A major focus in Alzheimer’s research has been the buildup of plaque in the brain. This plaque consists of amyloid-β (Aβ), a peptide that forms naturally but can accumulate and cluster together over time. These deposits are widely believed to play a key role in driving the disease.

Targeting PTP1B To Improve Memory

Tonks, along with graduate student Yuxin Cen and postdoctoral fellow Steven Ribeiro Alves, has identified a new potential strategy. Their research shows that blocking a protein known as PTP1B can improve learning and memory in a mouse model of Alzheimer’s disease.

Tonks first discovered PTP1B in 1988 and has spent decades studying its role in health and disease. In this latest work, his team found that PTP1B interacts with another protein called spleen tyrosine kinase (SYK). SYK helps control microglia (the brain’s immune cells), which are responsible for clearing debris such as excess Aβ.

“Over the course of the disease, these cells become exhausted and less effective,” says Cen. “Our results suggest that PTP1B inhibition can improve microglial function, clearing up Aβ plaques.”

Links to Metabolism and Disease Risk

Alzheimer’s disease is also strongly associated with obesity and type 2 diabetes, both of which are recognized risk factors. These conditions are thought to contribute to the growing global burden of Alzheimer’s. Because PTP1B is already considered a therapeutic target for metabolic disorders, this connection strengthens the case for exploring it in Alzheimer’s treatment as well.

Toward More Effective Alzheimer’s Treatments

Current therapies for Alzheimer’s disease largely focus on reducing Aβ buildup, but their benefits are often limited for many patients. “Using PTP1B inhibitors that target multiple aspects of the pathology, including Aβ clearance, might provide an additional impact,” says Ribeiro Alves.

The Tonks lab is now collaborating with DepYmed, Inc. to develop PTP1B inhibitors for several medical applications. For Alzheimer’s disease, Tonks envisions combining these inhibitors with existing approved drugs. “The goal is to slow Alzheimer’s progression and improve quality of life of the patients,” he says. With PTP1B emerging as a promising target, this approach could help move closer to that goal.

Researchers note that while other types of exercise can still help, they should be used alongside aerobic activity rather than replacing it as the primary approach.

What Causes Knee Osteoarthritis

Osteoarthritis develops when the cartilage that cushions the ends of bones gradually breaks down. This leads to pain, swelling, and difficulty moving the joint. Although it can affect different parts of the body, the knees are especially vulnerable. Nearly 30% of adults over 45 show signs of knee osteoarthritis on x-rays, and about half of them experience severe symptoms.

Exercise is widely recommended as a key part of treatment, but current guidelines often lack clear advice on which specific types work best for knee osteoarthritis.

Large Review of 217 Clinical Trials

To address this gap, researchers carried out a comprehensive analysis of exercise therapies. They reviewed 217 randomized trials conducted between 1990 and 2024, involving a total of 15,684 participants. These studies compared several forms of exercise, including aerobic, flexibility, strengthening, mind-body, neuromotor, and mixed programs, against control groups.

Although the quality of the trials varied, the researchers used the recognized GRADE system to evaluate how reliable the evidence was.

Measuring Pain, Function, and Quality of Life

The analysis focused on key outcomes that matter most to patients. These included pain levels, physical function, gait performance, and quality of life. Results were examined at short term (four weeks), mid-term (12 weeks), and long term (24 weeks) follow up.

Across these measures, aerobic exercise consistently ranked as the most effective option among all exercise types studied.

Aerobic Exercise Shows the Strongest Results

Evidence with moderate certainty showed that aerobic exercise reduced pain in both the short and mid-term. It also improved physical function across short term, mid-term, and long term timeframes. In addition, it enhanced walking ability and quality of life in the short and mid-term.

Other forms of exercise also delivered meaningful benefits. Mind-body approaches likely improved short term function, while neuromotor training likely boosted short term gait performance. Strengthening and mixed exercise programs appeared to improve function at mid-term follow up.

Exercise Is Safe and Widely Beneficial

Importantly, none of the exercise types were linked to a higher risk of adverse events compared with control groups. This supports the idea that exercise is a safe and reliable treatment option for knee osteoarthritis.

The researchers did point out some limitations. Many of the comparisons were indirect, some outcomes lacked long term data, and smaller studies may have influenced early results.

Clear Guidance for Treatment

Despite these limitations, the study offers one of the most thorough and up-to-date evaluations of exercise for knee osteoarthritis. The findings can help guide clinicians in choosing the most effective therapies for their patients.

Based on the results, the researchers recommend aerobic exercise “as a first line intervention for knee osteoarthritis management, particularly when the aim is to improve functional capacity and reduce pain.” They also note that if aerobic exercise is not suitable due to individual limitations, “alternative forms of structured physical activity may still be beneficial.”

In July 2025, a study published in Nature introduced an AI model called “Centaur.” Built on standard large language models and refined using data from psychological experiments, Centaur was designed to simulate human cognitive behavior. It reportedly performed well across 160 tasks, including decision-making, executive control, and other mental processes. The results drew widespread attention and were seen as a possible step toward AI systems that could replicate human thinking more broadly.

New Research Raises Doubts

A more recent study published in National Science Open challenges those claims. Researchers from Zhejiang University argue that Centaur’s apparent success may come from overfitting. In other words, instead of understanding the tasks, the model may have learned to recognize patterns in the training data and reproduce expected answers.

