Blogroll

If you've been eyeing the recently-announced OnePlus Watch 3 as your next smartwatch, you might need to wait a bit to get your hands on one. OnePlus is apparently delaying most shipments of its latest smartwatch, and it's all thanks to an issue with the watch's appearance that made it through all the way to production.

The Raspberry Pi is a small, single-board computer that has been hacked into nearly any type of device you can imagine. But what if you could run Linux on a Pi in the form factor of a BlackBerry smartphone? Someone did just that.

Robert De Niro is joined by Angela Bassett, Lizzie Caplan, Connie Britton, Dan Stevens, Matthew Modine and the the Zero Day showrunners to break down the show's hack and it's startling relevance to today.

Zero Day is now streaming on Netflix.

Microsoft now offers unlimited Copilot Voice and Think Deeper functionality for free users. These advanced features were already available to non-paying users, but they were bound by daily usage limits.

Hyperspace is a new Mac application designed to save storage space. It merges data used by duplicate files, using a trick implemented by Apple in the Mac’s modern file system, giving you more storage space without any changes to your visible files.

You don’t have to worry about too many hidden costs when buying an already-expensive MacBook. However, there are still a few extra purchases you’ll likely need to make in order to get everything you want out of your Apple laptop. Here are six vital add-ons to keep in mind.

Retro games tend to look great in our minds, but that always changes once we finally replay them. Even for games that you've sunk hundreds of hours into, the graphics will usually seem a lot worse than you remembered. But why do we always seem to misremember our favorite games?

It's so hard to make sense of money. One second, you're intent on saving; the next, you've somehow blown through an entire paycheck.

The internet's chock full of folks offering financial advice. And the information can be helpful. It can also be dangerous since not everyone has your best interests in mind. The best advice about financial advice is to be careful who you trust.

Mashable previously listed some of the savvy TikTok finance accounts that appear to have their audience's best interests in mind. But with the internet overflowing with financial tips, we also thought it'd be helpful to share some of the better advice we've seen. Here are four of the most useful money suggestions:

SEE ALSO: 4 savvy finance accounts to follow on TikTok 1. Negotiate your big expenses

This is a major tenant of Vivian Tu, also known as @YourRichBFF. Medical bills, the price of your car, your home purchase — try to get the number down, Tu advises. Getting a discount on a big bill has a much bigger, immediate effect than, say, cutting back on groceries.

"It's pretty normal to be able to say, 'I got my home $50,000 under asking,'" Tu recently told Yahoo. "Do you know how many lattes you need to not drink to save $50,000? Just put on your big boy or big girl pants for 10 seconds, have that uncomfortable conversation, and really, really work on negotiating."

One of Tu's videos discussing negotiating medical bills has racked up more than 2 million views.

2. Build an emergency fund before doing anything else

Stuff happens. Being ready for that stuff really matters. Tori Dunlap of @herfirst100k recommends building up three months of living expenses in a high yield savings account before even paying off debt. Her reasoning is that it helps prevent you from going into further debt during an emergency and it'll help you mentally to have something saved for a rainy day.

3. Keep an impulse list

Michela Allocca runs the TikTok account @BreakYourBudget which primarily focuses on helping folks create and keep a budget. Allocca noted they keep a list of things they'd like to impulse buy, which they return to at a later date. That helps you determine if you actually want or need something.

4. Loud budgeting

It can be awkward to tell friends "no" because you want to keep to your budget. TikTok creator Lukas Battle coined the idea of "loud budgeting" where you openly state you want to avoid purchasing something because of the cost.

"It's not, 'I don't have enough.' It's, 'I don't want to spend,'" Battle said in a viral TikTok.

It's a somewhat radical idea that you can be more open about money and be frank about what is worth your hard-earned dollar.

The idea behind all these tips, really, is to be more mindful about money. Taking care of your finances can feel so unwieldy and impossible, but the best advice makes it more manageable.

