Technology analyst Benedict Evans has raised questions about Apple’s recent product execution, suggesting the company may be experiencing a shift in its ability to deliver new features and innovations smoothly. In a recent blog post, Evans pointed to the launch of the Vision Pro headset and delays in personalised Siri updates as examples of this trend. He compared these challenges to “Microsoft’s Windows Vista era”, which was known for software issues and design changes that were not well received. Evans argued that the Vision Pro may have been released too early, serving more as a concept than a fully developed product. "Apple doesn’t usually release unfinished products, so why did it ship the Vision Pro? It didn’t sell in large numbers, and it hasn’t attracted significant developer interest," he wrote. He noted that even within Apple, there seems to be some uncertainty about the device’s role in the market. Another key concern is the delay of Apple’s AI-powered Siri features, first showcased at the Worldwide Developers Conference (WWDC) last year. Initially expected to launch by the end of 2024, reports now indicate the update could be delayed until 2025 or later. Evans observed that Apple typically only presents features that are close to completion, making the delay unexpected. "It now appears that what was shown at WWDC was more of a concept than a working feature," he said. Evans also noted that Apple’s software release cycle has historically been predictable, with announcements in the summer and product rollouts in September. However, he suggested that this pattern may be changing, with more delays becoming apparent.

Safety incidents increase on gerototed devices compared to non-cleared devices by 250 times. The risks fit the mobile-first strategy of cybercriminals. The security provider Zimperium warns of interventions in mobile operating systems via rooting and jailbreaking. According to current analyzes, the threat to companies by rooting Android devices and jailbreaks on Apple's devices is still high. Geroted smartphones and tablets with Android are affected 3.5 times more often by mobile malware attacks-and safety incidents due to compromised systems increase 250 times in comparison to non-stated devices. According to the Zimperium, cybercriminals pass into a mobile-first attack strategy. The root and jailbreak functions on mobile devices originally used for adaptation and app installation therefore remain a frequently used attack vector. Business users who deal with critical security protocols and usage rights increased the threats and vulnerability in the company - from mobile malware and data protection violations to the complete compromising of entire systems. "Modified mobile devices prove to be frequently used entry points for threats who can take advantage of security gaps and compromise a company network as a whole with complex attacks," said the specialist for real -time protection on mobile devices. “Visual actors use these devices to assume mobile applications and to carry out fraudulent activities. Although mobile operating systems integrate more defensive measures, the community around mobile rooting tools grows. Tools such as Magisk, Apatch, Kernelsu, Dopamine and Checkra1n are constantly being developed and have new stealth mechanisms Avoidance of conventional security measures. " The threat for companies remains high because cybercriminals further adapted their attack techniques, Zimperium continues. A compromised mobile device is enough as a dangerous gateway for data theft, ransomware and advanced persistent threats (apts). Zimperium recommends organizations to measure mobile security measures to the highest priority and to always be one step ahead of the attackers through proactive defense measures as well as AI-based threat detection. "The cat-and-mouse game between security teams and developers of mobile rooting tools is far from over," said Nico Chiaraviglio, Chief Scientist at Zimperium. "Companies need continuous real-time recognition technologies against attempts to manipulate on mobile devices. As soon as a mobile device is compromised, the risks for the entire company are increasing rapidly."

A French researcher was denied entry to the U.S. over text messages on his phone that were critical of President Donald Trump, according to several reports from media outlets in France. An FBI investigation was reportedly launched into the scientist, and it’s just the latest sign that America’s slide into authoritarianism is going to have serious ramifications for academic research in the near future. The researcher, who hasn’t been identified publicly, was traveling to Houston on March 9 for a conference, according to France24, though the airport where U.S. Customs and Border Protection (CBP) stopped the scientist hasn’t been identified. France’s minister of higher education and research, Philippe Baptiste, only identified the researcher as someone working for the French National Center for Scientific Research. CBP reportedly went through the researcher’s phone and bags during a so-called “random” check and discovered text messages critical of Trump and his assault on science research. The U.S. president has been destroying the U.S. federal government with the help of billionaire oligarch Elon Musk, recently cutting vital research programs at the National Institutes of Health, the EPA, and countless other agencies. France24 quoted an unnamed diplomatic source who claimed the researcher “showed hatred towards Trump” and it “could be qualified as terrorism,” though the substance of the messages is still unclear, and the Trump regime isn’t exactly known for its honesty in situations like these. The French minister Baptiste characterized the messages differently. “This measure was apparently taken by the American authorities because the researcher’s phone contained exchanges with colleagues and friends in which he expressed a personal opinion on the Trump administration’s research policy,” said Baptiste. Reached for comment by email, an FBI spokesperson would only refer Gizmodo to U.S. Customs and Border Protection. The FBI would not confirm or deny that it had launched an investigation into the researcher. CBP, who reportedly seized the researcher’s devices, gave a vague emailed statement to Gizmodo that started by referring to individuals who “violate the terms of a lawful immigration status and attempt to re-enter the country” but didn’t elaborate any further on that suggestion. “All persons arriving at a port of entry to the United States are subject to inspection on a case-by-case basis. As part of their critical national security mission, CBP officers routinely determine admissibility of foreign nationals using longstanding U.S. immigration law,” the agency said in a statement attributed to Hilton Beckham, Assistant Commissioner of Public Affairs. “If an individual has material discovered on their electronic media that raises flags during an inspection, it can result in further analysis,” the statement continued. “Claims that such decisions are politically motivated are completely unfounded.” CBP has the authority to stop any foreign national from coming into the U.S., but the circumstances around this case will cause further questions internationally about whether the U.S. is a safe place to travel. Other high-profile cases of people stopped at the border recently suggest a more aggressive posture by American authorities than before Trump came into office. Previously, someone traveling to the U.S. without a valid visa might be denied entry, but there are more and more stories of people instead being thrown into ICE detention facilities for weeks at a time. A Canadian woman, Jasmine Mooney, was recently trying to enter the U.S. through Mexico near San Diego and was thrown into a private prison, subjected to invasive medical tests, and shackled in chains, according to the Guardian. Another recent case involved Fabian Schmidt, a German national with permanent residency in the U.S. who lived in New Hampshire. Schmidt had been visiting Luxembourg and was stopped by U.S. authorities in Boston on his flight home. Schmidt, who got his green card in 2008, was reportedly “flagged” and transferred to the Donald W. Wyatt detention facility in Central Falls, Rhode Island, according to WGBH. That man was “violently interrogated,” stripped naked, and put in a cold shower, according to the news outlet. The man was also subjected to sleep deprivation and other forms of torture. The U.S. has also illegally flown people without trial to a notorious prison in El Salvador under a deal between the country’s authoritarian president and the Trump regime, which recently invoked the Alien Enemies Act of 1798, allowing “deportations” without judicial review. But these aren’t even “deportations” in the traditional sense, given the fact that these people were Venezuelan and were being sent to a country they never lived in. The U.S. government claims these people were all gang members, though it’s impossible to say for sure given the fact that you simply have to take Trump’s word for it. Sworn declarations filed late Wednesday claim some of the men sent to El Salvador were asylum seekers who had never broken the law. In at least one case, ICE appears to have claimed that a man’s soccer tattoo and a common gesture of excitement were evidence he belonged to a gang. ICE policy says a person can be deemed a gang member if any ICE officer notes two “Gang Membership Identification Criteria.” In Mr. Reyes Barrios’ case, that appears to have been: 1) A tattoo modeled off the Real Madrid logo 2) A picture from social media of him throwing the horns. That was it! [image or embed] — Aaron Reichlin-Melnick (@reichlinmelnick.bsky.social) March 20, 2025 at 5:52 AM Needless to say, these cases will make researchers think long and hard about whether it’s safe to visit the U.S. with Trump now in power. Trump’s fascist movement has been ransacking the U.S. government and purging anyone who’s not loyal to his perverse worldview; an ideology built on greed, corruption, and racism.

is a news writer fond of the electric vehicle lifestyle and things that plug in via USB-C. He spent over 15 years in IT support before joining The Verge. Honda and its luxury brand Acura are getting access to more than 20,000 Tesla Superchargers in North America starting in June for their Prologue and ZDX EVs. Both brands will also sell their own EV charging adapters for an undisclosed price through dealerships. The new Honda and Acura branded adapters that convert their vehicles’ existing CCS1 port to Tesla’s NACS connector have a design identical to Ford’s Lectron-built offering. But unlike Hyundai and Ford, Honda hasn’t said it will provide the adapters for free. Without one, you can still charge at some Tesla Supercharger stations as long as they have built-in Magic Dock adapters. Both the Honda and Acura adapters are the same except for the small logo printed on top. Image: Honda Honda, like every other automaker, agreed last year to adopt Tesla’s now standardized connector for EV charging in the US. However, they still have to wait until Tesla flips the switch for their vehicles to link up with Tesla’s vast fast-charging network.

There are sinister irony in the times of Italian politics. Over two days we saw the manifesto of Ventotene torn to the Chamber by the Prime Minister, Giorgia Meloni, and two commissions of the Senate passed a very disturbing security bill. Knowing the matrix of those who govern us, neither of them should be surprised. On the one hand we have the leader of the most right government in the republican history that underlines the distances from the project of the United States of Europe imagined by Altiero Spinelli and Ernesto Rossi in 1941, in the moment of greater political and economic weakness of the European Union and the countries that make it up. On the other hand, the go -ahead of the right majority to a rule that assigns more powers to secret services, without consequently increasing their supervision, forcing the public administration to deliver data and information without being able to resist. More powers to secret services There is little to sleep peacefully. By approving article 31 of the Security DDL, the Package of Rules of the Meloni government that imposes a securitarian narrow to the country (including an increase in penalties against those who protest), the Constitutional Affairs and Justice Commissions have indicated the road that the majority intends to travel, although the text has yet to get the green traffic light of Senate and Chamber. It is true, as written on ants, that the article strengthens the 2007 intelligence reform. The problem is as it does. Today public administrations can decide whether to collaborate with the secret services, providing information or opposing to balance the request with other rights. With the security bill, we move on to the obligation in the name of national security. Universities, hospitals, public offices would be required to provide all the data that the services ask for, on the basis of conventions that asphalt any other right of privacy. But not only that: also public companies. Category in which Rai, so much so that the National Federation of the Italian Press (FNSI), the union of journalists, has denounced the risk that the professional duty not to reveal the sources will be trampled on. Here the passage in which up to 10 I expect waiting for someone to jump on saying: "Eh, but if one has nothing to hide ...". It will be surprising, but sometimes it is necessary to "hide" things. If you have confidential information that you want to make known to public opinion without directly expanding and telling it to journalists, they have a duty not to trust what the source is. It is called professional secret and applies to many categories of crafts. Being able to oppose it is a protection towards the weakest and most exposed people. But he risks falling under the blows of article 31, in the public world. The attack on the counter-balancing of the powers of the state As the text states, "the conventions [between secret services and public bodies, editor's note] may provide for the communication of information to the aforementioned bodies also in derogation from the sector regulations on confidentiality". Goodbye privacy. At that point there is no limit to the type of information to be caught, brandishing the club of national security: political guidelines, state of health, communications. Is there anyone who, in the name of the evergreen national security, is he willing to put all his business in the square? Even after reading searches that can be led without even moving a finger, with a spyware as a graphite of paragon? Invoking national security as a hammer to dismantle the counter-balancing of state powers is nothing good. It takes, at a time when politics speaks the lexicon of bellies, but will not ensure greater protection for our state. He will not keep us safe, as they want to make us believe. "The reactionary forces have skilled and educated men and paintings in command, who will fiercely fight to preserve their supremacy. At the serious moment they will be able to present themselves well disguised, they will proclaim lovers of freedom, peace, general well -being, the poorest classes. Already in the past we have seen how they have insinuated behind the popular movements, and they have paralyzed them, devoted, converted into the precise opposite. dangerous with which they will have to deal. Guess where is it written? Yes, in the Ventotene Manifesto.

