LISTEN TO THIS ARTICLE

SANTA FE, NEW MEXICO— A new battle has broken out over Monte Verde, a coastal site in southern Chile that may provide some of the earliest evidence of humans in the Americas.

In March, a group of archaeologists claimed the site may be as young as 4200 years old —10,000 years younger than most researchers thought—challenging key evidence that people arrived in the Americas before the end of the last ice age. But the original excavator at Monte Verde, Vanderbilt University archaeologist Tom Dillehay, and loosely affiliated collaborators have launched a fiery rebuttal. At Paleoamerican Odyssey—a major international conference held here last week—they argued the March study, published in Science , contained fundamental errors in reinterpreting the site. They also described a new excavation, completed just 3 weeks ago, that they say reaffirms the early dates.

The Monte Verde work, first done in the 1970s and corroborated in the ’90s, was a major factor in overturning what was then the prevailing hypothesis about the peopling of the Americas. Known as “Clovis first,” it held that humans arrived no earlier than about 13,000 years ago, coinciding with the widespread appearance of the so-called Clovis-style projectile points in the archaeological record.

In the Science paper, skeptics of the Monte Verde dates, led by University of Wyoming archaeologist Todd Surovell, attempted to revive a version of that view. The team claimed a layer of volcanic ash known as tephra dated to between 8200 and 4200 years old can be found below the sedimentary layer that held artifacts, hearths, and the remains of a structure, meaning the site is no older than that ash layer.

Conference attendees had been whispering about possible fireworks since the meeting began. Dillehay kicked off his presentation uncontroversially, offering a staid description of how archaeological sites in South America reveal the cultural and organizational capacities of ancient peoples there. At the end of his talk, though, he switched gears to a defense of his original Monte Verde claims. He also revealed the just-completed independent excavation, in which about a dozen archaeologists from several South American countries visited Monte Verde and dug four new test pits near the original site. Evidence from these pits undermines Surovell and colleagues’ reinterpretation, Dillehay and colleagues say.

“‘Clovis first’ is dead,” Dillehay tells Science . “It’s been dead for 15 to 20 years.”

In a talk following Dillehay’s, geoarchaeologist Vance Holliday of the University of Arizona argued that the new test sites contradicted Surovell’s team’s interpretation, revealing two ash layers above the Monte Verde layer dating to 8500 and 11,000 years old. “Both tephras are clearly above the occupational layer, supporting an age older than the lowest tephra,” he told attendees.

Chilean geologist Mario Pino offered a similar rebuttal of Surovell’s claims that an older riverbank could have washed into the original site and deposited old wood debris into the Monte Verde site, throwing off radiocarbon dates and making the site look older than it is.

David Meltzer, an archaeologist at Southern Methodist University, finds the rebuttals convincing. He says Surovell incorrectly extrapolated from one spot at the site to others. “Archaeological sites are unusual places on the landscape,” he says. “You can’t assume that you can go 50 yards over and dig and find the same thing.”

Surovell tells Science the way Dillehay and company’s presentations unfolded felt unfair, though it wasn’t totally unexpected. He notes that Pino’s talk defending the original Monte Verde dating was added to the program after his March Science paper came out, but that he hadn’t been afforded an equivalent opportunity. Holliday’s talk felt like a “sneak attack,” he says, as it hadn’t been included on the program. “Did I feel ambushed? Yes. Did I expect it? Also, yes.”

Surovell says he can’t adequately evaluate the new evidence from the recent digs “until the methods, observations, and results are made available to scrutiny.” Dillehay noted that the independent committee of researchers who conducted the recent excavations is currently preparing its report for publication and most likely will submit it to Science .

Conference attendee Elizabeth Akiya Chestnut, who formerly directed the Indian Pueblo Cultural Center, says she found Dillehay’s defense convincing. “He knows his stuff. He did the excavations. How can you doubt him?”

However, Michael Rondeau, an independent archaeologist based in Sacramento, California, says the enduring debate between “Clovis first” and “pre-Clovis” camps is tiresome. “What I find interesting is why Clovis culture emerged,” he says. “Who was first in the Americas is beside the point.”

LISTEN TO THIS ARTICLE

There are glimmers of hope amid the relentless bad news about the explosive epidemic of the Ebola Bundibugyo virus in the Democratic Republic of the Congo (DRC). After months of debates and preparations, a vaccine trial, named Tumaini—Swahili for “hope”—is expected to clear the final regulatory and ethical hurdles any time, and the first doses could be administered this week. The trial will first test Ervebo, a “mismatched” vaccine developed for another Ebolavirus species named Zaire. But two other vaccines, designed specifically to protect against Bundibugyo, may be added to the study in November.

Meanwhile, after more than 9500 reported cases and more than 4600 deaths, the epidemic appears to be slowing in Ituri, its epicenter and the hardest hit province. Both the number of newly reported cases and the percentage of tests that come back positive are going down, says Jean-Jacques Muyembe-Tamfun, who heads the DRC’s National Institute of Biomedical Research: “It is not optimal, but it has improved.”

And finally, a trial testing two potential Ebola therapeutics—an antibody cocktail named MBP134 and the antiviral remdesivir—has already enrolled more than 600 patients. There may soon be enough data to see an effect, says Placide Mbala of the Africa Centres for Disease Control and Prevention, principal investigator on the trial. An effective treatment would be a big boost to the response, says epidemiologist Chikwe Ihekweazu, who heads the Health Emergencies Programme at the World Health Organization (WHO). “Every extra tool you have in the arsenal helps.”

Still, the challenges are enormous. The disease could still spread to more places. Earlier this month, Kenya announced its first Bundibugyo case, a 40-year-old Kenyan man who died in Nairobi after traveling from the DRC by way of Uganda. Four family members and four health care workers have also tested positive and are in isolation. On 20 October, Nigeria reported a case as well, a woman who had traveled to Uganda, which had 20 cases earlier this year but declared its outbreak over in July.

In the DRC, two-thirds of the people who die of Ebola still do so in their communities instead of dedicated treatment centers, creating the risk of further spread. Contact tracing is still not working well; about 40% of newly identified cases aren’t on the contact list of any known Ebola case. That’s down from about 70% early in the outbreak, says Marie Roseline Belizaire, incident manager for the outbreak at WHO’s Regional Office for Africa. “It’s progress, but it’s not enough.”

And although cases in Ituri have gone down, they have trended upward in North Kivu, the second most affected province. WHO and other groups have scrambled to move more resources there. “I think it will be very important to see where the trajectory goes,” Ihekweazu says. With dozens of new cases reported each day, there will be no quick end to the outbreak, he warns. “The one game changer would be if we have a highly efficacious vaccine that we can deploy at scale.”

That’s why the stakes in the vaccine trial are so high. Ervebo, developed more than 25 years ago, incorporates a surface protein of Ebola Zaire into a livestock virus that’s innocuous to humans. It was first used during the devastating West African Ebola Zaire epidemic of 2013–16, where it showed complete protection in a trial in 2015 . Animal studies suggest it may also work against Bundibugyo. The Tumaini trial will test that promise by giving the candidate vaccine to direct contacts of known Ebola cases, a “ring vaccination” design similar to the one used in the 2015 trial.

In that study, conducted in Guinea, entire rings of contacts were randomized to receive the Ebola vaccine either immediately or 3 weeks later. In the Tumaini trial, each person within a ring will be randomized. For now, half will receive the Ervebo vaccine on Day 1 and a typhoid vaccine 4 weeks later; for the other half, the sequence is reversed.

Randomizing individual people rather than entire rings should lead to efficacy data faster, says Ira Longini, a biostatistician at the University of Florida who was involved in designing both trials. But the typhoid vaccine given to the control group could make it a little harder to detect any benefit from the Ebola vaccine because the typhoid shot might give the immune system a boost. That could protect some people in the control group from infection with Ebola and make any benefit of the vaccine seem a little smaller than it is.

The team started to explain the setup of the trial to people in the affected area 2 weeks ago, says Jean-Pierre Van geertruyden, an epidemiologist at the University of Antwerp and one of the trial’s principal investigators. Complicating their outreach is the fact that Ervebo is already given on a “compassionate” basis to first responders in some parts of the country. It’s also used in health care workers in Ituri, in a separate trial called BRAVO that has no control group. Van geertruyden acknowledges that people might question why Ervebo is good enough to give to health care workers but needs testing on the rest of the population: “People might be a bit confused.”

Once Tumaini is up and running, the vaccine candidates specifically targeting the Ebola Bundibugyo virus will be added as they become available. Furthest along is a vaccine developed by researchers at the University of Oxford that contains a surface protein of Bundibugyo stitched into a chimpanzee adenovirus.

Data analyzed so far from ongoing phase 1 trials in the United Kingdom and Uganda suggest the vaccine can proceed to the next stage, says Oxford vaccinologist Teresa Lambe, who hopes it will “enter the trial as soon as possible.” A Moderna-made messenger RNA vaccine is closely behind. Van geertruyden believes both could be added to Tumaini in early November. It may take until February or March 2027 for efficacy data for the new vaccines to emerge, he says.

By then, the epicenter of the epidemic may have moved, and the trial would have to move with it. Organizers would have to shift their team, ensure they have housing and food, and keep the laboratory and refrigerators running in areas with poor infrastructure: “We have to follow the epidemic and that’s not easy,” Van geertruyden says. Perhaps by that point, the outbreak may be not just moving, but waning.

LISTEN TO THIS ARTICLE

Forget Ant-Man or Spider-Man. The newest marvel of biology is Spider-Ant—or at least an insect that acts like one. In a preprint published this month on bioRxiv, researchers describe a species of ant that launches a sticky, weblike substance from its stinger that glues rivals to the ground .

Ants typically attack their enemies by biting them with their pincers or injecting them with venom from their stingers. But Lordomyrma azumai , which lives in the forests of Japan, has repurposed its stinger as a web slinger.

When the scientists staged encounters between the insect and various other ant species in a simulated nest, L. azumai shot a viscous liquid at the invaders that quickly hardened into a sticky thread (as seen in the video above). Nest workers then quickly dismembered the enemies.

In the wild, these “glue gun ants”, as the team has christened them, might shoot webs because—unlike biting or injecting venom—it’s a strategy that doesn’t require the insect to get too close to a potential combatant. In any case, the approach is unique among insects. No other ant, wasp, or bee is known to use this supersticky superpower.

LISTEN TO THIS ARTICLE

A familiar pattern is beginning to emerge when artificial intelligence (AI) makes a scientific breakthrough. After an AI company announces the feat, a researcher who uses that company’s tool claims it may have ripped off their work.

On 8 September, for example, OpenAI said its agents had solved the Navier-Stokes problem, one of the biggest open challenges in math. But mathematician Tristan Buckmaster of New York University publicly suggested the company had benefited from extensive work he and a colleague, Levent Alpöge, had done using OpenAI’s coding assistant, Codex.

Two weeks later, Anthropic said its model, Claude, had found a new enzyme system. Mario Rodríguez Mestre, a computational biologist at the University of Copenhagen, soon told The New York Times his group had been studying the same systems since 2022, and had shared unpublished drafts with Claude.

OpenAI says an investigation confirmed Buckmaster’s Codex work from the prior 2 months could not “have influenced the model or the math results it produced in any way.” Anthropic says Claude was “not trained on any user transcripts.” Neither researcher says they can prove otherwise.

Still, scientists are now asking an uncomfortable question about tools many use daily. When you type an unpublished idea into a chatbot or coding agent, could it end up being used by the company that built the model—or a rival who asks the right question a year later? And what can you do to reduce the risk?

Many AI companies give researchers—who are also potential customers—free or reduced-price access to their tools, and some worry there is a quid pro quo. “People working in a university environment are a much more valuable segment of the market,” says Mohammad Hosseini, who studies the ethical implications of using AI in research at Northwestern University. “And perhaps their information is used more frequently to train these models.”

In principle, an unpublished proof or paper could become a part of a model’s next training run, he and others say. But can such a tiny drop in the vast ocean of training material still make a splash? There’s some evidence it can. For example, a preprint published in January showed chatbots could reproduce most of a Harry Potter novel nearly word for word.

But such memorization isn’t really the threat for research, says A. Feder Cooper, an AI researcher at Yale University and co-author of the study. “The data sets that these models are trained on are absolutely massive … too large from an information theoretic standpoint for the models to memorize all of it,” Cooper says. But he adds that an AI model doesn’t need to memorize a chat word for word for it to be useful. “It just needs to get enough of an understanding from the chat … to use that information.”

