Russian society is becoming increasingly militaristic since the war in Ukraine began. These soldiers marched in Moscow in a parade in 2023.

The Trump administration has taken government equity stakes in more than three dozen private companies.

(Image credit: Win McNamee)

Deep Blue took down Garry Kasparov at chess in 1997, AlphaGo beat Lee Sedol at Go in 2016, and poker bots have been beating professionals for years. But one classic game called Stratego held out. Even DeepMind, with its exceptional budget, couldn't build a machine that reliably beat the best human players.

Now, a team of researchers from Carnegie Mellon, MIT, New York University, and Stanford University has done it. Their AI, called Ataraxos, beat Pim Niemeijer, arguably the best Stratego player of all time, 15 games to one, with four draws. And it took just 16 GPUs and a few thousand dollars to train it.

Hidden armies

In Stratego, each player gets 40 pieces representing military ranks, from a marshal down to a spy, plus bombs and a flag. You win by capturing the opponent's flag. Your opponent knows where your pieces are, but not what they are. Identities are revealed only when two pieces collide in battle—the weaker one is removed, and the identity of the winner is revealed. That makes Stratego an imperfect-information game, just like poker, which computers cracked years ago. “There's something super distinctive about Stratego, which is that it is a massive amount of hidden information that unfolds over a very long time scale,” said Eugene Vinitsky, a researcher at NYU and co-author of the study.

In some forms of poker, the hidden information is tiny. In Texas Hold’em, “You only have two hidden cards,” said Gabriele Farina, an MIT computer scientist and another co-author. That leaves just 1,326 possible hands, few enough for a machine to weigh them all. “In Stratego, there's 40 pieces on the board that could be in any order,” Farina said. That's more than a decillion possible setups. Then there's the game's length.

“In chess, usually the game lasts 40 moves, but in Stratego, a game can easily last 2,000 moves,” Farina said. On top of that, Stratego is a game of bluffing. Sometimes you move a weak piece as if it were a marshal, just to scare the opponent off. When players bluff too often, their threats mean nothing; when they never bluff, they become predictable. That balancing act, the team explains, is what stumped earlier AIs like DeepMind’s DeepNash, introduced in 2022.

Learning to guess

Just like DeepNash, Ataraxos learned by playing against itself—163 million games in total. In these self-play sessions, moves that led to wins were reinforced and played more often in future matches, while moves that led to losses were played less, which was the same simple training idea. The difference was in how much Ataraxos adjusted after each game, because hidden information tends to send self-play learning algorithms around in circles. The team addressed this by making big, bold changes in strategy early in training and small, careful ones later.

The even bigger innovation was something DeepNash never had: thinking ahead before each move. AIs like AlphaGo refine their general strategy with a search just before acting. DeepMind couldn't make that work in Stratego because the search space was too large, leaving it an open question whether it was worth trying.

“This is one of the things that we did figure out how to do,” Farina said. The solution was a second neural network, a belief model, trained to guess the opponent's hidden pieces based on how they had been moving. This way, instead of iterating through every possible arrangement, Ataraxos samples plausible ones, plays out candidate moves in each, and picks based on how they turned out.

And it shows in its playstyle.

Calm and unbothered

The name Ataraxos comes from the ancient Greek word for a state of calm. “It means somebody that's calm and unbothered,” Farina explained. He suggests the structure of the AI and its lack of human emotions ensure it doesn’t react impulsively, “even in situations where a human would be losing their mind.” While the human might try big gambles to come back from a significant deficit, Ataraxos would work its way back into the game slowly and methodically.

The strategy it developed also avoids drawing attention to any problems it faces. When Ataraxos estimates its opponent has no reason to suspect a weak spot, it leaves that spot alone, even if it might look like a disaster waiting to happen to anyone who can see both sides of the board.

“For humans, it's very hard when you know a secret to make decisions ignoring the fact that you know that secret,” Farina said. “For machines, it's easy.” This, the team says, leads machines to make moves a human would only do while bluffing—and follow up on them much better than humans. “We would watch the bot ‘bluff’ its way back from like a two percent victory probability, very, very casually,” Vinitsky added.