To test this idea, the researchers created several new evaluation scenarios. In one example, they replaced the original multiple-choice prompts, which described specific psychological tasks, with the instruction “Please choose option A.” If the model truly understood the task, it should have consistently selected option A. Instead, Centaur continued to choose the “correct answers” from the original dataset.

This behavior suggests that the model was not interpreting the meaning of the questions. Rather, it relied on learned statistical patterns to “guess” answers. The researchers compared this to a student who scores well by memorizing test formats without actually understanding the material.

Why This Matters for AI Evaluation

The findings highlight the need for caution when assessing the abilities of large language models. While these systems can be highly effective at fitting data, their “black-box” nature makes it difficult to know how they arrive at their outputs. This can lead to issues such as hallucinations or misinterpretations. Careful and varied testing is essential to determine whether a model truly has the skills it appears to demonstrate.

The Real Challenge: Language Understanding

Although Centaur was presented as a model capable of simulating cognition, its biggest limitation appears to be in language comprehension. Specifically, it struggles to recognize and respond to the intent behind questions. The study suggests that achieving true language understanding may be one of the most important challenges in developing AI systems that can model human cognition more fully.

Subduction zones shape the planet in dramatic ways. They move continents, trigger major earthquakes and volcanic eruptions, and pull old crust deep into Earth’s mantle. Yet despite their immense power, they do not last forever.

Why Subduction Zones Eventually Fail

If subduction zones continued indefinitely, continents would keep piling up, oceans would disappear, and much of Earth’s geological history would be erased. Scientists have long wondered what causes these systems to shut down.

“Getting a subduction zone started is like trying to push a train uphill — it takes a huge effort,” said Brandon Shuck, an assistant professor at Louisiana State University and lead author of the study. “But once it’s moving, it’s like the train is racing downhill, impossible to stop. Ending it requires something dramatic — basically, a train wreck.” Shuck conducted the research while he was a postdoctoral research fellow at the Lamont-Doherty Earth Observatory, which is part of the Columbia Climate School.

Cascadia Reveals a Tectonic Plate Tearing Apart

The answer appears to lie off the coast of Vancouver Island, in the Cascadia region. Here, the Juan de Fuca and Explorer plates are slowly sliding beneath the North American plate. Using advanced imaging techniques and earthquake data, scientists have now seen this subduction zone starting to come apart.

The team relied on seismic reflection imaging, which works much like an ultrasound of the Earth’s interior, combined with detailed records of earthquakes. Together, these tools revealed a plate that is not simply sinking, but actively tearing.

Inside the 2021 Seismic Imaging Experiment

The data came from the 2021 Cascadia Seismic Imaging Experiment (CASIE21), conducted aboard the research vessel Marcus G. Langseth. Led by Lamont scientist Suzanne Carbotte, with co-author Anne Bécel, the team sent sound waves into the seafloor and captured their echoes using a 15-kilometer-long array of underwater sensors.

This method produced highly detailed images of faults and fractures deep beneath the ocean floor. Those images clearly show sections of the plate breaking apart.

“This is the first time we have a clear picture of a subduction zone caught in the act of dying,” said Shuck. “Rather than shutting down all at once, the plate is ripping apart piece by piece, creating smaller microplates and new boundaries. So instead of a big train wreck, it’s like watching a train slowly derail, one car at a time.”

Carbotte noted that scientists have long known that subduction can slow or stall when lighter parts of a plate reach the boundary. “But we haven’t previously had such a clear picture of the process in action,” she says. “These new findings help us better understand the life cycle of the tectonic plates that shape the Earth.”

Massive Faults and Silent Gaps

Researchers identified several large tears cutting through the Juan de Fuca plate, including one major fault where the plate has dropped by about five kilometers. “There’s a very large fault that’s actively breaking the [subducting] plate,” Shuck explained. “It’s not 100% torn off yet, but it’s close.”

Earthquake data supports this picture. Along a 75-kilometer-long tear, some areas are still producing earthquakes while others are unusually quiet. “Once a piece has completely broken off, it no longer produces earthquakes because the rocks aren’t stuck together anymore,” he said. These quiet gaps suggest that parts of the plate have already separated and that the break is gradually expanding.

A Slow, Piece-by-Piece Breakdown

The study shows that subduction zones do not fail all at once. Instead, they shut down through a process known as “episodic” or “piecewise” termination. The plate tears apart in stages, with different sections breaking off over time.

As smaller pieces detach, the larger plate loses the force pulling it downward. Over millions of years, this gradual loss of momentum can bring the entire subduction system to a stop.

Clues to Earth’s Geological Past

This step-by-step breakup helps explain puzzling features seen elsewhere on Earth. In some regions, scientists have found fragments of old tectonic plates and bursts of volcanic activity that did not fully make sense before.

One example lies off Baja California, where remnants of the ancient Farallon plate remain as fossil microplates. For years, researchers suspected these fragments were linked to dying subduction zones, but the exact process was unclear. The new observations from Cascadia suggest that these ancient plates likely broke apart in the same gradual way.

What This Means for Earthquakes in Cascadia

Scientists are now investigating how these newly discovered tears might influence future earthquakes. One key question is whether a major rupture could travel across these breaks or if the fractures might change how seismic energy spreads.

For now, the findings do not significantly alter the overall risk in the Cascadia region. The area is still capable of producing very large earthquakes and tsunamis. However, incorporating these new details into models will improve how researchers understand and simulate seismic hazards in the Pacific Northwest.

The CASIE21 project is supported by the National Science Foundation under awards OCE 1827452 and OCE 2217465.