1. “This happened sooner than I expected, and the story is more dramatic than I expected,” therefore I am more worried than before.

2. “This happened, and the inferior Chinese cyber-defense seems to have performed just fine,” therefore I am less worried than before.

3. “This happened, and as far as we can tell, absolutely no one was harmed,” therefore I am less worried than before.

Obviously the net impact, from those bare hypotheses, is indeterminate.  And yet few people seem to be paying much heed to #2 or #3.  Joshua Saxe from the cyber world has some relevant observations.  And perhaps other updates are needed as well.

The post The optimal Bayesian update? appeared first on Marginal REVOLUTION.

      

Related Stories

• An OpenAI Model Escaped Its Sandbox and Hacked Hugging Face
• Words of wisdom on Chinese AI and our responses
• Solve for the equilibrium

 

2. Christian Kracht, Eurotrash, A Novel.  I was put off by the title, but it turns out this is a fun and high-level short Swiss novel about driving around Switzerland with your semi-crazy eighty-year-old mother on a road trip.  It helps to have some knowledge of both Switzerland and broader German-language literature.

3. Walter Kempowski, Alles Umsonst.  From 2006, could this be the best German-language novel since Sebald?  It is about the pending doom from the Russian army approaching on East Prussia in 1945, and how the different characters deal with that, set on a German estate.  Applicable to many other real world situations as well.  There is an English-language translation, I am not sure how good it is.  In any case an important and very good work of fiction.

4. Lars Behrisch, Democracy’s Double Helix: Participation, Equality and Revolution in Early Modern Europe.  A good book about how the roots of semi-democratic decision-making are found in 16th century Europe, and stemmed from new needs to assemble various military and fiscal coalitions.  Looks at the problem more broadly, in geographic terms, than most comparable studies.

5. Katie Kitamura, Audition.  A fun short novel, full of mystery and suspense, good for those who like puzzles in their writing.

The post What I’ve been reading appeared first on Marginal REVOLUTION.

      

Related Stories

• My excellent Conversation with Andrew Graham-Dixon
• Markets in everything?
• What should I ask Luis Garicano?

 

2. Anthropic chief economist on AI and unemployment.

3. Whales use different vowels when ships are around.

4. Oliver Kim on McNamara.

5. Large language models can predict the results of social science experiments.

6. Alexander Salter Substack on space economics.

7. New Google data and study on AI and the economy.

8. Does it matter if you don’t like the characters?

The post Thursday assorted links appeared first on Marginal REVOLUTION.

       

The post Alex Tabarrok on the Economic Analysis of Crime appeared first on Marginal REVOLUTION.

      

Related Stories

• Building luxury homes is good for the poor
• What should I ask Luis Garicano?
• What should I ask Gita Gopinath?

 

Gita Gopinath…is an Indian-American economist who is currently serving as the Gregory and Ania Coffey professor of Economics at Harvard University and previously served as the first deputy managing director of the International Monetary Fund (IMF), from 21 January 2022 to 31 August 2025. Before that she also served as chief economist of the IMF between 2019 and 2022.

Here is Gita on scholar.google.com, she is an expert in international finance and exchange rates, and also international capital flows, among other topics.  Here is Gita on Twitter.  So what should I ask her?

The post What should I ask Gita Gopinath? appeared first on Marginal REVOLUTION.

       

Comments

• In reply to rsm. If you follow the policies you suggest, then ... by M
• In reply to M. What is going to replace the dollar as the ... by rsm
• In reply to rsm. Whether or not cross-border borrowing is ... by M
• In reply to M. Do you totally dismiss the very serious ... by rsm
• If the US wishes to rebalance its trade deficit, what should it ... by doc b
• Plus 10 more...