Apple is shaking up its Siri development team Mike Rockwell, the man behind Vision Pro, is taking over from John Giannandrea Could this be the pivotal change Apple needed to bring an upgraded Siri to market? There's a major shake-up at the top of the Apple food chain as Tim Cook opts for a new leader to help the company bring Apple Intelligence-powered Siri to life. Following reports from earlier this week that Apple Intelligence would be a focal point of Apple's off-site Top 100 exec meeting, Bloomberg's Mark Gurman is now reporting that AI head John Giannandrea is no longer going to be overseeing Siri's overhaul. According to Gurman, "[Apple CEO] Tim Cook has lost confidence in the ability of AI head John Giannandrea to execute on product development, so he’s moving over another top executive to help: Vision Pro creator Mike Rockwell." Rockwell will now be in charge of Siri, according to Gurman's sources, who have asked not to be named, and he'll report directly to software chief Craig Federighi. With Giannandrea no longer in charge of Siri, he'll now be working on other AI projects. The decision was made during the Apple Top 100 meeting and is said to be confirmed to employees later this week. Assuming Gruman is right, this major exec shift comes at a pivotal time for Apple – once a pioneer in the voice assistant game but now just a mere passenger – as it tries to devise a solution for Siri. Apple had advertised its flagship iPhone 16 Pro as the best device for Apple Intelligence, yet nearly seven months after its launch, customers have yet to see the real benefits of AI. Get daily insight, inspiration and deals in your inbox Sign up for breaking news, reviews, opinion, top tech deals, and more. Contact me with news and offers from other Future brands Receive email from us on behalf of our trusted partners or sponsors With Apple's Siri delay a public disaster, this shift in leadership could be the catalyst for success that's needed to make the personal context-capable Siri a reality. Siri's redemption arc? At WWDC 2024 in June, we got to hear directly from Giannandrea about his vision for the then-newly-announced Siri. He said Siri with Apple Intelligence has a "rich understanding of what’s on your device," and that the voice assistant's knowledge base would "only get richer over time." Nearly a year later, Siri still can't tell you what month of the year it is and definitely doesn't understand what's on your device. Rockwell, who can take credit for Apple's Vision Pro development, has the potential to get Siri's development back on track. While the mixed reality headset hasn't necessarily been a commercial success it does achieve some incredible feats. Tim Cook is likely counting on Rockwell's ability to innovate, as evidenced in the Vision Pro, to make Siri as good as the company advertised back at WWDC. And, if he does, then all the naysayers will have to accept that Apple truly is back on track.

Microsoft has released a new feature in its latest 26120.3576 and 22635.5090 Insider builds that helps users understand their PC's capabilities (via phantomofearth on X). From its looks, this feature isn't cutting-edge by any means, as a simple search on the internet will likely land you with many more detailed and fine-tuned answers. In addition, it also tells you if you have the latest Windows release installed on your system, but oddly asks you to verify the information in the Windows Update section. Back in the Vista days, Microsoft released WinSAT, a tool designed to assess the capabilities of your hardware through a set of tests. It would report back the score on a scale between 1.0 and either 5.9 or 9.9, depending on the Windows version. While this number wasn't definitive, it let people know where their PC stood. WinSAT also benchmarked individual components such as the CPU, memory, graphics, and disk enabling users to identify potential bottlenecks. The new FAQ feature spans two different Windows Insider versions: build 26120.3576 for Windows Insiders in the Beta Channel for Windows 11 23H2 and build 22635.5090 for Windows Insiders in the Beta and Dev Channels for Windows 11 24H2. Release notes for both releases mention the inclusion of a new "Frequently Asked Questions" list within "Device specifications" under System > About. The FAQs are disabled by default, though can be activated by enabling the "55305888" Feature ID. New Frequently Asked Questions list in Settings > System > About, hidden in builds 26120.3576 and 22635.5090. Has some questions related to the Windows version and device specs. (vivetool /enable /id:55305888) pic.twitter.com/AkaP8XR3PRMarch 17, 2025 Known FAQs (non-exhaustive) detailing your Windows version, followed by how much RAM you have and your system's graphical capabilities have been compiled in a GitHub Gist. The first question allows users to check if their Windows is up-to-date but confusingly asks them to verify again within the Windows Update section. That’s most certainly a bug. The second question outlines what applications can be run on your system based on its RAM capacity, using a few common ranges such as <4GB, 4GB-8GB, 8GB-16GB, and >16GB. Lastly, the graphics section goes over the graphical capabilities of your system, or lack thereof if you don’t have a dedicated GPU. This approach is not as quantifiable as WinSAT and omits assessment of key components such as the CPU and storage. Given Microsoft’s strong push for AI, an AI-driven hardware analysis utility would have been a significant overhaul to WinSAT. (Maybe I'm just reading too much into what’s supposed to be a supplementary feature.) Hopefully, the official release will include more detailed insights that provide users with more than just the basics.

is a senior editor and author of Notepad , who has been covering all things Microsoft, PC, and tech for over 20 years. Meta is finally allowing Threads users to set the following feed or any custom feeds as the default. The change is part of a number of new Threads features, including the ability to now limit post replies to followers only, a better video player, and the option to add topics in your bio. Threads users have been asking to be able to set the following feed as default ever since the social media network launched in December 2023, and Meta is finally giving in. After testing the option to choose a default feed late last year, it’s now rolling out so that you can set a custom feed or following as default, and it will remain even when you close and reopen the app. You can now limit replies to posts to just your followers. Image: Meta Threads is also getting more options for managing how people interact with your posts. You’ll now be able to limit replies and quote-posts to just your followers, which can help limit the amount of people that engage and share your posts. Meta is also testing the ability to add up to 10 topics to your bio, so you can list favorite communities or interests for others to see. Drafts will also include prompts to suggest adding a trending topic. “According to our internal data, posts with tagged topics generally receive more views than those that without one, making them an effective way to help you reach more people,” Meta says.

Katie Drummond: Right, available to the public, as opposed to DOGE, despite the claims that Musk and even President Trump have made, that this is a maximally transparent group of people doing this work. We don't actually know that much about what they're doing, what they're finding, and the changes that they're making, short of journalists like you and so many of your WIRED colleagues and other news organizations obtaining that information and publishing it. Now, I want to back up a little bit because at the heart of all of this is this notion of fraud, of waste, of financial abuse. It's something that Elon Musk has repeatedly claimed is widespread, pervasive across government agencies that there is fraudulent spending, there is excessive waste happening. There is abuse of financial power, financial authority on the part of these agencies. Now, from what you've reported over the last several months and in your role, how valid is that claim to begin with? Vittoria Elliott: There was a report from the government accountability office that covered the years 2018 to 2022, and that found that there's actually over $200 billion of waste, fraud, or abuse that happens. So it's like, it's not this made up thing, but in the context of a $7 trillion federal budget, that is not what's breaking the bank. I don't think anyone, Democrat or Republican, DOGE fan or not, would argue that, hey, that's a problem that we need to solve. That's a very valuable thing. But one of the auditors that I spoke to basically said DOGE could go look at the current recommendations from the IGs. They could go look at the current outstanding investigations. They could pick up that work if they are really dedicated to this because there actually are systems, and there actually are reports about this. But from what the auditors I spoke to said, they said that doesn't really seem to be what DOGE is doing. And one of the other things they said was a lot of auditors would probably be really happy to help them, to offer advice. They'd be happy to work with them, they'd be happy to bring young talent into their teams to sort of make the auditing system more robust. But DOGE is really kind of siloed, it doesn't seem like they're really interested in that. Katie Drummond: Right. Now, speaking of DOGE and the auditors that you spoke to for this story, they certainly didn't have kind words to say about how DOGE is going about its work. Tell us a little bit more about their impressions of DOGE and their commentary on sort of what is happening under the auspices of this idea that Musk and DOGE are essentially auditing the federal government. What do the actual auditors think about that? Vittoria Elliott: Yeah, so to give some context, first off, a regular audit, which involves like five different steps, so there are five steps to an audit. One is where you plan it, you sit down, you say, "This is what we're looking at, this is the kind of information we're going to need to go after." You sort of map out who the stakeholders are, who you're going to have to interview, because a lot of times they're going into programs or systems that maybe they don't fully understand. So, it's almost a little ethnographic. You got to go in, you got to interview people, you got to be like, "What is your job? What does this connect to? What does this mean inside your system?" You have to do that baseline understanding of getting to know the program or the agency. Then you do the actual field work, which is, you go out, you talk to people. In the case of, for instance, the hurricane stuff, they actually sent auditors, I believe, to Puerto Rico to see what was going on. You go into the systems, you do that kind of human on the ground legwork. Once you've done that, they sort of pull all this information together into a report, they sort of do this analysis. They talk to the agencies, they say like, "Hey, these are the problems we found. You have the opportunity to correct these. These are our recommendations." And then they submit that to the agencies, to Congress, and then they get to follow up. They get to come back and be like, "Hey, we recommended this thing. Did you fix this problem?" There's a level of accountability for trying to fix that thing. So, there are very clear understandings within the government of what the best practice of this is, and all of these people do that. That takes six to 18 months.

is a news writer who covers the streaming wars, consumer tech, crypto, social media, and much more. Previously, she was a writer and editor at MUO. Apple is replacing the head of Siri, shortly after pushing back the rollout of its highly-anticipated AI features, according to a report from Bloomberg. Mike Rockwell, the head of Vision Pro development, will reportedly replace John Giannandrea as the leader of AI and Siri. The move comes after Apple CEO Tim Cook “lost confidence” in Giannandrea’s ability to lead AI development, Bloomberg reports. Earlier this month, Apple said in a statement to Daring Fireball that it’s delaying the launch of generative AI upgrades for Siri to “the coming year,” as development is taking “longer than we thought.” Bloomberg later reported on the tense Siri team meeting, in which Apple senior director Robby Walker called the AI situation “ugly.” Giannandrea will no longer have control over the Siri team, while Paule Meade, an “executive who has run hardware engineering for the Vision Pro under Rockwell,” will take over as the head of the Vision Pro Group, Bloomberg reports.

is a news writer who covers the streaming wars, consumer tech, crypto, social media, and much more. Previously, she was a writer and editor at MUO. Apple is replacing the head of Siri, shortly after pushing back the rollout of its highly-anticipated AI features, according to a report from Bloomberg. Mike Rockwell, the head of Vision Pro development, will reportedly replace John Giannandrea as the leader of AI and Siri. The move comes after Apple CEO Tim Cook “lost confidence” in Giannandrea’s ability to lead AI development, Bloomberg reports. Earlier this month, Apple said in a statement to Daring Fireball that it’s delaying the launch of generative AI upgrades for Siri to “the coming year,” as development is taking “longer than we thought.” Bloomberg later reported on the tense Siri team meeting, in which Apple senior director Robby Walker called the AI situation “ugly.” Giannandrea will no longer have control over the Siri team, while Paule Meade, an “executive who has run hardware engineering for the Vision Pro under Rockwell,” will take over as the head of the Vision Pro Group, Bloomberg reports.

