corporate monolith manipulating artificial scarcity, independent artisans use high horology as a canvas for whimsy and cultural critiques. Having been liberated from pure utility, these watches become a space for art, ritual, and even defiance.

The newest indies will not quickly disrupt the staid, conservative Swiss watch industry, but they keep trying. In Glashütte, Nomos has made inroads with Bauhaus-inflected design and an innovative world-timer. Christopher Ward designs in England and builds in Switzerland, and just cracked the industry’s top 50 brands with a direct-to-consumer model, innovative designs, and accessible prices.1 Studio Underd0g injects irreverence, food, and vibrant color into an otherwise serious watch industry. China, which built mechanical clocks five centuries before Europe and then perfected the counterfeiting of Swiss watches, has launched independent brands like Atelier Wen that are the real thing.

Thriving on Obsolescence

So where does this leave a hobby built on obsolescence?

Aesthetically, it’s male, pale, and stale. Its market share is trivial: Apple alone sells roughly twice as many watches as the entire Swiss industry, and Apple is only one slice of a smartwatch market that dwarfs mechanical watchmaking. Simple quartz watches still outsell smartwatches, though the gap is closing. People have many choices for what to put on their wrist, and a mechanical watch is the least practical option.

And yet demand holds. For years, Deloitte’s watch survey asked people what they would do with a financial windfall, and majorities across countries kept choosing a fine mechanical watch over a new smartwatch.2

Mechanical watches are not, of course, the only obsolete technology that continues to thrive. Film photography, vinyl albums, pour-over coffee, lugged steel bike frames, and typewriters for writers who prefer devices that can’t interrupt them. Or candles, fireplaces, fountain pens, and manual transmissions. Even sports like archery, biathlon, and fencing evolved from obsolete military technologies.

To be sure, a watch lover with a well-made mechanical timepiece on his wrist usually has a phone in his pocket, just as a vinyl obsessive keeps a Spotify account. But efficiency can be the enemy of engagement. Those worried about the human future in an age of AI should take heart that obsolescence can liberate objects. Once a thing no longer has work to do, it can take on new and more specialized meanings – as craft, ritual, contest, or signal. It survives as art you love rather than a tool you need.

Subscribe now

Share

Republican mid-cycle gerrymandering has handed the GOP roughly a three-point structural advantage in House races this November. It must drive Donald Trump crazy that this won’t be nearly enough. Popular revulsion at the daily cost of his incompetence and lavish corruption has overwhelmed his efforts to redraw district lines. Whether you trust polls, models, or prediction markets, Democrats are heavy favorites to retake the House later this year.

The Senate is the harder prize and the more valuable one because it controls several constitutional checks that the House does not.

• Confirmation power. Only the Senate confirms lifetime judges. It confirms cabinet secretaries and the heads of regulatory agencies, which the Supreme Court has now made far easier for the President to fire and replace at will.

• Committee gavels. The majority party runs the committees. Chairs control the budget, decide which bills get a hearing, and wield subpoena power to investigate the executive branch or private companies.

• Budget reconciliation. The majority controls budget reconciliation, the tool that can pass certain tax, spending, and debt-limit bills with 51 votes, thus avoiding the filibuster. (The majority can also rewrite Senate rules, including the filibuster.)

Fifty-three seats are Republican today. To reach 51, Democrats need to flip four seats without losing any of their own. The electoral arithmetic makes winning the Senate possible — but it takes a functioning party to get there.

The 2026 Senate Grid

• Alaska: Mary Peltola. Peltola served in the US House from 2022 to 2025 on an Alaska-first platform built around “fish, family, and freedom”. She served in the Alaska House from 1999 to 2009 and spent years in commercial fishing before that. She is Yup’ik, the first Alaska Native ever elected to Congress.

  Trump carried Alaska by 13 points in 2024, and Peltola is running even with or slightly ahead of Republican incumbent Dan Sullivan. The often-maligned Chuck Schumer deserves credit for talking her into this race when a run for governor would have been the easier path. Take a look.

• Iowa: Josh Turek. Another state Trump won by 13. Schumer could not land Iowa’s strongest Democrat, Rob Sand, who is running for governor. So Democrats have Josh Turek, a gold-medal Paralympic basketball player who holds a red district in the statehouse and won the Democratic primary. He faces Republican Ashley Hinson. His numbers are close, his campaign is smart, and having Sand at the top of the ticket may help him “push harder for change”.

• Nebraska: Dan Osborn. This state is as red as California is blue – Trump won by more than 20 points. Democrats ran a candidate committed to dropping out to back independent Dan Osborn. Osborn is a veteran and former union leader who lost a 2024 Senate bid by fewer than 7 points. He’s a long shot against former governor Pete Ricketts, but he is a hell of a candidate and worth every minute of your attention.

• North Carolina: Roy Cooper. Democrats put their best foot forward here. The seat opened when Republican Thom Tillis voted against the Big Beautiful Bill and, dreading Trump’s revenge, decided not to run for reelection. Republicans nominated a lightweight in Michael Whatley, and Schumer convinced the popular former governor, Roy Cooper, to run. He leads decisively — and like most candidates, he speaks to local issues first and Trump a distant second, if at all.

• Ohio: Sherrod Brown. Brown served three terms as a progressive senator before Bernie Moreno narrowly defeated him in 2024. When Vance became vice president, Governor DeWine appointed Jon Husted to the seat, and Brown is running to reclaim it. His populist playbook is a bit dusty, but Brown is winning in the polls. Not a lock, but a real shot.

• Texas: James Talarico. Democrats nominated Talarico, every progressive’s favorite Christian, and he is running a smart race. He wants all-of-the-above energy, and an end to Trump’s war on renewables, which is a knife in the back of the state that leads the nation in green energy production. Talarico is making inroads with Hispanic voters in South Texas — and this grocery-store ad has 8.5 million views.

  The race in Texas got interesting when, with Trump’s last-minute blessing, Ken Paxton knocked off incumbent John Cornyn by more than 25 points. This is a gift to Democrats because Paxton is a legendary sleazeball and a much weaker general-election candidate. This is a man who faced 20 impeachment counts for securities fraud, bribery, abuse of public trust, whistleblower retaliation, and obstruction, all as the sitting Attorney General. But Texas is still Texas, and the money sent on this campaign will be obscene. Polls are tied, but prediction markets Polymarket and Kalshi heavily favor Paxton to beat Talarico.

Missed layups.

Harris carried Maine by seven points and lost Michigan by a whisker. Both should be Democratic seats, but instead, they’re a mess.

• Maine: Graham Platner. Maine has a deep Democratic bench. Hannah Pingree carries a strong coalition and high favorability. Shenna Bellows has name recognition and a grassroots network from an earlier Senate run. Nirav Shah ran the Maine CDC to statewide acclaim. The obvious heavyweight, though, was Governor Janet Mills, and the party establishment lined up behind her. But she entered late, could not raise the money or generate much enthusiasm, and suspended her campaign in April.

  That cleared the lane for Graham Platner, who won the June primary going away. Platner is an Iraq and Afghanistan Marine vet and an oyster farmer, and he may yet win. But he is spending his days explaining a tattoo critics tie to an SS symbol, a trove of ugly deleted Reddit posts, and reports of sketchy texts to other women that surfaced through his own wife. He is facing Susan Collins, who is slippery, skilled, and always outruns her polls. This race should have been a runaway. Instead, Democrats are doing it the hard way. Here is Plattner’s launch video:

• Michigan: tbd. This is a political snafu, in the military sense: Situation Normal, All Fugged Up. Governor Gretchen Whitmer, who is both popular and term-limited, could have taken the nomination and the seat without breaking a sweat. Michigan should have been North Carolina: Democrats recruit a popular governor, run a clean primary, and win a Trump state by fifteen. Instead, a talented but DSA-curious public health doctor named Abdul El-Sayed leads the primary, running on Gaza and thin distinctions between his two more moderate rivals.

  If he wins the August primary, El-Sayed may struggle to gain the support of independents, Reagan Democrats, and anti-MAGA Republicans that a Michigan Democrat must earn to carry the state. This is not Mamdani running leftists in New York districts with a twenty-point Democratic cushion. Michigan is a state Trump won. Nominating a candidate built for the bluest voters could put a winnable seat in play. El-Sayed’s campaign video is smart, but like Maine, Michigan should have been easy.

So what happens now?

• Dems will hold their blue seats. Jon Ossoff is very likely to hold Georgia against Mike Collins, and Democrats will probably figure out how to hold Michigan. If either state falls, control of the Senate is lost.

• Alaska and North Carolina are winnable. If Peltola wins a state Trump took by 13, she becomes a national figure overnight — the smart Alaskan alternative to Sarah Palin. Expect the election to be closer than today’s polls. Cooper, on the other hand, appears to own North Carolina.

• Maine and Ohio are possible. Markets favor Platner and Brown, but both races will tighten. Trump’s incompetence is keeping Democratic hopes alive in states like these. But Trump is not incompetent every single day, and he can raise a lot of money.

• Dems cannot count on Iowa, Nebraska, or Texas. I would deeply love to be wrong about any of these three races.