I love getting out in nature, but I hate getting lost. Fortunately, modern smartphones make that less likely than ever as long as you’ve got enough power and the right software.

More than 20 Department of Government Efficiency ("DOGE") employees resigned Tuesday (Feb. 25), saying they would no longer aid in leader Elon Musk's technological overhaul of federal services.

The group — made up of technologists previously employed at digital giants like Google and Amazon — comprised about one-third of the U.S. Digital Service, the government office subsumed and rebranded as DOGE by Musk following President Donald Trump's instatement. Musk previously fired around 40 digital service staffers after taking over the office.

SEE ALSO: Grok blocked sources accusing Elon Musk of spreading misinformation

The remaining members posted their call to action as part of WeTheBuilders, a recently-launched collective and website created by current and public federal workers who decry Trump's federal efficiency plans and share inside stories about DOGE's overhaul. The group calls themselves "The official resistance team of the U.S. Digital Service."

"We swore to serve the American people and uphold our oath to the Constitution across presidential administrations. However, it has become clear that we can no longer honor those commitments... We will not use our skills as technologists to compromise core government systems, jeopardize Americans’ sensitive data, or dismantle critical public services. We will not lend our expertise to carry out or legitimize DOGE’s actions," wrote the Digital Service employees in a joint resignation letter to Trump's chief of staff, obtained by the Associated Press. "These actions are not compatible with the mission we joined the United States Digital Service to carry out: to deliver better services to the American people through technology and design."

💪💻 We are the Builders. #altUSDS #wethebuilders @apnews.com

[image or embed]

— Alt US Digital Service (@altusds.bsky.social) February 25, 2025 at 12:30 PM

According to the employees, unaffiliated White House visitors who refused to identify themselves entered the Digital Services offices in January, aggressively interviewing current employees about their political stances and expertise, while hinting at Musk's overhaul. Digital Services staff said those actions, and more, presented an alarming breach of security protocol. "DOGE’s actions — firing technical experts, mishandling sensitive data and breaking critical systems — contradict their stated mission of ‘modernizing federal technology and software to maximize governmental efficiency and productivity," they wrote.

Last week, Musk sent out a mass email (and statement via social media platform X) to federal workers, telling them that they needed to send a documented report of all the work they had accomplished over the last work week or effectively resign from their positions. The Office of Personnel Management, the federal government's human resources department, announced yesterday that response to the email is "voluntary."

The White House quickly responded to the DOGE resignations, dismissing the group's criticisms. "Anyone who thinks protests, lawsuits, and lawfare will deter President Trump must have been sleeping under a rock for the past several years," said White House press secretary Karoline Leavitt. "President Trump will not be deterred from delivering on the promises he made to make our federal government more efficient and more accountable to the hardworking American taxpayers."

The iPhone 16e may be budget-friendly, but it still needs protection. Rugged cases offer the best defense, while slim and clear options provide style and function. And, even though it lacks MagSafe, third-party cases can add compatibility.

The Nissan GT-R is a performance powerhouse, rocketing from 0-60 mph in just 2.9 seconds and on to a top speed of 196 mph in R35 guise. They don't come cheap, though, so if a GT-R is out of budget, here are 10 cheap sports cars that would give the Nissan a serious run for its money.

Google’s Pixel 9a just got hit with a major leak, and this time, it’s a full-on hands-on video.

The yet-to-be-announced phone reportedly made an early appearance, showcasing its familiar minimalist design — but with a build that leans more budget than premium, according to 9to5Google.

Google’s A-series Pixel phones have always been about balancing cost savings with flagship-tier features, but the Pixel 9a might take a slightly different approach. If the latest leaks hold up, the device is expected to pack familiar specs: a Tensor G4 processor, up to 8GB of RAM, and a 120Hz OLED display, among other Pixel-specific features.

SEE ALSO: Google Pixel 9 review: Officially too much AI

But affordability comes with trade-offs. According to a hands-on video spotted on 9to5Google, the 9a might be built with cheaper materials for its chassis and frame. Despite its more budget-friendly appearance, the leaker describes the phone as "plastic feeling" — comparing it to a "Nokia Lumia" that’s surprisingly comfortable in hand.