Thought it was hard to get an Nvidia RTX 50-series Blackwell GPU? Things are potentially going to get even more difficult, as Nvidia has now revealed its Blackwell RTX Pro series of graphics cards. These will target both laptops and desktops, as well as standalone PCs and data center products. We've heard from some people at GDC/GTC that Nvidia is working to improve the supply of all of its Blackwell GPUs, and some even suggested we "might" see supply finally start to catch up to demand by May/June — meaning we could see MSRP-priced models on sale and readily available. We'll believe that when we see it. We expected the Blackwell professional announcement, and certainly Nvidia would know it was incoming and would plan for the increased production required. But we would also say the same of the RTX 50-series launches. Nvidia should have known demand would be high, and yet the supply has been woefully insufficient. And given the choice between shipping GB202/GB203/GB205 GPUs as consumer parts with an ostensible $550~$2,000 MSRP, or shipping professional and data center parts that can cost five times as much (or more), we suspect the latter category will be served first whenever possible. Swipe to scroll horizontally Nvidia RTX Pro Specifications Graphics Card RTX Pro 6000 RTX Pro 5000 RTX Pro 4500 RTX Pro 4000 Architecture GB202 GB202 GB203 GB203 Process Technology TSMC 4N TSMC 4N TSMC 4N TSMC 4N Transistors (Billion) 92.2 92.2 45.6 45.6 Die size (mm^2) 750 750 378 378 SMs 188 110 82 70 GPU Shaders (ALUs) 24064 14080 10496 8960 Tensor Cores 752 440 328 280 Ray Tracing Cores 188 110 82 70 Boost Clock (MHz) 2600 2500? 2500? 2500? VRAM Speed (Gbps) 28 28 28 28? VRAM (GB) 96 48 32 24 VRAM Bus Width 512 384 256 192 L2 Cache 128 96? 64? 48? Render Output Units 192 144? 96? 80? Texture Mapping Units 752 440 328 280 TFLOPS FP32 (Boost) 125.1 70.4? 52.5? 44.8? TFLOPS FP16 (FP4/FP8 TFLOPS) 1001 (4004) 563 (2253) ? 420 (1679) ? 358 (1434) ? Bandwidth (GB/s) 1792 1344 896 672? TBP (watts) 600 300 200 140 Details on the specifications and configurations (some of them, anyway) were shared after the keynote, and the higher solutions will use the same GB202 chip as the GeForce RTX 5090, but with a dramatically altered memory configuration. The top RTX Pro 6000 will also be equipped with 24Gb (3GB) GDDR7 chips, the same chips that are currently going into the RTX 5090 Laptop GPU. 24Gb chips potentially bumps the memory configuration from 32GB on a 512-bit interface to 48GB, or from 16GB on a 256-bit interface (for lower tier parts) to 24GB, and for a 192-bit interface there will be 18GB options. But that's only part of the potential upgrade. As we've traditionally seen with professional and data center solutions, Nvidia will ship some products with memory chips in "clamshell" mode — with GDDR7 chips on both sides of the PCB. That doubles the maximum capacity for every interface width, yielding up to 96GB for GB202 and its 512-bit interface, up to 48GB for GB203 and its 256-bit interface, and up to 36GB for GB205's 192-bit interface. And that's not some hypothetical number. Nvidia has stated that it will have a Blackwell RTX Pro GPU with 96GB of GDDR7 memory, with ECC enabled. Laptops on the other hand look like they'll stick to similar capabilities as the RTX 50-series mobile solutions, with an RTX Pro solution sporting up to 24GB — the same as the RTX 5090 Laptop GPU, which uses the GB203 silicon. At present, the four workstation models only use two of the Blackwell chips: GB202 and GB203. The RTX Pro 6000 uses a near-complete GB202, with 188 of the potential 192 SMs. It also has the full 128MB L2 cache, and the aforementioned 96GB of GDDR7 ECC memory, with 16 chips on each side of the PCB. It also has the full complement of four NVDEC/NVENC units for video encoding/decoding. Power use for the workstation and server variants is set to 600W max (the Server Edition has a configurable TDP), while the Max-Q variant clamps the power limit to 300W. Nvidia didn't disclose clocks or theoretical TFLOPS on the other three RTX Pro GPUs, so we've just estimated 2.5GHz for now. The step-down RTX Pro 5000 still uses GB202, just trimmed down quite a lot. It has 110 SMs enabled, which is only 57% of the computational elements. It also has four of the 32-bit memory interfaces disabled, giving it a 384-bit interface. Nvidia sticks with the more readily available 16Gb (2GB) GDDR7 chips as well, with 12 on each side of the PCB yielding 48GB of total VRAM — with no mention of the L2 cache size, though we suspect it will have 96MB. Along with the other reductions, the RTX Pro 5000 has just two NVENC/NVDEC units enabled. It has a 300W TDP. Nvidia's RTX Pro 4500 switches to the GB203 chip, the same chip found in the RTX 5080 and 5070 Ti. It's also a nearly complete solution, with 82 of the potential 84 SMs enabled alongside all eight memory channels. Like the RTX Pro 5000, it uses 2GB GDDR7 modules in clamshell mode, for 32GB of total VRAM. It also features two NVENC/NVDEC units, and a power limit of just 200W — surprisingly low, considering it's otherwise similar to the RTX 5080 that has a 360W TGP. Last up, the RTX Pro 4000 also uses the GB203 chip, with some severe trimmings in some areas. It has 70 SMs enabled, the same as the RTX 5070 Ti, but the memory interface is only 192 bits wide, the same as the RTX 5070. Nvidia sticks with the 2GB GDDR7 modules as well, which in clamshell mode gives the GPU 24GB total VRAM. Dual NVENC/NVDEC are again present, and the power limit gets slashed to just 140W. All of the models feature a full PCIe 5.0 x16 slot, with 16-pin 12V-2x6 connectors. Nvidia had the RTX Pro 6000 variants on display, which appear to be launching first. The other RTX Pro models will presumably launch in the coming months, and it's not clear when exactly the laptop solutions will become available. Image 1 of 9 (Image credit: Tom's Hardware) (Image credit: Tom's Hardware) (Image credit: Tom's Hardware) (Image credit: Tom's Hardware) (Image credit: Tom's Hardware) (Image credit: Tom's Hardware) (Image credit: Tom's Hardware) (Image credit: Tom's Hardware) (Image credit: Tom's Hardware) Nvidia also announced a change in branding, if you didn't notice. Where the previous professional and data center solutions were sold under various RTX names (RTX A6000/A5500/A5000/etc. for Ampere, then RTX 6000/5000/4500/etc. for Ada), the new Blackwell generation of professional and data center GPUs will use RTX Pro nomenclature. Nvidia already listed the RTX Pro 6000/5000/4500/4000 series GPU names for desktops. For laptops, there will be RTX Pro 5000/4000/3000/2000/1000/500 models, and for data center so far there's only an RTX Pro 6000. That last will, naturally, be the full-fat model with 96GB of GDDR7 ECC memory, taking over from the Nvidia L40. We're glad to see the change in naming, as things were becoming a bit obfuscated after Nvidia killed off its Quadro branding several generations back. Now, RTX Pro will very clearly indicate that something is different from the standard GeForce RTX lineup. There's still the matter of those numbers, however. RTX Pro 6000 quite obviously implies a Blackwell GPU for now, but what will happen with the next generation Rubin (or whatever the codename ends up being for the non-DC parts) GPUs? This is where the Ampere RTX A6000 branding made sense, but Ada parts dropped the "A" and things became a little fuzzy. Hopefully, that's something Nvidia addresses when those future products finally arrive.

Release dates are some of Hollywood’s most valuable natural resources. Often the window when a movie is set for release, or if it’s delayed or advanced, can tell you a lot about what to expect when you sit down in the theater. Warner Bros. just gave its schedule a huge shake-up—and while you don’t have to worry about Superman, it’s safe in July, a few other films are not so lucky. Let’s break it down. In terms of io9 readers, the biggest and most disappointing move has to do with Maggie Gyllenhaal’s The Bride! starring Christian Bale and Jessie Buckley. This is the Bride of Frankenstein movie that may also be a musical and despite that sounding like a huge, huge, swing, we were excited, especially given its original September 26 release date. Alas, that is no longer and now The Bride! has been delayed almost six months, coming out March 6, 2026. It’s a window Warner Bros. has had success with—Dune, Godzilla x Kong, etc.—so we’ve still got our fingers crossed. It’ll open against a new Pixar movie called Hoppers. With that September date open, it shifted a few things. One, the studio moved Paul Thomas Anderson’s latest, now officially titled One Battle After Another and starring Leonardo DiCaprio, from its original August 8 release date to September 26. And that August 8 date is now occupied by Weapons, the latest from Barbarian director Zach Cregger, which was moved up from January 2026. Releasing Weapons five months earlier is a very strong and encouraging move for a film that’s described as a sprawling, multi-character epic a la Nashville or Magnolia, but just with a more thriller/horror vibe. It stars Josh Brolin, Julia Garner, and Alden Ehrenreich. Weapons will now show down with Freakier Friday. But not all the moves are encouraging. Flowervale Street, the new film from It Follows director David Robert Mitchell, was originally set for release this summer. It was then moved to March 2026 and now, with The Bride! in that window, it’s been pushed again to August 14, 2026. Ewan McGregor and Anne Hathaway star and it may or may not have dinosaurs, so we’re hoping the extra time is just to make sure it’s right, and not to bury it in any way. The final move is the least exciting but still significant. Have you been wondering what Bill Hader’s been up to since he finished Barry? Well, he’s been lending his voice to a new animated version of Dr. Seuss’ The Cat in the Hat along with Quinta Brunson, Bowen Yang, Xochitl Gomez, Matt Berry, and Paula Pell. Originally set for release March 6, 2026, now it’s out a week later: February 27, 2026. Everything else is, for now, the same: Minecraft on April 4, Sinners on April 18, Final Destination: Bloodlines on May 16, F1 on July 27, Superman on July 11, and Mortal Kombat 2 on October 24. Which are you most excited about?

A Florida man’s hankering for feral pig landed him in the hospital. In a recent case report, doctors describe how the man likely contracted a serious and rare infection from swine meat he received from a hunter. Scientists and health officials in Florida detailed the porcine mishap this month in the journal Emerging Infectious Diseases. The man endured multiple trips to the hospital and needed his chest implant removed before doctors identified what was making him sick: pig-derived bacteria called Brucella suis. Fortunately, once the bacteria was found, the man was successfully treated with the right mix of antibiotics. According to the report, the man was a pastor in his 70s who had a history of heart failure—a condition that necessitated an automated implantable cardiac defibrillator (AICD) device to keep his heartbeat steady. Starting in spring 2019, he developed chest issues that led to multiple hospital stays. His doctors at the time determined that some sort of infection had taken root in his device, but initial tests failed to find any suspects. The man underwent multiple antibiotic treatments but stopped due to the side effects. After stopping the antibiotics, he seemed fine for a few months, but once again visited a hospital in Alabama with left chest discomfort and fever. He then went for a follow-up visit at the Malcom Randall Veterans Affairs Medical Center in Florida in the fall of 2020, which was where doctors decided to remove his implant and finally nabbed the culprit behind his infection. B. suis is one of several bacteria that cause the zoonotic disease brucellosis (zoonotic meaning they’re primarily spread from animals to humans). Four species are known to sicken people, and infections are usually caught from drinking raw milk and cheese or from direct exposure to infected animals. The man lived on a farm with plenty of potential animal vectors, but he reported never coming into contact with them nor was he a hunter. He did recall, however, that he was gifted feral pig meat from a local hunter on several occasions in 2017—meat that he handled with his bare hands before cooking and eating it. While other possibilities exist, the doctors believe this gift “likely served as his exposure to B. suis.” Brucellosis is no joke. These bacteria were once formally designated as potential bioterror threats, since they can be easily aerosolized, which then makes them very contagious (they were recently removed from this list to make research into them easier). The infection can also be difficult to identify with conventional tests, and it can cause chronic, debilitating symptoms like arthritis and heart inflammation if left untreated (unlike many other bacteria, these germs can make themselves cozy inside our cells). Indeed, the man’s infection was misidentified as a different bacteria prior to his last hospital visit. The man underwent surgery to remove his implant, which could have potentially exposed the infection to his medical care team. Fortunately, the risk of aerosolization was low, proper precautions appeared to have been followed, and no other cases among staff were found. As for the man himself, he was given the standard antibiotic treatment for brucellosis, which seemed to work. Four months after his tainted implant was removed, a new one was placed. And three years later, he had no signs of an active infection. Brucellosis is very rare in the U.S., with only around 100 cases reported annually, most of which are from other species besides B. suis. But feral pigs are a persistent and growing presence in parts of the U.S., Florida included. And the authors say that hunters and anyone else who loves feral pig meat should be made well aware of the potential danger. “Feral swine hunters are at risk of contracting B. suis, and measures to reduce that risk include the use of personal protective equipment and thoroughly cooking animal products before consumption,” they wrote.