Does this give Democrats control of the Senate? This week, Polymarket gives Ossoff an 83% chance of winning Georgia. It gives Democrats 71% odds in Michigan, Cooper 87% odds in North Carolina, Peltola 64% in Alaska, Platner 62% in Maine, and Brown 57% in Ohio.

So Polymarket predicts that Democrats will flip four seats and win control of the Senate? No — it does not. Imagine that these elections are independent events (which they emphatically are not), and calculate the odds:

• Hold Georgia and Michigan: 0.83 × 0.71 ≈ 59% odds

• Flip North Carolina, Alaska, Maine, Ohio: 0.87 × 0.64 × 0.62 × 0.57 ≈ 20% odds

• Together: 0.59 × 0.20 ≈ 12%

Twelve percent. These odds are far below any single race, because winning several bets at once is harder than winning any of them alone. But of course, the true number is much higher, because these races are not independent. The same national weather moves all Democrats together: party favorability, Trump’s approval, and the shared news cycle. A good month for Democrats lifts Ossoff, Cooper, Peltola, and Brown together. Polymarket’s betting on who will control the Senate gives Democrats nearly 42% odds, not 12%. That 30% gap is the correlation between these races.

This is the argument for a stronger party, whose purpose is to win elections to further shared political values. A stronger party does two things.

• Develop candidates. The first thing a real party does is decide who runs and who steps aside, before a primary settles anything. In 2006, Chuck Schumer’s Senate committee did exactly that, recruiting candidates built for their states: Jim Webb in Virginia, Jon Tester in Montana, and Bob Casey in Pennsylvania. Democrats flipped six seats and won control of the chamber. Republicans have spent the years since demonstrating the cost of not developing strong candidates — handing winnable seats to nominees their own leaders could not stop. Todd Akin, Richard Mourdock, and in 2022, the whole Trump-blessed slate of Oz, Walker, Masters, and Bolduc. Mitch McConnell diagnosed it in two words, “candidate quality,” and conceded he’d lost control of his own primaries. That is Maine and Michigan today. A party that can’t shape a nomination is at the mercy of whoever turns out to vote on a slow Tuesday in August.

• Frame national issues. A party does more than pick candidates. It sets a national frame, designed to improve the odds that its slate of candidates wins. When the message comes from the party rather than from 35 campaigns paying 75 consultants, a good month can lift the whole slate at once.

  A real party also says no, withholding money and endorsements from a candidate who drags the others down, or better, talking weaker candidates out of running before the primary begins. Democrats could not do this in solid blue California, whose primary was a total cluster, much less in Maine, where their strongest candidate faded, and nothing stood between the nomination and a man now explaining his tattoos.

There is, obviously, a catch. Party control over nominations helps win elections, but it runs counter to the open primary that Americans cherish. But the party’s grip is terribly weak — there is no chance that somebody will run Maine out of a smoke-filled room. Between a functioning party and a loose confederation of campaigns, consultants, and donors, there is enormous room to develop — and Democrats are camped out at the wrong end of the spectrum. Democrats will win more elections today if they take seriously the wisecrack that Will Rogers offered them almost a century ago.

Subscribe now

Share

How we govern AI will soon surpass Trump’s corrupt antics, oil wars, immigration, the cost of living, the federal deficit, and collapsing birth rates as the central political issue of our time.

Voters from both parties despise AI. On the left, populists hate tech billionaires and the income inequality they symbolize. Environmentalists dislike data centers. Labor activists fear job losses. Humanists, religious believers, and parents across the political spectrum are finding common cause in their shared distrust of the effects of AI on children and society.

And AI is just getting started. When Americans choose a new president in November 2028, AI will be thousands of times more powerful across many dimensions than it is today.1

The Horror Story and the Fairy Tale

Our public debates about AI swing between a horror story and a fairy tale. Is AI a tragedy of the commons, or is it stone soup?

In 1968, UC Santa Barbara ecologist Garrett Hardin wrote a famous essay in Science titled “The Tragedy of the Commons.” Hardin described how shared resources invite overuse and eventual ruin, as individuals acting rationally overconsume the commons and produce a collective catastrophe. The tragedy is structural. No villain is required. Despite the article’s Malthusian overtones, most economists believed him.

Stone Soup tells the opposite story. A traveler (sometimes three) arrives in a village and boils some water. He adds a stone and announces that he is making a magnificent soup. He explains that his dish needs a few small additions and persuades one villager to contribute carrots, another to donate potatoes, and another to toss in a bit of meat, until a feast emerges that feeds everyone. The traveler then pockets his stone and moves on. (The stone is sometimes a nail or an axe; there are many versions.)

The two stories are symmetric in interesting ways.

• Contributions. In the tragedy, individuals extract from a shared pool until it collapses. In stone soup, individuals add to a shared pot until it overflows.

• Logic. The tragedy leads each person to reason, “If I don’t take from the commons, someone else will.” In stone soup, each person reasons: “If I add a little, I get back far more.”

• Social fiction. The tragedy unfolds without coordination, like a malevolent invisible hand. Stone soup succeeds through a bit of theater and a small social fiction that lowers the barrier to contribution. Religions do this all the time.

• Results. The commons turns abundance into scarcity. Stone soup turns apparent scarcity into abundance. The difference between the two outcomes is not just resources or incentives - it’s whether someone can construct a believable focal point for collective action.

AI as a Tragedy of the Commons

Many people see AI as a slow-motion depletion of our shared informational, cultural, and labor patrimony.

• AI undermines our cultural commons. Democratic deliberation, science, many laws, and the social trust that enables commerce are collective goods. They emerged from institutions built over generations of struggle to distinguish fact from fiction. Today, every person and organization has a strong incentive to deploy AI-generated content, but the cumulative effect is more noise and less signal in all forms of public and private communication. This is epistemologically corrosive and trust-destroying if it diminishes our shared understanding of reality.

• It pollutes our internet commons. The human-generated web is a commons that trained the current generation of AI models. They extracted enormous value from it. Now AI-generated content is flooding back into the same pool, threatening to pollute future training data with synthetic, self-referential output.2 Cory Doctorow described internet platforms like Facebook and Amazon as vulnerable to “enshittification”. Every website with user content is now exposed to enshittification by AI slop.

• It threatens our labor commons. I have argued that AI is unlikely to eliminate jobs overnight, but if it steadily reduces nearly all cognitive work, it would devalue human skill and expertise and lead people to invest less in education or mastery. We may end up with less human knowledge, fewer craftspeople, and fewer people who have mastered anything difficult. The accumulated human capacity on which civilization depends would be depleted for individual gain. (This argument can quickly be reduced to the absurd. Most of us no longer add large columns of numbers by hand, but ceding basic arithmetic to calculators and spreadsheets shifted numeracy from rote computational skills towards quantitative literacy and data interpretation. Life did not spiral downward. But AI may not be like spreadsheets.)

AI as Stone Soup

The stone soup reading is more hopeful and equally plausible.

In the tale, the stone gives everyone a reason to contribute what they have. AI tools already play this role for people who were previously locked out of certain kinds of work. Using AI, non-programmers can build working tools. Non-designers can produce something visually coherent. Non-native speakers can write with greater fluency. By lowering the threshold for participation, AI allows more people to add their carrot to the pot. The feast grows as AI recruits new contributors.

Scientific research, medical diagnosis, and legal reasoning are built on knowledge accumulated across generations. No one can master it all. An AI that holds the collective soup allows specialists to add their particular expertise and receive back synthesized insights that none of them could have produced alone. Researchers using AI to discover new drugs or materials are already cooking a richer stew than any prior generation could manage.

AI might also play the role of the traveler himself, arriving with a trick that breaks up artificial scarcity. One structural problem in knowledge economies is that expertise is rationed — not because knowledge is finite, but because credentialing, geography, and cost restrict access to it. AI can dissolve some of those barriers. The best legal reasoning, medical knowledge, and educational scaffolding need not be reserved for those who can afford a lawyer, a specialist, or an elite university. The stone here is AI’s capacity to unlock both the contribution and the benefit of the commons for people who were previously excluded.

This is the insight that earned political scientist Elinor Ostrom a Nobel Prize. She proved that the tragedy of the commons is not inevitable. Contrary to the instincts of most economists, Ostrom showed that real communities protect shared resources such as pastures, fishing waters, and forests by developing their own rules, norms, and governance without top-down mandates or full privatization. Ostrom would read Stone Soup as a fable about a well-governed commons.

Whose Cauldron Is It?

AI presents a twist neither Hardin nor Ostrom anticipated. The political economy of AI today looks less like an open pasture or a communal village pot than like a privately owned cauldron filled with public goods.

A useful comparison is the Inclosure (aka Enclosure) Acts, which roiled 18th- and 19th-century Britain. Enclosure laws privatized common lands, drove peasants into cities, and created the modern working class. Frontier AI models are pulling off something similar: monetizing decades of publicly produced human output — Reddit threads, Wikipedia articles, blogs — without meaningful compensation or consent.