Indeed, a single user’s chats may not matter at all. What may matter more is how thousands of chats add up. AI companies routinely analyze conversations in bulk to see how people use their products. Anthropic, for example, has described a system called Anthropic Insights that summarizes millions of chats and sorts them into clusters. To protect privacy, it discards clusters with small numbers of users, but it doesn’t say what that threshold is.

Such analyses are the likeliest way sensitive research information could leak, says AI privacy researcher Niloofar Mireshghallah of Carnegie Mellon University. The risk isn’t copying, but a company noticing from aggregate chats that a problem is close to being solved and putting its resources behind it. “That gives them enough hints of: ‘Oh, this is solvable, let’s go do it.’”

The privacy safeguards on these analyses may also fail researchers, Mireshghallah adds. They protect users by hiding them among thousands of others with similar chats, but a small number of researchers sending many queries about the same niche problem can still stand out. “Average case privacy methods don’t protect the outliers,” Mireshghallah says.

No setting makes a chatbot fully private, but researchers say these steps make leaks much less likely:

Use an enterprise account

Universities that license ChatGPT, Claude, or Gemini typically sign contracts that bar the company from training on user data, whereas consumer accounts get no such protection. If “you’ve been using a personal account for your research, switch to your university’s enterprise account,” says Jennifer King, privacy and data policy fellow at Stanford University.

Even with an enterprise account, however, “no training” doesn’t always mean “no use.” California State University’s contract with OpenAI , shared by CalMatters, lets the company “process information derived from [customers] that has been deidentified, anonymized, and/or aggregated.” King suggests researchers check the fine print of their institution’s AI deals.

Turn off training on personal accounts

If you use a personal account, switch on the setting that stops your chats from being used for training.

Watch for exemptions

On ChatGPT, rating a response with a thumbs up or down “can make the entire associated conversation eligible for training,” even if you’ve opted out, Mireshghallah notes in a recent analysis of company policies. Anthropic keeps feedback data , too, including the related conversation, for 5 years, and may use it to train AI models. Other providers have similar policies.

Chats that AI systems flag as potentially unsafe, such as those touching on bioweapons or mental health, are another possible carve-out. For example, when Anthropic’s systems flag a chat, the company keeps the inputs and outputs for up to 2 years, and other kinds of safety-related data for up to 7 years

Practice data hygiene

Once a project is complete, King suggests downloading your chat record, “then maybe delete it.” A log of your AI use may be needed for replication or disputes, but it needn’t stay with your provider. Deleting isn’t watertight. Anthropic, for example, says deleted chats leave its servers within 30 days, but it can hold onto them longer for safety reasons. And deletion doesn’t remove data from models already trained on them.

Host your own model

For those savvy enough, the most private option is running an open-weight model on your own hardware. But that does mean “moving off of some of these platforms that host it for you, have nice tooling, and do a lot of things as part of the product,” Cooper says.

If it’s really sensitive, consider not using AI

The decision often boils down to weighing potential risks against the rewards. But even “not sending stuff [to AI] yourself will only get you so far,” Cooper says, pointing out that a peer reviewer or colleague could run a manuscript through AI even if you don’t.

As for Buckmaster and Mestre, both still use AI, but want much more control over where their work goes. “I still use AI extensively and believe it can be enormously useful,” Mestre says. “What has changed is how much control and visibility I want over the data, infrastructure, and actions of the agent.” Academics, Buckmaster says, “need to somehow push back and take some sort of ownership of our user data.”

LISTEN TO THIS ARTICLE

Biologist David Sabatini, whose prominent career unraveled 5 years ago after an investigation found he had violated sexual harassment policies, scored a major victory on 16 October when a Massachusetts jury awarded him $6 million—plus interest—in damages in his defamation case against Kristin Knouse, a former colleague at the Whitehead Institute for Biomedical Research. The physician-scientist’s claims led one research institution to fire Sabatini and precipitated his resignations from Whitehead and the Massachusetts Institute of Technology (MIT), where he was a tenured faculty member.

In a phone interview with Science , Sabatini described the hours leading up to the verdict as “the most nerve-wracking of my life.” To him, the jury’s decision represents “a clear vindication” that has brought him “a massive sense of relief.” The jury also rejected Knouse’s counterclaims of sexual harassment and retaliation.

The outcome came as a surprise to some legal experts, in part because Sabatini had earlier lost legal challenges against Whitehead and its director, Ruth Lehmann. After investigating Sabatini’s conduct, Whitehead in 2021 concluded he had breached the institute’s consensual relationships policy and its antiharassment policy. He had met Knouse while he was associate director at Whitehead and she an M.D.-Ph.D. student at Harvard University and MIT, where he was one of her instructors and served on her thesis committee. The two began a sexual relationship just before Knouse joined Whitehead as a fellow and started her own lab there.

Although Knouse said she was coerced into the relationship by Sabatini, who at the time served as director of the Whitehead Fellows Program, he maintained the relationship with Knouse was consensual. He contended Knouse only leveled harassment accusations after he broke up with her to pursue another woman. A Suffolk County jury has now ruled in his favor, finding that Knouse defamed Sabatini and interfered with his ability to work.

“That’s a pretty big finding,” says Ann Olivarius, a lawyer who specializes in cases of sexual harassment. “Obviously, the jury really went against her in a big way.” Olivarius was also surprised the jury awarded Sabatini such a hefty sum in damages. “That’s really rare. This is a standout case.”

The civil lawsuit verdict has no direct impact on the conclusions of Whitehead’s probe, although Sabatini and his supporters view it as an exoneration. And some, including billionaire Bill Ackman, who has used his large social media platform to defend Sabatini and bankrolled his recent research, called on the institute to reconsider its findings and demanded Lehmann resign .

Knouse’s attorney, Ellen Zucker, tells Science her client intends to appeal the decision. “The verdict is deeply disappointing,” she says, noting that several other women testified in court about “concerning interactions” with Sabatini. “It is really sad the jury’s verdict reflects none of this.” Requiring more than $6 million in damages from an early-career scientist, Zucker argues, “is proof positive that the jury was not well-equipped to make appropriate decisions in this case and that they did not do so.”

Sabatini rose to prominence while still a graduate student at Johns Hopkins University with his research into a key molecular signaling pathway, which regulates cell growth and metabolism and malfunctions in diseases such as cancer. He continued that work at Whitehead, later winning a prestigious position there as a Howard Hughes Medical Institute (HHMI) investigator. In 2021, after a Whitehead employee survey identified concerns in Sabatini’s lab, the institute hired an outside law firm to investigate.

In addition to finding that Sabatini breached the institute’s sexual relationships policy with Knouse, the investigation also concluded that Sabatini cultivated a sexualized lab environment and tried to influence the probe by misleading investigators and attempting to shape interviewees’ testimony. As his official employer, HHMI fired Sabatini , who also relinquished his position and lab at Whitehead. Sabatini later resigned from his professorship at MIT after three senior officials recommended his tenure be revoked.

The ramifications followed him beyond Massachusetts: In 2022, following student and faculty protests, he withdrew from consideration for a position at New York University’s Grossman School of Medicine. Sabatini’s lawyer, Lisa Arrowood, says the accusations meant the once high-flying biologist “became a complete pariah in the scientific community.”

Sabatini has since made efforts to resurrect his career. In February 2023, Ackman and an anonymous donor pledged a total of $25 million to restart his lab. Later that same year, Sabatini took a position at the Institute of Organic Chemistry and Biochemistry of the Czech Academy of Sciences. The institute recently opened a branch in the Boston area, where Sabatini now maintains a second lab space.

In 2024, however, the U.S. Department of Health and Human Services (HHS) announced it was seeking to debar him from receiving any federal research funding because of the accusation. Science has learned that in September 2025 the agency proceeded to debar him from receiving National Institutes of Health grants for a period of 1 year.

But that debarment was terminated early, in April, according to an HHS email provided to Science by Sabatini. It cited as a reason additional information the biologist had provided to the agency.

Sabatini has argued there was no power imbalance between himself and Knouse. “She had her own lab, she had her own money, she had her own students,” Arrowood says of Knouse, whom she describes as “someone who wanted revenge against her former lover.”

But the idea that Sabatini did not hold a position of power over Knouse, a much younger colleague early in her career, strikes Olivarius as “ridiculous.” “Look who he is, and look who she is. … That’s not an equal relationship.”

Sabatini also denies accusations that he created a sexualized or “toxic” environment in his lab, arguing that he comes from a family of “very accomplished women” and is proud of the many women he has trained in his lab. He contends that many individuals involved in the Whitehead probe were “manipulated and lied to” by Knouse, while also pointing to Harvey Weinstein, the #MeToo movement, and “gossip swirling” during the COVID-19 pandemic as aggravating factors.

Some in the scientific community, including MIT students and faculty, have come to Sabatini’s defense. Regina Barzilay, a computer scientist at MIT who describes herself as a friend and colleague of Sabatini, says the original Whitehead probe was “extremely one-sided” and she was “overjoyed” that Sabatini had won his case. “I find it very strange that MIT … couldn’t see the truth that the jury of seven women and six men unanimously decided.”

To Zucker, however, last week’s verdict against her client is emblematic of larger cultural issues: “At the end of the day, this trial and this jury’s verdict seems to me to be a sad marker of our times, where it is still hard for women to be believed and even harder to hold powerful men to account.”

Sabatini, according to Arrowood, plans to appeal the judge’s summary judgment that dismissed his claims against Whitehead and Lehmann, which included defamation and, for accepting Knouse’s version of events over his, gender discrimination. According to Olivarius, both Knouse and Sabatini will likely face an uphill battle. “Appeals are hard to win, always.”

LISTEN TO THIS ARTICLE

It’s a crime scene with no victim, no suspect, and no witnesses. All you have is a single severed tail embedded in sticky tree resin. And your only piece of evidence is 99 million years old.

That is the investigation Leonidas-Romanos Davranoglou, a zoologist at Tel Aviv University’s Steinhardt Museum of Natural History, and his team faced when they inspected the semitransparent sliver of fossilized resin, or amber. They had bought the amber from an online seller, thinking the 8-millimeter-long thing inside it was a twig or a piece of insect.

Then, Ph.D. student Dolev Fabrikant inspected the fossil. “Oh, it has scales! This is a lizard tail,” he told Davranoglou.

“I literally almost jumped off my chair,” Davranoglou recalls. “What has been sitting for … a few years now in a drawer actually resolves a very big enigma.”

As the team reports this month in Palaeontologia Electronica , it is likely a lizard purposefully let go of the tail —creating the oldest known direct fossil evidence of this dramatic defensive tactic.

When under attack, about two-thirds of modern-day lizards can snap off their tails at a built-in weak spot . Called caudal autotomy, the adaptation allows them to escape a predator and leave behind a wiggling, distracting decoy. “Autotomy is a dramatic but effective way of avoiding being predated,” says ecologist Bill Bateman of Curtin University, who was not involved in the study. “It seems costly losing a whole tail, but it is better than being dead.” Most lizards can regenerate the tail with cartilage instead of bones.

Researchers suspect the trick is ancient. The tuatara, a reptile native to New Zealand, can shed its tail, and its lineage goes back at least 240 million years to the common ancestors of tuataras, snakes, and lizards, so scientists assume the trait arose then. To find early examples of caudal autotomy, paleontologists have mostly tried to look for weak spots in fossilized tail vertebrae . But from Davranoglou’s point of view, these so-called fracture planes offer circumstantial evidence at best. “Finding a fracture [plane] does not equate with the ability to shed your tail.”

So, to find out what happened to their tail in amber, the team used micro-CT scanning of the 18-millimeter-wide amber piece to make detailed 3D images of the fossil (see video below). From the scales and other features, the team concluded the tail probably belonged to a slow worm—a type of legless lizard—or a wall lizard, although they couldn’t home in on the exact type.

But the tail tells a clear story: This was no accident, nor was it torn off by a predator. The fossilized tail matches very closely with museum specimens of shed tails from modern lizards. The top of the tail, for example, shows a clean cut ringed by flowerlike wedges of muscle, a characteristic of voluntarily shed tails.

“It’s very impressive because you can basically see the soft tissue,” says Juan Daza Vaca, a herpetologist at Sam Houston State University who was not involved in the new paper. “There’s evidence of the muscles … and the behavior stayed frozen in time. That’s cool.”

The CT scans also revealed more clues of a battle for survival. The tail also has a gash in the middle that didn’t come from shedding. The most likely scenario, the team says, is that the lizard got stuck in the resin and freed itself by shedding its tail. Then, some insect chewed a part of it. Another scenario is that a predator grabbed the animal, ended up with only the tail, and dropped it in the resin.