There’s not much else known about the Pixel 9a at this point, but if you’re particular about how a phone feels in hand, well, here’s something to consider. That said, the leak suggests the 9a will deliver an aesthetically pleasing design, opting for a flat profile over bulky camera bumps.

When can you actually get your hands on one? If the rumors are true, Google could be gearing up to launch the Pixel 9a sometime in late March.

Imagine an AI system capable of guiding a robot to manipulate physical objects as effortlessly as it navigates software menus. Such seamless integration of digital and physical tasks has long been the stuff of science fiction.  

Today, Microsoft researchers are bringing that vision closer to reality with Magma (opens in new tab), a multimodal AI foundation model designed to process information and generate action proposals across both digital and physical environments. It is designed to enable AI agents to interpret user interfaces and suggest actions like button clicks, while also orchestrating robotic movements and interactions in the physical world.  

Built on the foundation model paradigm, Magma is pretrained on an expansive and diverse dataset, allowing it to generalize better across tasks and environments than smaller, task-specific models. As illustrated in Figure 1, Magma synthesizes visual and textual inputs to generate meaningful actions—whether executing a command in software or grabbing a tool in the physical world. This new model represents a significant step toward AI agents that can serve as versatile, general-purpose assistants. 

Figure 1: Magma is one of the first foundation models that is capable of interpreting and grounding multimodal inputs within both digital and physical environments. Given a described goal, Magma can formulate plans and execute actions to achieve it. By effectively transferring knowledge from freely available visual and language data, Magma bridges verbal, spatial and temporal intelligence to navigate complex tasks and settings.

Vision-Language-Action (VLA) models integrate visual perception, language comprehension, and action reasoning to enable AI systems to interpret images, process textual instructions, and propose actions. These models bridge the gap between multimodal understanding and real-world interaction. Typically pretrained on large numbers of VLA datasets, they acquire the ability to understand visual content, process language, and perceive and interact with the spatial world, allowing them to perform a wide range of tasks. However, due to the dramatic difference among various digital and physical environments, separate VLA models are trained and used for different environments. As a result, these models struggle to generalize to new tasks and environments outside of their training data. Moreover, most of these models do not leverage pretrained vision-language (VL) models or diverse VL datasets, which hampers their understanding of VL relations and generalizability.  

Magma, to the best of our knowledge, is one of the first VLA foundation model that can adapt to new tasks in both digital and physical environments, which helps AI-powered assistants or robots understand their surroundings and suggest appropriate actions. For example, it could enable a home assistant robot to learn how to organize a new type of object it has never encountered or help a virtual assistant generate step-by-step user interface navigation instructions for an unfamiliar task. Through Magma, we demonstrate the advantages of pretraining a single VLA model for AI agents across multiple environments while still achieving state-of-the-art results on user interface navigation and robotic manipulation tasks, outperforming previous models that are tailored to these specific domains. On VL tasks, Magma also compares favorably to popular VL models that are trained on much larger datasets. 

Building a foundation model that spans such different modalities has required us to rethink how we train and supervise AI agents. Magma introduces a novel training paradigm centered on two key innovations: Set-of-Mark (SoM) and Trace-of-Mark (ToM) annotations. These techniques developed by Microsoft Research, imbue the model with a structured understanding of tasks in both user interface navigation and robotic manipulation domains. 

• Set-of-Mark (SoM): SoM is an annotated set of key objects, or interface elements that are relevant to achieving a given goal. For example, if the task is to navigate a web page, the SoM includes all the bounding boxes for clickable user interface elements. In a physical task like setting a table, the SoM could include the plate, the cup, and the position of each item on the table. By providing SoM, we give Magma a high-level hint of “what needs attention”—the essential elements of the task—without yet specifying the order or method.

