MiniMax has introduced MiniMax-M1, a new open-weight reasoning model built to handle extended contexts and complex problem-solving with high efficiency. Built on top of the earlier MiniMax-Text-01, M1 features a hybrid Mixture-of-Experts (MoE) architecture and a novel “lightning attention” mechanism.

By Robert Krzaczyński

Midjourney has launched its first video generation V1 model, a web-based tool that allows users to animate still images into 5-second video clips.

By Daniel Dominguez

Chalmers engineers built a pulse-driven qubit amplifier that’s ten times more efficient, stays cool, and safeguards quantum states—key for bigger, better quantum machines.

What if your old chest scans—taken years ago for something unrelated—held a secret warning about your heart? A new AI tool called AI-CAC, developed by Mass General Brigham and the VA, can now comb through routine CT scans to detect hidden signs of heart disease before symptoms strike.

Nature, Published online: 24 June 2025; doi:10.1038/d41586-025-01717-5

A gender gap in mathematical ability arises shortly after children begin school — irrespective of the type of school they attend and their socio-economic background.

Hundred-year storm tides could strike every decade in Bangladesh

Tropical cyclones can generate devastating storm tides—seawater heightened by the tides that causes catastrophic floods in coastal regions. An MIT study finds that as the planet warms, the recurrence of destructive storm tides will increase tenfold for one of the world’s hardest-hit regions.

New electronic “skin” could lead to lightweight night-vision glasses 

MIT engineers have developed a technique to grow and peel ultrathin “skins” of electronic material that could be used in wearable sensors, flexible transistors and computing elements, and sensitive compact imaging devices.

Technology makes pesticides stick to plant leaves

A new pesticide application system developed by MIT researchers and their spinoff company could significantly cut use of pesticides and fertilizers, saving farmers money and reducing polluting runoff.

New printable metamaterials that are both strong and stretchy could allow fabrication of bendable ceramics, glass, and metals.

These tough yet bendy materials could be made into tear-resistant textiles, flexible semiconductors, electronic chip packaging, and durable yet compliant scaffolds on which to grow cells for tissue repair.

Over the past decade, Institute Professor Paula Hammond ’84, PhD ’93, and her students have used a technique known as layer-by-layer assembly to create a variety of polymer-coated nanoparticles that can be loaded with cancer-fighting drugs. The particles, which could prevent many side effects of chemotherapy by targeting tumors directly, have proved effective in mouse studies. Now the researchers have come up with a technique that allows them to manufacture many more particles in much less time, moving them closer to human use.

“There’s a lot of promise with the nanoparticle systems we’ve been developing, and we’ve been really excited more recently with the successes that we’ve been seeing in animal models for our treatments for ovarian cancer in particular,” says Hammond, the senior author of a paper on the new technique along with Darrell Irvine, a professor at the Scripps Research Institute.

In the original production technique, layers with different properties can be laid down by alternately exposing a particle to positively and negatively charged polymers, with extensive purification to remove excess polymer after each application. Each layer can carry therapeutics as well as molecules that help the particles find and enter cancer cells. But the process is time-consuming and would be difficult to scale up. 

In the new work, the researchers used a microfluidic mixing device that allows them to sequentially add layers as the particles flow through a microchannel. For each layer, they can calculate exactly how much polymer is needed, which eliminates the slow and costly purification step and saves significantly on material costs.

This strategy also facilitates compliance with the FDA’s GMP (good manufacturing practice) requirements, which ensure that products meet safety standards and can be manufactured consistently. “There’s much less chance of any sort of operator mistake or mishaps,” says Ivan Pires, PhD ’24, a postdoc at Brigham and Women’s Hospital and a visiting scientist at the Koch Institute, who is the paper’s lead author along with Ezra Gordon ’24. “We can create an innovation within the layer-by-layer nanoparticles and quickly produce it in a manner that we could go into clinical trials with.”

In minutes, the researchers can generate 15 milligrams of nanoparticles (enough for about 50 doses for certain cargos), which would have taken close to an hour with the original process. They say this means it would be realistic to produce more than enough for clinical trials and patient use.

To demonstrate the technique, the researchers created layered nanoparticles loaded with the immune molecule ­interleukin-­12; they have previously shown that such particles can slow growth of ovarian tumors in mice. Those manufactured using the new technique performed similarly to the originals and managed to bind to cancer tissue without entering the cancer cells. This lets them serve as markers that activate the immune system in the tumor, which can delay tumor growth and even lead to cures in mouse models of ovarian cancer. 