For Susan Evans, an evolutionary morphologist at University College London who was not involved in the study, the new finding is not surprising. “Yes, it’s showing that the tail has [shed],” she says. “But it’s a bit oversold.” Older reptile fossils, she says, have fracture planes that indicate the animals could shed tails, and many have fossilized cartilage at the tip of their tails—instead of bones—suggesting they regenerated them after shedding them.

Daza Vaca and Evans also raise concerns over the provenance of this amber specimen: It comes from a region in Myanmar where the trade has been tied to armed conflict since 2017 . Davranoglou says his team bought the piece in 2022 from a Chinese seller, who provided a handwritten statement—which Science has seen—that it was mined before February 2017, months before a June 2017 cutoff date paleontologists use to prevent the study of fossils dug up after a Myanmar military campaign to seize the mines.

Davranoglou stands by how the fossil was obtained, as well as the verdict itself. “It’s the first irrefutable and strongest supported claim and evidence for tail shedding,” he says. “It’s the smoking gun.” Bateman agrees: “We would predict that extinct lizards would be able to do it, but this is the first clear evidence involving soft tissue.” In addition, the new specimen could help researchers calibrate when the genes necessary to regrow a tail originated in reptiles, Davranoglou adds.

The tiny tail is a reminder, Davranoglou says, that tail-shedding “is not something that [was] lost in time,” but an adaptation still helping lizards escape peril today. “Every time I see a lizard,” he says, “I now think of this discovery.”

LISTEN TO THIS ARTICLE

After a deadly blowout on the Deepwater Horizon rig in 2010 sent oil gushing into the Gulf of Mexico, an emergency response team gambled on an unprecedented tactic. It sent robotic submersibles to the wellhead, 1500 meters below the surface, to inject chemical dispersants to break up the oil into tiny droplets and dissolve it. The hope was that by diluting the oil, the approach would minimize harm to wildlife. The move likely spared many seabirds, turtles, and shallow-water fish from oil that would otherwise have pooled at the surface. But it may have taken an unexpected toll in the deep sea.

A 13-year survey, begun the year after the spill, has discovered a steep, sustained decline in fish, crustaceans, and cephalopods in the deep waters of the gulf . The most likely cause is the oil spill, the team concludes this week in the Proceedings of the National Academy of Sciences , although it doesn’t directly blame the dispersants. “They have witnessed a major change in an ecosystem that we thought was isolated and perhaps protected,” says marine ecologist Bruce Robison of the Monterey Bay Aquarium Research Institute, who was not involved in the study. “I just don’t know whether the system itself is resilient enough to recover,” he says. “That’s a big question mark.”

The study began 10 months after the disaster, which killed 11 crew members onboard the BP oil rig and released 5 million barrels of oil. Tracey Sutton, a biological oceanographer at Nova Southeastern University, organized a survey of the deep-sea ecosystem as part of a federal investigation into damages from the spill. Starting in February 2011, the team collected data from a research vessel equipped with sonar, which allowed it to estimate the abundance of schools of fish and crustaceans. The researchers also put down a trawling net that snagged marine life at various depths at 46 locations around the gulf. Working in day and night shifts, they pulled up and counted about 1200 species to create a snapshot of the ecosystem in the aftermath of the spill.

When the team returned 4 years later, it hoped to see signs of recovery. Instead, researchers found further, drastic declines. The populations of three major groups of animals in the survey—fish, crustaceans, and cephalopods—were far below levels recorded in 2011. The picture was no better in subsequent surveys. The numbers remained stubbornly low through to 2023. Fish populations registered the largest declines, dropping by 64% compared with 2011. Because the team has no baseline data from before the spill, the declines may have been even greater than it reported.

The team didn’t measure how much oil is still present in the ecosystem. But it suspects it continued to have an effect via the food web long after the spill. Years after the disaster, study co-author Isabel Romero, a chemical oceanographer at the University of South Carolina, found polycyclic aromatic hydrocarbons—a particularly toxic component of oil—in the eggs and tissue of deep fish.

The study offers “a sobering example of the unintended consequences of human activity at sea,” says Cynthia Smith, a conservation veterinarian who leads the National Marine Mammal Foundation and was not part of the effort. Jeffrey Drazen, a marine ecologist at the University of Hawaii at Manoa, is particularly impressed by the scale of the survey, which involved collecting samples with a trawl 2000 times, often at depths of more than 1 kilometer. “This is one of the most comprehensive sampling efforts ever conducted in the mesopelagic realm.”

It also has implications for life closer to the surface. Lanternfish—small fish with large eyes and glowing patches of blue, yellow, and green—experienced some of the steepest declines. The small fish spend their days in the ocean depths and migrate upward at night to feed on zooplankton, where they are in turn eaten by tuna, swordfish, and whales. A scarcity of lanternfish, with many of the survivors possibly contaminated with hydrocarbons, could help explain a recent decline in whales in the gulf. Sperm whales are now about 30% less common there than before the spill, for example.

Steven Murawski, a marine ecologist at the University of South Florida, is surprised by the lack of recovery. He thinks an influx of animals from nearby populations, for example in the western or southern gulf, should have bolstered the depleted region. “There must be some cascading effect from the footprint of the oil spill to the larger community in general in the gulf,” he says.

Smith says the new study reinforces the need for a broader look at when and whether to use dispersants in oil spills. “Those decisions should consider all animals within the oil spill footprint, from the surface to the sea floor.”

It is also a cautionary tale about other industrial activities in the deep sea such as seabed mining, the authors say. “There’s this idea that populations in deep-sea environments are abundant and can’t be harmed,” says co-author Heather Bracken-Grissom, an evolutionary biologist at Florida International University. But the study indicates that’s not the case. “The deep sea is a much more fragile system than we thought,” Sutton says.

Do you have what it takes to win Science ’s “Dance Your Ph.D.” competition ? We want you to top this year’s winning entry from a scientist who translated physics’ piezoelectric effect into dance .

As always, we’re challenging scientists to explain their research obsession with fancy footwork, but no PowerPoint slides or jargon. It doesn’t matter whether you’re just starting your Ph.D. or you completed it decades ago; you just need imagination and the ability to keep a beat.

The annual contest, in its 19th year and again sponsored by the AI and quantum technology company SandboxAQ , features four traditional categories (physics, biology, chemistry, and social science). Winners in each category take home $750. The overall winner of those categories gets an extra $2000. You can see the full slate of past winners here and the rules and other key information here .

Beyond that, it’s simple: Go dance some science, create a video of it, and upload it to YouTube. The deadline for submissions is 29 January 2027 .

To enter the contest, submit here . We’ll announce the winner by March 2027.

LISTEN TO THIS ARTICLE

The National Institute of Justice (NIJ) has announced it plans to fund research into a possible link between gender dysphoria and violent crime, raising concern among scientists and activists that the initiative is driven by political bias and proposes work that can’t be done rigorously with available data.

The 9 October funding notice from NIJ, the Department of Justice’s research arm, says it will grant up to $1 million for studies of gender dysphoria, or the feeling of distress when a person’s gender identity doesn’t align with their sex assigned at birth. Applications should focus “on gender dysphoria, its treatment, and the perpetration of violence and violent behavior,” including the prevalence rates of violent crime among people diagnosed with gender dysphoria. The funding will support retrospective analysis of existing data sets, not the collection of new data.

“It looks to me that the administration’s already decided what it wants the science to say,” says Meghan Maury, policy director at the Transgender Law Center, a nonprofit organization. They worry that if poorly conducted, the funded research could be used to justify policy actions that harm transgender people.

NIJ did not respond to a request for comment by the time of publication.

Researchers studying gender identity say they are puzzled by the funding notice, in part because moves by President Donald Trump’s administration may make it harder to address NIJ’s proposed research questions. The administration has cut federal funding for research on gender minorities and directed federal agencies to define people according to biological sex assigned at birth. Early in his second administration, Trump moved to strip self-described gender identity and sexual orientation from national surveys.

“If you’re instructed from the beginning that you can’t actually recognize the humanity of a population or their self-determined language, then it automatically sets the research off on the wrong foot,” says Elana Redfield, federal policy director at the University of California, Los Angeles’s Williams Institute for sexual orientation and gender identity law and policy.

Further complicating the proposed research, the term “gender dysphoria” doesn’t show up in most data sets, as it is a clinical diagnosis reserved for medical records subject to privacy protections, Redfield notes. “It’s very unclear” what data NIJ intends for researchers to use, she says, or how those can be obtained ethically.

The announcement specifically mentions “post-conviction” analysis and incarcerated people, which seems to point researchers toward data sources “likely to produce estimates of elevated crime rates,” says Alexander Tsai, a psychiatrist and researcher at Harvard Medical School.

“I’m not aware of a literature on gender dysphoria and violence perpetration,” Tsai adds, but existing studies suggest transgender people are at increased risk of being victims rather than perpetrators of violent crime. The NIJ notice of funding did not plainly invite research on crimes against people with gender dysphoria, though it encourages the analysis of “offender-victim overlap” and other “contextual factors to interpret trends.”

In August, the Department of Health and Human Services released a report titled Where the Clinic Meets the Movement: Does Gender Ideology Arm Left-Wing Authoritarian Psychology? that found an association between “gender-affirming” attitudes and the justification of political violence. But many scientists have highlighted design flaws in the report and a lack of peer review , NPR reported.

NIJ has given interested researchers until 30 November to submit their research proposals. It’s not uncommon for an agency to release a funding notice with specific researchers in mind, Tsai says. “My guess though is that most serious researchers are not looking at this.”

LISTEN TO THIS ARTICLE

A lavish 1919 dinner party that ended up killing five people. An attempt to save the canning industry in the 1920s. A World War II biological weapons program and a CIA plot against Fidel Castro. A cosmetic product, best known as Botox, that erases facial wrinkles and exploded into a multibillion-dollar business.

A new analysis of more than 1000 bacterial genomes connects these disparate episodes in U.S. history in a piece of microbiological detective work. Detailed in a new paper published yesterday in the Proceedings of the National Academy of Sciences ( PNAS ), the study shows how scientists unwittingly triggered the evolution of a dangerous bacterium into one with major medical and commercial advantages.

That microbe is Clostridium botulinum , whose neurotoxin disrupts nerve communication, paralyzing muscles and causing victims to suffocate. Gram for gram, the toxin, known as BoNT, is one of the most lethal natural substances known. Under oxygen-free conditions—in preserved food, for example— C. botulinum spores can “wake up,” developing into toxin-producing bacteria.

That’s what happened with spores trapped in jarred Californian olives during the “Great Olive Poisoning of 1919,” which included a notorious dinner hosted by Detroit industrialist Murray Sales and his wife that killed two of the couple’s children, a guest, and two maids. (“How Death Came, Unbidden, to Mrs. Sales Dinner Party,” read a headline in The Philadelphia Inquirer .) Nine other people in two other states died.

The tragedies posed a major threat to the burgeoning canned food industry and triggered a new wave of microbiological research, as well as a California law that compelled the industry to treat its products at higher temperatures. Among the researchers who joined the fight was Ivan Clifford Hall, a bacteriologist at the University of California who amassed a collection of C. botulinum isolates.

One, named Hall A, became a laboratory workhorse, propagated over and over. In the process, it became “domesticated”—reshaped by human-driven evolution, like countless other plants, animals, and microbes. Hall A lost its ability to form spores, which made it easier for scientists to purify BoNT, and churned out exceptionally high levels of the toxin. Those characteristics made Hall’s strain a favorite of U.S. military researchers, who in 1942 started to use it to create biological weapons and a botulinum vaccine.

The military shared Hall A with others. In the early 1960s, CIA launched a plot —ultimately scuttled—to assassinate Cuban leader Fidel Castro by contaminating a box of his cigars with BoNT. (The United States renounced its bioweapons program in 1969 and signed the Biological Weapons Convention 3 years later.) Around the same time, a military researcher took a Hall A subculture with him when he moved to the University of Wisconsin and shared it with a Californian ophthalmologist , who began to conduct monkey experiments to see whether BoNT could relax eye muscles, in hopes of finding treatments for misaligned eyes, a condition called strabismus.

“This really is a swords into plowshares type of story,” says microbiologist Paul Keim of Northern Arizona University (NAU), the new study’s first author.

In the 1980s, U.K. microbiologist Andy Pickett and others, working with drug companies, found that tiny doses of BoNT—mere billionths of a gram—produced by descendants of Hall A could treat strabismus in humans. In 1989, the U.S. Food and Drug Administration approved the purified toxin as a treatment for both strabismus and eyelid spasms.

But something interesting happened: “Some of these patients would say, ‘Oh, and by the way, doctor, my migraine is gone, and the wrinkles on my face are gone,’” Pickett recalls. The tiny amounts of Botox, as the drug is now called, turned out to also temporarily paralyze facial muscles that create lines around the eyes, forehead, and nose.

