Dear Aventine Readers,

The polar ice caps are melting fast enough to deliver an ice-free Arctic before 2030, an eventuality that would forever change the earth as we know it. Scientists are divided over what to do: Try out geoengineering alternatives to preserve the ice while we still can, or push harder for CO₂ reductions to slow global warming? This week we speak to five polar ice experts who have starkly opposing opinions on the best course of action.    

Also in this issue: 

Until next week! 

Danielle Mattoon 
Executive Director, Aventine

Subscribe

Subscribe to our newsletter and be kept up to date on upcoming Aventine projects

Since 1979, the Arctic has been warming four times faster than the rest of the world. To put that in starker terms, the first ice-free day in the Arctic Ocean could occur before 2030. In worst-case scenarios in which the polar ice sheets shrink rapidly, estimates suggest that the planet could experience sea-level rises of up to 5 meters by 2150, causing global flooding that could displace hundreds of millions of people. In parts of the Arctic, some rivers are starting to run orange, as permafrost thaws and releases iron into waterways. 

Reducing carbon emissions is the default long-term strategy to bring global temperatures down and reduce ice melt. But CO₂ emissions are expected to hit a record high of 38 gigatons in 2025 — up 1.1 percent from 2024 — so that plan might not work on its own. This reality has caused a rift in the scientific community over whether to pursue geoengineering as a possible Plan B. 

There are a range of proposals for what this might look like but the most frequently discussed include: targeting the atmosphere by injecting aerosols into the stratosphere to reflect sunlight back into space; dumping iron dust into oceans to provoke phytoplankton blooms that absorb CO₂; slowing the flows of water that accelerate the erosion of ice sheets; building huge underwater curtains to prevent warm ocean currents from melting polar ice from below; and drilling holes through ice sheets and pumping up water from beneath so that it freezes and thickens the ice mass. 

None of these options have been tested and all are controversial. Last fall, a pair of papers published in the journal Frontiers in Science laid out the two prevailing points of view. The first was an analysis by more than 40 researchers of the five approaches described above that found them all to be infeasible and dangerous. The other was a commentary by three researchers who argued that it’s essential to understand the potential impact of these approaches if we want a chance at preserving polar ice.

Opponents of polar geoengineering argue that none of the approaches have been shown to work, and that even if they did work and were deployed at scale, they would be hard to govern and could lead to unintended consequences. Finally, opponents argue, the very discussion of such approaches distracts from the mission of decarbonization. Proponents contend that much of the world has been decarbonizing for decades and the effort has failed to keep up with increased CO₂ emissions, that failing to explore all available options is itself risky and that it should be possible to carry out safe proof-of-concept experiments. The debate also raises significant questions around governance and ethics, including who gets to decide whether and how to deliberately alter the Earth's climate systems. 

To get a handle on the debate, Aventine spoke with experts in polar science and geoengineering, including authors from the Frontiers papers. Their comments have been edited for clarity and length.

China is all-in on Alzheimer's research. China is confronting a looming Alzheimer’s crisis: Nearly 17 million people in the country are living with the disease today, a number that is predicted to climb to as many as 100 million by 2050 as the population ages. In response, Nature reports, the government has dramatically expanded funding for Alzheimer’s research and launched a national push to battle the disease. Several experimental treatments are already in development. One, called BrAD-R13, mimics a protein involved in brain growth and repair; another, derived from Chinese celery and known as DL-3-n-butylphthalide (NBP), builds on ingredients long used in traditional Chinese medicine. Researchers have also explored a controversial surgical technique known as cervical shunting, designed to improve the brain’s waste clearance system. Early results were promising, but the procedure was banned for routine use in 2025 after concerns about premature clinical adoption. At the same time, China is investing heavily in early detection. One effort, built on a huge national genetic database, has produced a blood test that Chinese researchers say can detect Alzheimer’s disease with up to 96 percent accuracy, and may potentially identify patients more than a decade before symptoms appear. China’s Alzheimer’s ecosystem still lags the US in terms of funding and maturity. But the gap is narrowing quickly, with research output rising faster than in the US, signaling that it could  become an increasingly important player in the global effort to understand and treat the disease.

Digital twins are helping people manage diabetes and obesity. A startup called Twin Health has built custom AI models that can be used to help coach people toward healthier lifestyles. Trained on individual patient data, including blood sugar, weight, sleep, stress, activity and blood pressure, the models simulate how a person’s metabolism responds to food, exercise and daily habits, allowing the system to generate highly personalized lifestyle recommendations. Wired reports that users can scan food labels or photograph meals and the app will predict how their blood sugar will respond. It can then suggest alternatives — recommending different foods, adjusting meal timing or proposing the best time to exercise. Over time, the system learns individual preferences, tailoring advice to behaviors and preferences. In a Cleveland Clinic study of patients with Type 2 diabetes who used the technology, 71 percent reached their target blood sugar levels while reducing medication use, compared with just 2.4 percent in a control group. The approach could prove particularly appealing to employers and insurers as an alternative to expensive GLP-1 weight-loss drugs. Twin Health’s business model reflects that: It charges customers only when users achieve measurable clinical outcomes, such as improved glucose control or weight loss. If the results hold up at a larger scale, digital metabolic twins could help treat some chronic diseases through behavioral intervention instead of pharmaceuticals.