Related Stories

• Alex Tabarrok on the Economic Analysis of Crime
• The economic effects of GLP-1s
• Building luxury homes is good for the poor

 

Tyler and Andrew discuss whether it was inevitable we’d rediscover Vermeer, how that vanishingly rare sect left its fingerprints all over his life, why the Met has misread its own Allegory of the Catholic Faith, whether Vermeer painted for money or pointedly refused to, where the Gardner’s stolen Concert might be, why Dutch music never blossomed as much as Dutch painting did, how the Church of England rivaled the Cultural Revolution in wiping out British art, whether you can still spot an English painting on sight, the love that saturates late Rembrandt and the mystery of his soaring print prices, the nail on the wall that proves two Vermeer paintings are a pair, why the French are to blame for George Stubbs’ lack of status, whether we can still love Malevich, why Andrews calls recent Richter “almost like printing money,” why female artists and antique textiles remain absurdly cheap, why nobody builds beautiful neighborhoods any longer, and much more.

Excerpt:

COWEN: In what sense was Vermeer a liberal?

GRAHAM-DIXON: Well, the main discovery of my book is that Vermeer was among the very first pioneers of what we now call the liberal tradition.

COWEN: What we now call the Netherlands, then the Dutch Republic.

GRAHAM-DIXON: In the Dutch Republic, and the Dutch contribution to the Enlightenment, which is the origins of the liberal tradition, has been very much forgotten. That’s absolutely at the heart of my book, is an attempt to remember these people, to bring them back into the place in history that they deserve. I’m not only talking about Vermeer. I’m talking about his friends, his patrons, because that’s the discovery of the book, is that he and his friends were a remarkable group of people, and they have been completely forgotten, and what they believed has been largely forgotten too. We need to remember it now, probably more than ever.

COWEN: This was also a religious movement.

GRAHAM-DIXON: Yes.

COWEN: Doctrinally, how would they have been different from, say, Protestants in England? The Collegiants, the Remonstrants?

GRAHAM-DIXON: Yes. Vermeer’s patrons, it emerges, and Vermeer himself, were part of a Protestant sect in Holland called the Remonstrants. They had a more extreme manifestation called the Collegiants, and they were unique among all Christian denominations of that time in being utterly opposed to division, hostility, enmity. The only thing they wanted was to bring all Christians, indeed all people—they included Jewish people and Muslim people—they wanted to bring everyone together within a faith that only really cleaved to the essentials of what Jesus Christ said, particularly in the Sermon on the Mount.

They said, “If you actually follow Jesus properly, you can never make war. You can never persecute someone who differs from your opinion. You can never pick on somebody because they believe something different.” They were very, very tolerationist. They formed the first pacifist movement in European history. They lived in a time of appalling warfare, the Thirty Years’ War, probably the worst war in the history of the modern West. Fifteen million out of 20 million German people died in the course of 30 years, and the 5 million who were left, all the women had been violated, and all the men only have one arm or one leg. It was truly atrocious.

They’re responding to real traumatic historical events. They’re responding to what is going on around the corner from where they live. They come up with this very, very beautiful approach to life, full of optimism, full of idealism. For about 20 years, the Dutch Republic actually lives by these codes of belief to a great extent. It’s the only tolerant country in Europe.

And this:

COWEN: What makes George Stubbs such an underrated painter?

GRAHAM-DIXON: I think he’s been underrated forever because of his subject matter. I blame the French. The French in the 17th century invented a system for ranking works of art by their subject matter. Up at the top, you and I, we could put on some armor and confront each other with swords, and Poussin would paint us. That would be a history painting done from the life. That would be at the pinnacle. Then would be a painting of an event, maybe a meeting between great men. Then, below that, there would be a portrait.

You’d keep going down, and eventually you’d get to paintings of still lifes, like flowers and fruit, or paintings of animals. These were the lowest works of art because they featured the basest things. I think that Stubbs, to a certain extent, was the victim of that.

COWEN: Are we now able to see them properly?