Botox today is but one of several dozen products that contain BoNT, and legal battles around ownership and manufacturing secrets have sprung up between companies that produce them. Pickett was hired as an expert witness in several of these cases, as was Keim, who had previously worked with FBI to investigate the origin of the anthrax used in a series of attacks in 2001.

To help unravel the differences between strains in dispute, Keim suggested lawyers also recruit Richard Lenski, an evolutionary biologist at Michigan State University who in 1988 launched a now-famous experiment to follow 12 identical populations of Escherichia coli and document how they evolved over decades. (The study is still running, nearly 85,000 generations later.)

As part of their work on the case, the researchers constructed family trees of C. botulinum based on publicly available genome sequences and ones they sequenced themselves. They found that the Army Hall A strain and its commercial descendants were on an “unexpectedly long branch” of the family tree, says Lenski, a co-author on the PNAS paper, suggesting strains in that lineage mutated far more rapidly than their cousins.

“That brought to mind what we have seen in our experiments” with E. coli , Lenski explains. In those studies, some lab-propagated strains that started to mutate faster than normal had a defect in a gene involved with DNA repair. At Lenski’s suggestion, Keim and NAU microbial genomicist Jason Sahl looked for similar defects in the Hall A relatives. Sure enough, they found one, a mutation in a gene named mutS .

Keim says the mutation likely arose while Hall was working with the strain. In the wild, it might soon have disappeared, he says, but it thrived in the lab, in part because scientists favored C. botulinum cultures that displayed the mutation’s effects. MutS prevents C. botulinum from wasting energy making spores, allowing them to grow well and produce unusually large amounts of the toxin. The bacteria’s metabolic pathways are “fine-tuned now for living in the laboratory as opposed to living out in nature,” Keim says.

“This work really fills in a fantastic hole of knowledge,” says Pickett, who’s now retired and was not directly involved in the study. The research offers more than just “a fascinating glimpse” into the history of the Hall strain and human-driven selection, says Hong Lei, an ophthalmologist and neurologist who specializes in movement disorders and has used BoNT for various medical applications. “The findings underscore that biologic therapeutics originate from living organisms whose continued evolution may influence manufacturing, product consistency, and future drug development,” Lei says.

Pickett is delighted that the study also spotlights Hall, whom he sees as an unsung hero. Although some of Hall’s “incredible” collection of anaerobic organisms made it to other labs, much of it no longer exists. “No one wanted it,” Pickett says. Unraveling the genetics of the Hall A strain, he says, is “a good tribute to someone who is basically an enigma.”

LISTEN TO THIS ARTICLE

Last year, the U.S. Department of Agriculture (USDA) announced it was closing its largest research campus and radically reorganizing research at the U.S. Forest Service (USFS). The moves were controversial because they involved long-distance job transfers that threatened to uproot many scientists and derail research programs. This week, a federal judge declared the plans were likely illegal and ordered a halt until at least 11 December, when the current federal budget runs out.

Despite the decision, researchers remain wary of the Trump administration’s intentions. One who works at the campus slated for closure, the Beltsville Agricultural Research Center (BARC), and who asked to remain anonymous because he did not have permission to speak publicly, commented, “If you have an administration that does not believe in execution of the mission of an agency, you’re never out of the woods.” A USDA spokesperson told Science the agency is reviewing the decision and will submit a response to the court.

USDA announced in July 2025 that it planned to close BARC, its largest and most diverse research facility. The 2800-hectare campus runs a major bee research lab, the National Fungus Collections, decadeslong soil experiments, and other projects. USDA cited old infrastructure and a maintenance backlog as a main reason to shut down the Maryland site, near Washington, D.C. The move is also part of a larger plan to move USDA headquarters employees out of the Washington, D.C., area. At a congressional hearing, lawmakers said they had concerns, and the subsequent federal budget, passed in November 2025, prohibited USDA from spending any money on the reorganization without permission from Congress.

But in July 2026, the agency announced it was accelerating the closure and issued letters to staff about where they might be relocated, if they chose to remain employed. USDA was also proceeding with the relocation of USFS headquarters to Salt Lake City and a consolidation of research facilities, which would close many labs and require staff to move to a new location.

Labor unions and others sued that month, and U.S. District Judge Vince Chhabria took up the case in September. He quickly issued a temporary stay of the USDA reorganization, then heard legal arguments. This week he ordered the stay extended in a strongly worded decision. “Congress did not contemplate USDA using its appropriated funds to shut down offices or relocate employees,” Chhabria wrote. “To the contrary, Congress was actively working to restrict the agency’s ability to do so. … USDA has almost certainly been violating those restrictions.”

Chhabria left open a path for USDA to proceed with its plans. “If the agency wishes to move forward with its plans to close offices and relocate employees, it will soon have the opportunity to convince Congress of the merits of those plans,” he wrote. That’s because the prohibition on reorganizing, including shuttering BARC, expires with the federal appropriations on 11 December. Congress will likely extend that funding with a continuing resolution, which could omit the ban. But there are no signs that lawmakers have changed their minds about allowing USDA to proceed.

The USDA scientist says morale remains low at BARC. Uncertainty is a key factor: Before last month’s temporary stay on reorganization, there were many all-staff meetings about decommissioning BARC and relocating employees, which left many questions unanswered—“all the markings of a very rushed and poorly prepared program,” the scientist says. Another reason is attrition. Some employees have taken early retirement, and USDA continues a “de facto” hiring freeze, which makes it hard to accomplish research goals.

Ann Bartuska, a former USDA administrator now at Resources for the Future, worries the agency’s research capacity may have already been damaged through loss of staff. And she feels bad for those who have stayed and endured the ups and downs of the past year and a half. “The whipsaw that’s been happening is really terrible,” she says. “This is torture for the employees.”

Cusco and Choquehuanca, Peru— Fifteen years ago, University of North Carolina at Charlotte archaeologist Dennis Ogburn found himself looking at some impossible radiocarbon dates. They came from Chamical, a small site in southern Ecuador where Ogburn had found Inca ceramics and weapons, such as bone arrow points and a sling stone, but no domestic trash or any other sign people had lived there long term. He concluded the site was an Inca military outpost, constructed and briefly used during the expansion of the vast Andean empire out of its capital of Cusco, in what is now Peru.

According to the traditional chronology archaeologists had been using for more than half a century, the Inca conquered southern Ecuador between 1463 and 1471 C.E., during the reign of Emperor Pachacuti and under the military leadership of his son, Topa Inca. But when Ogburn radiocarbon dated a seed, wood, and charcoal from Chamical’s earliest layers, he found the outpost was likely built decades earlier, between 1432 and 1451.

That puzzling discovery echoed others from the opposite end of the empire. In Argentina, a handful of archaeologists had found Inca radiocarbon dates that were more than 50 years earlier than they were supposed to be. Their presentations had been met with skepticism and hostility, and their data were largely ignored.

But Ogburn thought the discrepancy was worth taking seriously. The traditional chronology of the Inca, a short-lived empire that left no written records of its own, was based on a Spanish colonial chronicle that claimed to capture earlier Inca oral histories. It asserted that the empire had existed for just 94 years from its founding until the Spanish conquest began in 1532 and met little resistance as it swept through the Andes. Adding just a few decades to the start of that compressed time frame, Ogburn realized, would mean the still-mysterious empire influenced the politics and cultures of South America for much longer than archaeologists understood—and that the people the Inca encountered had more time to react to, accommodate, or resist imperial expansion. In 2012, he published a paper in the journal Radiocarbon calling for Andean archaeologists to start to build a new Inca chronology from radiocarbon dates alone.

Now, thanks to statistical methods that are shrinking radiocarbon dating’s inherent uncertainties and an increasing number of reliable dates from sites all over the Inca empire, archaeologists are closer than ever to that goal. A critical mass of researchers is on board. At a meeting of Andean researchers at the Pontifical Catholic University of Peru last year, attendees unanimously voted to discard the traditional chronology and start again.

The challenge remains daunting, says Erik Marsh, an archaeologist and radiocarbon expert at Argentina’s National Scientific and Technical Research Council (CONICET). “We don’t have something that completely replaces [the traditional chronology] yet,” he says. But already, what he and others have learned confirms some of the empire’s most surprising claims about its sudden founding while also undermining its propaganda in crucial ways. The new dates—for both the Inca and the societies in their path—are transforming a top-down story of inevitable conquest into a bottom-up patchwork of negotiation, cooperation, and resistance. “The next few years in Inca research are going to be transformative,” says Julia Earle, an archaeologist at the University of Warsaw (UW).

When pioneering Andean archaeologist John Rowe developed the traditional Inca chronology in the 1940s, written records were the only reliable way to tell how old anything was. That limited dating efforts to cultures with deciphered scripts, such as ancient Egypt, or at least decoded calendars, such as the ancient Maya. The Inca, like other Andean cultures, did not use writing at all; instead, bureaucrats kept track of the empire’s inner workings with khipus, series of knotted strings that recorded numbers and likely a wide range of other information.

It’s possible an Inca historical calendar remains undeciphered in the more than 1000 khipus currently known from sites across the Andes. But Terence D’Altroy, an archaeologist at Columbia University and one of the researchers who first found early Inca dates in Argentina, thinks it’s unlikely. The Inca kept track of monthly lunar cycles, annual agricultural cycles, and the sequence of rulers, he explains, but they didn’t combine that information into linear, chronological history. There was no need, because for the Inca, the past wasn’t fixed. By interacting with—and sometimes destroying—the material of the past, such as ancient stone monoliths, the ruins of past societies, and ancestral mummies, the Inca could alter their history to better suit the politics of the present.

For example, when the last Inca Emperor Atahualpa defeated his brother Huascar in a civil war for the throne in 1532, just months before being captured by conquistador Francisco Pizarro, his troops burned the mummy of their shared grandfather, Topa Inca. Atahualpa intended to erase Topa Inca from imperial history and destroy the power of his kin group, which supported Huascar’s claim. “With a new emperor, there was always a new story,” says Dominika Sieczkowska-Jacyna, a radiocarbon expert at UW and the Silesian University of Technology.

That malleability confused colonial chroniclers who recorded Inca oral histories in the 16th century. Rowe deemed the most reliable to be Miguel Cabello de Balboa, one of the few chroniclers to assign dates to the reigns of Inca emperors. Whereas other colonial sources charted a gradual expansion of Inca power over several centuries, Cabello de Balboa, a 16th century Spanish priest, claimed the empire began abruptly in 1438, when Pachacuti deposed his father and ascended to the throne.

Cabello de Balboa’s account underpinned Rowe’s traditional chronology. It describes how Pachacuti quickly took control of the Cusco heartland and territories stretching south to Lake Titicaca, then sent three of his sons, including Topa Inca, to conquer the north, including the region around Chamical. Next, Topa Inca incorporated southern territories in Argentina, Bolivia, and Chile during his own reign, from 1471 to ’93. His successor Huayna Capac would add comparatively little land between 1493 and 1525, when power passed to Huascar, whose imperial ambitions were derailed by civil war.

“We have, of course, no way of knowing where Cabello got these dates,” Rowe wrote in a 1945 American Antiquity paper outlining his Inca chronology. But he considered the compressed timespan “reasonable” and congruent with how shallow—and, therefore, short-lived—Inca archaeological deposits tended to be.

The year after Rowe’s paper, physical chemist Willard Libby proposed directly dating wood, bone, and other organic artifacts by measuring their carbon-14, a radioactive isotope that is taken up by living things from the atmosphere and slowly decays after death. But the technique seemed unpromising for Inca artifacts. Aside from a few rare cases , radiocarbon dates are expressed as ranges that can span decades or even more than a century. If the empire only lasted 94 years, as Rowe argued, its entire existence could fall inside radiocarbon dating’s error bars.

Rowe remained a vocal defender of his chronology until his death in 2004. When D’Altroy presented his early dates from Argentina at a conference in 1997, Rowe stood up and said, as D’Altroy recalls, “You can’t trust the radiocarbon dates. Historical dates are more reliable, and I think you should think twice before publishing anything like this.”

Three decades later, archaeologists have managed to reduce—though not eliminate—radiocarbon dating’s uncertainties. Advanced statistical models constrain error ranges to as little as 20 years. Better tree-ring records have improved radiocarbon calibration, which involves matching radiocarbon dates from tree rings with the rings’ ages in calendar years. Because radiocarbon readings depend on how the atmosphere and oceans circulate the isotope around the planet, researchers must calibrate dates from the Northern and Southern hemispheres differently. Air from both hemispheres mixes over the Andes, however, so atmospheric scientists are helping archaeologists create bespoke calibration curves for major Inca sites.