Figure 2: Set-of-Mark (SoM) for Action Grounding. Set-of-Mark prompting enables effective action grounding in images for both UI screenshot (left), robot manipulation (middle) and human video (right) by having the model predict numeric marks for clickable buttons or robot arms in image space. These marks give Magma a high-level hint of “what needs attention” – the essential elements of the task 

• Trace-of-Mark (ToM): In ToM we extend the strategy of “overlaying marks” from static images to dynamic videos, by incorporating tracing lines following object movements over time. While SoM highlights key objects or interface elements relevant to a task, ToM captures how these elements change or move throughout an interaction. For example, in a physical task like moving an object on a table, ToM might illustrate the motion of a hand placing the object and adjusting its position. By providing these temporal traces, ToM offers Magma a richer understanding of how actions unfold, complementing SoM’s focus on what needs attention.

Figure 3: Trace-of-Mark (ToM) for Action Planning. Trace-of-Mark supervisions for robot manipulation (left) and human action (right). It compels the model to comprehend temporal video dynamics and anticipate future states before acting, while using fewer tokens than next-frame prediction to capture longer temporal horizons and action-related dynamics without ambient distractions.  Performance and evaluation Zero-shot agentic intelligence Table 1: Zero-shot evaluation on agentic intelligence. We report the results for pretrained Magma without any domain-specific finetuning. In this experiment, Magma is the only model that can conduct the full task spectrum. Figure 4: Zero-shot evaluation on Google Robots and Bridge with SimplerEnv. Magma shows strong zero-shot cross-domain robustness and demonstrates impressive results in cross-embodiment manipulation simulation tasks. Efficient finetuning Table 2: Efficient finetuning on Mind2Web for web UI navigation. Figure 5: Few-shot finetuning on Widow-X robot (left) and LIBERO (right). Magma achieves a significantly higher average success rate in all task suites. Additionally, removing SoM and ToM during pretraining has a negative impact on model performance. Table 3: Without task-specific data, Magma performs competitively and even outperforms some state-of-the-art approaches such as Video-Llama2 and ShareGPT4Video on most benchmarks, despite using much fewer video instruction tuning data. Relation to broader research

Magma is one component of a much larger vision within Microsoft Research for the future of agentic AI systems. Across various teams and projects at Microsoft, we are collectively exploring how AI systems can detect, analyze, and respond in the world to amplify human capabilities.

Earlier this month, we announced AutoGen v0.4, a fully reimagined open-source library for building advanced agentic AI systems. While AutoGen focuses on the structure and management of AI agents, Magma enhances those agents by empowering them with a new level of capability. Developers can already use AutoGen to set up an AI assistant that leverages an LLM for planning and dialogue using conventional LLMs. Now with MAGMA, if developers want to build agents that execute both physical or user interface/browser tasks, that same assistant would call upon Magma to understand the environment, perform reasoning, and take a sequence of actions to complete the task. 

The reasoning ability of Magma can be further developed by incorporating test-time search and reinforcement learning, as described in ExACT. ExACT shows an approach for teaching AI agents to explore more effectively, enabling them to intelligently navigate their environments, gather valuable information, evaluate options, and identify optimal decision-making and planning strategies.

At the application level, we are also exploring new user experience (UX) powered by foundation models for the next generation of agentic AI systems. Data Formulator is a prime example. Announced late last year, Data Formulator, is an AI-driven visualization tool developed by Microsoft Research that translates high-level analytical intents into rich visual representations by handling complex data transformations behind the scenes​.  

Looking ahead, the integration of reasoning, exploration and action capabilities will pave the way for highly capable, robust agentic AI systems.

Magma is available on Azure AI Foundry Labs (opens in new tab) as well as on HuggingFace (opens in new tab) with an MIT license. Please refer to the Magma project page (opens in new tab) for more technical details. We invite you to test and explore these cutting-edge agentic model innovations from Microsoft Research.

Opens in a new tab

The post Magma: A foundation model for multimodal AI agents across digital and physical worlds appeared first on Microsoft Research.