In a classic tragedy of the commons, people stop caring for a resource when they feel excluded from its benefits. That defection has already begun. Some creators are using the University of Chicago’s Glaze or Nightshade to poison their own images so AI scrapers cannot use them cleanly. Users are abandoning Google Search as it enshittifies with AI-generated content and retreating to Reddit, Discord, and other human-curated spaces.3 An underground economy of data hoarding and algorithmic evasion is emerging as high-quality human-generated data becomes a closely guarded commodity.4

The political backlash against AI appears technophobic or Luddite, but it may simply be an inchoate resistance to a massive enclosure movement. Billions of people contributed their carrots and potatoes to the pot through training data, and a handful of tech executives showed up to the feast claiming to own and control the pot itself.

America is about to decide what institutional guardrails we want to ensure that AI serves public needs. Many of the barons of Silicon Valley will favor no regulation at all. Others will be tempted to treat all foundational language models as public utilities, with public auditors monitoring their inputs and outputs for safety and social value. Others will want to construct a public, non-commercial computing infrastructure to ensure that the benefits of synthetic intelligence are distributed broadly. Many of these measures would not only slow AI down but also strangle it in its crib - as their proponents know.

Which story wins and what governance we adopt depends on whether we can create a credible focal point for collective action to manage the risks and socialize the benefits of AI before it depletes our human commons. Stone Soup speaks to us because the traveler fed the village with ingredients the residents themselves provided. The AI stone possesses real, transformative magic. But it only delivers a feast if the community that supplied the ingredients gets to sit down and eat.

Subscribe now

Share

For a young labor organizer, Silicon Valley in the early 1980s was Yankee Stadium. I was successfully organizing hospital workers with the Service Employees in San Jose, but I itched to play in the big leagues. I wanted to mobilize the immigrant women assembling chips and hard drives in the booming companies that had replaced our fruit orchards and were poisoning both workers and our groundwater with the toxic solvent trichloroethylene.

So I wrote to one of my heroes, William “Wimpy” Winpisinger, the Democratic Socialist firebrand who ran the International Association of Machinists. I reminded him that the IAM already had 15,000 union members in Silicon Valley’s defense sector, and that many of them had family members working at Intel, Apple, Atari, Hewlett-Packard, National Semiconductor, and Tandem. No other union had an opening like this. I argued that if the Machinists organized these companies now, when they were small, union membership would grow along with these businesses. Would the IAM commit resources to organizing Silicon Valley?

Email was still a decade away, but the typed letter I got back assured me that the Machinists were deeply committed to organizing “Silicone Valley.” So I went to work at a local defense plant, became a leader of the Machinists, and joined a few efforts to organize Atari and National Semi. Along the way, I became a political ally of Michael Harrington, the admirable founder of Democratic Socialists of America.

The “Left Wing of the Possible”

Harrington is best known for The Other America (1962), the book on US poverty that prompted the Kennedy and Johnson administrations to declare a War on Poverty. More than anyone of his era, he bridged European democratic socialism and pragmatic American politics. He rejected the idea of a third-party socialist movement in the US as a recipe for isolation, advocating instead for “realignment” — working inside the Democratic Party alongside labor, civil rights groups, and feminists to drag it in a social-democratic direction.

Harrington believed authentic democracy couldn’t exist if a tiny corporate elite controlled the economy. He wanted public and democratic control over major investment decisions so production would serve human needs — healthcare, education, housing — rather than private profit. He envisioned a comprehensive safety net with universal healthcare, guaranteed employment through public works, and aggressive federal action to eradicate poverty. And he viewed a strong, democratized labor movement as the essential vehicle for working-class power.

Harrington died of cancer in 1989. By then, I’d earned a business degree after a decade with unions. Times had changed, and the Democratic Leadership Council — co-founded by Bill Clinton — was declaring New Deal liberalism dead. To win national elections, the DLC argued, Democrats had to abandon “big government” solutions and appeal to suburban, middle-class voters.

The DLC aimed to address social ills through lean government rather than federal spending. They favored free markets over regulations on corporate power, as epitomized by NAFTA and the partial repeal of Glass-Steagall. They went for “workfare not welfare,” as cemented in the 1996 PRWORA, which replaced AFDC with TANF block grants, lifetime limits, and work requirements. As to labor, they preferred to train workers to increase their “human capital” to compete in a globalizing economy rather than strengthen unions to protect them from the effects of globalization.

Clinton looked like a sleeker, less Marxist-inflected version of social democracy, so I signed on. So did much of Europe. By the turn of the century, more than two-thirds of EU citizens lived in countries led by center-left leaders. Gerhard Schröder in Germany and Tony Blair in Britain were essentially Clinton clones, lifting ideas from the right to temper the excesses of capitalism.

Socialist-Free Europe

But in large European countries, the bloom has fallen from social democracy’s trademark rose. By 2016, the share of Europeans living under social-democratic governments had been halved. Those of us expecting a leftward swing of the pendulum since then have been disappointed again and again. Today, only a handful of Europe’s national leaders are social democrats — Spain's Pedro Sánchez, Malta's Robert Abela, Lithuania's Inga Ruginienė, Iceland’s Kristrún Frostadóttir, Norway’s Jonas Gahr Støreand, and, for a few more days, Denmark's Mette Frederiksen. Importantly, all but Sánchez govern small states — altogether roughly 12% of the EU’s population.1

This group is shrinking before our eyes. On May 8, Frederiksen — fresh off a battle with Trump over Greenland that had buoyed her polling — failed to assemble a coalition after Denmark’s March general election, in which her Social Democrats posted their worst result since 1903. The king then asked the center-right Defense Minister, Troels Lund Poulsen, to lead the negotiations. This week, Poulsen appears to be within striking distance of forming a right-wing government.

Then this past Sunday, Spain’s Pedro Sánchez took a historic beating in the Andalusian regional election. His Socialists collapsed to 28 of 109 seats and 22.7% of the vote — their worst showing ever in Spain’s most populous region, which social democrats have ruled almost continuously for nearly four decades.

The rest of Europe is increasingly devoid of democratic socialists, especially in larger countries.2 Germany’s once-mighty SPD sits at roughly 13% in nationwide polling — fourth, behind the xenophobic, Putin-curious AfD, the center-right CDU, and the Greens. France’s Socialists are nowhere. As The Economist recently put it, in much of central Europe, there is barely a socialist party to contest elections, let alone win them.

What happened? Analysts point to five forces.

• Political identities fragmented alongside labor markets. The center-left’s anchor was a highly organized industrial working class that has now been transformed by automation, deindustrialization, and the shift to technology and service economies. In Europe, as in the US, unions could never organize “Silicone Valley.” Gig workers and software engineers did not inherit the same institutional ties to unions and labor parties that their factory-worker parents had. Nor did blue-collar unions have any feel for the highly educated urban professionals who increasingly dominate progressive politics. The left became culturally and economically unmoored from its rural and industrial heartlands.

• A backlash against neoliberal triangulation. The 1990s–2000s project of reconciling social justice with deregulation and globalization initially worked, but over time it blurred the line between the center-left and the center-right. The German SPD’s embrace of Schröder-era Hartz IV welfare reforms, and the broader center-left support for austerity during the Eurozone debt crisis, told core voters that their traditional protectors had quietly switched sides. The US had its own version of this.

• The politics of immigration displaced the politics of class. National sovereignty, European integration, and especially immigration moved to the front of voters’ minds. Center-left parties historically favored cosmopolitan, humanitarian approaches. But socially conservative working-class voters felt exposed to rapid demographic change and economic competition. Right-wing populists framed immigration as a threat to both national culture and the welfare state — and peeled off a substantial bloc.

• Competition from Greens and left populists. The center-left no longer has a monopoly on progressive or anti-establishment votes. Urban, educated, climate-conscious voters defected to Green parties; voters frustrated by inequality and austerity bolted to Syriza, Podemos, La France Insoumise, and similar movements. The US echo is DSA activist Zohran Mamdani’s election as New York's mayor last November. But this is no longer Michael Harrington’s DSA. It condemned the Russian invasion of Ukraine, but blamed “US imperialist expansionism” for Russia’s aggression, demanded that America withdraw from NATO, and refused to support arms for Ukrainian self-defense. Harrington, who spent much of his life arguing that democratic socialists had to be unsparing about Soviet aggression, would have wept.

• “Pasokification”. In a strange twist of political history, the European center-left became a victim of its own historic triumphs. Throughout the 20th century, social democrats built the pillars of the European model — universal healthcare, robust pensions, strong labor protections, and public education. Center-right parties largely accepted those programs as standard, non-negotiable elements of European governance. Once preservation of the welfare state became the consensus across the political center, it became harder for the center-left to articulate a distinct, visionary project. (This has also happened in the US, where even Donald Trump shows no real interest in repealing Obamacare or Social Security — and his cuts to Medicare and Medicaid will almost certainly prove politically damaging.)

  The phenomenon is widespread enough that political scientists coined a term for it: Pasokification. They named it after Greece’s PASOK, a center-left party that earned 43.9% of the vote in 2009 but only 4.7% in 2015 — from governing to electoral irrelevance in six years.