Because radiocarbon dates from one site can’t be extrapolated to any other, rewriting Inca history with radiocarbon means moving city by city, monument by monument, village by village through one of the premodern world’s largest empires. “It’s one of those things we could never be finished with,” Ogburn says. “There’s always another site we could date.”

On a pleasantly sunny day in October of 2025 at 3600 meters above sea level, Peruvian archaeologist Abimael Choqque Torres was preparing to finish excavating the find of his life. He and his team were digging at Saqsaywaman, a sprawling Inca ceremonial site overlooking Cusco as part of a project supported by UW. A few weeks earlier, while excavating on top of an overgrown hill tucked away from the site’s most impressive monuments, they had caught their first glimpse of a long, pointed metal rod. Since then, they had uncovered a total of 18 metal objects.

As Choqque Torres and his team carefully removed the final metal artifacts from their resting place, he felt the weather shift. Storm clouds were gathering over the surrounding mountains. Soon, the sky cracked and hailstones pelted down. It felt like confirmation that the team had found something special—and it had.

Preliminary analyses later revealed most of the objects were made of bronze, and some of copper. They looked like tools but showed no signs of having been used before being buried. Most were rods of various lengths, some shaped into points on both ends. Others took forms archaeologists had never seen in Inca sites before, such as a small, curved spatula with serrated teeth along its edge and a flat triangle with a handle extending from one of its points.

Although organic material excavated with the offering hasn’t yet been radiocarbon dated, stratigraphy suggests the mysterious implements were buried during a sharp transition at Saqsaywaman. The site was originally dominated by large, open terraces bordered by rough walls of local rock, which were likely used to host lavish feasts intended to elevate the Incas’ status among their neighbors, says Sabino Quispe Serrano, a Peruvian archaeologist working with the UW team. But around the time of the offering, the earlier terraces were abandoned or even dismantled in favor of perfectly fitted masonry, including three zigzag walls, built in part from giant stones that had to be hauled from dozens of kilometers away. The Inca had evidently shifted from feasting with local allies to commanding the labor of a far-flung network of subordinates.

Combined with the architectural transition, the extraordinary offering suggests that at least in the Cusco heartland, the birth of the empire was as sudden as Cabello de Balboa and Rowe believed. “With this offering, they’re trying to inaugurate a new time,” Earle says. “There really is a moment where everything changes.” That pivotal period could correspond to the rise of Pachacuti, as Rowe argued. Earle points out that the name Pachacuti, which the emperor adopted upon taking the throne, means “one who turns over a new universal cycle.”

But when the UW team radiocarbon dated charcoal from its excavations at Saqsaywaman, it discovered that the site’s—and perhaps the empire’s—transformation likely took place several decades before Rowe’s start date of 1438. The researchers, including Earle, Choqque Torres, and Quispe Serrano, found that Saqsaywaman began to be remodeled around 1400, with construction kicking into high gear starting around 1418. The project’s first radiocarbon results are in press at Antiquity .

The Inca could have been on the rise even earlier, before Pachacuti was born, radiocarbon dates from beyond Cusco suggest. Marsh and fellow CONICET archaeologist Kevin Lane argue that to launch his coup as a young man in the early 1400s, Pachacuti had to be born between the 1370s and ’90s. But new radiocarbon dates from the southern part of the empire indicate Inca ceramics and architecture had already appeared at many sites in Argentina, Bolivia, and Chile by that time, more than a century before Rowe’s chronology claimed, and before Saqsaywaman’s transformation. Marsh and Lane think Pachacuti’s little-known grandfather, Yahuar Huacac, may have been the true leader of the Inca’s first push southward . Perhaps Yahuar Huacac was among the unfortunate kings largely erased from Inca history to heap more glory on his descendants.

The new early dates from the edges of the empire also change archaeologists’ vision of what Inca conquest looked like. Rowe wrote that the Inca wielded “a small but efficient army [that] could overrun vast areas and apparently powerful states quickly, and with little opposition.” But researchers can now see the expansion was far more piecemeal. Inca artifacts, infrastructure, and architecture often appeared at nearby sites decades apart and endured for significantly different lengths of time. “It’s not an army that advances town by town,” says Francisco Garrido, an archaeologist at the National Museum of Natural History in Santiago, Chile, who has excavated and dated Inca sites in northern Chile’s Copiapó Valley. “It’s more like seeds that disperse at the same time but produce something different in each place.”

The stories of the people the Inca aimed to subjugate, invisible for centuries, are also coming to light thanks to new dates—and it turns out the Inca faced far more sustained resistance than the empire would ever admit.

“Brush, brush, brush!” Elizabeth Arkush, an archaeologist at the University of Pittsburgh, clambers over a collapsed stone wall and crouches in the ruins of a small, square room. A team member hands her a paintbrush, and she gently clears earth away from a broken piece of a ceramic plate. A geometric pattern of grouped vertical lines emerges against a dark background, edged by a border of triangles.

“Certified Inca,” Pontifical Catholic University of Chile archaeologist Francisco García-Albarido declares. He found the sherd Arkush is now cleaning on his first day excavating at the site of Sicllani Incacancha this summer, more than 4200 meters above sea level on a mountaintop overlooking the Peruvian altiplano, the high plateau north of Lake Titicaca.

“I’m so jealous!” Arkush says.

At a typical excavation, one ceramic sherd wouldn’t inspire much excitement. Archaeologists can find hundreds per day in the Andes, with its high density of archaeological sites and dry, artifact-preserving climate. Arkush and co–principal investigator Sarah Kennedy, an archaeologist at Carleton College, expected the stone buildings at Sicllani Incacancha and nearby mountaintop sites to brim with plates, cups, beer jugs, and other remains of the Incas’ famous feasts, like the ones they threw at Saqsaywaman to consolidate their power. But the structures turned out to be shockingly empty. Day after day, the team dug up nothing but dirt. Without artifacts, the building’s purpose remained a mystery.

The lone ceramic, at least, emphasized who was behind its construction: the invading Inca empire. With their sharp corners, wall niches, and rectangular layouts, Sicllani Incacancha and other nearby Inca sites couldn’t be more different from the local architecture. Starting around 1300, the people of the altiplano began to live inside rambling, seemingly disorganized hilltop forts now called pukaras . Human remains from this period show a spike in violent deaths, reflecting the long tail of political turmoil ignited by an earlier drought and the collapse of the Wari and Tiwanaku states about 4 centuries before the Inca arose. Communities, including a people known as the Colla, retreated behind imposing walls to protect themselves.

Pukara walls were no match for the invading Inca, however—or at least that’s what archaeologists thought. The colonial chronicles claim the Inca quickly conquered the Colla and the other altiplano communities during Pachacuti’s reign, forcing them down from the pukaras “to live not like savages, but like men of reason,” in the words of Pedro Cieza de León, a 16th century Spanish conquistador and chronicler. Archaeologists who held to the traditional chronology assumed the lack of Inca artifacts in otherwise densely occupied pukaras confirmed the forts were hastily abandoned when the empire invaded.

But when Arkush recalibrated radiocarbon dates from both pukaras and Inca sites on the altiplano, she found the two cultures overlapped for between 30 and 60 years . Some of the largest hillforts were actively constructing defensive walls decades after the first Inca sites appear around 1400. “It’s not pre-Inca. It’s non-Inca,” Arkush says. Many pukaras represented “ongoing resistance.”

In Copiapó Valley, meanwhile, Garrido found evidence of conquered communities pushing back against imperial domination. His new radiocarbon dates for a site known as Iglesia Colorada revealed that between 1400 and ’40, the Inca aggressively remodeled the village and seemingly killed anyone who didn’t like it. Garrido studied decapitated human skulls from the site that had holes bored into them, suggesting they had been displayed as trophy heads before being discarded in a trash heap. Local people expressed their displeasure by deliberately excluding Inca ceramics from their cemeteries. “It was a form of disobedience, of not collaborating with the Inca project,” Garrido says.

The resistance may have been enough to push the Inca to rethink their approach. Garrido’s new dates from slightly later sites in the valley reveal that after 1440, the Inca concentrated their presence near a site dedicated to copper smelting, their main economic interest in the region, and left local villages to their own devices.

Perhaps a similar timeline of conflict and negotiation will emerge in the altiplano, as archaeologists continue to build the region’s chronology. For now, it seems the Inca did eventually wrest control away from the Colla. Arkush’s radiocarbon dates show people left the last pukaras around 1450, whereas altiplano Inca sites continued to be occupied. If they turn out to have been built after the Incas’ eventual victory, Sicllani Incacancha and the other mysteriously empty structures may have served to commemorate the conquest and make the empire’s unchallenged presence known across the plain below.

“Maybe building it was the point,” rather than occupying it, Kennedy proposes. “Maybe the Colla were forced to construct it.” Many of the Spanish chroniclers noted that the hill where Sicllani Incacancha is located was the site of the Collas’ last stand. Each morning on their way to the site, the archaeologists hike past a rocky pukara bristling with irregular walls that may have been the Collas’ final stronghold. Meanwhile, Sicllani Incacancha’s entryway is lined with trapezoidal niches large enough to fit a seated human body; in other parts of the empire, the Inca are known to have taken local mummies hostage and displayed them.

But it’s also possible radiocarbon dating will reveal that Sicllani Incacancha was built decades earlier, soon after the Inca arrived on the altiplano. Perhaps they did intend to use the structure for feasts as Arkush and Kennedy initially expected. Maybe it was the Colla who refused to cooperate with this imperial script, rendering the ghostly structures useless as soon as they were completed.

Only radiocarbon dates can distinguish between these possibilities, or reveal others. The day after the discovery of the Inca plate at Sicllani Incacancha, Arkush and Arizona State University archaeologist Ryan Smith gather around a niche in an interior wall. Arkush digs her trowel into the mortar between stones and dislodges dried stalks of local ichu grass, which the site’s builders added to the mixture during construction. Smith catches the plant remains in a pocket of aluminum foil, destined for a radiocarbon lab. Unlike so many Inca archaeologists before them, they have no idea what the answer will be.

LISTEN TO THIS ARTICLE

When the National Science Foundation (NSF) chooses a new contractor to provide logistical support for the U.S. Antarctic research program, it’s almost a given that one of the losing contenders will file a protest. The most recent bidding cycle for an $8 billion contract, NSF’s largest single expenditure, was no exception.

But what happened next is unprecedented. The Government Accountability Office (GAO), the congressional watchdog agency that handles such protests, concluded that one of the objections raised by Battelle Memorial Institute was valid—the first time that has happened, sources tell Science —and recommended NSF revisit its decision. Researchers were also already worried the new contractor’s plans to hire foreign workers at below-market wages could degrade the quality of support to scientists and pose a potential threat to national security, issues not addressed by GAO. But NSF appears to be sidestepping both concerns.

On 2 June the agency awarded a 20-year contract to KBR, a large science and technology contractor. KBR said a new spinoff company would do the work, providing all services needed to operate the three U.S. research stations in Antarctica, including food, housing, and communications as well as logistical support for scientists traveling to and from Antarctica from New Zealand or Chile and those working in field camps on the continent. Battelle immediately filed a protest claiming NSF’s decision was badly flawed.

In a 26-page report issued on 10 September, GAO agreed with one of Battelle’s contentions. It found NSF had “unreasonably evaluated KBR’s prior experience and past performance because it improperly credited the [new] firm with the experience of the affiliated company.” GAO said NSF should “make a new selection decision” after conducting “a new analysis” of all the bids submitted. That process could take several months, says a GAO official who was permitted to speak anonymously to explain its findings.

“We’re not telling NSF who should get the award,” the GAO official says. “We’re just saying redo the evaluation the right way and then make a decision that’s consistent with that analysis.” NSF had 60 days to tell GAO what it planned to do.

But NSF doesn’t appear to have done the re-evaluation GAO recommended. The agency paused the contract as soon as Battelle filed a protest, according to an NSF spokesperson, but “that stay has since been lifted now that NSF has resolved GAO’s recommendations.”

The spokesperson declined to comment on how NSF had addressed GAO’s concerns, saying only that “NSF reviewed the GAO protest decision and fully complied with the recommendations.” NSF recently added 1 year to the current contract, the spokesperson noted, “to allow for a successful transition of work.” That contract, held by Leidos, now expires in September 2027.