For years, iPhone and iPad owners were unable to change their default app settings. But times have changed, and Apple has loosened its grip with a bit of pressure from the courts for good measure. Here’s how to change app defaults, and some ideas of what to replace them with.

While more and more phones are slowly getting satellite text or cellular calling, carriers aren't stopping there. This week, Verizon and AT&T separately announced that they had successfully completed and tested satellite-to-phone video calls. And no, it's not through Starlink.

SPEND $25, SAVE $5: Through 3 a.m. ET (12 a.m. PT) on Feb. 25, you can save $5 when you spend $25 on over 4,000 eligible paperback and hardcover books at Target.

Opens in a new window Credit: Red Tower Books / Entangled Publishing Spend $25 on books, save $5 Shop Now

E-book deals are popping up pretty much every week these days.

But if physical page turning is what really gets your dopamine dopamine-ing, Target's current book sale is about to do a real number on your 2025 reading list — and perhaps provide inspo for a spring refresh of your coffee table decor. Through 3 a.m. ET (midnight PT) on Feb. 25, Target will shave off $5 when you spend $25 on books.

SEE ALSO: I read over 50 books a year. Here are my favorite e-readers.

The discount is applied at checkout for Target Circle members, which is a quick, free membership you'll want to join — Target offers little coupons like this all the time.

The list of eligible books is more than 4,000 long — sifting through all 50 pages is essentially reading a book in itself. Shop BookTok heavy hitters like the Court of Thorns and Roses series or titles from Freida McFadden and Colleen Hoover, nonfiction gems like Matthew McConaughey's Greenlights, shadow work journals, or even the occasional coloring book.

Mashable curated a list of TikTok's favorite fiction books from 2024, in case you don't feel like spending hours on the app yourself. You know, because you like to spend your free time... reading. Even if you're not a big book person yourself, this sale is a gold mine for gift ideas or donation options.

With the launch of a brand new version of Photoshop for smartphones, Adobe has also launched what is now its cheapest premium Photoshop plan. We're not sure if it's a good replacement for that 20GB Photography Plan that Adobe killed off a couple of months ago, though.

TL;DR: Enjoy a lifetime subscription to Skoove Premium Piano Lessons for just A$235 (reg. A$470).

Anyone interested in learning to play the piano should take a closer look at Skoove. This app is more like your personal piano coach. Using cutting-edge AI technology, Skoove listens to the notes you play and offers real-time feedback to help you improve. A lifetime subscription is on sale for just A$235 (reg. A$470).

The lessons are tailored to your pace and skill level, ensuring a customized experience. Whether you’re diving into classical compositions or belting out pop hits, Skoove’s extensive library keeps learning fresh and exciting.

With compatibility across all pianos and keyboards — USB/MIDI or acoustic — you can learn wherever you are. Plus, Skoove’s monthly updates mean you’ll always have new songs and lessons to explore. And if you ever hit a snag, their one-on-one support from expert instructors has you covered.

Learning piano doesn’t have to be intimidating. With Skoove, you can start at your own pace and build confidence through intuitive lessons. Want to play Imagine by John Lennon or dive into Beethoven’s Für Elise? Skoove makes it possible, breaking each piece into digestible steps that make learning fun and achievable.

For seasoned players, Skoove offers advanced lessons and techniques to refine your skills and broaden your repertoire. And since it works on tablets, smartphones, and personal computers across multiple operating systems, your piano practice is always at your fingertips.

Say goodbye to expensive piano tutors and rigid schedules. For just A$235, you can unlock a lifetime of Skoove lessons and resources designed to grow with you.

StackSocial prices subject to change.

Opens in a new window Credit: Skoove Skoove Premium Piano Lessons: Lifetime Subscription A$235.00 at the Mashable Shop
A$470.00 Save A$235.00 Get Deal

Activision's Call of Duty: Black Ops 6 is advertising directly to players through the killcam. This new feature lets players add weapon blueprints and skins from the killer to a wishlist.