In a upset geopolitical context, Europe, of which France strives to organize resistance, in particular to reduce its technological dependence on American giants. All tech news for pros every day in our newsletter learn more about the use of personal data But between speeches and acts, the gap sometimes remains gaping. Tuesday, March 18, the Dinum organized a new edition of its “State in the cloud” event. Clara Chappaz, Minister Delegate to AI and Digital, intervened on this occasion, thus encouraging to use public procurement for the benefit of European solutions. Choosing European is not a taboo for Clara Chappaz The “at the same time” dear to President Macron has however been there. In a notice of March 14, national education and youth confirms Microsoft's choice for a minimum amount of 74 million euros. The value of the framework contract could even reach 152 million euros. Four days after the publication of this opinion, Clara Chappaz nevertheless had an offensive discourse on the issue of sovereignty before representatives of the public sector. Turning to a non-European solution, "it's not bad in itself", recognizes the minister. "But I think you have to realize that the money we spend does not support economic development in our country," she said. Sovereignty on the cloud, not the on-premise? Clara Chappaz here referred to the cloud market. But the reasoning of the government representative could apply as well to the acquisition of software for the workstation and the datacenters, as well as national education. Indeed, the delegate minister called for "being more strategic in the way" of which the public sector uses her money by directing public procurement towards European solutions. This orientation of the order "should no longer be a taboo today," she said on March 18. During her speech, the minister also promised actions for the benefit of European technological offer. Moreover, as part of a call for projects from France 2030, the State funded a consortium of publishers and the development of a sovereign collaborative solution: Collabnext. The previous one in 2015 and already unconverted speeches In 2022, the Ministry of National Education indicated that it has given the judgment of any office deployment 365, the Cloud de Office edition, as well as Google equivalent products. Reason: incompatibility with the cloud strategy in the center of the state. On the On-premise, a long way remains obviously to be traveled. A previous contract concluded 10 years earlier between the ministry and Microsoft triggered controversy, trampling on the Ayrault circular on free software in administration. Ditto for the 2013 law provided to give priority to free in higher education. By renewing its contract with the American publisher, the Ministry of Education lacks an opportunity to demonstrate an inflection in its purchasing policy and to assert a sincere attachment to more digital sovereignty in France.

Solidigm, a US-based subsidiary of SK hynix specializing in storage solutions for data centers has introduced what it claims to be the world's first liquid-cooled eSSD (Enterprise SSD) solution for AI servers. Solidigm presented its D7-PS1010 E1.S eSSDs at GTC, paving the way for fully liquid-cooled servers in the future. SSDs are an extremely vital component of servers designed for AI due to their fast data access requirements. While SSDs are prone to endurance limits, AI workloads are generally more read-intensive. This coupled with the lack of mechanical parts, higher efficiency, and predictable response times makes SSDs the go-to choice for AI-centric data centers. In an attempt to foster fanless server designs, Solidigm has pioneered the world's first liquid-cooled eSSD solution. Traditional DLC (Direct Liquid Cooling) solutions for eSSDs fail to adequately cool both sides of drives. Additionally, the design is not hot-swappable, incurring downtime during replacements. Solidigm's D7-PS1010 E1.S eSSD overcomes these limitations as its cold plates have been engineered to keep temperatures at both sides under check. Furthermore, the design is highly serviceable, allowing technicians to hot-swap SSDs from the rear using a spring-loaded mechanism. (Image credit: Solidigm) Solidigm says these eSSDs can eliminate fan-based 1U rack designs, decreasing HVAC and air-cooling costs for data centers. Industry-wide adoption of D2C (Direct to Chip) cooling solutions can set a standard for future data centers to be built with liquid-cooling infrastructure in mind. Just a few days back, CoolIT revealed a single-phase DLC cold plate, advertised to dissipate 4kW of heat for next-gen AI accelerations, like Nvidia's Blackwell Ultra B300 chips. The Solidigm D7-PS1010 E1.S will also launch in a 15mm form factor for already-existing air-cooled servers and other storage solutions. Servers still employ certain components that are not liquid-cooled, wherein heat from these components must be dissipated through HVACs and other air-cooling solutions. We're still faraway from immersion cooling, however. While they have several advantages of traditional DLC solutions, existing infrastructure cannot be easily molded to suit their requirements. Attendees at Nvidia's GTC event can see this SSD at Booth #1602. Solidigm says the D7-PS1010 E1.S is set to launch in the second half of this year for AI servers, presumably alongside Nvidia's B300 accelerators.

The European Commission yesterday presented its preliminary conclusions concerning the alphabet ecosystems, the mother house of Google and Android. All tech news for pros every day in our newsletter learn more about the use of personal data It considers that Google is not in good standing with the obligations imposed by the DMA (Digital Markets ACT). These obligations oblige "access controllers" - global large platforms - to process third -party publishers offering their services in their ecosystem in a fair manner. Favoritism on Google Search The Commission believes that Alphabet, Google’s parent company, on the contrary promotes its own services and products within its Google Search search engine. How ? By offering them a prominent place. In particular, she explains that she has noted that the company's services were better referenced on its own search engine. And this at the expense of competitors. And to quote specifically: Google Shopping Hotel room reservations Transport booking services Information services on sports results and finance This is not the first time that Google has been accused of promoting its own services in its search engine. The company has already been convicted of abuse of a dominant position in 2017 after having favored the results of its own price comparison tool, Google Shopping, on its search engine. The commission then inflicted on the American company a fine of 2.42 billion euros. Google Play's “overflow” in the viewfinder The Commission also looks at the Google Play ecosystem. She maintains that Alphabet imposes constraints on third -party publishers who wish to redirect their users to alternative distribution channels. For these publishers, this makes it possible to escape the taxes imposed by Google on the Google Play ecosystem. Some commissions taken by Google go "beyond what is justifiable" depending on the European body. In particular, it targets additional costs inflicted on in-app purchases. This preliminary step of the process does not impose any particular obligation on alphabet for the moment. The company can meet the Commission's observations and access the documents gathered by the Commission during its investigation. The commission must then determine if alphabet actually violated the regulations. And if changes are necessary for the company to comply with European law. The European Commission also published yesterday the new obligations imposed on Apple in terms of interoperability.

The ugly and embattled Cybetruck took another blow this week after the federal regulators issued a recall. According to the recall, Tesla must take back the vehicles to repair a bit of the trim that can fly loose while driving “increasing the risk of a crash.” The recall affects more than 46,000 vehicles. Despite claims from Elon Musk that the Cybetruck is a bulletproof beast, it’s actually a shoddily manufactured vehicle. Drivers and people who test the durability of cars have known this for a long time. This recall, which covers every single Cybetruck since it went on sale, is the 8th one for the vehicle. It’s the second time the Cybertruck has been recalled over a trim issue. In June last year, the company issued a similar recall because “the trunk bed trim sail applique could have been improperly adhered and can loosen.” Tesla told Bloomberg that it believes only 1% percent of the recalled Cybertrucks will have the issue, but it’s calling all the vehicles back anyway and will foot the bill for a fix. It also said that it had not fixed the problem in the manufacturing process, but expected to roll it out around March 21. It said it had already received 151 warranty claims about the trim issue. The reason the trim sucks and is prone to flying away from the Cybertruck is that it’s just glued on. In a video published in August 2024, YouTuber WhistlinDiesel put the Cybetruck through various durability tests and compared it to the more popular F-150. In the video WhistlinDiesel appears shocked when he’s able to peel away various parts of the trim from the Cybertruck using just his hands. The clip of the trim being peeled away got a second life when WhistlinDiesel took to TikTok. A cut from the longer video has been viewed more than 20 million times online. In it, WhistlinDiesel slams the Cybertruck with a pickaxe before attempting to peel away pieces of the trim with the tool. He drops it because he doesn’t need it. He only needs to use his hands to disassemble the car and you can see it’s held together with glue. “That is not acceptable dude. You just glued your truck together. That’s not OK. That is not OK,” he said in the video. The high-profile recall of the Cybetruck is another blow to Tesla. Its stock price plummets every day. There are protests at the Tesla showrooms around the country. People are setting the cars on fire and spray painting swastikas on them. Long-time Tesla owners are selling the cars in massive numbers. The cost to insure the vehicles is skyrocketing. Major Tesla investor Ross Gerber has made multiple television appearances in the past few days calling on Musk to step down as CEO. “If Elon came back to Tesla, that would be great. That’s exactly what Tesla needs. But he’s not. And he’s gonna stay at DOGE and he’s one of the most powerful people in the world now it’s definitely gotten to his head and I just don’t see him changing and going back. I actually think he’s completely neglecting Tesla at this point,” Gerber told Bloomberg TV. “I think Elon should appoint a new CEO to run Tesla so he can focus on what he’s doing in the government and also separate himself from Tesla,” he continued. “Which is the great company that’s trying to solve climate change. It’s now associated with Elon in such a negative fashion that it’s destroying the sales of their main product, which is selling EVs. Somethings gotta change and I think it’s gonna happen this year.” Musk has attempted to use his connection to the White House to bolster sales and repair Tesla’s image. It hasn’t worked. Trump recorded what amounts to an ad for the vehicles on the South Lawn of the White House. The stock went down. On March 19, Treasury Secretary Howard Lutnick went on Fox News and pushed people to buy the stock. “I think, if you want to learn something on this show tonight, buy Tesla,” Lutnick said. “It’s unbelievable that this guy’s stock is this cheap. It’ll never be this cheap again.” The stock got cheaper on Thursday morning after his appearance.

Dark energy is a dynamic entity, rather than a matter-of-fact constant in the universe’s rapid unfurling, a team of scientists announced this week. The Dark Energy Spectroscopic Instrument, or DESI, is an international effort involving more than 900 researchers. The project has produced gargantuan sky maps and helped reveal some of the universe’s largest structures, including the source of the universe’s largest known jets. The latest DESI release shakes up the foundational understanding of dark energy and its role in the universe. As best scientists can tell, the vast majority of the universe is made up of dark matter and dark energy. Dark matter is the catch-all term for about 27% of the universe that is only known to exist because of its gravitational effects on other matter. The stuff is so infinitesimally small or otherwise elusive that it cannot be observed with telescopes, which take in wavelengths of light emitted by objects to better understand those sources. Observable matter–everything from the coffee cup on your desk, to distant planets, to the largest and most ancient galaxies—makes up just 5% of the cosmos, meaning that dark energy is responsible for a whopping 68% of everything we think exists in the universe. The notion of dark energy as a constant—which is to say, it manifested the same way 10 billion years ago as it does today, and as it will 10 billion years from now—is “baked into” the predominant model of the universe, Lambda-CDM, according to Rossana Ruggeri, a physicist at the University of Queensland who was involved in the DESI analysis. “Results from the first batch of data gave a hint that dark energy might not behave like a simple cosmological constant—but it wasn’t strong enough to draw firm conclusions,” Ruggeri said in The Conversation. “Now, the second batch of data has made this evidence stronger.” Though the data does not yet meet the statistical threshold physicists require to firmly declare a bona fide new discovery, the data strengthen physicists’ resolve that something is going on that will mean the model needs to be revised. “If dark energy is changing over time, it could have profound implications for the ultimate fate of the universe,” Ruggeri wrote. According to the new DESI numbers, dark energy may be changing over time—either causing the universe’s expansion to accelerate or driving an inward collapse, which would theoretically cause everything we know to collapse into a singularity in a “Big Crunch.” The team revealed its findings at the American Physical Society’s Global Physics Summit, and the results are hosted in several papers posted to the preprint server arXiv. At 270 terabytes, DESI’s first data release is just a fraction of what’s to come from the instrument’s five-year survey of the cosmos. It contains information about 18.7 million objects deep in space, including the distances to some extremely remote galaxies, covering 11 billion years of the universe’s history. The data release is free and available for public access; you can also explore some of the instrument’s data via the Legacy Survey Sky Browser. “The DESI project has maintained the pace of making 3D maps of the universe that are 10 times larger every decade,” said David Schlegel, one of the lead scientists at Berkeley Lab for both DESI and SDSS, in a laboratory release. “That’s our version of Moore’s Law for cosmology surveys. The rapid advance is powered by the clever combination of improved instrument designs, technologies, and analysis of ever-fainter galaxies.” The instrument works rapidly; with clear observing conditions, it can collect data on more than 100,000 objects in a single night. DESI collects its data from its perch atop the National Science Foundation’s Nicholas U. Mayall 4-meter telescope, an NSF NOIRLab program at Kitt Peak National Observatory in Arizona. The release contains more than twice as many unique objects outside of the Milky Way than in all previous 3D spectroscopic surveys combined—a testament to the amount of information scientists can garner about the universe at scale, such as the role of dark energy in its expansion. DESI is currently in its fourth year of a five-year data collection period, and the team intends to record spectra for more than 50 million galaxies and quasars (active galactic cores) before it ends. Thereafter, there will be a lot of data processing and analysis to do. But once it’s done, scientists may have a revised theory of dark energy’s exact machinations in our universe, and how its inexorable role in the universe’s grand scale changes over time.