Former NSF polar officials had hoped the re-evaluation would also lead the agency to fix what they believe are other flaws in the winning bid. One is KBR’s apparent intention to use foreign workers to provide logistical support. “We suggested hiring a small number of other country nationals from friendly foreign governments. We do this all over the world, including at U.S. military installations,” says Doug Hill, president of explorations and logistics at KBR. That policy allowed it to submit a bid of $8.4 million for the estimated cost of providing two components of the contract—logistics at Christchurch, New Zealand, and operations of the Long Duration Balloon facility on the Ross Ice Shelf. That bid was the lowest of the four competitors and some 40% less that Battelle’s bid of $13.8 million.

“Using foreign workers in Antarctica can be very cost-effective,” says a former NSF senior polar manager who requested anonymity to protect his relationship with the agency. “But it’s not necessarily a smart way to run the program. Motivation, morale, and performance are highly linked to pay. Low morale could lead to people slow walking their jobs or not showing up for work. And that could be a serious problem if it’s a job requiring unique skills.”

GAO concluded NSF had “determined that all the proposed rates were sufficiently substantiated and, therefore, realistic.” GAO also accepted NSF’s assessment “that KBR has a unique advantage of having low direct labor rates due to its experience employing” non-U.S. workers.

Former NSF officials are also troubled by KBR’s plan to hire a recently retired senior NSF official to be project director for the new contract. “One thing that’s drilled into all of us when we leave is to avoid apparent conflicts of interest,” one says. “So, I was shocked that KBR would do that.”

The company’s bid noted that the intended program leader was currently working “as an NSF employee providing oversight for multidisciplinary research teams.” In its protest, Battelle argued that person “would not be able to negotiate or communicate directly with NSF” because of a mandatory 1-year “cooling off” period for those leaving the agency. But GAO concluded that “the solicitation did not state that the program director must be the sole point of communication between the agency and the contractor” and that KBR had proposed a workaround using another senior manager.

The use of foreign workers also poses a potential threat to national security, researchers say. Although NSF doesn’t fund any classified research in Antarctica, all communication to and from the frozen continent must go through government websites. “Can you imagine the Chinese or Russian stations employing foreign workers?” one researcher asks.

Battelle declined to say whether it plans to file a second protest based on NSF’s latest decision. “We do not comment on pending litigation,” the spokesperson said. The GAO official says they expect such a request, which would prompt a second GAO investigation. But at press time, the official says, NSF still had not notified GAO of its response to the report.

LISTEN TO THIS ARTICLE

When coral bleaches during a heat wave and ejects its symbiotic algae, it usually means either a death sentence or a slow recovery over months or years. But scientists working separately at two labs an ocean apart have found a faster acting shortcut, one that has scientists hoping it could eventually help coral recover from bleaching and endure future heat waves.

The approach, reported in a 2 October paper in Environmental Microbiology Reports and one published today in Science Advances , is deceptively simple. By touching a piece of bleached coral to a healthy piece, a kind of algal transfusion can occur. Symbionts flow through a bridge of tissue and then rapidly reproduce, repopulating the bleached piece in as little as 4 weeks.

“People are always looking at ways to increase or accelerate recovery from coral bleaching,” says Andrew Baker, a coral biologist at the University of Miami who wasn’t involved in the research. “This is one of those methods that had never occurred to people.”

Stanford University marine biologist Stephen Palumbi, an author of the Science Advances paper, stumbled upon the phenomenon by accident. In his lab, a chunk of coral intentionally bleached for an experiment wound up resting against a healthy chunk from the same coral colony, which is made from genetically identical polyp clones. He noticed that the bleached piece soon began to regain its brownish-green color—a sign the symbionts were back.

Around the same time, Ph.D. student Danny Collins was in another Stanford lab exploring whether algal symbionts could be delivered to coral through contact with some kind of substance. When Palumbi mentioned his coral observation, Collins launched a series of experiments. After bleaching fingertip-size coral fragments of cauliflower coral ( Pocillopora damicornis ) by exposing them to warm water, Collins placed them in an aquarium in contact with healthy bits of coral from either the same colony or a different colony.

The method didn’t work when the healthy coral came from a different colony. In those situations, the coral polyps sent out tiny stinging appendages as if attacking an intruder. Collins likened it to a person’s immune system rejecting an organ transplant from a stranger. Eventually, those bleached corals died.

But when corals from the same colony were physically touching, tissue from polyps in the two chunks merged to create a kind of algal highway. If the bits stayed in contact for a week, symbionts poured in and began to replicate, doubling approximately every 4 days, as much as 10 times faster than what has been witnessed on bleached reefs, which normally recover by reabsorbing free-floating algae. All 11 of the corals in this arrangement fully recovered within 8 weeks.

These tantalizing findings, however, don’t immediately translate to a way to repair bleached coral in the wild, Baker notes. He ticks off several challenges: Even in cases where it could work, it’s not realistic to think about such treatment for every colony on a reef. Also, reinserting the same symbionts means the coral could be just as vulnerable to bleaching during the next heat wave, because different algal strains affect how a coral copes with heat.

The other paper, led by Bede Johnston, a coral researcher at the Australian Institute of Marine Sciences, addresses one of those problems. There, scientists bleached sections of two species of coral, Galaxea fascicularis , commonly known as octopus coral, and Psammocora columna . They let polyps in some of the fragments absorb lab-raised symbionts bred for tolerance to extreme temperatures. When the inoculated fragments were joined to bleached fragments, some of the heat-tolerant symbionts found their way into the neighboring polyps.

This process happened faster than it did with fragments left to catch stray symbionts floating in the water. But the recovery wasn’t as quick as that witnessed by Palumbi’s team. Madeleine van Oppen, a coral geneticist at the University of Melbourne and the Australian Institute of Marine Science and author on the Australian study, says she “was quite blown away how quickly the symbionts were transmitted [in the Stanford experiment].”

Palumbi thinks the new findings open the door to coral recovery strategies that hadn’t previously been considered. Collins is already looking into ways to deliver packets of healthy symbionts in a way that doesn’t require genetically identical donors, while also being more concentrated than free-floating algae.

“It would be totally cool if what it would do is spark 100 little experiments on finding some clever way of getting it all to happen,” Palumbi says of the new findings. “Some coral symbiont enema or something.”

LISTEN TO THIS ARTICLE

In May, Health and Human Services (HHS) Secretary Robert F. Kennedy Jr. launched an effort to combat what he called a “dependency crisis” caused by the overuse of antidepressants, now taken by roughly 15% of U.S. adults. The initiative emphasized deprescribing, or tapering patients off the drugs, and HHS is expected to release clinical guidance soon on how best to do so.

Doctors and researchers know antidepressants don’t work for everyone, and some even agree the drugs are overprescribed. But the central issue, many argue, is that finding the right treatment is often a frustrating matter of trial and error, leaving many people using drugs that aren’t working for them. Rather than focusing only on getting people off the drugs, many researchers say a more useful approach is to predict who is likely to benefit from them in the first place.

“People are spending weeks, months, sometimes years trying to find what will be most helpful, all while bearing the burden” of depression symptoms, says Jordan Smoller, a psychiatric geneticist at Massachusetts General Hospital. “If we could do a better job of stratifying people, either clinically or biologically, we could do much better at treatment.”

Studies now underway are using genetic data, brain scans, and artificial intelligence (AI) analysis of health records to determine which patients might respond best to a given drug—and which would be better served by psychotherapy or another form of treatment. One method for doing so, based on brain imaging, is already moving toward approval from the U.S. Food and Drug Administration (FDA) as a tool to guide prescribing.

The biological roots of depression remain murky, and treatments for the disorder are poorly understood. Physicians usually take their best guess at prescribing either psychotherapy such as cognitive behavioral therapy (CBT) or one of several “first-line” antidepressants such as selective serotonin reuptake inhibitors. But either treatment only works about 30% to 40% of the time, and for about 30% of people, depression persists after they’ve tried two or three different drugs.

Part of the problem, Smoller and many others believe, is that depression comes in a variety of forms that respond differently to intervention. Researchers have long hoped to find clues to these subtypes in the genome—antidepressant response is known to be largely heritable—and giant research teams, such as the PsycheMERGE study that spans 12 health systems across the country, are amassing medical records and genetic data to search for links. So far, few specific genetic predictors have emerged. Still, Smoller says that if genetic data can “change that probability [of success] even modestly, you could be saving a lot of real suffering.”

Adding in other types of data might improve predictions, especially with the help of AI. Earlier this year, University of Oxford psychiatrist Andrea Cipriani and colleagues published a statistical and AI tool called PETRUSHKA , which they trained on the sociodemographic details and health histories of more than 120,000 people who’d taken at least one of 21 antidepressants. By linking patterns identified in these data to drug responses, the tool is designed to recommend which drug a patient should try first.

In a trial with about 500 additional people, those who received the drug recommended by PETRUSHKA were 38% less likely to quit the antidepressant after 8 weeks than those who just received the doctor’s recommendation, the team reported in March. “We’re not identifying classes of drugs that are better, but the best drug for the individual,” Cipriani says. His team is now adding genetic data to the model, which could further improve its predictions.

Future iterations of PETRUSHKA may also be able to predict who will benefit more from psychotherapies such as CBT than from any drug. Cipriani says this could help clinicians as they consider patients’ preferences—many people do not want to take antidepressants, and factors such as insurance coverage and life circumstances can affect whether a person will stick to a given treatment.

In India, a team led by Harvard University psychiatrist Vikram Patel is staging real-world tests of a machine-learning tool designed to sort people between drugs and a simple form of psychotherapy that can be administered with minimal training. The researchers are tracking 1500 people with depression who filled out a questionnaire about their symptoms, socioeconomic status, and other health data. The researchers will feed the data to the AI tool, then assign each person to receive one or the other treatment at random. After 3 months, the team will assess whether people who received the treatment recommended by the algorithm recover better. Patel says the study will release its first results in the coming months.

More nuanced analyses might emerge from larger data sets, like the collection of electronic health records from 200,000 people that insurance giant Kaiser Permanente is using to develop its AI-driven tool to guide antidepressant prescribing.

The hope is that AI models will pick up subtle factors that humans miss. “We’re not as good as we think we are at predicting ahead of time what variables or symptoms are most meaningful,” says clinical psychologist Devon LoParo at Emory University. He and his team have trained a machine learning model on patient records from a clinical trial and used it to predict who would do better on CBT or one of two antidepressants. They found that people in a separate trial did better if they had happened to receive the algorithm’s recommended treatment.

A few patterns emerged from their analysis: Some antidepressants worked better in people with more severe physiological symptoms of depression such as insomnia, whereas CBT tended to help people with stronger social support networks. Still, LoParo and others say AI predictions can be misleading and should be considered alongside patients’ preferences and doctors’ judgments.

Other researchers are looking for clues to treatment response in the brain activity thought to underlie depression. “That’s really what we’re treating,” says Leanne Williams, a psychiatrist at Stanford University. Her research has identified six distinct “biotypes” of depression based on functional magnetic resonance imaging (fMRI). These biotypes seem to react differently to depression treatments: People who have hyperactive circuits associated with negative emotions, for example, may respond better to behavioral therapy, whereas those with underactive circuits might respond to a specific antidepressant.

In a recent study, Williams found that prescribing based on biotype improved symptom remission rate from 16% to 70% for certain treatments. Her team is now testing the fMRI technique in 20 clinics across the United States, and in September FDA designated it as a “breakthrough device,” speeding its path toward commercial approval.

The tool might also identify people who will likely never respond to either psychotherapy or first-line antidepressants, Williams says—potentially allowing them to skip straight to second-line therapies such as transcranial brain stimulation, electroconvulsive therapy, or ketamine.

Kennedy isn’t talking about predictive tools, Williams says, but they fit with his priorities. At a visit to the White House in April, she urged him to support research into the biological roots of depression and its treatment. “It really is about identifying early who is not likely to benefit from an antidepressant—but also who is,” she says.

LISTEN TO THIS ARTICLE

From a perch high in Chile’s Atacama Desert, a cluster of telescopes is taking aim at the oldest light in the universe, the cosmic microwave background (CMB), emitted just 380,000 years after the Big Bang. With greater sensitivity and sharper resolution than earlier CMB surveys, the Simons Observatory will deepen cosmologists’ understanding of the universe’s makeup and look for signs of “cosmic inflation,” a hypothesized burst of exponential expansion in the universe’s first moments. “To a first approximation, it’s going to do what those [earlier surveys] did, but better,” says theoretical physicist Marc Kamionkowski of Johns Hopkins University.

But the observatory is much more than a simple CMB experiment, says co-director Mark Devlin of the University of Pennsylvania. By using the CMB as a backlight, the observatory will also study how the universe has evolved since the light was created. It will discover tens of thousands of galaxy clusters and observe the distribution of matter across an evolving universe. A community of thousands of scientists is waiting for the data, Devlin says. “You have to have a broad appeal.”