The researchers have filed for a patent and are working with MIT’s Deshpande Center for Technological Innovation in hopes of forming a company to commercialize the technology, which they say could also be applied to glioblastoma and other types of cancer. 

Google DeepMind has released a new language model called Gemini Robotics On-Device that can run tasks locally on robots without an internet connection.

Security researchers at Wiz have uncovered a sophisticated crypto-jacking attack targeting publically accessible API servers for several popular DevOps tools. Similarly, researchers at Sysdig have uncovered an attack on the popular AI tool Open WebUI using many of the same techniques and crypto-miners.

By Matt Saunders

Despite widespread fears, early research suggests AI might actually be improving some aspects of work life. A major new study examining 20 years of worker data in Germany found no signs that AI exposure is hurting job satisfaction or mental health. In fact, there s evidence that it may be subtly improving physical health especially for workers without college degrees by reducing physically demanding tasks. However, researchers caution that it s still early days.

In a remarkable twist of science, researchers have transformed a fungus long associated with death into a potential weapon against cancer. Found in tombs like that of King Tut, Aspergillus flavus was once feared for its deadly spores. Now, scientists at Penn and several partner institutions have extracted a new class of molecules from it—called asperigimycins—that show powerful effects against leukemia cells. These compounds, part of a rare group known as fungal RiPPs, were bioengineered for potency and appear to disrupt cancer cell division with high specificity.

Nature, Published online: 23 June 2025; doi:10.1038/d41586-025-01946-8

The uncontrolled deployment of machine learning in medicine can distort patient information and sacrifice long-term data reliability for short-term benefits.

When testing an AI model, it’s hard to tell if it is reasoning or just regurgitating answers from its training data. Xbench, a new benchmark developed by the Chinese venture capital firm HSG, or HongShan Capital Group, might help to sidestep that issue. That’s thanks to the way it evaluates models not only on the ability to pass arbitrary tests, like most other benchmarks, but also on the ability to execute real-world tasks, which is more unusual. It will be updated on a regular basis to try to keep it evergreen. 

This week the company is making part of its question set open-source and letting anyone use for free. The team has also released a leaderboard comparing how mainstream AI models stack up when tested on Xbench. (ChatGPT o3 ranked first across all categories, though ByteDance’s Doubao, Gemini 2.5 Pro, and Grok all still did pretty well, as did Claude Sonnet.) 

Development of the benchmark at HongShan began in 2022, following ChatGPT’s breakout success, as an internal tool for assessing which models are worth investing in. Since then, led by partner Gong Yuan, the team has steadily expanded the system, bringing in outside researchers and professionals to help refine it. As the project grew more sophisticated, they decided to release it to the public.

Xbench approached the problem with two different systems. One is similar to traditional benchmarking: an academic test that gauges a model’s aptitude on various subjects. The other is more like a technical interview round for a job, assessing how much real-world economic value a model might deliver.

Xbench’s methods for assessing raw intelligence currently include two components: Xbench-ScienceQA and Xbench-DeepResearch. ScienceQA isn’t a radical departure from existing postgraduate-level STEM benchmarks like GPQA and SuperGPQA. It includes questions spanning fields from biochemistry to orbital mechanics, drafted by graduate students and double-checked by professors. Scoring rewards not only the right answer but also the reasoning chain that leads to it.

DeepResearch, by contrast, focuses on a model’s ability to navigate the Chinese-language web. Ten subject-matter experts created 100 questions in music, history, finance, and literature—questions that can’t just be googled but require significant research to answer. Scoring favors breadth of sources, factual consistency, and a model’s willingness to admit when there isn’t enough data. A question in the publicized collection is “How many Chinese cities in the three northwestern provinces border a foreign country?” (It’s 12, and only 33% of models tested got it right, if you are wondering.)

On the company’s website, the researchers said they want to add more dimensions to the test—for example, aspects like how creative a model is in its problem solving, how collaborative it is when working with other models, and how reliable it is.

The team has committed to updating the test questions once a quarter and to maintain a half-public, half-private data set.

To assess models’ real-world readiness, the team worked with experts to develop tasks modeled on actual workflows, initially in recruitment and marketing. For example, one task asks a model to source five qualified battery engineer candidates and justify each pick. Another asks it to match advertisers with appropriate short-video creators from a pool of over 800 influencers.

The website also teases upcoming categories, including finance, legal, accounting, and design. The question sets for these categories have not yet been open-sourced.

ChatGPT-o3 again ranks first in both of the current professional categories. For recruiting, Perplexity Search and Claude 3.5 Sonnet take second and third place, respectively. For marketing, Claude, Grok, and Gemini all perform well.