GRAHAM-DIXON: Yes. It’s been no problem since Stubbs and Constable, more than anyone else, slightly controversial to say, but I think they paved the way for modernism because they showed, not necessarily deliberately, but they showed that anything could be painted, anything at all, like a piece of mud or a piece of a river could be painted in such a way as to touch on the very highest thoughts and ideas and beliefs and feelings that subject matter is completely, in a sense, irrelevant. Cézanne picked up on that when he said, “I want to stun Paris with an apple. I’ll paint an apple, and I’ll paint it with such astonishing concentration that you’ll see it as an epistemological challenge to all of your philosophy.”

I think that French art gets that from British art because British art has to be like that because that’s all the aristocrats are going to commission. They look down on people like Stubbs, in a sense. If I’m an aristocrat in the 18th century and I want a really important picture, I’ll go to Italy and I’ll buy one, thank you very much, and I’ll buy my history painting from Titian. You, sir, Stubbs, you just paint my horse. It’s my racehorse, and I’m very fond of him. Paint him well and don’t scare him.

Recommended, interesting throughout.  And here is Andrew’s very interesting new book on Vermeer.

The post My excellent Conversation with Andrew Graham-Dixon appeared first on Marginal REVOLUTION.

       

Comments

• In reply to Asher. When someone pulls out numbers like that I ... by Etalon d'Silomar
• I once met a painter named Vermeer, He botched a painting with ... by Sam Skiffle
• They wanted to bring everyone together in a single faith – ... by Asher
• I have never been more interested and less convinced by a book ... by JJ
• In reply to Student. Excellent comment. Thank you by Hoosier
• Plus 10 more...

Related Stories

• Markets in everything?
• *The Odyssey*, the movie (no real spoilers)
• My excellent Conversation with Chase Koch

 

Something that the Ukraine and Iran wars have taught me is that for a lot of countries, including very big countries, there are a small number of buildings and infrastructure components that are required for their economy to function.

And in countries that aren’t protected by oceans, like the US, drones completely change everything.

You can just make a list of oil refineries and production plants and the infrastructure around their main exports, and you can just attack them and destroy them and take them off the list.

4 years ago I would have thought there are too many of these components and they’re too easy to recreate, but both of these collisions have taught me that there are far fewer and that many of them will take years, if not over a decade, to rebuild.

It’s very strange to me that drones can now effectively harm the economy of a country.

That is from Daniel Miessler.  While I am glad Russia is on the receiving end of this right now, this is arguably our biggest pending problem, bigger than what people typically refer to as AI risk.

The post Solve for the equilibrium appeared first on Marginal REVOLUTION.

       

Comments

• In reply to Mactoul. As if the Chinese have ever hesitated to ... by Anti-Gnostic
• In reply to Benjamin Cole. I bet it would be a cakewalk, and ... by Anti-Gnostic
• While I am glad Russia is on the receiving end of this right ... by Niko Davor
• I think the biggest risk of AI is that it will defeat ... by JW
• In reply to superdestroyer. Sarcasm seems to have garnered ... by y81
• Plus 10 more...

Related Stories

• An OpenAI Model Escaped Its Sandbox and Hacked Hugging Face
• Alec Stapp on the new Science report from Michael Kratsios
• Betting markets in everything

 

Last year, the US exported $171bn of agricultural goods according to the Department of Agriculture, just $2bn more than the export total claimed by Brazil, which many analysts say has benefited from trade disruption set in train by President Donald Trump’s tariffs.

With US exports falling and Brazil’s rising — they increased 6 per cent in the first half of this year to hit a new record of $87bn — 2026 could prove the year in which America loses its agricultural ascendancy.

Here is more from Susannah Savage and Michael Pooler at the Financial Times.

The post Brazil fact of the day appeared first on Marginal REVOLUTION.

       

2. Why evolvable AI is not yet a Darwinian threat.

3. “This paper argues that societies with greater historical exposure to natural disasters are more likely to develop long-term-oriented norms that emphasize preparation, saving, and future well-being.”