Moisture in Earth’s atmosphere blocks microwaves, so observers have mapped the CMB with satellites and balloons, or from the ground at high, dry spots. Prime locations are the South Pole and the Atacama, where the $120 million project, funded mostly by the Simons Foundation, has built three small-aperture telescopes, each 0.5 meters across, and a 6-meter Large Aperture Telescope that is set to begin its survey of the sky this month.

Simons is starting up as CMB researchers adjust to the loss of a much larger planned experiment. Last year, the U.S. National Science Foundation and Department of Energy ended support for a “CMB-S4” experiment , a next-generation project in Chile and at the South Pole that was designed to seek signals of cosmic inflation. “We’re not as sensitive as the original CMB-S4 was planned to be, but we’re covering all of the same science … and we’re doing it now,” says Simons Observatory spokesperson Jo Dunkley of Princeton University.

The CMB, the glow of the superhot infant universe, was stretched by cosmic expansion and now looks like a faint microwave haze from all directions. It is mottled with tiny temperature fluctuations that correspond to concentrations of matter, including unseen dark matter, in the infant universe—seeds that would grow under gravity into the cosmic web of galaxies, filaments, and voids. By precisely measuring those fluctuations, cosmologists have reconstructed the basic recipe of the universe: roughly 5% ordinary matter, 26% dark matter, and 69% dark energy, a mysterious force that is accelerating the expansion of the universe.

Simons will not only refine those numbers, but also use its map of CMB fluctuations to address other questions, including the mass of the neutrino, a ghostly particle that streams through space at high speeds. Neutrinos don’t clump like slow-moving matter and dark matter, so they suppress the growth of the cosmic web. By comparing the early-universe pattern encoded in the CMB with the later distribution of matter, Simons researchers can constrain the mass of these elusive particles. “We should be able to actually have a positive measurement of this mass,” says cosmologist Seshadri Nadathur of the University of Portsmouth.

The observatory will also probe an event thought to have preceded even the CMB: a brief spurt of exponential growth during a fraction of the universe’s first second. In that instant of “cosmic inflation” the universe would have ballooned from less than one-trillionth the size of a proton to about as big as a beach ball. Inflation would have ironed most unevenness out of the newborn universe, explaining why today it is remarkably smooth and homogeneous on large scales yet contains a web of galaxies on smaller ones.

The growth spurt would have caused spacetime to reverberate with gravitational waves that might have left an imprint on the CMB: swirls in the polarization of the light. Researchers operating a CMB telescope at the South Pole called Background Imaging of Cosmic Extragalactic Polarization 2 declared they had found such swirls in 2014 , only to retract the claim a year later after discovering that much of the signal came from dust in our Galaxy.

Simons hopes to do better. Its small-aperture telescopes are optimized to see fluctuations on scales wider than 1°, where the primordial polarization signal is expected to be strongest. They also observe at several wavelengths, helping distinguish the swirls from polarized emission from galactic dust and other signals from the nearer universe. The first two of the scopes began observations almost 2 years ago, with the third joining in more recently. “From here on out, we should be digging down,” Dunkley says. “After 5 years or so, we might be able to see [the swirls].” Some models suggest it might take longer.

While the small telescopes chase inflation’s polarization signal, the 6-meter mirror of the Large Aperture Telescope will create fine-scale maps of a large chunk of the CMB. It will funnel light onto a receiver packed with 60,000 detectors, which are cooled to one-tenth of a degree above absolute zero to register the faint CMB signal.

The maps could hold clues to what triggered inflation, Nadathur says. The standard version of inflation is driven by a single field mediated by a particle called an inflaton. But “maybe there was more than one type of field that was interacting in the early universe,” Nadathur says. The two scenarios would result in different distributions of fluctuation sizes in the CMB.

Simons researchers will also use the CMB as a backlight. Hot electrons in the thin gas between galaxies scatter CMB photons, changing their energy in a process known as the Sunyaev-Zeldovich effect. The distortions act as signposts for galaxy clusters, even when the galaxies are too distant or faint to see. An earlier telescope at the Atacama site found thousands of galaxy clusters this way. With its wide field of view, Simons should detect as many as 40,000, including clusters in the early universe that raise questions about cosmic structure formation, Devlin says. “How did you form such a big thing that early on? And what are the physics of that?”

Researchers also hope to exploit a related phenomenon called the kinematic Sunyaev-Zeldovich effect in which moving bodies of gas shift the wavelengths of CMB photons streaming toward Earth. Such “redshifts” or “blueshifts” could reveal the presence of dark matter without the usual signpost of visible stars. “We always assume that dark matter and [normal matter] are distributed in the same way in the early universe. Now we’ll be able to test that,” Kamionkowski says.

Devlin is looking forward to using the facility for so much more than studying the CMB. “I feel like we’re serving a community now that we weren’t accessing before. And that also is a lot of fun.”

LISTEN TO THIS ARTICLE

Declan Bulwa is rushing to complete an underwater rescue mission before the full force of El Niño hits the California coast.

The University of Southern California Dornsife ecologist has been scuba diving in the state’s kelp beds in recent months, gathering reproductive tissue from the temperature-sensitive giant algae in a bid to bank as much genetic diversity as possible before expected underwater heat waves wipe them out. Many beds still haven’t recovered from a marine heat wave in 2014 and ’15, fueled partly by an El Niño. Already warmer water is gnawing at the remaining beds, home to a diverse array of marine species, Bulwa says. “It is having an effect, and having an effect fast.”

Bulwa’s dives are part of a range of urgent scientific work around the world spurred by the arrival of what has been dubbed a “super” El Niño. Marked by unusually warm surface water in the central and eastern reaches of the tropical Pacific Ocean, El Niños set off weather patterns that reverberate planetwide—fueling droughts and blazing heat in some places and heavy rains in others.

This year’s event is shaping up to be one for the record books. Sea surface temperatures in the central tropical Pacific in July were 1.7°C above the current average temperature for tropical oceans worldwide, according to the U.S. National Oceanic and Atmospheric Administration (NOAA), which began using that metric earlier this year to monitor the ocean phenomenon. That figure was only slightly below the highest recorded July level of 1.8°C in 1997. Surface temperatures in some parts of the eastern Pacific are now 4°C above the 30-year average, and NOAA recently warned of a 90% likelihood of a “very strong” El Niño.

The extreme pattern could lead to crop failures, floods, and fires, as well as devastating consequences for ecosystems. But researchers say it’s also an opportunity to study how the natural world responds to weather extremes that could become more common as the planet warms.

“Most of us climate change ecology people are kind of perversely excited about it,” says John Bruno, a marine ecologist at the University of North Carolina at Chapel Hill. “It’s an experimental treatment that nature is providing to us.”

Bruno was the lead author of a recent letter in the Proceedings of the National Academy of Sciences , in which a group of U.S. and European marine scientists urged researchers and science-focused agencies to take immediate steps to prepare. Among their recommendations: Consider preemptively limiting fishing where fisheries are known to suffer downturns during El Niños, prepare rescue missions for particularly vulnerable species, and work to expand ecosystem monitoring so the impacts of the event can be documented and used to make predictions in the future.

More than 180 Indian ecologists issued a similar plea in a September letter . “We actually do not have a good sense of how Indian ecosystems are going to respond to this event,” says Rohan Arthur, a marine biologist for the Nature Conservation Foundation and an author of the letter. He said he has been “overwhelmed” by the response from other scientists. Many have said they are modifying their monitoring programs to track El Niño–related changes spurred by what is expected to be high temperatures in India on land and the nearby ocean.

In a lowland rainforest in Panama, University of Wyoming ecologist Corey Tarwater and collaborators are deploying insect traps and acoustic recorders that can capture the sounds of birds and bats across 100 hectares of jungle. Tarwater has spent years studying why many bird populations there have been declining . A leading suspect is rising temperatures and drought—conditions the El Niño is expected to intensify. Even the wet season has been extraordinarily dry this year, she says, and the dry season, starting in January 2027, “could get really bad.” The data her team collects in the coming months could offer a preview of how birds, bats, and insects will fare in a hotter world.

Meanwhile, along Australia’s Great Barrier Reef, coral scientists are making plans to monitor the reef during what could be a brutal summer of marine heat waves. Overheated coral polyps often eject their algal companions, leading to “bleaching” that can starve and even kill the coral. The El Niño summer could provide a critical test for coral inoculated with experimental symbiotic algae bred for heat tolerance. “We expect this is going to be one of the first serious opportunities to assess the efficacy of that intervention,” says Matthew Nitschke, a coral biologist at the Australian Institute of Marine Science.

Exactly how this El Niño will play out in specific places isn’t clear, but better forecasting models are helping scientists prepare. Alistair Hobday, a marine researcher for Australia’s Commonwealth Scientific and Industrial Research Organisation, the country’s main national research agency, has overseen the creation of a tool for predicting marine heat waves up to 4 months in advance in waters around the continent. Public forecasts began last year. “It’s put us in a lot better position for avoiding some of the surprises,” Hobday says.

Right now, the model shows severe conditions are likely near Tasmania and the southeastern coast in October and November. The prediction has already prompted scientists to postpone work restoring seagrass beds, because the warm water could kill the new plants, Hobday says. Researchers are also getting ready for a possible replay of the 2024 rescue of red handfish, in which divers captured 25 of the fewer than 100 remaining individuals in this highly endangered Tasmanian species. They were kept in tanks until the heat wave passed.

In the United States, scientists are speeding the rollout of a computer program to forecast ocean conditions along the Pacific coast, which has been repeatedly rocked by marine heat waves. The new high-resolution model, part of an NOAA program to improve fisheries management in a changing ocean, was supposed to be released in 2027, says Mike Jacox, a physical oceanographer at the agency. But signs that an El Niño was coming prompted them to rush its release later this month, with an eye toward helping fisheries managers respond. So far, the forecast is alarming. “We’re seeing exceptionally warm water all the way across the west coast and into Alaska,” Jacox says.

Whether this “super” El Niño will translate into superdamage around the world remains to be seen. Other ocean phenomena, such as water temperatures in the Indian Ocean, can either moderate or amplify an El Niño’s effects, says Chris Funk, an applied climatologist at the University of California, Santa Barbara whose work focuses on developing weather forecasting tools to predict drought, famine, and other natural disasters. He says the indicators so far point to a wild ride for people and ecosystems. “The pretty obvious conclusion is that we should buckle up.”

The shock was palpable. On the evening of 6 October, some of the world’s leading mathematicians gathered in Princeton, New Jersey, to mark the opening of another academic year at the Institute for Advanced Study’s (IAS’s) School of Mathematics, a renowned research center. For weeks, rumors had been circulating that OpenAI, the San Francisco–based artificial intelligence (AI) company, was close to releasing scores of solutions to advanced math problems. But the news that broke during the gathering left many of the mathematicians agape: OpenAI had dumped an extraordinary mass of 722 preprints covering 372 distinct results on a website.

For some, the preprint avalanche held exciting opportunities to make progress on solving some of the field’s most famous and intractable problems, such as the 166-year-old Riemann hypothesis . But for Constantin Kogler, an IAS researcher early in his career, the news landed closer to home. Scrolling through the papers on his phone while still at the party, he found a proof of the so-called exact overlaps conjecture—a result he and a colleague had been working flat out to write up since cracking it 10 days earlier. “It was really big,” Kogler says. “I would say a big breakthrough in the area.” When he got home that night, he decided to post his own paper. It went online just 4 hours after OpenAI’s.

Now, as the shock begins to wear off, the global math community is starting to grapple with a panoply of questions. How much of OpenAI’s work is actually right? How can anyone begin to understand it? And what does it even mean to do math in the age of AI?

If nothing else, the effort has highlighted that the price of AI math is coming down—fast. To obtain its 372 results, the company says it used an unreleased ChatGPT model to take a swing at about 4000 problems. The model made quick work of many: OpenAI says reaching the vast majority of the solutions took, on average, the equivalent of just 3 hours of ChatGPT Pro “thinking.” In comparison, the company’s headline-grabbing Navier-Stokes breakthrough last month took a swarm of 10,000 agents working continuously for 88 hours, likely costing millions of dollars.

The work is often impressive, mathematicians say, but the presentation a mess. “I’m quite confident that a human who understands these results could write them in a more understandable way,” says Edward Witten, a mathematical physicist at IAS who has been examining a cluster of papers claiming to resolve a 1975 conjecture by physicist Alexander Polyakov.