“It is really difficult for benchmarks to include things that are so hard to quantify,” says Zihan Zheng, the lead researcher on a new benchmark called LiveCodeBench Pro and a student at NYU. “But Xbench represents a promising start.”

Around the world, countries are facing the challenges of aging populations, growing rates of chronic disease, and workforce shortages, leading to a growing burden on health care systems. From diagnosis to treatment, AI and other digital solutions can enhance the efficiency and effectiveness of health care, easing the burden on straining systems. According to the World Health Organization (WHO), spending an additional $0.24 per patient per year on digital health interventions could save more than two million lives from non-communicable diseases over the next decade.

To work most effectively, digital solutions need to be scaled and embedded in an ecosystem that ensures a high degree of interoperability, data security, and governance. If not, the proliferation of point solutions— where specialized software or tools focus on just one specific area or function—could lead to silos and digital canyons, complicating rather than easing the workloads of health care professionals, and potentially impacting patient treatment. Importantly, technologies that enhance workforce productivity should keep humans in the loop, aiming to augment their capabilities, rather than replace them. 

Through a survey of 300 health care executives and a program of interviews with industry experts, startup leaders, and academic researchers, this report explores the best practices for success when implementing integrated digital solutions into health care, and how these can support decision-makers in a range of settings, including laboratories and hospitals. 

Key findings include: 

Health care is primed for digital adoption. The global pandemic underscored the benefits of value-based care and accelerated the adoption of digital and AI-powered technologies in health care. Overwhelmingly, 96% of the survey respondents say they are “ready and resourced” to use digital health, while one in four say they are “very ready.” However, 91% of executives agree interoperability is a challenge, with a majority (59%) saying it will be “tough” to solve. Two in five leaders say balancing security with usability is the biggest challenge for digital health. With the adoption of cloud solutions, organizations can enjoy the benefits of modernized IT infrastructure: 36% of the survey respondents believe scalability is the main benefit, followed by improved security (28%). 

Digital health care can help health care institutions transform patient outcomes—if built on the right foundations. Solutions like AI-powered diagnostics, telemedicine, and remote monitoring can offer measurable impact across the patient journey, from improving early disease detection to reducing hospital readmission rates. However, these technologies can only support fully connected health care when scaled up and embedded in ecosystems with robust data governance, interoperability, and security. 

Health care data has immense potential—but fragmentation and poor interoperability hinder impact. Health care systems generate vast quantities of data, yet much of it remains siloed or unusable due to inconsistent formats and incompatible IT systems, limiting scalability. 

Digital tools must augment, not overload, the workforce. With global health care workforce shortages worsening, digital solutions like clinical decision support tools, patient prediction, and remote monitoring can be seen as essential aids rather than threats to the workforce. Successful deployment depends on usability, clinician engagement, and training. 

Regulatory evolution, open data policies, and economic sustainability are key to scaling digital health. Even the best digital tools struggle to scale without reimbursement frameworks, regulatory support, and viable business models. Open data ecosystems are needed to unleash the clinical and economic value of innovation. Regulatory and reimbursement innovation is also critical to transitioning from pilot projects to high-impact, system-wide adoption.

Download the full report.

This content was produced by Insights, the custom content arm of MIT Technology Review. It was not written by MIT Technology Review’s editorial staff.

This content was researched, designed, and written entirely by human writers, editors, analysts, and illustrators. This includes the writing of surveys and collection of data for surveys. AI tools that may have been used were limited to secondary production processes that passed thorough human review.

Early-stage founders, listen up! You will want a front row seat at the Builders Stage on October 27 at 1:00 p.m. PT. This session at TechCrunch Disrupt 2025 brings together Nina Achadjian, partner, Index Ventures; Jerry Chen, general partner, Greylock; and Viviana Faga, general partner, Felicis, all of whom will share their 2026 investment priorities […]

Privacy only matters to those with something to hide. So goes one of the more inane and disingenuous justifications for mass government and corporate surveillance. There are others, of course, but the “nothing to hide” argument remains a popular way to rationalize or excuse what’s become standard practice in our digital age: the widespread and invasive collection of vast amounts of personal data.

One common response to this line of reasoning is that everyone, in fact, has something to hide, whether they realize it or not. If you’re unsure of whether this holds true for you, I encourage you to read Means of Control by Byron Tau. 