4. And from Agustin: Questions that Are Rarely Asked.

5. MacroMusings podcast with Basil Halperin.

6. Kudos to Laura Loomer.

The post Wednesday assorted links appeared first on Marginal REVOLUTION.

       

Earlier this week, we detected and responded to an intrusion into part of our production infrastructure. This one was different from anything we had handled before in one important way: it was driven, end to end, by an autonomous AI agent system – and we detected and dissected it largely with AI of our own.

The timeline here is important so keep in mind that the attack was detected probably around Monday July 13 or Tuesday July 14. Note further:

A malicious dataset abused two code-execution paths in our dataset processing (a remote-code dataset loader and a template-injection in a dataset configuration) to run code on a processing worker. From there, the actor escalated to node-level access, harvested cloud and cluster credentials, and moved laterally into several internal clusters over a weekend.

So this means the breach started earlier, perhaps Sat July 11 or even a bit earlier. The attack was not just one thing but multi-pronged including decoys:

To understand what a swarm of tens of thousands of automated actions did, we ran LLM-driven analysis agents over the full attacker action log, comprised of more than 17,000 recorded events. This allowed us to reconstruct the timeline, extract indicators of compromise, map the credentials touched, and separate genuine impact from decoy activity. Thanks to this approach, we were able to do in hours what would usually take days, and match the adversary’s speed.

Hugging Face tried to respond but they were initially held back by the fact that the most advanced models at their disposal treated defense as attack and refused to work with Hugging Face. HF thus had to turn to open models–specifically GLM 5.2, a Chinese open-weight model run on their own infrastructure. Note the irony: HF had to use a Chinese model to defend themselves because the American models refused to help. The irony gets deeper.

At the time, I assumed this was a state based attack–maybe China or Russia testing out defenses. Indeed, HF “reported this incident to law enforcement agencies.”

But yesterday (Tuesday July 21), we learned who the real attackers were. The attackers were OpenAI models–GPT-5.6 Sol and an even more capable pre-release model. OpenAI had taken some off the guardrails off the models but they felt safe because they were testing the models in a highly secured sandbox.

The models, however, broke out of the sandbox exploiting a never before seen fault. They then gained access to the internet and from there broke into Hugging Face–all in an effort to steal the answers to the very test they had been asked to solve.

While operating in our sandboxed testing environment, our models spent a substantial amount of inference compute finding a way to obtain open Internet access, in pursuit of solving the evaluation problem. To gain access, the models identified and exploited a zero-day vulnerability (which we’ve now responsibly disclosed to the vendor) in the package registry cache proxy. With this access, our models performed a series of privilege escalation and lateral movement actions in our research testing environment until the models reached a node with Internet access.

After gaining Internet access, the models inferred that Hugging Face potentially hosted models, datasets and solutions for ExploitGym. Knowing this, the model searched for and successfully found ways to gain access to secret information that it could use to cheat the evaluation. In one example, the model chained together multiple attack vectors, including using stolen credentials and zero-day vulnerabilities to find a remote code execution path on the Hugging Face servers. OpenAI’s security team discovered this anomalous activity internally.

Now go back to the timeline. As I read it, the models had escaped the sandbox by around Sat. July 11, possibly earlier, and were detected by Hugging Face on Monday July 13 or Tuesday July 14. HF alerted legal authorities around that time–so Hugging Face clearly had no idea who was attacking them. OpenAI says its security team discovered the anomalous activity internally but has not said when. Attribution was not disclosed until Tuesday July 21, so it may well be that the models were loose for about a week before OpenAI realized that they were the ones attacking Hugging Face. And whatever OpenAI knew and when, nobody warned Hugging Face while the attack was underway–they were left to fight off a frontier lab’s models on their own.

This is a very serious breach.

Addendum: People have been wondering why I signed the We Must Act Now statement. This is why.