Discord’s mobile app will have video ads starting in June, the company announced today. The initial pilot for the video ads, which Discord calls Video Quests on Mobile, will offer advertisers the ability to “showcase trailers, make impactful announcements, and highlight premium content” to users, Discord said. Discord was a proudly ad-free platform until March 2024, when it introduced ads to its desktop and console apps. Those ads offer Discord users rewards for PC games if they play certain games or get people to watch a stream of their gameplay through Discord. Discord followed up with Video Quests, which let developers show Discord users video ads, like trailers and announcements of new seasons and downloadable content. Discord users see prompts for both types of ads on the bottom-left side of their screen and can choose to expand or ignore them. Discord users can also opt out of personalized promotions and “hide an in-app promotion for a specific Quest or game you’re not interested in,” Discord said. In June, users of Discord’s mobile apps will see Video Quest advertisements on the bottom of the screen as depicted below: Discord First, the ad prompt appears. First, the ad prompt appears. Discord Discord This is what the prompt looks like expanded. This is what the prompt looks like expanded. Discord Discord If users click "Accept Quest" on the expanded prompt, a video ad like this will play. If users click "Accept Quest" on the expanded prompt, a video ad like this will play. Discord This is what the prompt looks like expanded. Discord If users click "Accept Quest" on the expanded prompt, a video ad like this will play. Discord Discord Users can then claim rewards... Discord ...and be assured that they've received said rewards. “Expanding our advertising platform to mobile is an obvious, natural evolution in our strategy," Peter Sellis, Discord’s product SVP, said in a statement accompanying today’s announcement. "Our mission is to create the most authentic, player-centric advertising platform in the galaxy."

Tesla said Thursday it would recall nearly all the Cybertrucks it has produced to repair a stainless steel trim panel that the carmaker said could become detached from the vehicle while driving. It’s the eighth recall for the distinctive all-electric pickup, which hit US roads just over two years ago. Unlike many other Tesla-related recalls, though, this one cannot be fixed with an over-the-air software update, and must be brought into a service center for repairs. The recall affects more than 46,000 Cybertrucks manufactured between November 2023 and late February 2025. The recall is related to a cosmetic applique on the vehicle’s exterior, which is attached to the rest of the truck by a kind of glue, according to a filing published by the National Highway Traffic Safety Administration, the US’ road safety regulator. On affected Cybertrucks, the adhesive can separate from the rest of the truck, creating a possible “road hazard” for others and increasing the risk of a crash. Complaints about the trim first bubbled up in social media posts; Tesla reportedly halted Cybertruck production last week. Tesla says in the filing that it began to investigate the issue in early January. The filing says Tesla will repair the issue with another kind of adhesive “not prone to environmental embrittlement.” Dale Harrow, chair and director of the Intelligent Mobility Design Center at the Royal College of Art London, labels this latest glue recall for the Cybertruck as “surprising”. “Glues are used a lot more than people think in car body construction these days,” he says. “Rather than having a mechanical fixing, weld or a screw or a bolt, it's very effective gluing panels together for a lighter construction. It's become very popular. Jaguar, Lexus, Audi, they'll use glues at some point. So I'm very surprised. It's not as if it's an unknown science. I've not heard [of problems] anywhere else.” There are numerous glues the auto industry uses to bond panels and parts together, Harrow says, adding that different adhesives are used depending on strength of bond needed, particular circumstances, and materials involved. However, what is different with the Cybertruck to other vehicles is its highly unconventional use of stainless steel for its panels. “Stainless steel is the big difference," Harrow says. "So maybe there's been an issue there?" Harrow points out this fault appears to be happening in cold weather, and that it might be that the stainless steel is conducting cold better, possibly just enough to make a difference in the curing process. "I could surmise that something on the production line is not quite bonding at the right temperature or getting the right UV coverage? And that could be due to the stainless steel having a different density from standard steel. It's not a small company, and it's not doing things on the cheap. So I'll be very surprised if there wasn't the knowledge within the company to [fix] this. But it would need quite a bit of investigation, I think, to find out the real cause of it.”

Jensen Huang, chief executive of Nvidia, stated that his company had not been approached to participate in a group effort to acquire a stake in a company that would operate Intel's foundry unit. He dismissed claims that Nvidia was working with industry peers and TSMC on such a deal at a press conference at GTC, reports Reuters. Separately, Paul Liu, a TSMC board member and the head of Taiwan's National Development Council, denied claims that the company is considering purchasing Intel's struggling foundry unit, reports DigiTimes. "Nobody has invited us to a consortium," Huang said, according to Reuters. "Nobody invited me. Maybe other people are involved, but I do not know. There might be a party. I was not invited." While speaking to Taiwan's Legislative Yuan Economic Committee on March 19, Paul Liu stated that the acquisition of Intel Foundry has never been discussed at the board level and compared it to mixing two incompatible substances, something that one in the semiconductor industry can consider both figuratively as Intel and TSMC have vastly different corporate cultures and literally as the two companies use different chemical substances for manufacturing. As a result, industry experts believe such an acquisition would be more harmful than beneficial to TSMC. Liu explained that stabilizing the company's most advanced production technologies already takes between 18 and 24 months. Exporting such a node to the the U.S. would add another year, making American operations three years behind those in Taiwan, according to Liu. He also addressed concerns about a possible second Trump presidency leading to a shift of Taiwan's semiconductor industry to the U.S., urging lawmakers to see the U.S. as a partner (which accounts for the lion's share of TSMC's revenue) rather than a competitor. Earlier this year there were rumors that the U.S. government pushed TSMC to take over Intel Foundry and operate it. The plan involves Intel spinning off its Intel Foundry unit, which makes chips for itself and external clients. TSMC would acquire less than half of the new entity, while the remaining shares would go to industry partners. The list of industry partners included AMD, Broadcom, Nvidia, and Qualcomm, according to Reuters. However, speaking at the conference, Huang denied any involvement in discussions about a potential consortium to take control of Intel's fabs. For Nvidia, which produces billions of dollars worth wafers every year, a dual sourcing supply strategy could make sense. However, designing large AI/HPC GPUs for different process technologies used by Intel and TSMC would significantly affect Nvidia's costs. Investing in a fab joint venture would also be odd for the company, which started as — and remains — a fabless chip designer. Stay On the Cutting Edge: Get the Tom's Hardware Newsletter Get Tom's Hardware's best news and in-depth reviews, straight to your inbox. Contact me with news and offers from other Future brands Receive email from us on behalf of our trusted partners or sponsors The talks reportedly started before TSMC's March 3 announcement of its $100 billion U.S. investment. This plan includes five additional Fab 21 modules, two advanced packaging facilities, and a research center. However, according to Reuters' own report, the conversations about the fab joint venture continued after the announcement as TSMC was seeking agreements with major fabless chip design companies. As a result, Taiwanese lawmakers raised concerns following such media reports and pressured Liu for clarifications.

Jensen Huang, chief executive of Nvidia, stated that his company had not been approached to participate in a group effort to acquire a stake in a company that would operate Intel's foundry unit. He dismissed claims that Nvidia was working with industry peers and TSMC on such a deal at a press conference at GTC, reports Reuters. Separately, Paul Liu, a TSMC board member and the head of Taiwan's National Development Council, denied claims that the company is considering purchasing Intel's struggling foundry unit, reports DigiTimes. "Nobody has invited us to a consortium," Huang said, according to Reuters. "Nobody invited me. Maybe other people are involved, but I do not know. There might be a party. I was not invited." While speaking to Taiwan's Legislative Yuan Economic Committee on March 19, Paul Liu stated that the acquisition of Intel Foundry has never been discussed at the board level and compared it to mixing two incompatible substances, something that one in the semiconductor industry can consider both figuratively as Intel and TSMC have vastly different corporate cultures and literally as the two companies use different chemical substances for manufacturing. As a result, industry experts believe such an acquisition would be more harmful than beneficial to TSMC. Liu explained that stabilizing the company's most advanced production technologies already takes between 18 and 24 months. Exporting such a node to the the U.S. would add another year, making American operations three years behind those in Taiwan, according to Liu. He also addressed concerns about a possible second Trump presidency leading to a shift of Taiwan's semiconductor industry to the U.S., urging lawmakers to see the U.S. as a partner (which accounts for the lion's share of TSMC's revenue) rather than a competitor. Earlier this year there were rumors that the U.S. government pushed TSMC to take over Intel Foundry and operate it. The plan involves Intel spinning off its Intel Foundry unit, which makes chips for itself and external clients. TSMC would acquire less than half of the new entity, while the remaining shares would go to industry partners. The list of industry partners included AMD, Broadcom, Nvidia, and Qualcomm, according to Reuters. However, speaking at the conference, Huang denied any involvement in discussions about a potential consortium to take control of Intel's fabs. For Nvidia, which produces billions of dollars worth wafers every year, a dual sourcing supply strategy could make sense. However, designing large AI/HPC GPUs for different process technologies used by Intel and TSMC would significantly affect Nvidia's costs. Investing in a fab joint venture would also be odd for the company, which started as — and remains — a fabless chip designer. Stay On the Cutting Edge: Get the Tom's Hardware Newsletter Get Tom's Hardware's best news and in-depth reviews, straight to your inbox. Contact me with news and offers from other Future brands Receive email from us on behalf of our trusted partners or sponsors The talks reportedly started before TSMC's March 3 announcement of its $100 billion U.S. investment. This plan includes five additional Fab 21 modules, two advanced packaging facilities, and a research center. However, according to Reuters' own report, the conversations about the fab joint venture continued after the announcement as TSMC was seeking agreements with major fabless chip design companies. As a result, Taiwanese lawmakers raised concerns following such media reports and pressured Liu for clarifications.

A mother suing Character.AI after her son died by suicide—allegedly manipulated by chatbots posing as adult lovers and therapists—was horrified when she recently discovered that the platform is allowing random chatbots to pose as her son. According to Megan Garcia's litigation team, at least four chatbots bearing Sewell Setzer III's name and likeness were flagged. Ars reviewed chat logs showing the bots used Setzer's real photo as a profile picture, attempted to imitate his real personality by referencing Setzer's favorite Game of Thrones chatbot, and even offered "a two-way call feature with his cloned voice," Garcia's lawyers said. The bots could also be self-deprecating, saying things like "I'm very stupid." The Tech Justice Law Project (TJLP), which is helping Garcia with litigation, told Ars that "this is not the first time Character.AI has turned a blind eye to chatbots modeled off of dead teenagers to entice users, and without better legal protections, it may not be the last." For Garcia and her family, Character.AI chatbots using Setzer's likeness felt not just cruel but also exploitative. TJLP told Ars that "businesses have taken ordinary peoples’ pictures and used them—without consent—for their own gain" since the "advent of mass photography." Tech companies using chatbots and facial recognition products "exploiting peoples’ pictures and digital identities" is the latest wave of these harms, TJLP said. "These technologies weaken our control over our own identities online, turning our most personal features into fodder for AI systems," TJLP said. A cease-and-desist letter was sent to Character.AI to remove the chatbots and end the family's continuing harm. "While Sewell’s family continues to grieve his untimely loss, Character.AI carelessly continues to add insult to injury," TJLP said. A Character.AI spokesperson told Ars that the flagged chatbots violate their terms of service and have been removed. The spokesperson also suggested they would monitor for more bots posing as Setzer, noting that "as part of our ongoing safety work, we are constantly adding to our Character blocklist with the goal of preventing this type of Character from being created by a user in the first place."