Midway through his book, Tau, an investigative journalist, recalls meeting with a disgruntled former employee of a data broker—a shady company that collects, bundles, and sells your personal data to other (often shadier) third parties, including the government. This ex-employee had managed to make off with several gigabytes of location data representing the precise movements of tens of thousands of people over the course of a few weeks. “What could I learn with this [data]—­theoretically?” Tau asks the former employee. The answer includes a laundry list of possibilities that I suspect would make even the most enthusiastic oversharer uncomfortable.

“If information is power, and America is a society that’s still interested in the guarantee of liberty, personal dignity, and the individual freedom of its citizens, a serious conversation is needed.”

Bryon Tau, author of Means of Control

Did someone in this group recently visit an abortion clinic? That would be easy to figure out, says the ex-employee. Anyone attend an AA meeting or check into inpatient drug rehab? Again, pretty simple to discern. Is someone being treated for erectile dysfunction at a sexual health clinic? If so, that would probably be gleanable from the data too. Tau never opts to go down that road, but as Means of Control makes very clear, others certainly have done so and will.

While most of us are at least vaguely aware that our phones and apps are a vector for data collection and tracking, both the way in which this is accomplished and the extent to which it happens often remain murky. Purposely so, argues Tau. In fact, one of the great myths Means of Control takes aim at is the very idea that what we do with our devices can ever truly be anonymized. Each of us has habits and routines that are completely unique, he says, and if an advertiser knows you only as an alphanumeric string provided by your phone as you move about the world, and not by your real name, that still offers you virtually no real privacy protection. (You’ll perhaps not be surprised to learn that such “anonymized ad IDs” are relatively easy to crack.)

“I’m here to tell you if you’ve ever been on a dating app that wanted your location, or if you ever granted a weather app permission to know where you are 24/7, there’s a good chance a detailed log of your precise movement patterns has been vacuumed up and saved in some data bank somewhere that tens of thousands of total strangers have access to,” writes Tau.

Unraveling the story of how these strangers—everyone from government intelligence agents and local law enforcement officers to private investigators and employees of ad tech companies—gained access to our personal information is the ambitious task Tau sets for himself, and he begins where you might expect: the immediate aftermath of 9/11.

At no other point in US history was the government’s appetite for data more voracious than in the days after the attacks, says Tau. It was a hunger that just so happened to coincide with the advent of new technologies, devices, and platforms that excelled at harvesting and serving up personal information that had zero legal privacy protections. 

Over the course of 22 chapters, Tau gives readers a rare glimpse inside the shadowy industry, “built by corporate America and blessed by government lawyers,” that emerged in the years and decades following the 9/11 attacks. In the hands of a less skilled reporter, this labyrinthine world of shell companies, data vendors, and intelligence agencies could easily become overwhelming or incomprehensible. But Tau goes to great lengths to connect dots and plots, explaining how a perfect storm of business motivations, technological breakthroughs, government paranoia, and lax or nonexistent privacy laws combined to produce the “digital panopticon” we are all now living in.

Means of Control doesn’t offer much comfort or reassurance for privacy­-minded readers, but that’s arguably the point. As Tau notes repeatedly throughout his book, this now massive system of persistent and ubiquitous surveillance works only because the public is largely unaware of it. “If information is power, and America is a society that’s still interested in the guarantee of liberty, personal dignity, and the individual freedom of its citizens, a serious conversation is needed,” he writes. 

As another new book makes clear, this conversation also needs to include student data. Lindsay Weinberg’s Smart University: Student Surveillance in the Digital Age reveals how the motivations and interests of Big Tech are transforming higher education in ways that are increasingly detrimental to student privacy and, arguably, education as a whole.

By “smart university,” Weinberg means the growing number of public universities across the country that are being restructured around “the production and capture of digital data.” Similar in vision and application to so-called “smart cities,” these big-data-pilled institutions are increasingly turning to technologies that can track students’ movements around campus, monitor how much time they spend on learning management systems, flag those who seem to need special “advising,” and “nudge” others toward specific courses and majors. “What makes these digital technologies so seductive to higher education administrators, in addition to promises of cost cutting, individualized student services, and improved school rankings, is the notion that the integration of digital technology on their campuses will position universities to keep pace with technological innovation,” Weinberg writes. 

Readers of Smart University will likely recognize a familiar logic at play here. Driving many of these academic tracking and data-gathering initiatives is a growing obsession with efficiency, productivity, and convenience. The result is a kind of Silicon Valley optimization mindset, but applied to higher education at scale. Get students in and out of university as fast as possible, minimize attrition, relentlessly track performance, and do it all under the guise of campus modernization and increased personalization. 

Under this emerging system, students are viewed less as self-empowered individuals and more as “consumers to be courted, future workers to be made employable for increasingly smart workplaces, sources of user-generated content for marketing and outreach, and resources to be mined for making campuses even smarter,” writes Weinberg. 

