Economics: Weekly Summary (December 22-28, 2025)

I skimmed a week of economics writing and came away with a few clear threads. They keep popping up like those persistent weeds in the garden — you pull one out and three more show up nearby. Some pieces felt like sharp little arguments. Others were more like long, slow-grooving reflections. I would describe them as a mix of worry, curiosity, and a fair bit of argument about what models actually buy us these days.

AI, models, and the limits of prediction

The machine-brained corner of the blog world was loud this week. You had technical digging into whether modern transformers can actually learn economic relationships and policy responses. Arpit Gupta lays out the Lucas critique and then asks: can a transformer, fed enough history, learn the data generating process well enough to predict what happens when policy changes? To me, it feels like someone asking whether a GPS can replace a mapmaker. The GPS may get you to town, sure, but will it explain why the bridge collapsed? Gupta’s point — that transformers are promising for forecasting but not a cure for structural-model problems — is a patient, modest pushback on overconfidence in black-box methods.

That theme bounces off a few other posts. Gary Marcus shows up twice this week with two different moods. One post is sceptical about the LLM frenzy and asks whether the AI boom is overheating. Another says the AI bubble is fading and warns about missing world models — basically the same worry as Gupta but framed more like an alarm bell. I’d say Marcus is the grumpy uncle at the family dinner. He’s loud, and he repeats the point until you notice it: shiny models don’t equal economic returns if they can’t reason about the world reliably.

Then there’s the more systemic, governance take from Grigory Sapunov. He argues we should think about AGI safety in distributional terms. Not, "can we align one model?" but "how do we govern a swarm of specialized agents trading services, competing, colluding, cheating?" That’s a different question. It’s not just accuracy. It’s rules, incentives, leakage, and market design. To me, this reads like zoning rules for a new suburb. You can’t let developers build whatever they want and expect the neighborhood to turn out fine.

Other writers bring complementary worries. Dave Friedman writes about how advanced AI might erode the institutions that give people purpose. It’s a softer scare. More cultural. It’s the kind of piece that makes you think of empty storefronts downtown after a big mall opens. The tone varies, but the recurring idea is the same: the tech is fast. Our understanding and institutions are slower. Trust and real economic benefit are not the same thing.

And a small, slightly odd note: a post asked whether transformers can actually learn policy responses in the presence of structural shifts. It’s a neat intellectual puzzle. It’s also a practical question for folks spending big on data centers. Some papers say companies are not yet seeing the returns they were promised. Others suspect overinvestment. It’s a classic boom-time smell. Like at an auction where bidders confuse beauty for value.

Housing, wealth illusions, and the filtering story

This week had a steady stream on housing and what rising home prices actually mean for living standards. Kevin Erdmann runs a multi-part tour through housing: part two argues we're not as wealthy as the headline prices suggest; part three talks about filtering — and how we’ve moved from downward filtering to upward filtering. If you don’t follow this daily, here’s the short version: once homes slowly become cheaper as they age and wear, people can move down the ladder and afford a place. That used to be normal. Now, homes seem to keep getting pricier even as investment in new residential construction falls. It’s like having fewer seats on the bus while the fare keeps going up. The result is local displacement and the weird sense that wealth on paper doesn’t buy much at the kitchen table.

There’s also an angle on subsidies and mortgage access. More money in the form of subsidies and looser mortgage terms can lift prices without adding the supply needed for affordability. It’s a bit of window dressing. You get the numbers to look okay, but families still feel the squeeze. I’d describe these posts as quietly furious about how policy often treats symptoms rather than causes.

Link that with the old monetary debates and you get a richer picture. One piece on the Continental currency by Quoth the Raven reminds us that money’s value has always been partly about promises — about whether someone will honor that later redemption in gold, or in taxes, or in credibility. That lesson is a small echo of the housing pieces: lots of value is social belief, not actual usable space.