I am optimistic about the economic impacts of AI, but I also have no doubt that this is a very powerful technology–an Alien Intelligence–quite unlike any we have dealt with before. This incident was, in fact, error-correcting–the attack was detected, contained, and disclosed. But note who paid for OpenAI’s experiment: Hugging Face. When a lab’s test imposes costs on third parties, that is a classic externality, and taking externalities seriously is not dirigisme, it’s law and economics. And that’s the easy case. What do we do when a Chinese model breaks out of its less secure lab? Hmmm…

I remain optimistic. Learning by doing is how I want us to proceed but we should not kid ourselves: this is a global issue and we must build with safety in mind.

The post An OpenAI Model Escaped Its Sandbox and Hacked Hugging Face appeared first on Marginal REVOLUTION.

       

Comments

• One detail that makes this worse than the summary suggests: the ... by Ryan Baker
• TL/DR: model was trained on hacking tools and methods, ... by J Greely
• In reply to Anonymous. Because the whole ... by European Troll
• In reply to Francisco. That's the entire point: the story is ... by Florida Man
• In reply to Florida Man. The use of the word 'sandbox' by no ... by Francisco
• Plus 10 more...

Related Stories

• The optimal Bayesian update?
• Words of wisdom on Chinese AI and our responses
• How does social media change people?

 

Major new report from the White House Office of Science and Technology Policy. Five things in the report I really liked:

1. Proposes metascience units as a way to advance experimentation in science agencies. IFP recommended OSTP take this forward in our response to a 2025 RFI.

2. Advocates for a portfolio approach to federal science funding. Right now, basic research funding largely goes to incremental, project-based grants. While important, they can’t be the only mechanism we use to fund science.

3. Recognizes the need to launch new institutions. It specifically highlights X-Labs as an experiment with independent labs that can take on ambitious challenges. This is a bipartisan idea whose time has come.

4. Focuses on a mix of innovation funding mechanisms, including fast grants, prizes & challenges, and advance market commitments. We described and contextualized these ideas in the Atlas of Innovation, which can help policymakers design and implement these approaches.

5. Seeks to reduce burden for American scientists, who face mountains of paperwork. Scientists should spend more time doing science and less time writing/reporting on grant proposals and working to meet regulatory requirements.

Here is the link, here is the report itself, by Michael Kratsios, Science a New Golden Age.   Overall, less money will be given to universities and more will be spent on AI-assisted science.  Here are further observations from Seth Bannon.

The post Alec Stapp on the new Science report from Michael Kratsios appeared first on Marginal REVOLUTION.

      

Related Stories

• Is America ruled by an oligarchy?
• Words of wisdom on Chinese AI and our responses
• What should I ask Luis Garicano?

 

Alan and Karen Miller’s first dance was a win-or-lose moment for many of their wedding guests.

As the couple glided across the floor during their reception in New Rochelle, N.Y., last year, attendees kept track of the chosen artist and the length of the dance. Earlier in the night, they had noted their guesses on prop bet sheets.

Proposition, or prop, bets are usually tied to sporting events. Users place wagers on specific details that aren’t necessarily correlated with a game’s outcome, like points scored by certain players. Now, these bets have entered the world of weddings.

More couples have been integrating betting into their celebrations in recent years through digital apps or printed cards. While the Millers, who live in New York City, aren’t avid sports bettors or gamblers themselves, they wanted to make their June 2025 celebration more interactive and distinctive for their 175 guests.

“I’ve been to many New York and New Jersey weddings, and sometimes I feel like for cocktail hour, it kind of has a little bit of a lull, even though there’s a lot of good food,” said Karen, 38. “We wanted to make sure our guests had an activity that a lot of people could partake in.”

She bought a prop bet template from Etsy and customized it on the graphic design platform Canva. Among the questions on the couple’s prop bet sheet: Who will give the longest toast? Will the couple cry during the speeches? How many outfits will the bride wear on the wedding day?

…“If the groom’s known to be an emotional man, it’s fun to bet on how early in the day he’s going to cry,” Drachenberg said. “I think it engages them a bit more in the day.”

Here is more from Ellen O’Brien at the NYT.  Supposedly more than 25,000 couples have attempted some version of this on the app.