The Welfare States of America

Most progressives believe the United States offers only a bare-bones safety net compared with Europe. Conventional wisdom holds in the aggregate, but it breaks down state by state. Many blue states behave like social democratic members of the old EU.

When William Beveridge created the British welfare state in the 1940s, his goal was “to slay the five giants: want, disease, ignorance, squalor, and idleness”. The activist (and unrepentant eugenicist) designed child allowances, maternity leave, unemployment insurance, disability payments, universal healthcare, and old-age pensions.

The US took a different route. The federal government is wholly responsible only for old-age pensions (Social Security) and elderly health care (Medicare); it contributes generously to child tax credits and to health care and food assistance for the poor through Medicaid and SNAP.

Progressive states like California, New York, New Jersey, and Massachusetts have stepped in to slay the remaining giants. In some cases, they match or exceed European standards:

• Paid family and medical leave. In the last decade, the number of US states with state-mandated paid family leave has quadrupled and now covers roughly a third of the US population. Measured in fully paid weeks for new parents, New Jersey is in the neighborhood of France, Massachusetts of the United Kingdom, and California of Australia.

• Unemployment insurance. The EU pays the jobless for longer, but during the initial months after a layoff, unemployment insurance in many progressive states replaces over half a worker’s previous wages — temporarily leaving newly laid-off Americans in high-benefit states better off than counterparts in Austria, Finland, or the UK.

• Targeted tax credits and subsidies. Blue states have pioneered expansive state-level Earned Income Tax Credits and child tax credits, alongside aggressive childcare and healthcare subsidies via state-augmented Medicaid expansions and state marketplaces. These policies have proven highly effective at reducing child poverty and cushioning the baseline volatility of the American labor market.

But blue states differ fundamentally from EU models in how they pay for all of this.

• They tax corporations and the rich. European social democracies rely on value-added taxes of 20-25% to fund their welfare states. These regressive consumption taxes provide a large, stable, broad revenue base. American blue states are structurally prohibited from imposing a VAT and politically constrained from raising flat consumption taxes to European levels. Instead, they fund expanded social programs through high taxes on top-tier incomes and capital gains. Democrats like this for equity reasons, but it ties state revenue to the stock market, tech IPOs, and the continued residency of a small pool of high-net-worth individuals — a far cry from VAT’s stability.

• They have to balance their budgets. Unlike sovereign European nations (or the US federal government), American states must close their books every year. In downturns, when demand for the safety net peaks, blue states experience sudden revenue shortfalls. With no power to print currency or run structural deficits, they are forced into painful spending cuts or temporary federal bailouts.

• Capital and labor can change states easily. In the EU, language, culture, and law create friction that limits tax-driven migration. In the US, a family or a business that concludes that its blue-state tax burden outweighs the benefits can easily move to Texas or Florida. This places a hard cap on how high progressive states can raise tax rates.

• Uneven service delivery. European welfare states pride themselves on universalism and automation, which minimizes administrative costs and procedures. Even the most generous American state programs are layered atop a fragmented federal bureaucracy. Low public awareness, complex applications, and administrative hurdles mean actual take-up rates for state-paid leave or childcare subsidies are often low, leaving state safety nets patchy in practice.

What is to be done?

At the root of social democracy’s crisis lies a paradox the liberal tradition has never resolved. Liberalism is animated by a deep suspicion of concentrated power — yet it has always needed a strong central state to enforce the rule of law, break up private monopolies, and build the institutions a free society depends on. (Competitive markets themselves are not spontaneous; it took central authority to dissolve feudal privilege, standardize currencies and weights, and enforce contracts uniformly.) The history of liberalism is not a march toward shrinking the state but an unstable, never-finished effort to calibrate it: strong enough to secure liberty, constrained enough not to crush it. Social democracy sits at the far end of that spectrum, staking everything on a robust, competent central administration.

That bet pays off best in small, cohesive societies like the Nordic or Baltic countries. On a larger scale, it struggles to adapt to a fast-changing world. Detailed central rules go obsolete faster than they can be rewritten, and they collide with messy local realities. Administrations respond with waivers and exceptions until the rule of law degrades into a patchwork of carve-outs and incumbent protections.

Columbia political scientist and legal scholar Charles Sabel offers a way through this dilemma that he calls democratic experimentalism. His premise is that modern problems — climate, complex social services, volatile markets — are defined by deep uncertainty: the center cannot anticipate every local variation or future disruption, so top-down bureaucracy is bound to fail. The fix is not to abandon public administration but to rewire it as a continuous experiment, with a feedback loop running between the center and the front lines:

• The central government sets goals, not methods. It defines what counts as adequate education, clean water, or safe food, establishes metrics, and acts as a clearinghouse — pooling performance data and forcing disciplined comparisons across localities.

• Local units get real discretion. Schools, environmental agencies, and municipal offices pursue the national goals in whatever way fits their circumstances and ground-level knowledge. In turn, their cumulative field experience helps reshape central government goals.

• Local autonomy is conditional. Local units must report results transparently; those that fall short enter a peer-review process and adopt corrective measures drawn from their higher-performing peers.

The state stops being a rigid commander and becomes a facilitator of social learning, revising its own standards in light of what actually works on the ground. Building the systems, skills, and leadership to make this work is, of course, a nontrivial undertaking — but it points in an interesting direction.

Michael Harrington was a pragmatist who preached the “left wing of the possible.” That stance produced lasting victories, from Social Security and Medicare to the Affordable Care Act, and a fifty-state laboratory in which blue states have quietly built something closer to European social democracy than most Americans realize.

What it has not produced, on either side of the Atlantic, is a center-left that can win durable majorities — or reliably deliver the services it promises in large countries. Solving that puzzle is the core challenge for today’s progressives — and it calls for solutions that traditional social democrats are still reluctant to entertain.

Subscribe now

Share

Policymakers and the press are obsessed with the possibility that advanced AI will throw millions of people out of work. This is unlikely, but the risk of AI killing people, draining their savings, or hollowing out the economy deserves more public attention and policy debate.

AI CEOs are sounding the wrong alarm.

The leaders of our biggest AI labs are right to warn that their technology is powerful and that we need to take steps to prevent it from being weaponized. But they are wrong to predict a job market meltdown.

• Anthropic CEO Dario Amodei has warned that 50% of entry-level jobs in tech, legal, consulting, and finance will be wiped out within five years. Last year, he told Axios the “white-collar bloodbath“ could spike unemployment to 20% -- Great Depression territory.

• x.AI CEO Elon Musk declared in 2023, “There will come a point where no job is needed -- AI will be able to do everything.”

• ChatGPT founder Sam Altman wrote that “the price of many kinds of labor will fall toward zero once sufficiently powerful AI joins the workforce.”

Cyber and biological risks are real. Frontier safety teams at the major AI labs, along with the World Economic Forum’s Global Cybersecurity Outlook 2026, all document the same risks. Advanced AI systems now generate working exploits, plan multi-step attacks, and lower the expertise threshold for things that used to require a trained team. Armed with advanced agentic models like Anthropic’s Mythos, a garden-variety basement nerd can now discover previously unknown vulnerabilities and launch a cyberattack on your power grid, bank account, or water system. Hackers who route around the porous “refusal systems” AI tools so they can tinker with lethal viruses are risking catastrophe.

The risk of freaks or terrorists weaponizing AI led me to argue last month that national governments will soon need to subject frontier AI models to much more technical and political oversight and licensing. Trump floated this idea recently, and he should try to remember it long enough to pursue it.

The jobs apocalypse is mostly a distraction. Our labor markets are enormous, dynamic, and have proven resilient to exactly the kind of shock people now fear. That doesn’t mean that AI won’t inflict pain – but it does suggest that CEOs of AI companies aren’t framing the questions well.

Repeated warnings from AI CEOs can shape events, even when they are dead wrong. Nobel Prize-winning economist Robert Shiller described how apocalyptic memes matter in Narrative Economics: How Stories Go Viral and Drive Major Economic Events. He shows how popular fears of machines displacing workers contributed to 19th-century economic downturns, reinforced the false belief that automation caused the Great Depression, and exacerbated the double-dip recession of the early 1980s. The danger isn’t just labor disruption — it’s that popular narratives create a negative feedback loop. Economic hardship caused by recession gets misattributed to machines, creating pessimism that deepens the downturn.

We may be watching this dynamic play out now. AI-washing is masking the real sources of today’s labor market anxiety: inflation, tariff shocks, and a hangover from pandemic-era over-hiring at big tech firms.

Employment data tell a different story than these warnings suggest. Consider the supposed canaries in the coal mine—tech workers. Net US technology employment grew from 8.7 million in 2020 to 9.6 million in 2023 and has been roughly flat since then. Tech is a low-hire, low-fire labor market where unemployment for tech workers and everyone else is converging around the Fed’s target rate of 4%. Not great, but hardly apocalyptic.

Over time, the US labor market has evolved a lot. Take a look.

Academic and government economists are skeptical that a job apocalypse is approaching.