At the heart of Smart University seems to be a relatively straightforward question: What is an education for? Although Weinberg doesn’t provide a direct answer, she shows that how a university (or society) decides to answer that question can have profound impacts on how it treats its students and teachers. Indeed, as the goal of education becomes less to produce well-rounded humans capable of thinking critically and more to produce “data subjects capable of being managed and who can fill roles in the digital economy,” it’s no wonder we’re increasingly turning to the dumb idea of smart universities to get the job done.  

If books like Means of Control and Smart University do an excellent job exposing the extent to which our privacy has been compromised, commodified, and weaponized (which they undoubtedly do), they can also start to feel a bit predictable in their final chapters. Familiar codas include calls for collective action, buttressed by a hopeful anecdote or two detailing previously successful pro-privacy wins; nods toward a bipartisan privacy bill in the works or other pieces of legislation that could potentially close some glaring surveillance loophole; and, most often, technical guides that explain how each of us, individually, might better secure or otherwise take control and “ownership” of our personal data.

The motivations behind these exhortations and privacy-centric how-to guides are understandable. After all, it’s natural for readers to want answers, advice, or at least some suggestion that things could be different—especially after reading about the growing list of degradations suffered under surveillance capitalism. But it doesn’t take a skeptic to start to wonder if they’re actually advancing the fight for privacy in the way that its advocates truly want.

For one thing, technology tends to move much faster than any one smartphone privacy guide or individual law could ever hope to keep up with. Similarly, framing rampant privacy abuses as a problem we each have to be responsible for addressing individually seems a lot like framing the plastic pollution crisis as something Americans could have somehow solved by recycling. It’s both a misdirection and a misunderstanding of the problem.     

It’s to his credit, then, that Lowry Pressly doesn’t include a “What is to be done” section at the end of The Right to Oblivion: Privacy and the Good Life. In lieu of offering up any concrete technical or political solutions, he simply reiterates an argument he has carefully and convincingly built over the course of his book: that privacy is important “not because it empowers us to exercise control over our information, but because it protects against the creation of such information in the first place.” 

For Pressly, a Stanford instructor, the way we currently understand and value privacy has been tainted by what he calls “the ideology of information.” “This is the idea that information has a natural existence in human affairs,” he writes, “and that there are no aspects of human life which cannot be translated somehow into data.” This way of thinking not only leads to an impoverished sense of our own humanity—it also forces us into the conceptual trap of debating privacy’s value using a framework (control, consent, access) established by the companies whose business model is to exploit it.

The way out of this trap is to embrace what Pressly calls “oblivion,” a kind of state of unknowing, ambiguity, and potential—or, as he puts it, a realm “where there is no information or knowledge one way or the other.” While he understands that it’s impossible to fully escape a modern world intent on turning us into data subjects, Pressly’s book suggests we can and should support the idea that certain aspects of our (and others’) subjective interior lives can never be captured by information. Privacy is important because it helps to both protect and produce these ineffable parts of our lives, which in turn gives them a sense of dignity, depth, and the possibility for change and surprise. 

Reserving or cultivating a space for oblivion in our own lives means resisting the logic that drives much of the modern world. Our inclination to “join the conversation,” share our thoughts, and do whatever it is we do when we create and curate a personal brand has become so normalized that it’s practically invisible to us. According to Pressly, all that effort has only made our lives and relationships shallower, less meaningful, and less trusting.

Calls for putting our screens down and stepping away from the internet are certainly nothing new. And while The Right to Oblivion isn’t necessarily prescriptive about such things, Pressly does offer a beautiful and compelling vision of what can be gained when we retreat not just from the digital world but from the idea that we are somehow knowable to that world in any authentic or meaningful way. 

If all this sounds a bit philosophical, well, it is. But it would be a mistake to think of The Right to Oblivion as a mere thought exercise on privacy. Part of what makes the book so engaging and persuasive is the way in which Pressly combines a philosopher’s knack for uncovering hidden assumptions with a historian’s interest in and sensitivity to older (often abandoned) ways of thinking, and how they can often enlighten and inform modern problems.

Pressly isn’t against efforts to pass more robust privacy legislation, or even to learn how to better protect our devices against surveillance. His argument is that in order to guide such efforts, you have to both ask the right questions and frame the problem in a way that gives you and others the moral clarity and urgency to act. Your phone’s privacy settings are important, but so is understanding what you’re protecting when you change them. 

Bryan Gardiner is a writer based in Oakland, California.