Macro puzzles, GDP vs GDI, and consumer mood

There were a few deep-but-not-showy posts about how we measure the economy. Mike "Mish" Shedlock had multiple short updates noting oddities. One flagged a 12-quarter run where Real GDP exceeded Real GDI. Another reported GDP up 4.3% but GDI only up 2.4% in Q3. Numbers like that matter because GDP and GDI are supposed to be two sides of the same coin. But they aren’t always. It’s like balancing your checkbook and finding an unexplained deposit on one sheet and a missing one on another.

Why mention this? Because the gap changes how we read the economy. Is growth broad-based? Is demand real or statistical noise from export swings? These are the kind of small, technical things that make big policy choices harder. Mish also noted that consumer sentiment hit a 50-year low. That’s dramatic. If sentiment stays crushed, the psychology of spending slows, investment cools, and politics gets spicy.

There’s a neat little interaction here with the AI debate. If firms fund big AI projects and the headline GDP numbers look good, but income doesn't trickle down as GDI suggests, then you get the weird mixture of proud firms and anxious households. That old split between top-line production and real income on the ground keeps returning.

Geopolitics, trade, and the human cost

Three posts pulled in geopolitics. Sam Cooper ran an op-ed sharing Dimon Liu’s testimony on China's growth and its human cost. This is blunt stuff: the miracle involved massive displacement and hardship. It’s a reminder that big numbers often hide big human stories. The piece urges Canada — and by implication any trading partner — to look harder at the supply chain and the social price of cheap goods. It reads like a suggested reading for anyone who still thinks trade is tidy and neat.

Michael Hudson (via Naked Capitalism) took a different tack. He argues the U.S. has shifted from collaboration to coercion in its economic relations. I’d describe his tone as both furious and nostalgic. He sees an older international order being rewritten into something more extractive. It’s a warning that the rules of the game are being rewritten in ways that can boomerang.

There are also domestic political notes. A number of posts — one explicitly imagining policy under a future Trump presidency — trace how political change can reshape economic life. The Trump pieces, including one that jokes about his view of the Fed chair, are loud and political. They make the obvious point: politics and economics are one messy bundle. You can’t study supply curves without also watching the fight over who gets to run the stadium.

Innovation, jobs, and the economy of ideas

On a lighter note, the innovation beat had helpful lists and resources. Matt Clancy put together job-market papers and notes about a PhD course on the economics of ideas. There’s a genuine hunger here. A smart economy needs new talent. Those posts felt like the teacher leaving a stack of reading on the table. They made the case for more eyes on how patents travel, how research careers unfold, and how cities and firms push new tech into the wild.

Irwin Collier dropped an archival curiosity: a 1944 budget proposal from the University of Chicago’s Economics Department. It felt like a postcard from the past. It’s useful because it shows how institutional choices about faculty and fields have a long tail. Decisions on what to fund in 1944 ripple into what questions are deemed important today.

And Gonçalo Valério posted a personal list of books he liked in 2025. He highlights one on Bell Labs and one on Portugal’s economic stagnation. That’s a nice human touch. Books, after all, are how many of us borrow other people’s long thinking. They connect historical pattern to the present.

Money, fraud, piracy, and strange economics

There were a few smaller, but oddly interesting, posts that feel like the kind of thing you share at a cafe.

• Josh Beckman wrote about software piracy and suggested there might be an optimal level of piracy for revenue. That’s provocative. It’s an economist’s version of a wry shrug: too strict and you kill distribution; too lax and you kill the market. He compares it to fraud having an optimal level. It’s the kind of perverse logic that sticks in your head.

• Nominal News looked at gifts in the workplace and argued that gifting can boost productivity. Not just because of dollar value, but because of perceived effort. I’d say it’s a reminder that not everything is a number. Some things are social glue.

• Zev Shalev dug into predatory lending and the old Manhattan financial machine that helped cause the housing crash. He paints Wall Street as looking and acting like sharks in a feeding frenzy. Those older scandals keep shaping trust in financial institutions.

Politics, policy campaigns, and targeted funding

A couple of small-but-important threads about policy advocacy and funding popped up.

thezvi.wordpress.com reported on Balsa Research’s fundraising and their work on the Jones Act. This is niche stuff, but it matters. Regulations like the Jones Act have distributional effects on shipping costs and regional economies. The piece is earnest: it’s about a small organization trying to move a specific policy needle. That kind of work often matters more in the long run than sweeping speeches.