Niemeijer, the human player Ataraxos pulled these miraculous comebacks against, has four world championships and more than 600 weeks as the world's top-ranked player.

The match

Over three weeks, Niemeijer played 20 online games against Ataraxos, earning $100 for each win. He knew the AI would not adapt to him, which gave him time to hunt for weaknesses. He managed to win just once.

That loss, researchers claim, wasn't really a flaw. Playing Stratego well requires randomizing the arrangement of your pieces, so luck always plays a role. “Even a perfect strategy, sometimes it will just lose,” Farina said.

The human champion apparently got lucky, but it went both ways. “Sometimes we got lucky,” Vinitsky admitted.

At the 2025 Stratego World Championship, attendees who challenged Ataraxos fared even worse. The AI won 38 of 40 games. In the process, it also changed how people play. “I think this bot has kind of skewed the metagame a little bit,” Farina said.

Players were surprised, for example, by how often it tucked its flag into a corner behind just two bombs, a rarely played setup.

But Ataraxos's best trick was arguably its price tag.

The price to pay

DeepNash was trained for two to three months on 1,024 of Google's specialized chips, a run the Ataraxos team estimates would cost $3 million to $4.5 million at 2025 prices. Ataraxos, in contrast, needed 16 GPUs for a week, plus an additional four GPUs for four days to train the belief model.

Farina and lead author Samuel Sokota achieved this efficiency by writing a simulator that runs millions of moves per second on graphics cards. “At the scale that we are in academia, we don't really have access to an entire field of GPUs,” Farina said.

The algorithm also learned far faster—it played about 34 times fewer games than DeepNash, and still ended up much stronger.

The Ataraxos architecture also worked in learning other games. The same approach beat three world champions at Barrage Stratego, a faster eight-piece variant of Stratego, mastered the cooperative card game Hanabi, and beat the best bots at the Chinese card game dou dizhu. But the team has its sights set on scenarios far more complex than board or card games.

Beyond the board

Board games have fixed rules and clear winners, while real-world problems like negotiations, financial markets, or military conflicts usually don't. But the Ataraxos team argues the gap is smaller than it looks, since tackling any real problem starts with building a simplified model of it.

“War gaming is a common thing that people do,” Vinitsky said. “You can use the techniques that were derived here to play it forward and see how a strong opponent might respond to what you do.”

Farina and his colleagues are now interested in making their AI more understandable, since Ataraxos, in its current state, can't explain why it makes the moves it makes. “We work on machines that produce strong but also interpretable and explainable strategies. I think we're not quite there yet," Farina said.

Nature, 2026. DOI: 10.1038/s41586-026-11036-y

Read full article

Comments

Researchers looked back at the baby-boom era and found that federal mortgage programs played a part in the country’s population upswing after World War II.

We are mere days away from the launch of Googlebooks, the long-awaited Android laptops that Google hopes will compete with macOS and Windows. But what of Chromebooks? There have been hints that Google plans to sunset its web-first ChromeOS platform, and a new support page lays out some of the specifics.

The support page, spotted by 9to5Google, aims to explain the role of Googlebooks to Chrome Enterprise and Education customers. The page starts off by noting that Googlebook OS, which is Android even if Google doesn't like to say that, will offer an enhanced experience with deeper system capabilities and lots of AI.

Like Chromebooks, Googlebooks are guaranteed 10 years of software support. However, the latest Chromebook models won't get that in the traditional way. Google's support page says that it plans to continue ChromeOS support through "mid-2034," so there are already devices with support windows beyond that.

This 2034 timeline was previously mentioned in a court filing, so this is not something Google has just decided to do. However, putting it on a support page is different than including it on page 18 of a 30-page legal document. Even if Googlebooks are aimed mainly at consumers right now, Google plans to make it a replacement for the Chromebooks used by its business and education customers.

Rather than get into a Windows XP situation in which it had to keep extending update support, the support page says that Google intends to upgrade newer Chromebooks to the Googlebook OS. It doesn't have specifics on the timeline or which models will be included, but the company does note that, in many cases, this will be a "direct migration."