Steven Sivek, a member of the Association for Human Mathematics, an organization that views AI as a threat to the field, was up at 4 a.m. in London when the papers dropped. The one closest to his own work ran 140 pages and was “completely unreadable,” he says, acknowledging the early hour may not have helped. But “I came back to it and it didn’t get any better.”

Some mathematicians are already forming groups to work through the claims, says Olga Holtz of the University of California, Berkeley. “Their goal, of course, won’t be just to certify pages, but to discover what, if anything, is new and useful in the mathematics,” she says.

Mathematicians are also hunting for mistakes. As Science went to press, OpenAI said it had withdrawn three of the manuscripts and revised 14. Of the 719 remaining papers, only 300 have been formalized—translated into Lean, a programming language that lets a computer check every logical step of a proof.

But formalization is no guarantee of truth. Lean can confirm a proof establishes the statement it is given in code, but it can’t confirm the statement that the paper claims in English. In some cases, Sivek says, that means what was formalized is “something weaker” than the main theorem. And a preprint posted last week found the Lean formalization of OpenAI’s Navier-Stokes result doesn’t match the proof described in the accompanying paper. That doesn’t mean the result is wrong, researchers say, but it suggests computer verification may not carry the weight often attributed to it.

“One must distinguish a claimed result, an independently verified result, and a result that has been understood and absorbed into mathematical culture,” Holtz says. “Those are not the same.”

OpenAI tells Science , “We are extremely supportive of mathematicians working on our results and welcome if they want to publish papers on them.” It plans to fund workshops and other events to further understanding and confirmed it’s already talking with potential collaborating institutions. Witten, who belongs to an independent advisory group that has counseled OpenAI on how to release such results, says he would consider writing up one or two of the papers himself. And Kogler says his recently posted paper is “much better” than OpenAI’s.

What the field does next will partly depend on negotiations with AI companies. The advisory group has called on OpenAI to stop testing open problems on unreleased models that mathematicians can’t access. So far the firm has rejected that request. But it agrees “on the spirit” of some of the group’s other recommendations, such as depositing results in independent repositories and disclosing the prompts behind each result.

Some mathematicians fear AI’s capabilities will cause the field to close ranks. “People are a lot more hesitant to share work in progress,” Sivek says, worried “that someone will come in and press a button and solve [the problem], and there goes their research program.”

But Max Lieblich, chair of the math department at the University of Washington, says the paper dump could catalyze a useful rethink of the field. “I never thought that in my lifetime I would have a chance to work with others to think about what our discipline even is, what we’re trying to accomplish, how we should be working,” he says. “Who belongs there? Is it all people? Does it also include some machines?”

Lieblich envisions a “mathematician mode” in which researchers share their ideas with an AI model on the condition that they are credited, and the system acts as a matchmaker when several people are working on the same problem. “Connect us all,” he says. “Let’s all work on this together.”

Whatever happens to math will reverberate across science. Open-AI says it is evaluating its models against difficult frontier problems in multiple fields, and other companies are applying similar tactics.

Sivek advises researchers in nonmath fields to prepare now. “It’s never too early to start thinking about what it is that you value most in your chosen field,” he says, and to make sure that the values of AI firms “are aligned with yours.”

Kogler, meanwhile, doesn’t sound like someone who lost a race, but rather someone who glimpsed the future of his field. “All of us have to rethink what it means to be a mathematician and what is valuable to do,” he says. And he’s certainly not ready to leave. “I’m more excited about staying in math than a few months ago, because it’s actually so much more interesting now. It’s all much faster. You can see amazing mathematics everywhere.”

LISTEN TO THIS ARTICLE

On 6 October, James Maynard was nervously checking his phone under the table throughout dinner. He’d heard rumors of a tsunami heading for the shores of math. When he checked in one last time from bed, the swell struck: OpenAI had released more than 700 math papers, generated with an advanced version of its artificial intelligence (AI) model. They appeared to topple hundreds of long-standing problems, forcing researchers to face the reality that no problem in math is off-limits for machines .

At first, Maynard, a mathematician at the University of Oxford and recipient of the Fields Medal, math’s most prestigious prize, feared for colleagues whose work would be upended. But then he spotted a paper that offered progress on one of his favorite pastimes: the Riemann hypothesis, a puzzle about prime numbers and one of seven famed Millennium Prize Problems, often considered the most formidable in all of math. “I was like a kid at Christmas,” Maynard says. “Even if it was some scary monster giving me a gift, I was just so excited.”

The release comes less than 1 month after OpenAI announced a solution to the Navier-Stokes equations , another Millennium Prize Problem. OpenAI’s latest proof solves only a weaker version of the Riemann hypothesis. But Maynard, who made the most recent notable Riemann advance in 2024 , says it is “groundbreaking” and likely to have payoffs for other problems across math. “Within analytic number theory, it’s obviously the best headline result for the past 100 years in the field.”

Prime numbers—divisible only by themselves and one—seem to be randomly scattered about the number line. But as a teenage prodigy in the 1790s, Carl Friedrich Gauss guessed they follow a simple curve, becoming less frequent as the numbers grow. Their actual distribution fluctuates around that curve. In 1859, Gauss’s protégé Bernhard Riemann showed those fluctuations could be encoded by a complex function—specifically, by inputs between zero and one on the x-axis that drive the function to zero. “Between zero and one is where all the mystery lies,” says Youness Lamzouri, a mathematician at the University of Lorraine.

Riemann’s famous hypothesis was that in this critical region, all of the function’s zeros should lie on the x-axis at exactly one-half. If true, it would show the fluctuations of primes around Gauss’s guess are tightly constrained. Although mathematicians have now verified that more than 10 trillion zeros in the critical region sit at exactly one-half, a single outlier could completely skew the overall distribution of primes.

Proving Riemann’s hypothesis by ruling out zeros everywhere between zero and one except at one-half is so tough that mathematicians are mocked for working on it. Even marginal improvements—including Maynard’s 2024 advance, which lowered the limit for how many zeros could exist at three-quarters on the x-axis—are hailed as major breakthroughs, because the efforts often unlock totally new branches of math. The Riemann hypothesis “definitely holds special status,” says Bryna Kra, a mathematician at Northwestern University. “It’s like a lighthouse in mathematics.”

That’s why mathematicians were stunned to find a Riemann advance in the flood of OpenAI papers. The AI model proved there are no zeros between seven-eighths and one , validating what mathematicians call the quasi-Riemann hypothesis: that a fixed-width region next to the right edge of the critical strip contains no zeros. “I didn’t think I would live to see this result, and I’m not that old a guy,” says Alex Kontorovich, a mathematician at Rutgers University.

The result is especially significant because it generalizes to a larger family of functions beyond just Riemann’s. In doing so, it rules out “Siegel zeros,” possible zeros near one that have vexed mathematicians for nearly a century. OpenAI released an additional brief paper that independently ruled out the Siegel zeros. “It’s completely mind-blowing that there were two separate proofs of no Siegel zeros,” Maynard says. That result alone helps with other problems in number theory, such as estimating how many primes end in seven below a certain number.

Of the hundreds of papers released in the dump, only the quasi-Riemann paper was edited by humans before release. OpenAI hasn’t named the collaborators, frustrating mathematicians eager to question them. But the human factor seems to have paid off. Unlike many AI forays into math, including last month’s convoluted Navier-Stokes result, the new paper “is actually remarkably well-written,” Maynard says. “This doesn’t feel like it’s some impenetrable, incomprehensible proof. … There’s definitely gold here.”

He says the AI model drew from research in a technical corner of math that’s difficult to connect to the Riemann hypothesis. “It’s putting together some ideas that were not terribly mainstream in an exceptionally clever way and making a jump,” Maynard says. Ironically, that leap renews his faith in human mathematicians, by showing how far they can progress by combining existing ideas in clever ways. “We can go further than we expect with the tools we already have,” he says.

The OpenAI model’s solution still needs to be confirmed and deciphered. Lamzouri, for one, has some relevant experience. In August, an advanced version of Anthropic’s Claude model improved the bound on the minimum proportion of zeros that must lie at one-half, from 41.7% to 67.2%. Lamzouri found the paper to be opaque “AI slop.” By tracing its references, however, he uncovered an unorthodox approach, and reproduced and built on the AI’s results in a much tidier fashion . The experience gave him some hope for human-AI collaboration. “AI shows you the door, and then humans open it,” he says.

But AI’s insurgence has also made his research more stressful. Since posting his preprint, Lamzouri has spotted a handful of papers from amateur mathematicians who have milked his method and prompted AI models to churn out incremental results they can’t explain. He now feels pressure to publish quickly to avoid being scooped—both by AI companies and individual users. “I feel like I’m working in a toxic research environment,” Lamzouri says. “It’s all just too fast.”

LISTEN TO THIS ARTICLE

Last week the White House held a daylong science symposium that included a Nobel laureate, six billionaire tech executives, two Cabinet secretaries, and the heads of several research agencies. But it was President Donald Trump who took center stage at the 8 October event, billed as a “summit” that highlighted a July report on how to maintain U.S. global leadership in science and that featured Trump handing out his first national science medals.

“This is the golden age of science in America,” Trump told a small, invitation-only audience at the former Institute of Peace, which he dismantled last year and renamed the Donald J. Trump Institute of Peace. “It wasn’t [a golden age] 2 years ago, as some of you may know.”

Trump ticked off 2 centuries of U.S. innovation, from harnessing electricity to air transportation, atomic energy, and “superintelligence,” as well as inventing “almost everything else of genius and great importance anywhere in the world.” He then segued to his father’s brother, John Trump, an electrical engineer and longtime faculty member at the Massachusetts Institute of Technology.

“He was a great scientist,” Trump said about his uncle, who received the National Medal of Science posthumously from then-President Ronald Reagan in 1985 for work in developing radar and cancer radiation therapy. White House staffers then showed a 2-minute video extolling John Trump’s achievements and presented their boss with the medal, suitably framed, that Reagan had given to the electrical engineer’s  son.

Before that, Donald Trump himself handed out science and technology medals to Tesla and SpaceX founder Elon Musk, Google co-founder Sergey Brin, Nvidia CEO Jensen Huang, and AMD CEO Lisa Su—all major donors to the Republican Party. The medals were the first he has given in his two terms as president. By contrast, former President Joe Biden honored 20 researchers during his 4 years in office.

The symposium showcased a 72-page document that Trump’s science adviser, venture capitalist Michael Kratsios, has trumpeted as the most important statement on science since Science: The Endless Frontier , a 1945 report by Vannevar Bush that laid the foundation for the modern U.S. research enterprise. To discuss the report, called Science: A New Golden Age , Kratsios enlisted the likes of physicist John Martinis, who won the Nobel Prize in Physics last year, and NASA Administrator Jared Isaacman. Martinis now runs a company striving to meet Trump’s goal of building a functional quantum computer before he leaves office, while Isaacman has pledged to set up a habitable moon base by 2030 as a steppingstone to Mars.

In his short speech, Trump highlighted those projects and the multibillion-dollar Genesis Mission at the Department of Energy to double research productivity through high-performance computing and artificial intelligence (AI) as research landmarks since he assumed office in January 2025. But many U.S. scientists would probably take issue with another action Trump touted: “We ended DEI [diversity, equity, and inclusion] as a criterion for scientific grants from the federal government,” he said. That has meant canceling billions of dollars in programs to attract and retain students from racial and ethnic groups traditionally underrepresented in science and engineering.

DEI programs have also benefited women, a demographic notably scarce among participants at last week’s symposium. “Oh, so you noticed that, too,” electrical engineer Deirdre Hanford, president of Micron Research Laboratories, remarked to Science during the event. Hanford was the only woman among a dozen senior executives appearing together on stage whose companies are donating billions in computing time and hardware to researchers funded by the Genesis Mission. And the only woman who participated in a panel discussion was Anastasia Gamick, CEO of Convergent Research.

As in a previous interview with Science , Martinis declined to comment on Trump initiatives that have drawn strong criticism from many in the research community, including proposing rules that would install political appointees to oversee funding decisions. “I haven’t followed those issues,” he said.

But Martinis is unhappy with the president’s broader restrictive immigration policies, including a recently announced plan to charge $100,000 for non-U.S. citizens who want to stay in the country after earning advanced degrees in scientific and technical fields. “We’re not a big company, so we can’t afford that,” he told Science during the event. “But I would like to be able to hire whoever I want. So I’d like to talk to Kratsios about finding a middle ground for the most talented scientists.”

Many in the audience seemed to share Martinis’s views. The only speaker who drew applause midspeech was Carnegie Mellon University President Farnam Jahanian. “It is impossible to win the AI race without winning the race for talent,” he asserted. “And that may require rethinking our immigration policies.”