Naked Capitalism’s link roundups kept returning to stories that knit politics, energy, and climate together. They have a kind of steady hum of links that make you feel like you’re sitting on a porch listening to a neighbor give the news. Practical, sometimes cranky, often urgent.

Long-term worries: resources, “peak everything”, and social fabric

A darker vein runs through a few essays. One called 2025 the Year of Peak Everything. The argument: resource depletion, rising energy costs, and collapsing industry in places like Europe create a real risk of deindustrialization. The author frames 2025 as a hinge year. It’s the sort of essay that will make some people nod and others roll their eyes. To me, it feels like watching clouds gather before a storm. The methods vary. But the sense that limits matter is not trivial.

Those worries about limits bump into the AI and institutional fears. If automation and AI advance while energy and materials tighten, what happens to expectations about growth? The two anxieties—running out of energy and running into machines that replace old forms of work—make for a strange pairing. It’s like two housemates arguing over the thermostat and the rent at the same time.

Little ideas that keep poking through

A few more pins on the board:

• Linch Zhang put up a list of five “unknown knowns” that change how arguments go. Things like Net Present Value and the Intermediate Value Theorem. Boring names. Useful powers.

• Tim Harford took an economist’s look at dating and marriage. He uses simple economic logic — economies of scale in consumption and production within households — to explain why people lower standards if options are scarce. It’s charming. It makes dating seem like shopping for a used car, with test drives and shaky warranties.

• Tree of Woe and Ryan Stohl both refused to make crisp predictions about 2026. That’s actually refreshing in a season of hot takes. They note that sometimes the obvious gets missed — the human tendency to reframe events after they happen. People love predictions because they make us feel smart. But the week’s posts remind you that history usually arrives sideways.

Strange tangents and cultural notes

A few pieces wander in unexpectedly but tie back. Hrvoje Morić talks about tokenization and the demolition of nation and constitution. It’s political, slightly conspiratorial, and a trip into how new tech stories become cultural myths. It’s like reading a thriller after finishing a government white paper.

And then there’s a very practical fundraising piece about the Jones Act. It’s oddly comforting. You remember that not everything is big theory. Some people work on the small levers.

Patterns and disagreements you can feel in the room

If you sit all these posts in a room together, a few things start to buzz:

• Skepticism about tech hype. Multiple voices ask if AI’s massive investment is matched by real economic returns. Some fear overinvestment. Some call for better governance. They disagree on tone and remedy, but they agree that bright lights do not equal steady incomes.

• Measurement matters. The GDP vs GDI gap and the consumer sentiment low show how fragile our reading of the economy can be. Numbers are helpful, but they are not gospel.

• Distribution and human cost keep returning. China’s mass displacement, the housing filtering story, predatory lending — these are different具体s but they carry the same moral note. Growth is not a neutral force. Who gets the benefits matters.

• Policy and small actors still matter. Balsa Research, university budgets from 1944, gift-giving at work — small things change big things, often in ways that get ignored in the loud debates.

• History informs the present. Books on Bell Labs and Portugal’s long stagnation, or the Continental currency argument, are not just nostalgia. They are warnings. They show how institutions, promises, and expectations shape outcomes.

Final, slightly messy thought

I’d say this week in economics felt a little like walking through a street market. There are shiny stalls with bragging signs about AI and growth. There are quieter tables selling sensible things — housing data, archival memos, funding plans. There are some folks loudly hawking doom, and others calmly pointing out measurement errors. You get both the showmanship and the small craftwork. It’s messy. It’s human. It’s often useful.

If one were to pick a mood for the week, it would be cautious curiosity. These writers aren’t all in agreement. They argue. They repeat the same warnings. They circle around the same problems. But that’s how you learn. Read them. The details are where the heat is. The summaries are only the smell of the cooking. If you like arguments, if you like odd facts, or if you like slow, careful dismantling of a comfortable story, click through the pieces. They reward that kind of patience. There are links and notes and sometimes an archival gem or two waiting for someone who will sit and read.