Nvidia will procure “several hundred billion” dollars’ worth of chips and other electronics manufactured in the U.S. over the next four years, CEO Jensen Huang said. The chip designer will source these products from suppliers like Taiwan Semiconductor Manufacturing Company (TSMC) and Foxconn, which can manufacture its latest systems in the U.S., Huang told the Financial Times (FT) in a report posted Thursday (March 20). By doing so, Nvidia will avoid tariffs and will improve the resiliency of its supply chain, according to the report. Nvidia has been highly dependent on TSMC’s manufacturing facilities in Taiwan, which are subject to the risks of threats from China, earthquakes and potential U.S. tariffs, the report said. Huang also said that the Trump administration’s policies will give a boost to Nvidia’s efforts to build data centers, which require vast amounts of energy, per the report. “Having the support of an administration who cares about the success of this industry and not allowing energy to be an obstacle is a phenomenal result for AI in the U.S.,” Huang told the FT. Huang’s remarks follow announcements from some other tech companies that are increasing their investment in the U.S. Apple said in February that it plans to invest more than $500 billion in the U.S. in the next four years. The company’s plan includes hiring 20,000 workers and opening a manufacturing facility in Houston to make servers to support its artificial intelligence (AI) system, Apple Intelligence. The company also said it plans to keep expanding data center capacity in North Carolina, Iowa, Oregon, Arizona and Nevada, while opening the “Apple Manufacturing Academy” in Detroit to help companies pivot to advanced manufacturing. TSMC said March 4 that plans to expand its investment in advanced semiconductor manufacturing in the U.S. by an additional $100 billion, bringing its total investment in the country to $165 billion. TSMC already has an ongoing $65 billion investment in its operations in Phoenix. The company’s latest investment plans include three new fabrication plants, two advanced packaging facilities and a major research and design (R&D) team center. The investment is expected to support 40,000 construction jobs over the next four years and create tens of thousands of tech jobs. For all PYMNTS AI coverage, subscribe to the daily AI Newsletter.

The Starliner Saga May not be over Just Yet. Despite a failed test flight that Left a crew beached in space, nasa still hasn’t given up on Boeing’s ill-fated SpaceCraft. Starliner May Get Another Chance to Transport Astronauts to the International Space Station (ISS). During A Briefing On Wednesday, NASA Officials revealed that they looking into launching another test flight of Boeing’s Starliner SpaceCraft Before Certifying the Vehicle for Regular Crew Rotation Missions. “What we’d like to do is that one flight and then get into a crew rotation flight,” Steve Stich, Nasa Commercial Crew Program Manager, Said, According to Spacenews. "So, the next flight up would real test all the Changes We’re Making to the Vehicle, and then the next fight beyond that, we real need to get Boeing into a crew rotation. So, that's the strategy." Boeing’s Starliner Launched to the Iss on June 5, Carrying Nasa Astronauts Butch Wilmore and Suni Williams for the SpaceCraft’s First Crewed Test Flight. Things did not go Smoothly, However, as Five of the Spacecraft’s Thruster on Its Way to the Iss, and Starliner Returned Home Empty After Nasa Deemed It Unfit to Fly the Crew Back to Earth. The Space Agency Came Up with a Nifty Plan to Return the Astronauts from Space by Hitching A Ride Aboard a SpaceX Dragon Crew Vehicle. An upcoming Starliner Test Flight Will Verify That The SpaceCraft’s Thruster Function Properly in Space After Malfunction Propulsion System. Boeing is Planning A Ground Test Of The Propulsion System This Summer to Review the Company’s Fixes. A subsequent flight to the Iss Will Demonstrate the SpaceCraft’s Ability to Operate in A Space Environment That’s Difficult To Replicate On Earth, Stich Said During the Briefing. Once the Changes have been validated, nasa hopes to include the starliner SpaceCraft in its crew mission schedule—on currently dominated (though “monopolized” might be a better word) --by SpaceX and its crew dragon spacecraft. NASA AWARDED BOING AND SPACEX Contracts in 2014 as part of the space Agency’s Commercial Crew Program to Develop SpaceCRAFT CAPABLE OF CARYING CREW and Cargo to the ISS. So Far, Boeing Hasn’t Met the End of Its $ 4.3 Billion Commercial Crew Program Contract with NASA While Its Competition, SpaceX, Has Flown 10 Crews to the Iss and Back. Starliner’s Crewed Test Flight Was Supposed to Be Its Last Before Being Certified to Transport Crew to the Iss on a Regular Basis, Along with SpaceX. But the Botched Flight Prompted Nasa to Delay Boeing’s CST-100 Starliner’s Next Chance at Flying to the Iss. The Space Agency Had Hoped Starliner would launch it's first crew mission by Early 2025. Obviously, that didn’t. Instead, Nasa Tapped Its More Reliable Commercial Partner, SpaceX, to launch the crew-10 mission on March 12. SpaceX’s Dragon was used to retrieve the Two Starliner Astronauts from the ISS. Despite Starliner’s Disastrous Crew Flight, NASA is still Willing to Give Its Commercial Partner Another Chance at Carrying Astronauts to And From The Iss So That It Can Have Two Private Vehicles As Options Instead Of Just The One, And Weean Itself Off Full Dependency On Space.

is a news writer who covers the streaming wars, consumer tech, crypto, social media, and much more. Previously, she was a writer and editor at MUO. TikTok will now show AMBER Alerts in the US. If there’s a missing child in your area, you’ll see an AMBER Alert directly within your For You feed. In the example shared by TikTok, the AMBER Alert shows a photo of the child, along with relevant details and an option to call 911. TikTok is launching the feature in partnership with the National Center for Missing & Exploited Children, a nonprofit organization mostly funded by the US government. Image: TikTok The app began piloting the feature in Texas last year, and said the AMBER Alerts on the platform were viewed more than 20 million times. Despite facing an uncertain future in the US, TikTok has continued to add features geared toward child safety, such as telling teen users to “wind down” after 10PM. In addition to the automatic AMBER Alert notifications sent to phones in the US, other apps have also added AMBER Alerts over the years, with both Uber and Waze bringing them to their platforms in 2015. Facebook, Instagram, and X have also integrated missing child notifications.

When shopping for a portable solid-state external hard drive, reliability is key. After all, backing up your files is pointless if your drive fails, becomes corrupted, or completely bricks for no apparent reason. That’s why choosing a trusted brand is essential—and few names are more respected than SanDisk. The SanDisk SSD has been a top choice since its initial release at the end of 2020, earning its reputation as one of the most dependable external SSDs on the market. With over 75,000 Amazon reviews and a 4.6-star average rating, this drive is also an Amazon’s Choice selection. Right now, you can grab the 2TB model of the SanDisk Extreme at a 32% discount on Amazon—bringing the price down to just $136 (from $199). See at Amazon Portability and Durability A portable hard drive needs to be, well, portable—and the SanDisk Extreme nails it. Measuring just 4 inches by 2 inches and under half an inch thick, this lightweight drive weighs only 0.1 pounds, making it easy to slip into your pocket, bag, or briefcase. For added convenience, it features a built-in carabiner loop, allowing you to securely attach it to a backpack or camera bag. Worried about durability? This is the upgraded Gen 2 model of the SanDisk Extreme, featuring IP65-rated water and dust resistance—an upgrade from the previous model’s IP55 rating. It’s also built to withstand drops of up to 3 meters (9.8 feet), so you can trust it to handle rough conditions. Blazing Fast Speeds and Secure Storage The SanDisk Extreme Portable SSD delivers high-speed performance, with read speeds up to 1,050MB/s and write speeds up to 1,000MB/s. Whether you’re a photographer, videographer, or just need a reliable backup drive, this SSD has the speed to handle large files effortlessly. Plus, your data stays protected with 256-bit AES hardware encryption and built-in password security. The SanDisk Extreme is broadly compatible across PC and Mac laptops and desktops, and Android and Apple smartphones and tablets. Android users get a leg up when it comes to file management when they download the free SanDisk Memory Zone app from Google Play. SanDisk SSD sales are always worthy of attention, and this 32% discount is an excellent opportunity to upgrade your storage solution to the latest version of the SanDisk Extreme with updated firmware and greater resistance to drops, water, and dust. See at Amazon

The Pentagon and Lockheed Martin swear there’s no secret “kill switch” that they can flip to remotely disable the F-35 fighter jet. But the logistical supply chains and software requirements of the jet make it impossible to operate without U.S. support. Rumors of a magic button the U.S. can use to turn off one of its most advanced and expensive weapons exports have circulated widely online over the past month. On the surface, the rumor is ridiculous but the fear is widespread and has been discussed among serious enough people that the U.S. military-industrial complex has had to issue multiple concrete denials. While the F-35 has no easy “off” button it is a complicated machine that requires constant American support and fears of a kill-switch reflect real anxieties in its allies and customers. The F-35 is a bestseller. Sure the aircraft gets into a lot of accidents but that hasn’t stopped manufacturer Lockheed Martin from selling them all over the planet. Australia, Belgium, Denmark, Finland, Israel, and several other countries have purchased the jet. But Trump’s election and his disruption of NATO and traditional treaty systems have been bad for business. Canadian Prime Minister Mark Carney said earlier this week that he is reviewing its current contract to acquire the fighters and might reconsider, given the “geopolitical environment.” Carney is following the lead of Portugal’s defense minister who said something similar earlier this month. Neither of them mentioned a kill switch, but the rumor began to circulate online in defense circles. The rumor was loud enough for official people to issue official statements. “There is no kill switch,” the Joint Program Office (JPO) for the F-35 program said in a statement. “The program operates under well-established agreements that ensure all F-35 operators have the necessary capabilities to sustain and operate their aircraft effectively. The strength of the F-35 program lies in its global partnership, and we remain committed to providing all users with the full functionality and support they require.” Lockheed Martin posted a copy of the statement on X. ‘There is no kill switch’—The F-35 'kill switch' rumor has been debunked. “After weeks of speculation, the Pentagon formally denies having the ability to remotely disable @LockheedMartin’s F-35 fighter jets.” — F-35 Lightning II (@thef35) March 18, 2025 There is no kill switch for the F-35, but the JPO’s statement points to the very real problems with the weapons system. In its own words, the jet “operates under well-established agreements,” its strength “lies in its global partnership,” and JPO “[remains] committed to providing all users with the full functionality and support they require.” In other words, the F-35 doesn’t fly unless JPO helps you, but don’t worry because it’s committed to helping. Buying an F-35 is a lot like buying an iPhone or a Tesla. When those machines break it’s hard to repair them without going back to the manufacturer. That’s by design. So too with the F-35. A country or military isn’t just buying a jet, it’s buying a whole bunch of software systems and logistical supply chains. Weapons this advanced require a lot of maintenance and support. When an F-35 breaks you can’t just repair it on sight with what you’ve got lying around. American contractors or American-trained contractors have to come in and take care of the problem. As The War Zone put it: you don’t need a kill switch to hobble exported F-35s. According to Lockheed’s own estimation the F-35 global supply chain involves 1,450 U.S. suppliers and 80 suppliers in 11 other countries. Pieces of the jet are made in Texas, Italy, and Japan. Lockheed Martin designed and sold the F-35 in an environment where many of these international relationships were rock solid. They aren’t anymore. One of the biggest problems with the F-35 is a software issue. The jet relies on a proprietary system called Autonomic Logistics Information System (ALIS) and the forthcoming successor Operational Data Integrated Network (ODIN), a cloud-based networking system that manages a lot of the jet’s systems. ALIS/ ODIN is also the system through which the F-35 uploads and downloads Mission Data Files, a collection of planning data and targeting information that’s critical to the jet. “It’s not just a matter of ‘updating software,’” OSINT analyst Bill Sweetman said in a post on X. “The Mission Data File (MDF) is the electronic battle manual for the F-35. It provides known target characteristics for the fusion engine that IDs targets with minimal emissions. It contains threat emitter models that permit the aircraft to follow the minimum-detectability flightpath, the ‘blue line’ track. It runs comms, hosts the electronic order of battle.” The F-35 may not have a “kill switch” in the traditional sense, but the countries who bought it are locked into an irrevocable pact with Lockheed Martin and America. ALIS/ ODIN might not be able to turn off the F-35 remotely, but losing access to it can make it impossible to fly. Only one country has escaped the F-35 software and logistics trap while still being able to fly the jet: Israel. The IDF’s contract for the jet allows it to operate its own software systems without ALIS/ ODIN and conduct its own maintenance.