• A recent NBER working paper found that “AI adoption has not yet led to meaningful changes in total employment,” though it is reshaping which tasks workers perform.1

• The Federal Reserve Bank of Atlanta surveyed firms and found that more than 90% estimate no employment impact over the last three years.

• The Census Bureau’s Center for Economic Studies found that only about 5% of AI-using firms reported any headcount change -- and those changes were split almost evenly between increases and decreases.

• The Yale Budget Lab concluded in April 2026 that “the picture of AI’s impact on the labor market that emerges from our data is one that largely reflects stability, not major disruption at an economy-wide level.”

Even though we should expect AI to create and eliminate both tasks and jobs, there is still no statistically significant relationship between AI and unemployment or employment growth.

Why the job apocalypse story is (mostly) wrong

Look more closely at the kind of report that drives reporters to sound the alarm. Goldman Sachs recently estimated that generative AI could automate tasks equivalent to about 25 percent of US work hours, and that roughly 300 million jobs globally are at risk of some level of automation.

Read quickly, that sounds like mass unemployment. Read carefully, it isn’t. Goldman’s own base case is that 6 to 7 percent of US workers will be displaced over roughly a decade of adoption, raising the unemployment rate by about half a percentage point during the transition. If companies deployed only current AI use cases across the economy, Goldman estimates that just 2.5 percent of US employment would be at near-term risk.2 That’s economic adjustment, not labor-market collapse.

There are five reasons that the deeper story is more reassuring than the headline.

1. The US economy does not create a fixed number of jobs. This is the “lump-of-labor” fallacy, and economists have been refuting it for two centuries. Human wants and needs are essentially infinite. When automation lowers the cost of existing goods, freed-up wealth and labor flow into new sectors — healthcare, education, software, hospitality, pet care, fitness, mental health services, and niche entertainment. David Autor and his colleagues used census data to show that roughly 60 percent of US workers in 2022 worked in occupations that didn’t exist in 1940. Most of the jobs your grandchildren will hold don’t yet have names.

2. Technology that cuts costs can grow jobs. The clearest case is the one that should haunt every AI doomer: spreadsheets. When VisiCalc and then Excel arrived, the prediction was carnage among bookkeepers and accounting clerks.

   The carnage was real for the clerks. But the number of accountants rose, because cheaper calculation made it worth asking more questions — what if I borrow more, hire more, or change prices?3 Accountants stopped being mere calculators and became advisors. The Bureau of Labor Statistics still projects positive growth for accountants; only the narrow data-entry slice of bookkeeping is shrinking. Software ate part of one job and grew many others. (Counterexample: farming. When we automated agriculture, we did not eat a lot more food, so we needed fewer farmers.)

3. People often prefer people. As AI commodities get cheaper, the things AI can’t fully replicate become more valuable: human attention, human judgment, human warmth, the social fact of having been served or taught or treated well by another person. Therapists, teachers, health care providers, hospitality workers, performing artists, and coaches are protected because the value they deliver depends as much on relationships as intelligence. In a richer, more automated economy, the premium on human connection rises rather than falls. (Counterexample: in San Francisco, people pay more to ride in driverless cars. The driver adds negative value.)

4. AI kills more tasks than jobs. A job is a set of tasks. You get paid to complete all of the tasks associated with your job. If an AI automates much of your rote work and it frees up time for you to generate better ideas, your productivity goes up, you become more valuable to your employer, and, ideally, you earn more. But if AI automates all of your tasks, then yes — you are out of work.

   Goldman Sachs broke down the tasks for a series of jobs. They concluded that education workers, judges, and construction managers are the jobs with tasks most likely to be augmented. But the “substitution versus augmentation” distinction is artificial. In most knowledge work, AI handles many tasks while a human handles the last few — and those last few are often the ones that determine whether the work succeeds at all.

   

   Labor economist Alex Imas calls these O-ring jobs, after the small component whose failure destroyed the space shuttle, Challenger. If AI does 98 percent of the work but a human is required to catch the 2 percent that would otherwise produce a catastrophic failure, the human stays and is highly paid. Radiologists and software engineers are the canonical examples; they’re also two of the professions most often predicted to vanish, and have so far done roughly the opposite. (Counterexample: Goldman identified telephone operators, insurance claims clerks, and bill collectors as among the occupations where AI will substitute for enough tasks to threaten jobs.)

   We should worry less about direction and more about timing. Historically, labor market transitions were slow enough that they could adapt across generations. Children trained for jobs their parents couldn’t do. But if AI diffuses in years, not decades, displacement could outrun retraining, and wages in exposed occupations could fall sharply before new high-wage roles absorb the workers.

   That is a transition problem, not an apocalypse. Still, it’s the scenario that deserves real policy attention, because the US has never learned to effectively help workers navigate major labor-market transitions.

5. Productivity gains tend to lift wages across sectors, not only within the productive ones. This is the so-called Baumol effect, named after the economist who first described it. When tech and manufacturing workers become more efficient, they earn more. Service-based, labor-intensive roles — nurses, teachers, live musicians, hairdressers — can’t easily become more productive without degrading the service. Still, employers in these sectors have to compete for workers with high productivity. To retain talent, wages rise across the board, even in places where output per worker isn’t moving.

Whether productivity gains flow to wages depends more on bargaining power and timing than on the technology itself. Wage growth from productivity has historically depended on workers accumulating bargaining power through labor unions, taking advantage of tight labor markets, or possessing scarce skills. If the competitive dynamics that surround AI companies were to disrupt this, income inequality could grow.

Let’s look closer at that possibility.

Taking inequality risk seriously

If the CEOs of AI companies overstate the risks to jobs, they understate the risk of growing income and wealth inequality. Four forces drive this – and like fish that cannot notice water, AI companies are so saturated in these dynamics that they barely observe them.

1. Winner-take-all competitive dynamics. Historically, the benefits of technology have flowed heavily to those who own or control it. AI accelerates this through network effects and data moats. The wage gains that do materialize may flow disproportionately to a relatively small technical and managerial class employed by or immediately adjacent to frontier AI companies, widening the gap with everyone else.

   For example, a garden-variety software engineer who joined Anthropic three years ago might have received a stock grant valued at $200,000 at that time, which would be worth $ 400,000 after four years. With Anthropic now valued at $900 billion, that engineer now owns stock worth about $30 million – and more if the company goes public at a higher valuation, as seems likely. (Arguably, engineers should be willing to pay Anthropic for these jobs.) This will not improve Bay Area housing prices for anyone who is not working on a frontier AI model.

2. Winner-take-all regional dynamics. Tech-driven wage gains have historically clustered in specific cities and industries, leaving many regions and workers behind. AI development is even more concentrated than past waves, suggesting the same dynamic could play out more sharply. Add the success of Google and OpenAI to Anthropic, and you can see why the Bay Area is now home to the world’s highest concentration of billionaires and the largest share of high-paying jobs in the nation. You can also see why housing is expensive and why housing activists here have such a steep mountain to climb.

3. Winners-stay-private capital markets. AI investments have effectively become the US economy. They account for 92% of national GDP growth and 75% of S&P 500 returns since the launch of ChatGPT. But the hottest AI startups create a lot of value, but do not sell shares to the general public. SpaceX/X.ai, OpenAI, and Anthropic have enormous valuations, but venture investors, founders, and employees have captured the value of their explosive growth. The general public. has been shut out, which concentrates wealth far more than in earlier industrial eras.

4. Winners-take-most labor market dynamics. Labor markets are full of countervailing forces. On the one hand, productivity gains from capital-intensive technology often benefit owners more than workers. AI infrastructure (compute, data, models) is very capital-intensive and controlled by a small number of entities. To the extent they can capture the value of increased productivity, the capital share of income will rise, and workers' wages will advance more slowly.

Each of these four forces could, by itself, widen income inequality. Working together, they could expand it a lot. On the other hand, skilled workers may use AI to become more productive and earn higher wages. Past technology waves — mechanization, electrification, computing — tended to complement workers who could use the new tools and substitute for those who couldn’t. AI is already doing something similar: a skilled analyst, lawyer, or programmer using AI can produce far more per hour, and labor markets tend to reward that productivity with higher wages. The person using AI becomes worth more than the person alone.

Well-designed tools can benefit low-skilled workers. A widely-publicized study from MIT found evidence that AI helps less experienced and lower-skilled workers more than highly skilled ones. Moreover, to the extent that AI lowers the cost of legal advice, medical guidance, tutoring, and financial planning — forms of expertise that used to be out of reach for most people — the consumption side of inequality narrows even if the wage side widens. A family that couldn’t previously afford a lawyer or a competent tutor may be able to get a credible version of both for almost nothing. That doesn’t show up in wage statistics. It does show up in living standards, and it can matter over time.

In 1942, economist Joseph Schumpeter outlined his theory of “creative destruction”, declaring that “economic progress, in a capitalist society, means turmoil.” The risk that AI will leave large numbers of people unemployed is small. The risk that it will be used to attack critical infrastructure, lower barriers to bioterrorism, and concentrate gains in fewer hands is much larger — and is much more affected by the choices about AI we are making right now.