AI might actually be intensifying work. Researchers at the University of California, Berkeley, have been studying how AI is being used inside a US tech company and found that workers who adopted the tools reported working longer hours and at greater intensity. The findings, described in Harvard Business Review, are preliminary and based on a study of about 200 employees. The researchers identified three main drivers of the increased hours and intensity. First, AI helps workers fill knowledge gaps, allowing them to take on a broader range of tasks. Second, it lowers the friction involved in starting work, making it easier to squeeze in additional tasks at the beginning, end or margins of the day. And third, it enables greater multitasking, with workers juggling more parallel streams of activity. The result is a feedback loop: “AI accelerated certain tasks, which raised expectations for speed,” the researchers wrote. “Higher speed made workers more reliant on AI. Increased reliance widened the scope of what workers attempted, and a wider scope further expanded the quantity and density of work.” The company had not mandated AI use, suggesting the shift emerged organically rather than through top-down pressure. The study is small and ongoing, but the pattern may sound familiar to economists. Technologies that make work more efficient often increase expectations — and with them, the volume, pace and intensity of work itself.

An entirely new kind of flu is being monitored as a potential pandemic threat. Meet the latest influenza strain worrying scientists: Flu D. First identified in pigs in 2011, the virus has since been found widely in cattle populations around the world. As Science reports, Flu D shares several concerning traits with influenza A, the family responsible for human pandemics and H5N1: It infects multiple species, circulates globally and can undergo “reassortment,” the technical term for its ability to swap genetic material between strains to create new and potentially more dangerous variants. Although cattle appear to be its primary host, antibodies to Flu D have also been detected in sheep, camels, horses, dogs and cats. Farm workers, too, show signs of exposure. One recent preprint study found that the virus can readily infect human airway cells in laboratory experiments, which could be a key step toward possible human transmission. So far, however, researchers have not isolated a fully intact Flu D virus from a human. Some studies from China suggest antibodies may already be widespread in people, but those findings remain controversial, and Western researchers have urged caution in interpreting the data. The biggest concern may be how little attention Flu D has received. Compared with other influenza strains, it remains poorly studied, leaving major gaps in understanding its true prevalence and pandemic potential. For now, it is less an immediate threat than something to observe in case it evolves into something more dangerous.

How to deter biothreats in the age of gene synthesis, from Big Think
3,400 words, or about 14 minutes

Advances in gene synthesis have made it easier than ever to design strands of DNA on a computer to be manufactured by specialist biotech firms. That holds huge promise for medicine, but also creates new risks: In theory, the same tools could be used to create dangerous pathogens, potentially even designed with the help of AI. This article argues that managing that risk will require applying the “Swiss cheese model” of security — layering multiple imperfect safeguards so that together they form a robust defense. DNA synthesis companies can vet both customers and the genetic sequences they are asked to produce, flagging requests linked to known pathogens or suspicious actors. Governments, in turn, can collaborate to maintain shared databases of high-risk sequences and actors. And companies that manufacture gene synthesis machines could screen buyers of the equipment before selling them, limiting access to trusted institutions. Meanwhile, researchers should develop advanced screening techniques capable of detecting entirely novel or AI-generated DNA sequences that might evade existing safeguards. The piece argues that these measures are not sufficient on their own and that traditional public-health defenses — disease surveillance systems, resilient healthcare infrastructure, rapid diagnostics and vaccine manufacturing capacity — remain essential backstops.

Wanna Bet?, from Bloomberg Businessweek
6,000 words, or about 24 minutes

What will the highest temperature in Miami be a month from now? What exactly will Jerome Powell say in his next speech? When will the Second Coming occur? On platforms like Kalshi and Polymarket, you could place a wager on all of it. These so-called prediction markets allow users to buy and sell contracts tied to real-world outcomes, effectively turning events in geopolitics, economics, theology and whatever else you care to think about into speculative assets. Supporters argue that such markets generate valuable information: By putting money behind forecasts, they claim, participants reveal collective beliefs about the future more accurately than polls or pundits. Critics see them as a not-very-lightly disguised form of gambling. Whatever you think of them, the growth has been dramatic. In just a few years, Kalshi and Polymarket have evolved from niche curiosities into multibillion-dollar platforms, valued at around $11 billion and $8 billion respectively, and they are attracting interest from major banks, including Goldman Sachs. This Bloomberg Businessweek feature explores what fueled that ascent and what comes next. Lawsuits, looming regulatory battles and insider-trading concerns — including over $330,000 in profit shared by 12 suspicious accounts over predictions about US air strikes on Iran — will test whether prediction markets are the future of information discovery or a form of gambling to be cracked down on.

Child’s Play, from Harper’s Magazine
9,300 words, or about 37 minutes

In San Francisco, the concept of “agency” has become a central virtue: the ability of some humans to bend the world to their will through sheer force of just getting stuff done, in order to triumph in a world being rewritten by AI. This essay by Sam Kriss pointedly skewers this ethos through the lens of Roy Lee, the “simultaneously baby-faced and creatine-swollen” cofounder of Cluely, a startup whose buggy software overlays AI assistance onto Zoom calls. The technology grew out of tools Lee developed and used while a student at Columbia University (before dropping out) to cheat on job interviews. On paper, he embodies the founder stereotype: relentlessly driven, highly productive and highly skilled at capturing attention. But Kriss underscores how crippling this persona — fully realized — can be. In addition to creating a piece of software that doesn’t work very well, Lee barely seems to engage with culture or intellectual life. His stated goals — “to hang out with friends, to do something meaningful, and to go on lots of dates,” — seem likely beyond his reach, turning his quest to optimize his own trajectory within the startup ecosystem into a sad joke.