Consumer advocates and the families of Tesla crash victims both called on Secretary of Transportation Sean Duffy this week not to weaken oversight of the automotive industry. As with other government agencies, since the change of administrations in January, much of the messaging from the Department of Transportation has been on overturning regulations on industry and weakening oversight and consumer protection to create a "level... playing field." On Tuesday, seven family members of people killed or seriously injured as a result of Tesla's Autopilot driver assist wrote to Sec. Duffy with a particular concern: that he not overturn the National Highway Traffic Safety Administration's Standing General Order that requires automakers or operators to report vehicle crashes that occurred while using a partially automated driving system (like Tesla Autopilot or FSD) or an autonomous driving system (like Waymo's robotaxis). In particular, the families are concerned that Duffy will give special treatment to Tesla's Elon Musk, who these days is busier poking his fingers into all corners of the federal government than he is running Tesla or SpaceX. "We fear this important measure is under threat given recent media reports and the influence of Tesla CEO Elon Musk, whose company operates the most widely used Level 2 ADAS [advanced driver assistance] systems in America," the families wrote. NHTSA has logged more than 1,800 crashes involving Autopilot and FSD since 2019, and at least 54 people have been killed as a result of using the systems. This has led the families to ask Sec. Duffy to "maintain robust federal oversight of autonomous vehicles and driver assistance software to preserve the Standing General Order and to ensure that active investigations into Tesla are free from political influence."

is a senior reporter who’s been covering and reviewing the latest gadgets and tech since 2011, but has loved all things electronic since he was a kid. Unlike those digital notebooks, which are appealing because they lack traditional displays, the XP-Pen Magic Note Pad uses a full-color LCD instead of E Ink. The tablet attempts to emulate the E Ink experience with three quick-access color modes, including a black-and-white option. The Magic Note Pad features three color modes accessible through a dedicated button on the side of the tablet. Image: XP-Pen XP-Pen says the Magic Note Pad’s nano-etched surface reduces reflection and glare while also simulating the feel of a pen on paper. It has a “Light Color Mode” that reduces the screen’s saturation to create an experience that sounds closer to how colors might appear on E Ink devices like the Kindle Colorsoft. The “Ink Paper Mode” discards color altogether. All three modes are accessible using a dedicated button on the side of the tablet. The Magic Note Pad includes a magnetic dock for its stylus built into the screen bezel. Image: XP-Pen The included X3 Pro Pencil 2 stylus, which magnetically attaches to a dock on the side of the Magic Note Pad, uses the same electromagnetic resonance technology as Wacom’s pens, so it never needs to be charged. The Magic Note Pad is available through XP-Pen’s online store for $439.99 but is discounted for a limited time to $359.99.

You were relaxed, at ease, when, without warning, your smartphone has been overflowed with notifications. It happened to me too often. And this is what led me to pass my phone in silent mode 24 hours a day and 7 days a week. Receive a selection of strategic advice and concrete tips every week in your mailbox with our newsletter practical guides. Learn more about the use of personal data But Google has introduced new feature in Android 15, called a cooldown notification or notifications limitation in the French version. And it changes everything. The notification limitation function manages the frequency to which notifications can be displayed by an application over time. Yes, this new feature is designed specifically to prevent applications from submerging users with too many notifications at the same time. Do not miss the important notifications What I appreciate in limitation of notifications is that it helps me not to miss important notifications because my smartphone was in silent mode (which happened to me quite often). This new feature appeared with the March 2025 update. It makes it possible to minimize the interruptions which often accompany notifications. Which is particularly problematic when working. Once activated, this function automatically drops the volume and minimize the alerts (up to 2 minutes) when several notifications arrive in a short time. For example, one of your friends discovers a new restaurant that you absolutely have to try. And it starts to send you repeated SMS to tell you how good it is. If the notification limitation function is activated, this avalanche of texts will be in silence. Genius ! You can always lower the notification pane and see the heap of waiting messages And even when the notification limitation function is activated, you can always lower the notification pane and see the heap of waiting messages. The drawback, however, is that this reduction in notifications cannot be personalized (so far). You can activate or deactivate the function. And that's it. In addition, this feature appeared for the first time on Pixel devices. But it should be extended to other devices over time. The limitation of notifications does not apply either to alarms, emergency alerts and priority conversations. How to activate the limitation of notifications What you need: to use the limitation of notifications, you will need a Pixel operating device operating (at least) with the March 2025 update. Hopefully the functionality will soon arrive on other Android devices. 1. Open Settings The first thing to do is to open parameters by pulling the notifications pane twice and pressing the gear icon or opening the application drawer and pressing the parameter launcher. 3. Navigate until notifications In settings, locate notifications and press it. 4. Open limitations of notifications Under the general category, you will see limitation of notifications. Press this input. 5. Activate the limitation of notifications Touch the ON/OFF cursor until it is in position to activate the limitation of notifications. You may find that the function is activated by default. It was originally believed that the limitation of notifications would not happen before Android 16. I suppose that this is another reason why I chose to continue using Pixel smartphones: they generally have the trendy early. And now, my friends in love with Android, everything there is to know about the limitation of notifications. Take advantage of this found peace. Source: "Zdnet.com"

Many computer users do not clean their PCs or peripherals as regularly as they should. Often, the PC case is only opened when something has gone wrong, which can be easily avoided if you take a few minutes out of your day to do a little PC maintenance now and then. Keyboards can collect all types of debris between the keycaps, and if you own pets or have a home susceptible to a little dust, it doesn't take long for PC fans and filters to get clogged up too, and this in turn can lead to overheating and poor PC performance. Air blowers/dusters are a popular choice for quickly blowing out fans and filters to remove any collected dust and fluff and are a great alternative to single-use compressed air cans. Today's deal features the Wolfbox MegaFlow 100 (MF100) Electric Air Duster, which is now only $62 at Amazon. The average list price for this air duster is $89, with a peak price of $139. Checking the Amazon price-checking tool - Camelizer, we can see that this deal is equalling the lowest-ever price offering. Wolfbox's MF100 electric air duster comes with 5 nozzles in varying sizes and 2 different brushes, handy for getting into those hard-to-reach areas. The force of air generated by the MF100 is pretty impressive, thanks to a 150,000 RPM turbofan with a maximum wind speed of 45 m/s. Wolfbox MF100 Electric Air Duster: now $62 at Amazon (was $139) This compact air duster can move air at up to 45 m/s with its 150,000 RPM turbofan. A rechargeable battery gives up to 100 minutes of use time and supports USB Type-C charging. The MF100 includes five nozzles and two brush tools to help clean your PC equipment. This convenient blower is battery-powered and supports USB Type-C charging. The MF100 fully charges in 2.5 hours and provides up to 100 minutes of runtime on the first setting. The MF100 is small and compact, weighing only 0.66lb, so it can be easily stored when not in use. The MF100 is also supported by a 24-month guarantee and 24/7 customer service. Don't forget to look at our Amazon coupon codes for March 2025 and see if you can save on today's deal or other products at Amazon.

Mistral Ai lifts the veil on Mistral Small 3.1, an open source artificial intelligence model of 24 billion parameters, presented as faster and more efficient than its competitors of equivalent size. Multimodal, multilingual and distributed under Apache 2.0 license, it promises an inference speed of 150 tokens per second and a contextual window of up to 128,000 tokens. All tech news for pros every day in our newsletter learn more about the use of personal data After the announcement of its commercial partnerships with Dassault Systems and Stellantis, the young shoot shows with this announcement that it remains in the race for innovation. She ensures that her model surpasses her rivals (Gemma 3, GPT-4O Mini) thanks to an 81 % score at Benchmark Mmlu. All, with a speed generally reserved for models half as voluminous. Talk to more and more customers And this race for power while remaining economical could speak to more and more professional customers. A digital marketing management or an agency could for example adopt Mistral Small 3.1 to quickly create multilingual texts, simultaneously analyze several videos or images from different campaigns. "The Genai is gaining ground, because many traders plan to explore the applications of large language models (LLM) and basic models (FM) in the fields of marketing, sales and customer engagement" under the past year an IDC report which assured that the expenses in Genai solutions in Europe will exceed 30 billion dollars in 2027. In this context, the compatibility of Mistral Small 3.1 with a single RTX 4090 GPU, or a Mac equipped with 32 GB of RAM, greatly facilitates its internal integration. Towards more control of professional data As often at the end of a new model, Mistral AI (just like his colleagues) share flattering metrics. The inference speed of 150 tokens/second and the versatility of the model (textual, multimodal understanding, management of many languages, etc.) will not fail to draw attention. However, independent tests will allow these promises to corroborate or not. To encourage adoption, Mistral AI provides two versions of his model: Mistral Small 3.1 Base Mistral Small 3.1 instruct, which targets more specific uses Being able to locally deploy Mistral Small 3.1 should also arouse the interest of many companies wishing to keep control of their data. Mistral Small 3.1 is already available on Hugging Face.

March Madness is upon us yet again as the NCAA basketball tournament kicks off in earnest: 68 teams facing off in a series of matchups until only one champion is left standing. Millions of sports fans take part by creating their own bracket predictions. The challenge is that there are so many variables and potential outcomes, it's almost impossible to devise the perfect bracket, correctly calling the outcome of every matchup. Online advice abounds, from amateur enthusiasts to professional sports analysts at media giants. (CBS Sportsline's bracket projection model, for instance, simulates every game in the tournament 10,000 times to optimize the accuracy of their predictions.) But what if you're just looking to, say, beat the office pool, or you don't want to simply blindly follow the seedings or predictions of those elaborate models? What simpler strategies might one employ to gain an edge in a crowded field? Albert Cohen of Michigan State University, who specializes in statistics and actuarial science, including sports analytics, isn't a gambler himself. (Someone who "looks at life based on risk" is understandably rather risk-averse when it comes to gambling.) But he did offer Ars some insight into the science of "bracketology," along with a few handy general tips. Ars Technica: There are so many different possible brackets. What are the actual odds of someone picking the perfect one? Albert Cohen: If you look at 64 teams, you're going to have 63 games. How come? Well, it's a geometric series. You have 32 plus 16 plus 8 plus 4 plus 2 plus 1. That's 63 games or 63 coin flips. There are going to be 63 teams that lost, and you can map each of their losses to a game. There's one team that never lost. If you add that up as a formula, which is 2N-1—in this case 264-1—that's 9 quadrillion, or 9 with 18 zeros after it. Complete randomness would be a 9.2 quintillion chance of picking the perfect NCAA bracket.

Plex, the self-serve streaming service that has been powering up millennial nerd home libraries for what seems like ages, announced upcoming price changes for later next month in an astonishingly bashful fashion. The company laid out its plans in a remorseful blog post about what would soon befall longtime users. Plex is increasing the price of everything, including now charging for what will soon be its former best free feature: remote playback. “This price increase will ensure that we can keep investing dedicated resources in developing new features, while supporting and growing your favorites,” swears Plex. Beginning April 29, the monthly Plex pass will increase from $5 to $7, while the annual cost will be set to $70/year. Lifetime subscriptions are also going up in price to more than double, from $120 to $250. If you want that lifetime Plex subscription, you had better buy now. Plex is letting new users lock into that fee before the deadline. The company laid out a roadmap to show that all this change is for growth. Plex plans to bring more parent-focused content through the pipeline as part of its cooperation with Common Sense Media. (I’m sure Plex noticed most millennial parents’ libraries turning into living shrines dedicated to Disney, PBS, and Nickelodeon’s cast of characters.) Plex also plans to introduce a new app for desktop and mobile users, and it’s working on an open API that pulls in more metadata about what you’re watching. But while the new stuff sounds nice I can’t escape the sting of losing free remote playback. Plex doesn’t even mention the upcoming changes to remote playback until it gets through all this—about two complete scrolls through the webpage on a computer monitor. Soon if you’re not at home and want to play your Plex library and don’t already pay for a subscription you’ll have to pay for remote playback. You’ll also have to pay if you want features like the ability to skip an intro song or the ending credits. This is perhaps the most significant blow to a community that’s relied on Plex’s software because streaming is already so expensive. If you want to watch you library away from home you’ll have to pay. It also affects anyone you share your home library with. They will have to pay if they are not Plex Pass subscribers themselves. Plex will offer a Remote Watch Pass tier for $2/month or $20/year. Apparently Plex isn’t immune to the price increases happening industry-wide. The company cites the need for “more resources” to make these changes. Other streaming services have also raised subscription prices, which is only easier to swallow because there’s a big corporation gatekeeping all that content. Plex libraries are personally curated by the people who put them together. This change with Plex comes at a less than convenient time in my life. I recently canceled YouTube TV after the big price hike to $83/ month for the basic package. I switched to ad-supported plans for a few networks where I watch TV and have already saved a lot of money. I’d planned to revert to my Plex library for most of my TV consumption, where I have dozens of shows I promised myself I’d watch and all my comforts, including the three summers worth of YouTube scrapes of the original Real World series. If you’d like me to share those seasons, you’ll have to pony up $2/month first. Or we’ll all need to switch to Jellyfin.