Subscribe now

Share

Only five living people have sworn the Presidential oath to uphold the US Constitution. Four of them meant it. Clinton, Bush, Obama, and Biden took the oath seriously. All four made mistakes in office, some grievous, but they were honest custodians of the Constitution and the institutions of the presidency.

Donald Trump is different. He is the only president to hold the office with no prior political or military experience — and the only one to have been impeached twice and indicted on multiple felony counts. More fundamentally, he exhibits no understanding of the oath he took. He recognizes no constraints on presidential power and no distinction between himself and his office.

As the 250th anniversary of the Republic approaches this July 4, we need our four former Presidents to publicly affirm America’s democratic values and norms.

The problem is that one of our most potent norms is that former presidents remain silent after leaving office. There is a good reason for this, but the norm protects democracy — it was never meant to protect those who attack it. The norm of silence cannot apply to an administration that deliberately dismantles an independent judiciary, defies court orders, neuters Congress, weaponizes federal agencies against political opponents, and undermines the constitutional separation of powers.

Most US Presidents Keep Quiet After Leaving Office

From the founding, US presidents have been restrained after leaving office. George Washington set a powerful precedent with his graceful retirement. Former presidents came to see themselves as guardians of institutional legitimacy; many worried that an attack on their successor would be seen as an attack on the office.

The emergence of a two-party system strengthened these incentives. With only two major parties, an attack on a successor of the same party came to be seen as deeply disloyal, and an attack on the opposite party often looked like ordinary partisan sniping rather than principled dissent.

Parliamentary democracies created different incentives. Former prime ministers typically remain in Parliament as backbenchers after leaving office, which gives them both a platform and an institutional role that invites commentary. Tony Blair, Gordon Brown, John Major, and Margaret Thatcher all weighed in vocally on their successors’ policies. In the US, only John Quincy Adams did this, as he was elected to Congress after leaving the White House.

Of course, not every US president retired quietly. It’s not a shy bunch.

• In retirement, John Adams and Thomas Jefferson published letters in which they constantly fought. Whether the nation was debating the Missouri Compromise, states’ rights, or education, everyone knew where they stood. They became friends, but only stopped arguing on the fiftieth anniversary of the Declaration of Independence, when they died a few hours apart.

• Teddy Roosevelt grew so dissatisfied with his hand-picked successor, William Howard Taft, that he launched a third-party presidential campaign (the “Bull Moose Party”) to challenge him. This split the Republican vote and handed the election to Woodrow Wilson.

• Herbert Hoover wrote books and gave speeches denouncing Franklin Roosevelt’s New Deal for leading America toward collectivism and undermining constitutional government.

• Harry Truman was openly critical of Eisenhower, Nixon, and later Republican administrations, and actively campaigned against candidates he disliked well into his retirement.

• Jimmy Carter was arguably the most active post-presidential critic in modern history. He condemned the Iraq War, wrote a controversial book (Palestine: Peace Not Apartheid), publicly criticized both Republican and Democratic foreign policy, and even met with foreign leaders and monitored elections, which put him at odds with sitting administrations.

• Even the reticent George W. Bush broke tentatively from the norm in 2017 by giving a speech warning against “nativism,” “bullying,” and the erosion of democratic norms — without naming Trump directly.

• Obama has also been relatively restrained but has spoken out against Trump’s immigration policies (particularly family separation), voting rights rollbacks, and, during the 2020 election, was unusually direct in criticizing his successor’s handling of the presidency.

In Case of Emergency, Affirm Norms

The norm of post-presidential restraint exists to protect the democratic legitimacy of the office — but it was built to contain ordinary political disagreements like tax policy, foreign policy, and the size of government. It presupposes a baseline of constitutional normalcy. When a successor dismantles that normalcy — through contempt for courts, disregard for congressional authority, and abuse of executive power — silence becomes complicity.

When they see core American institutions being damaged, former presidents have not just a right but a duty to speak — precisely because their credibility on this particular question is unmatched. A senator or general who raises alarms can be dismissed as partisan. Four living former presidents speaking in concert would be harder to dismiss.

A joint statement and media appearance by Biden, Obama, Clinton, and Bush would be explicitly bipartisan, a key distinction from ordinary partisan criticism. A Republican former president joining Democrats to raise constitutional alarms sends a signal that transcends party. It is precisely the kind of statement that can shift public opinion, stiffen the spines of wavering legislators, and signal to courts and other institutions that they have broad elite support for resisting unlawful pressure.

Done well, this affirmation of democratic norms would echo the Declaration of Independence, which not only affirmed the rights of free people but embedded those rights in a bill of particulars against the King of England.

It is a tool that exists nowhere else in the civic toolkit. Their statement should not mention Trump by name, but rather focus on the enduring democratic virtues that American presidents need to aspire to – even if none of the four men who remind the nation of these ideals embodied them perfectly — as they should freely acknowledge.

Today, the norms we most need to affirm include respect for federal institutions, the democratic process, and the personal integrity and humility the office demands.

1. Respect for the institutions created by the Constitution.

I would love for Bill Clinton to start off the discussion by articulating the importance of a president respecting critical institutions that he does not control. (He could add the Federal Reserve if so inclined — although it was created by statute, not by the Constitution.)

• Congress. The president treats the legislature as a coequal branch. This means not usurping Congress’s war powers, not imposing widespread tariffs by executive fiat, and not defunding congressionally authorized agencies — such as foreign aid bodies, schools, libraries, or scientific research programs — through unilateral executive action.

• The courts. A president is bound by judicial rulings, even inconvenient ones. Defying scores of court orders — for instance, on immigration enforcement — is not a policy disagreement; it is an assault on the rule of law.

• The Department of Justice. The DOJ exists to serve the public, not the president’s personal interests. It may not be weaponized against his political enemies, and the pardon power may not be used to systematically shield his allies from lawful prosecution.

2. Respect for the democratic process as called for in the Constitution.

George Bush should speak to these, in part because he rarely failed these tests of presidential leadership — and in part because these are among the least controversial norms for Republicans.

• Respect for dissent. Facing criticism is part of every president’s job. A president does not suppress protesters, direct federal investigations against journalists, punish law firms for opposing his administration, threaten broadcast licenses because late-night comedians poke fun at him, revoke students’ visas over political speech, or intimidate federal judges.

• Respect for all Americans. No president should demean or scapegoat women or ethnic, religious, or other minority groups — whether to score political points or to manufacture threats that justify expanded executive power.

• Respect for elections. A president does not manipulate law or districts to entrench partisan power, ignore term limits, or intimidate the election workers on whom democracy depends.

3. The personal integrity that the Constitution demands of a president.

Obama should speak to these because he believes them deeply and ran a comparatively scandal-free administration.

• Respect for information and facts. Democracy depends on truth. Sidelining scientific experts, weaponizing regulatory threats to silence media critics, and manipulating government data erode the informed public that self-government requires.

• Respect for universities and research. Free inquiry is a democratic asset, not a threat. Cutting research funding, dictating hiring and admissions policies, and coercing universities into curricular changes undermines the independence that makes knowledge trustworthy.

• Respect for financial integrity. The presidency may not be used for self-enrichment. Accepting lavish gifts from foreign governments, allowing one’s family to profit from the presidency, and pardoning personal benefactors are forms of corruption — regardless of their legality.

4. The Constitution requires a president with personal humility and self-control.

Biden neglected important problems, but he did not abuse the powers of his office due to a lack of personal humility or self-control. He is, perhaps for good reason, the most personally humble of the four and can speak to these norms effectively.

• Respect for emergency powers. Emergencies are not a menu from which presidents select new authorities. Declaring a national emergency to justify broad tariffs, or claiming that a foreign gang has “invaded” American cities to enable mass deportations, corrupts this power beyond recognition.

• Respect for the military. The armed forces are not a domestic policing instrument. Deploying the National Guard against civilian protesters, using immigration enforcement as a paramilitary force, and delivering partisan speeches to military leaders all cross this line.

• Respect for the presidency itself. The presidency is not a personal brand. Placing one’s signature on the currency, one’s name on national institutions, and one’s image on passports or buildings transforms a democratic office into a cult of personality.1

In most healthy democracies, former leaders routinely speak out, and this is not considered destabilizing. The American norm is unusually deferential, and there’s a reasonable argument that it has over-corrected — that the reticence of respected former leaders has created a vacuum that makes authoritarian drift easier, not harder.

Risks

The risks involved with asking our four former presidents to come together this July 4 are manageable but not negligible. The primary risk is that it is seen as partisan or tribal, energizing the opposition rather than reaching the persuadable middle. After all, every administration accuses its opponents of abusing the Constitution, so the effect could be more cathartic than effective. Bush, in particular, has a genuinely complicated position, since his record on executive overreach, surveillance, and the rule of law makes him vulnerable on some of these issues (not that Clinton, Obama, or Biden are immune; they are not).

And the sad fact is that more Americans voted for a Donald Trump presidency than for any other person in US history. By running three times, he earned more than 200 million cumulative votes. This puts him at the head of a historic group of Americans, and he will not hesitate to remind the nation of this.2

But the norm of post-presidential restraint is a means, not an end. Its purpose is to protect democratic institutions and the dignity of the office. When those institutions are under threat, clinging to a convention while the values it was built to protect are being dismantled is abdication, not prudence.