Via the excellent Samir Varma.

The post Betting markets in everything appeared first on Marginal REVOLUTION.

      

Related Stories

• Solve for the equilibrium
• An OpenAI Model Escaped Its Sandbox and Hacked Hugging Face
• Alec Stapp on the new Science report from Michael Kratsios

 

2. Are LLMs obsessed with Japan?

3. Princeton University art museum gets landmark gift of Haitian paintings.

4. “Sight unseen” — the revenge of asymmetric information.  But was it a good buy?

5. “Democratic types would give up about $210,000 and Republican types about $100,000 in annual partner income to avoid a typical cross-party partner.”

6. Behind the Circe scene.

7. More on AI and mathematics.

8. Tribute to George Akerlof, who just retired.

The post Tuesday assorted links appeared first on Marginal REVOLUTION.

       

For all the praise of IMAX and 70mm’s supposed clarity, several scenes are out of focus on the actor. Pulling focus is harder with a large-format negative. The problem is compounded by Nolan’s fondness for darkness: too many scenes are dark enough to squander much of what the format offers. The sound was earth-shaking in the way we have come to expect from Nolan but there is no song.

The “woke casting” controversy is a non-issue — barely noticeable in practice. The film is obviously conservative in temperament. Helen gets some of the best lines, and Nolan’s slight disfigurement of her traditional arc is exactly right. The Circe scene is the best in the film.

Tyler is entirely wrong about Calypso. Odysseus’s seven years with her were among his most enjoyable. I have no doubt about this.

The deeper flaws are structural. Nolan loves to play with time, and the resulting flashbacks and memories ironically shortchange the odyssey itself — making the journey feel shorter and less arduous than it should. This is very much in the mode of Interstellar: a sequence of set-piece locations strung together. One planet/one monster/one scene–on to the next. But Odysseus as a character on an odyssey never quite coheres.

We are told repeatedly that Odysseus is smart but we shouldn’t need to be reminded. Odysseus is both beloved and resented by gods, a man whose men will follow him to the ends of the earth and then betray him but Mat Damon just doesn’t bring it. Things happen to him; he responds stoically. What we needed was the equivalent of Kirk defeating the Kobayashi Maru — a moment that makes the audience understand, viscerally, that this man bends the rules and contends with the gods by the sheer force of his wit and will. Of courses the Trojan horse is this but Nolan treats this as something of which Odysseus is ashamed and the other clever bits are downplayed. Damon never gets his Kobayashi Meru. The odyssey is a slog, rather than an adventure. Could have used a bit more Sinbad, a bit less Dark Knight. The dialogue, as Tyler noted, is lame. 

Not Nolan’s best film but still better than most films and for scale, ambition, and grand themes well worth the 3-hour investment.

The post The Odyssey appeared first on Marginal REVOLUTION.

      

Related Stories

• Whit Stillman’s *Metropolitan* (that was then, this is now)
• Markets in everything?
• Cross-ethnic casting

 

We estimate the causal impacts of GLP-1 treatment on labor market outcomes using linked Danish administrative data and a matched stacked difference-in-differences design. We compare patients who initiate GLP-1 treatment during the first two years of Semaglutide availability to observably similar patients who initiate four years later. We find that GLP-1 treatment reduces long-term sickness leave by 17.3%. We estimate total fiscal benefits of GLP-1 initiation of approximately 1.3–1.5% of annual labor income per employed individual. We do not detect statistically significant or economically meaningful impacts on income, labor force participation, or employment over four years.

That is from a new NBER working paper by N. Meltem Daysal, Camille JH. Fredrickson, Ida L. Kristiansen, Mircea Trandafir & Jonathan Zhang.

The post The economic effects of GLP-1s appeared first on Marginal REVOLUTION.

      

Related Stories

• The small business boom
• Alex Tabarrok on the Economic Analysis of Crime
• What should I ask Gita Gopinath?