Googlebooks are launching at much higher prices than Chromebooks. Credit: HP

It's understandable that Google doesn't want to make this more clear—Chromebooks may not have caught on for consumers, but they're big in schools and business. Googlebooks won't have comparable management tools for a year or more, according to the support page. If Google is more open about ditching ChromeOS, its Chromebook business could take a dive before Googlebooks are even theoretically in a position to take over.

In its public statements, Google representatives have only said that Chromebooks will continue to exist. However, this guidance certainly makes it sounds like the platform's days are numbered. That may present issues for organizations like schools that became accustomed to getting capable, easily managed PCs for dirt cheap. As Google has explained on numerous occasions, Googlebooks are intended to be a premium product.

Maybe Google's partners will explore budget-friendly Googlebooks eventually, but the initial lineup starts at $899. It's impossible to spend that kind of money on a Chromebook today.

Read full article

Comments

The cost of healthcare in general is a debilitating, pre-existing condition for Americans. But the high prices of prescription drugs usually stand out as a pain point. While there are many insidious reasons why Americans pay more—often far more—for their medicines than people in peer countries, exploitation of the US patent system is an obvious one.

A study published Monday in JAMA highlights just how much patent exploitation has grown since 1990. In that time, researchers found that the number of patents on small-molecule drugs has more than tripled, going from an average of 2.1 patents per drug approved in 1990 to 6.9 for those approved in 2019.

Most of the growth was in "nonprimary" patents—patents that generally aren't related to a drug's active ingredient, but are instead for things like minor tweaks to a drug's nonactive ingredients, updates to the way the drug is used, or the design of specialty delivery devices, such as auto-injectors. Together, those extra patents on an individual drug can create what's called a "patent thicket," which delays the release of affordable generics on the market, keeping drug prices higher for longer without actual clinical advancements.

The study—led by S. Sean Tu, an expert in drug and patent law at the University of Alabama—found that the increase in patents per drug extended the time in which a drug was patented from an average of two years in 1990 to an average of 6.1 years in 2019.

"Because patent protection typically determines how long brand-name firms can charge monopoly prices, the rapid growth of nonprimary patents may contribute to limited price competition that benefits patients and the health care system by helping avoid unnecessary spending," Tu and colleagues write.

Patent overgrowth

For the study, Tu and colleagues used publicly available data to look at small-molecule drugs approved by the Food and Drug Administration and the patents filed on those drugs. (Other types of approved drugs, like biologics, are not systematically listed by the FDA in a publicly available database.) The researchers categorized the types of patents associated with each drug and how they affected the term in which the drug remained under patent. The researchers focused on drugs granted FDA approval between 1990 to 2019, giving a five-year follow-up period for patents. Still, this likely underestimates the current sizes of patent thickets, as patent activity is now extending up to nine years after FDA approval, the authors write.

Between 1990 and 2019, Americans saw their spending on prescription drugs soar. According to a Peterson-KFF analysis, the per capita, inflation-adjusted spending on prescription drugs in the US was $291 in 1990. By 2019, spending had risen to $1,084. Earlier this year, an analysis by the Commonwealth Fund found that Americans spent nearly twice as much on prescription drugs as the average spending of other high-income countries.

During the study period, "primary patents" on drugs—those usually related to the active ingredient—changed little. But nonprimary patents more than tripled. Overall, the FDA approved 1,981 small-molecule drugs between 1990 and 2019. For those 1,981 drugs, there were 10,940 patents total. Nonprimary patents made up 84 percent of them.

A key limitation of the study is that it didn't look directly at delays to the availability of generic drugs, which are likely linked to patent thicket growth. Tu and colleagues note that drug makers often argue that additional patents don't delay generic drugs. But the examples often used in those arguments are drugs approved before 2010, when patent thicket growth was lower than it is today.

The authors call for various reforms to cut back patent thickets, including more scrutiny by the US Patent and Trademark Office (USPTO), as well as laws that could limit minor add-on patents. They also suggest allowing courts to force disgorgement of drug companies found to have abused patents.

Without regulatory reforms soon, this boom in patent thickets "will likely entrench extended exclusivity periods that are disconnected from meaningful therapeutic innovation and continue to complicate and delay generic entry," Tu and colleagues write.

Read full article

Comments