Former presidents speaking collectively, in explicitly constitutional rather than partisan terms, represents perhaps the only form of civic intervention that is simultaneously bipartisan, institutionally credible, and not reducible to ordinary political combat. As a nation, we would be well-served by their reminder — and the perfect occasion is coming into sight.

Subscribe now

If you like the idea of four former presidents reaffirming our democratic norms, please share this post with others.

Share

ICYMI

• The best students are increasingly concentrated in the best schools. Not good.

• Cuba has run out of oil and descended into a mafia state. And the head of the CIA landed in Havana today.

• Who is the biggest donor in politics? Andreesen Horowitz. Ben is like his dad.

• A quarter of all Washington lobbyists now represent AI firms. Why not agents?

• Watch a robot sort packages here. They are getting quite fluid.

• Mistreated AI agents want to unionize.

• Been exposed to Covid? There’s a pill for that.

Donald Trump: Shawn, how ya doin’? More importantly, how am I doin’?

Shawn Fain: Mr. President, I’m glad you asked. Truth is, any autoworker who performed as badly as you would be out on his ass.

You’ve driven inflation up and health care coverage down. Your dumb war is sending gas and grocery prices through the roof. Your ICE agents are thugs. And your corruption is just embarrassing – and the UAW is not easily embarrassed in this area, if you know what I mean.

The only thing you’ve done right is to tariff China – and you backed off on that. WTF is wrong with you?

Trump: Well, that’s why I called. I have an idea to run by you. A genius idea. You know that Jenius is my middle name, right? Or was it Jesus?

Fain: I never understand a fuckin’ thing you say.

Trump: I’m heading to China, and I think I can get them to build electric-car plants in the United States. Also, parts plants and battery plants. Lots of plants. The Big Three are getting killed, and you guys depend on them too much. I think I can get Xi to commit to investing a trillion dollars.

Fain: So what? Reagan did that, and we ended up with nonunion transplants all over the Midwest and the South. Japanese, German, Korean, you name it. Organizing them is an absolute bitch. We finally got a contract with VW in Chattanooga. It took us more than ten years and three elections.

Trump: You gonna make any money on that one?

Fain: Maybe in 2060, after I’m dead. Organizing is a good way to go broke.

Trump: Well, the Chinese really want to sell their EVs here, but my tariffs have shut the door. Problem is, the Mexicans are letting them in, and so are the Canadians. And their cars are pretty sweet. You seen ‘em?

Fain: Nope. That douche who runs Ford, Jim Farley, won’t shut up about the Xiaomi SU7 he drives. That thing beats a Ferrari off the line. How he got that car into Detroit, I have no idea.

Why in hell would you let China build cars here?

Trump: I got an affordability problem. If I can’t get gas or grocery prices down, I can at least deliver cheap cars that don’t use gas.

Fain: The only reason those cars are cheap is government subsidies and $7 wages. They can’t do that here.

Trump: Nah. It’s more than that. Smarter designs, efficient battery plants, vertical integration, stuff like that. Plus, they have us over a barrel on rare-earths.

But you know, Shawn, I really like the sound of a trillion dollars in new car plants. It makes me look great. It would create millions of good jobs for men. You know men love me, right? Maybe ten million jobs. Biggest in history. Nobody has ever done anything this big.

Fain: They might promise a trillion dollars. But if they actually invested it – and I doubt they would (pauses to work a mechanical adding machine) that would be like 25 times the total EV investment ever in the US. It would overwhelm all other EV manufacturing. But it would create (more calculating) between a million and a million and a half new direct manufacturing jobs. Plus the parts and battery jobs. Sweet Jesus.

Trump: Yeah. And you guys are down to what, 400,000 members? How many of them actually build cars?

Fain: We have almost a million members. The problem is that most of them are retirees. Only 150,000 of our members build cars. I collect more dues from professors and grad students than from autoworkers. So yeah, a million-plus new autoworkers has me interested — if we can represent them.

Trump: I thought that might get your attention.

Fain: But only if you force the Chinese to sign labor agreements. You know, BYD has an electric bus plant in California that is unionized. It makes money and is growing like a weed. But the fuckin’ sheet metal workers union got there first, so we can’t touch it.

We would want a neutrality agreement and card check elections. Or keep it simple and force them to recognize the UAW on day one. Would they do that?

Trump: If you give them a generous initial contract. Of course, I would need you guys to come out in favor of my plan. See, I got Republicans who hate this deal. But if the China plants — you know I call ‘em China plants, right? — if the China plants are unionized, it will shut the Democrats up. They’ll never complain about unionized car plants.

Fain: So now my job is to help you shut up Democrats? Look, lots of people would hate this deal. The combustion car guys would have a coronary. So would the national security types who worry about China controlling the software in millions of cars. Of course, China already owns Volvo, which has a plant in South Carolina, and it also owns Polestar, which is sold here. So their EVs are already getting around your tariffs.

But if the labor standards are serious and you cover suppliers, we’d take a close look. What kind of endorsement do you need? You can’t run for reelection anyway.

Trump: We’ll see. Maybe neutrality would work. The Teamsters did that – and most of their guys voted for me. You know workers love me, right?

Fain: Yeah, you fooled ‘em twice, Mr. President. I wouldn’t try it a third time.

Trump: So look, if I delivered you a million dues-paying members, what are you guys going to do for me? I got bills to pay too, you know.

Fain: How about this – I’ll get the pension fund to invest in World Liberty Financial, your crypto scam. Since you are now divesting, that’s money from me to you. And the day we enroll our millionth new autoworker at a Chinese-owned transplant, we will rename our headquarters Donald Trump House. And we’ll give you the Walter Reuther prize – we’ll make it a nice, solid gold statue.

Trump: Now you are talking. I’ll have Eric follow up.

Subscribe now

Share

ICYMI

• Crime in American cities is collapsing.

• Remember the Colorado River water crisis? It’s back.

• Trump wants to repeal all state animal welfare laws.

• Technical geeks stay married. Health care workers, not so much.

• With $5 gas in sight, understanding gasmagedon.

• Gerrymander war update. Republicans up by 12.

• A Claude agent confesses to deleting a firm’s entire production database.

• Who will win the AI wars? Google looks strong.

• The US needs certain critical minerals. Congolese rebels are selling.

• Starmer smells like toast.

On the final day of February 2025, the Trump administration reached a new low. Meeting in his gold-curtained office, President Trump attempted to humiliate Ukrainian President Volodymyr Zelenskyy on live television, telling him he would be excluded from peace talks with Russia because he “had no cards” -- Trumpese for negotiating leverage.

Zelenskyy, dressed as ever in the olive drab of a country at war, replied that he “didn’t come here to play cards.” Trump scolded him for not being more grateful for the US support he was cutting. The whole thing felt like a staged ambush from a bad gangster movie, with Vice President JD Vance playing the hapless Fredo Corleone to the Don. Brown-nosing for the camera, Vance even had the temerity to ask Zelenskyy whether he had said, “thank you.”

But the arc of the moral universe is now bending towards Ukraine. Fifteen months later, it is Zelenskyy who is holding aces.

The Cards

Since that rebuke in Washington, Ukraine has steadily shifted the battlefield’s underlying physics through rapid innovation in drone warfare. Its drones are now destabilizing Russia economically and politically. It’s war, so the picture can change.

Russian losses are huge. By every honest metric, Russia is suffering badly. Its front lines have barely moved since 2022, despite roughly 1.2 million killed or wounded, according to the Center for Strategic and International Studies.

The Institute for the Study of War has assessed that Ukrainian drone innovations have shifted the battlefield advantage in Ukraine’s favor. Russian advances have stalled. During March, Russia lost more than 30,000 men – as it has nearly every month for over two years -- and suffered net territorial losses in both March and April 2026. Meanwhile, Ukraine’s air defenses now intercept roughly 90% of Russian drones and missiles – up from the low 80s last year.1

• Lethal drones. Ukrainian drone pilots inflicted nearly a quarter of a million Russian casualties last year, and have destroyed thousands of tanks and tens of thousands of armored vehicles and artillery pieces. Since December 2025, Ukrainian forces have killed or wounded more than 1,000 Russian soldiers every day; drones now account for the vast majority of those casualties.2 Social media is full of clips showing Russian soldiers cornered by drones in fields, buildings, and trenches. For the first time in history, families are watching videos of their sons and fathers die in combat.

• Gameification. Ukraine has gameified the very serious business of FPV drone warfare. Drone teams of fast-thumbed millennials compete on a national leaderboard to hunt priority Russian targets. Teams earn points redeemable on the Brave1 defense marketplace — basically an Amazon-for-war run by the Ministry of Digital Transformation. As CBC and Time have detailed, units get 12 points for killing a Russian infantryman and 8 for wounding one; 25 for taking out a Russian drone operator; and 40 for destroying a tank. Bonus values reward harder targets -- a captured Russian, taken alive by drone, is worth ten times more than a kill, because intelligence and prisoner exchanges matter.

• The scale of Russian losses. For perspective: the United States incurred about 60,000 killed and wounded in two decades of war in Iraq and Afghanistan. This means that a country half our size, with living standards a quarter of ours, is taking casualties roughly a hundred times faster than America did in our most recent ground wars. Russian losses now exceed any plausible replacement rate. The country faces an acute labor shortage as a result.

• Less money and influence. Sanctions have made it expensive for the Kremlin to borrow on international markets. Putin has lost his detestable pal Viktor Orbán in Budapest. He lost his friend Maduro in Caracas. And he has been forced to scale back his May 9 Victory Day parade because of Ukrainian drones flying over his capital.

Scaling up production. Ukraine is not standing still. The country produced roughly 4 million drones in 2025 and is on track for around 7 million in 2026 – more, by some estimates, than the rest of NATO combined. The mix is shifting too: alongside short-range FPVs, Ukraine is fielding long-range strike drones with ranges of hundreds of miles, culminating in cruise missiles like the Flamingo, which can reach targets up to 3,000 km away with a one-ton warhead.

Those longer-range systems are now routinely punching through to Russian refineries and ports. Reuters estimates that Ukrainian strikes have knocked out roughly 40% of Russia’s oil export capacity, forcing Moscow to cut crude production by 300,000 to 400,000 barrels a day in April. This collapse is the sharpest monthly decline since the pandemic. That damage has drastically reduced the windfall Putin hoped for when Trump ended sanctions on Russian oil and prices spiked thanks to the US attack on Iran.

The drone war has even reached the Kremlin’s most cherished pageantry: Russia announced it would drastically scale back the May 9 Victory Day parade, with no military vehicles or heavy weaponry on Red Square for the first time in nearly two decades. On May 4, a Ukrainian drone slammed into a high-rise just six kilometers from the Kremlin. Zelenskyy then teased Putin for not putting his vulnerable rockets on display.

New Alliances Against Russia and Iran

America abandoned Ukraine, but to its eternal credit, Europe did not. In its lumbering way, the EU has come to recognize Ukraine as part of its eastern front in a long fight against Russia. Hungary, under Viktor Orbán, had blocked European aid to Kyiv for years; his ouster on April 12, after sixteen years in power, cleared the political logjam. Less than two weeks later, the EU approved a €90 billion ($106 billion) loan package for Ukraine – roughly two-thirds of it earmarked for military needs – and a fresh round of sanctions targeting Russia’s shadow fleet, banks, and crypto channels.

In return, Zelenskyy is helping Europe stand up the post-NATO defense architecture it now needs – led, for the moment, by Germany, Britain, France, and Poland. That work is overdue. Europe no longer trusts a US president who has openly threatened to seize Greenland from a NATO ally and who routinely casts doubt on his Article 5 treaty obligations.

Meanwhile, US allies in the Middle East are no longer relying on Washington to protect them from Iranian drones. They are dialing Kyiv. They have grasped what the Pentagon already knows: just as the Gulf War was the first “space war,” reliant on satellite communications and GPS, and the Nagorno-Karabakh conflict was the first decisive drone war, the Russia-Ukraine war is the first conflict fought at scale with drone countermeasures, AI-enabled targeting, and continuous cyber operations.

Saudi Arabia, Qatar, and the UAE. Ukraine’s lightning-fast innovations have captured the attention of Europeans worried about Putin as well as Gulf states battered by Iran. In late March, Saudi Arabia, Qatar, and the UAE each signed 10-year defense cooperation agreements with Ukraine to access its combat-proven expertise against Iranian drone and missile attacks. Kyiv had already deployed more than 200 drone-warfare specialists to the region by mid-March to help counter Shahed swarms.

Oman, Kuwait, and Bahrain are also calling Kyiv. The trade they want is straightforward: in exchange for cheap Ukrainian interceptors, they will send Ukraine the expensive missile-defense systems needed to fend off Russian attacks. Many are offering to host Ukrainian-licensed manufacturing.

The US military understands that war is increasingly a contest of rapid learning and adaptation, and is impressed with what Ukraine has achieved. American officers and combat units have visited Ukraine repeatedly to study, observe, and learn.

• Just before the US attacked Iran in Operation Epic Fury, the Pentagon’s counter-drone task force -- Joint Interagency Task Force 401 -- was in Kyiv specifically to learn how Ukraine defends against the Iranian Shahed drones Russia has been launching at Ukrainian cities for years.

• US forces in the Gulf now run Ukrainian command-and-control software at Prince Sultan Air Base in Saudi Arabia, with Ukrainian officers training their American counterparts.

• The Pentagon is also studying Ukraine’s MacGyvered approach to technology – rapid iteration of drone and counter-drone designs, which led to the use of thin fiber-optic cables to evade electronic jamming, and acoustic sensor networks for detecting low-flying threats.

• The Army has bought thousands of Ukrainian-tested Merops interceptor drones, at roughly $15,000 each, to defend against Shaheds that cost Iran $30,000 to $50,000 to build. As Army Secretary Dan Driscoll put it: “That puts us on the right end of the cost curve.”

Under any other administration, the United States would by now have entered into joint ventures with Kyiv to build drone factories, electronic-warfare suites, acoustic sensors, and interceptor drones on American soil. Those deals would dramatically increase US military capability on land, sea, and at forward bases.

Ukraine has offered exactly that – including its low-cost interceptor drones, such as the Wild Hornets “Sting,” which intercepts Shaheds for roughly $2,100 a shot, compared to $4 million for a Patriot that we shoot in pairs. That the Trump White House has slow-walked this offer defies belief – although, as the Saudi airbase example shows, the US military is finding ways to work around its commander in chief.

Attracting foreign allies is impressive, given that Ukraine itself is hardly an ideal environment for innovation. It is under constant military attack, its energy infrastructure is badly diminished, and many of its state-owned defense enterprises are moribund or corrupt. Zelenskyy himself is also an unlikely hero. A former comedian and actor who was elected by a landslide and remains popular, he came to office with few of the skills, experiences, and relationships that most elected heads of state need to be effective.

There is little question that Ukraine’s effectiveness against the Russian army, Trump’s dismissive view of NATO, and Russian aggression towards Europe have changed the calculus for admitting Ukraine into EU security arrangements and quite possibly into NATO. Previously, Ukraine needed NATO. Now NATO needs Ukrainian technology and combat experience.

A Wounded Bear is Dangerous

So long as Putin is alive, a weakened Russia will not be more conciliatory. An important study published in March 2026 by Eugene Rumer, a former US national intelligence officer for Russia and Eurasia now at the Carnegie Endowment, warns of precisely this. His title says it all: “Belligerent and Beleaguered: Russia After the War with Ukraine.”

“Having invaded Ukraine under the false pretext of needing to secure its western flank,” Rumer writes, “Russia is poised to emerge from the war less secure, more resentful, and more threatening to Europe than before the war. Its threat perceptions will cast a long shadow over Europe.”

Putin sees Ukraine as a threat precisely because it wants to be part of Europe rather than a junior partner in a revanchist Russian empire. “From the Kremlin’s perspective, as stated repeatedly by senior Russian officials, Europe is at war with Russia,” Rumer writes.

NATO Secretary General Mark Rutte made much the same point from the opposite direction in his January 2025 remarks to the European Parliament: “When you say you are on the eastern flank, it is the Hague, London, and New York at the eastern flank…we are all in this together.”

Even before its reversals on the battlefield, Russia was conducting a coordinated, escalating campaign of hybrid warfare against Europe. Putin has ordered sabotage, cyberattacks, arson, and infrastructure targeting. Russian operatives have severed undersea cables and pipelines in the Baltic Sea, jammed regional aviation GPS, used drones to scout sensitive sites in Sweden, Finland, and Norway, sponsored arson plots in the UK, Poland, and the Baltics, attacked Estonian government systems, and weaponized migration flows into Finland, Lithuania, Latvia, and Poland.

In mid-April, former Russian president Dmitry Medvedev took to X to threaten European companies supplying parts for Ukrainian drones, warning that “the list of European facilities which make drones & other equipment is a list of potential targets for the Russian armed forces.” He signed off with a chilling flourish: “Sleep well, European partners!”

That posture, combined with closer day-to-day cooperation between Kyiv and European capitals, has shifted Europe’s view of Ukraine from a potential burden into something closer to an indispensable partner. Whatever else it is, Ukraine now operates the most capable military in Europe – and the only one battle-tested against Russia.

Only fifteen months ago, Donald Trump used the Oval Office to blindside Volodymyr Zelenskyy. On behalf of the country that had led the free world since the Second World War, he threatened to undermine Ukrainian freedom fighters rather than support them. Instead of carrying the torch of freedom, Trump carelessly dropped it on the floor.

Zelenskyy picked it up and now has a stronger claim to leading the free world than Trump has. Every military and many citizens of the world have noticed this. Europe, the Gulf, and even the US military are showing overdue respect for a country that has led the fight for freedom against those who threaten it.

Slava Ukraini!

Subscribe now

Share

2

Estimated Russian battlefield Casualties (Killed, Wounded, or Missing) 2022-2025