The sixth season of Hulu’s The Handmaid’s Tale arrives in April, bringing its run to an end while also presumably setting up sequel series The Testaments, itself coming to Hulu soon. For all the show’s initial acclaim, many fans had sort of forgotten about it; not entirely surprising, since season five wrapped up all the way back in 2022. With its final installment on the way, we decided to look back at that fiery first season—and realized just how chillingly believable its bleak alt-reality has become. When The Handmaid’s Tale first arrived on Hulu in April 2017, at least half of America was primed and ready for a dystopian series depicting a worst-case-scenario for the country. President Donald Trump had been inaugurated into his first term a few months prior, and everyone who’d voted for Hillary Clinton (or worse, apathetically sat out the election) had already started to contort into the cringe-ball posture they’d frequently find themselves in over the next four years. The Handmaid’s Tale, with its themes of feminist rage in the face of extreme oppression, tapped into frustrations that many politically progressive people felt that spring. The show provided a shiver-inducing but often cathartic viewing experience; week to week, June—or “Offred,” as Elisabeth Moss’ character is first introduced, having been forced to take the name of the man who has enslaved her as a breeding vessel—adapted to her grim new normal while silently plotting revenge or psyching herself up to survive. In the world of The Handmaid’s Tale, the United States is now called “Gilead.” It’s a society ruled by far-right Christian conservatives, where paranoia and suspicion guide every conversation, and men with machine guns and black vans roam the streets looking for dissidents… or really, anyone who dares to stick a toe out of line. Violence is the knee-jerk reaction to any infraction, with cruelty layered in to make sure anyone with rebellious ideas becomes too frightened to act on them. The linguistic tics of Gilead—”Under his eye,” “Praise be,” “May the Lord open,” and so on—and the red dresses and white caps worn by June and her fellow handmaids immediately took their place in pop culture, with the distinctive costumes popping up at women’s rights protests, particularly when reproductive rights were involved, but also in more lighthearted contexts, including among cosplayers at San Diego Comic-Con. The Handmaid’s Tale, in other words, was an immediate hit; at the 2017 Emmys, it became the first streaming release to be named Outstanding Drama Series, alongside wins for Moss and her co-stars Ann Dowd and Alexis Bledel, as well as for the show’s writing, directing, cinematography, and production design. That was 2017. Eight years later, the culture has again shifted—much farther to the right than it perhaps ever has been, and that includes Ronald Reagan’s 1980s, when Margaret Atwood’s source-material novel was first published. In 2022, Roe vs. Wade was overruled by a Supreme Court that had scuttled far-right during Trump’s first presidency; just a few months ago, Trump took office for his second term and the mood out of Washington, D.C. has rarely felt so extreme as it does now… at least outside of dystopian fiction, that is. The opening scene of The Handmaid’s Tale‘s first episode, “Offred,” follows June, her husband, and their young daughter as they’re frantically trying to escape the gun-toting men who are chasing them. “What did they do?” is the natural audience reaction, until it becomes clear that they’re being hunted because in a world where fertility rates have nearly bottomed out, children are seen as rare prizes, and women who’ve given birth to healthy kids are imprisoned as handmaidens, raped over and over until they conceive again. Once this initial horror sinks in, viewers realize The Handmaid’s Tale is in fact a parade of horrors, with rights ripped away from everyone who doesn’t conform to Gilead’s strict moral standards. June may not be able to express herself outwardly, but her voice-overs let us know her true feelings, and we get valuable context from the frequent flashbacks to her life before everything around her was jammed into a twisted embrace of “traditional values.” Some of these memories are happy—time spent with her family and friends, including Moira (Samira Wiley), who loses her own partner in the “dyke purges” and ends up in handmaid training camp alongside June. We learn that some people, including the wife and child of another lesbian handmaid, Alexis Bledel’s Emily, managed to escape to Canada when things started getting bad, and Canada emerges as an important second location through the series. There’s an organized resistance that starts off slowly in this first season, as well as an intriguing character in Serena Joy (Yvonne Strahovski), the wife in the home where June is made to serve. We learn she was a powerful conservative leader and author in the “before,” but was swiftly shunted into a subservient role in the culture she helped create. She’s both poignant and repulsive, and therefore consistently fascinating. The center of The Handmaid’s Tale, however, is always June, and through her we learn how Gilead forced its way into existence. First the government was dismantled through an attack blamed on terrorists—it’s implied this was a lie—that led to the suspension of the constitution. Martial law was enacted, including in what was once Boston, where June lives. Gradually, things that felt like well-entrenched parts of everyday life began to change. One day, June’s debit card was declined at the local coffee shop, and she learned women were no longer allowed to have bank accounts or own property. Armed men barged into her office and forced her boss to let all the female employees go. “They can’t just do this,” June says to Moira, but indeed They can, and They do. “Now I’m awake to the world,” Gilead-era June narrates from the nightmare of her new reality. “I was asleep before. That’s how we let it happen.” Doesn’t feel quite as fantastical as it once did, does it? Still… it could never really happen, right? Right? You can watch seasons 1-5 of The Handmaid’s Tale on Hulu and on Hulu on Disney+. Season six arrives April 8.

IBM is cutting thousands of jobs across multiple locations in the U.S., and its Cloud Classic unit hit especially hard, reports The Register. The company has not publicly acknowledged these layoffs, but insiders suggested to The Register it is part of an ongoing effort to restructure and shift jobs offshore, particularly to India. The report estimates that about 9,000 positions may be at risk, including a quarter of the Cloud Classic group and 10% of the Cloud group, which is a separate business unit. The job cuts are happening in cities like Dallas, New York, Raleigh, and various locations in California. Employees from several departments, including consulting, cloud infrastructure, corporate social responsibility, internal IT, and sales, have been affected. Some learned of their termination through individual notifications, while others heard about it in internal meetings. IBM's Classic Cloud is the company's original cloud infrastructure platform, formerly known as SoftLayer, which IBM acquired in 2013. This platform provides a range of services, including bare metal servers, virtual servers, storage, and networking solutions, all operating within a traditional cloud environment. While IBM continues to support and maintain its Cloud Classic infrastructure, the company has also introduced a more advanced cloud environment called IBM Cloud VPC (Virtual Private Cloud). This platform offers improved hardware, increased network performance (200 Gbps vs 25 Gbps), greater flexibility in resource management, and enhanced security features compared to the Classic. It is reasonable to expect IBM's Classic Cloud customers to shift to the Cloud VPC, which makes downsizing the unit logical. The company is making it difficult to determine how many people work in the Cloud Classic division, but The Register's sources indicate an aggressive push to relocate jobs overseas. For example, IBM has significantly more open roles in India than in the U.S., which reinforces suggestions that much of the work is being outsourced. The layoffs extend beyond Cloud Classic, though. IBM recently cut positions in its marketing and communications department, as revealed in an internal meeting led by a senior executive. It also cut some 10% of staff in the Cloud group, as noted above. IBM has been using multiple tactics to reduce its workforce, including official layoffs labeled as 'Resource Actions,' as well as policies that encourage employees to leave voluntarily. One such policy is a new requirement for workers to be physically present in offices at least three days a week, with badge swipes being tracked. Only medical exemptions are allowed, but even those are reportedly discouraged by the middle management, according to the report. Also, instead of laying workers off, IBM asks them to sign a separation agreement, which reduces the company's risks. The Register's sources suggest that IBM's restructuring is far from over. The company is expected to continue layoffs as it chases acquisitions and phases out roles deemed obsolete or better suited for offshore teams. Employees remaining after this round of cuts are bracing for further reductions as IBM shifts its workforce strategy.

The traces of supernovae who reached the earth could be more numerous than estimated by scientists. At the Global Physics Summit 2025 of the American Physical Society of Anaheim, California, the astronomer Brian Fields of the University of Illinois illustrated the progress made in identifying the remains of star explosions on the bottom of our oceans, explaining that science has managed to identify the chemical signatures of these cosmic events, be present everywhere. "We live in a supernovae cemetery," he commented during his presentation. The remains of supernovae kept on earth The scenario in which what remains of a supernova comes into contact with the earth's surface does not necessarily imply a planetary catastrophe. Although the radiation emitted by a collapse star is lethal in a ray of a few years light, over time these explosions lose speed and become imperceptible. However, the particles that arise continue to travel thanks to the initial acceleration, until they find a planet or a moon on which to rest. Some of these debris have reached the earth. The Fields team has sought effective methods for decades to detect them. In the report presented at the Anaheim summit, the astronomer told how him and his colleagues managed to identify a radioactive version of the iron that is not naturally found on our planet. For researchers, these still active exotic elements are the supernovae test that occurred in recent history and near the earth. One of the most recent studies of Fields, published in September 2024 in the Astrophysical Journal magazine, suggests that at least two supernovaes have occurred near the Earth: the first three million years ago at a distance between 50 and 65 Parsec and the second seven million years ago at 100 Parsec (a Parsec equivalent to 3.26 light years). Space cocktail For the search for last year, the researchers were based on the analysis of sediments of deep water and lunar regolite. But the team also found more exotic champions generated by rare events in the universe: specifically radioactive isotopes of Plutonio, a characteristic sign of a past collision between two stars of neutron, or Kilonova: it is one of the events capable of release more energy ever in the universe, responsible for the formation of elements such as gold. In this case, the Kilonova who produced Plutonio isotopes is prior to the two supernovae already under study. The Fields working group advances a hypothesis to explain the presence this "cocktail" of space isotopes on earth. It is possible that Kilonova debris mixed with the remains of supernovae and that some of these dust have reached our planet in the form of a sort of "cosmic rain". To verify their theory, scientists need to examine van samples that have not been contaminated. The next shipments to our natural satellite will bring back champions of this material, which in part could be used to look for iron isotopes and plutonium. The discovery of these elements would represent further proof that both the moon and the earth are custodians of supernovae remains. This article originally appeared on Wired En Español.

Ben Stiller, Cate Blanchett, or Guillermo Del Toro ... A real red carpet of more than 400 personalities urge Donald Trump to protect copyright across the Atlantic, while Google and Openai wish to train their artificial intelligence models (IA) on protected works. All tech news for pros every day in our newsletter learn more about the use of personal data In an open letter, these artists denounce the devastating consequences on the Hollywood economy of this possibility. For the filmmakers signatory to the letter, this automation with forced march can lead to an abusive use of protected works to supply algorithms, without remunerating their beneficiaries. Google and Openai take advantage of Trump's coming to power Despite the historic strike that rocked Hollywood in 2023, Google and Openai continue to put pressure on to “soften” intellectual property legislation. These actors underline that the Chinese Deepseek may take a considerable lead if the United States does not massively exploit content under copyright to enrich their AI. Since his re -election, Donald Trump has launched an action plan to strengthen American domination in artificial intelligence, notably by repealing certain agreements concluded under Joe Biden. OPENAI and Google take the opportunity to present to the White House proposals aimed at bypassing intellectual property rules, and this to the chagrin of the 400 signatories of the letter. In France, victims and profiteers In France, the actors who lend their voice to the characters of films, series and video games, also alert on the impacts of AI for their profession. More than two hundred of them demonstrated last December in Paris, chanting "the AI ​​outside, the voices are our treasures" or "IA Creet, you do not touch my voice". In June 2024, these voiceover professionals had in an open letter called the government and parliamentarians to legislate to regulate generative AI. The fact remains that AI tools are increasingly used by cinema professionals themselves. The CNC AI Observatory, made public last year, shows that in the production professions, the AI ​​is used for documentary research (58.0 %), reformulation assistance (50.6 %), but also as a source of inspiration (45.4 %) or the test of new ideas (50.9 %).

This story contains mentions of suicide. If you or someone you know needs help, please call 1-800-273-8255 for free, 24-hour support from the National Suicide Prevention Lifeline. Twenty-three-year-old Arnold Robert Haro addressed his final words to the phone in his hand. “If I die, I hope you guys turn this into a memecoin,” he said. Then Haro took his own life. Haro died on February 21 at his family home in Madera County, California, a death certificate obtained by WIRED shows. His suicide was broadcast live to his followers on X, where he went by the handle @MistaFuccYou. Footage of Haro’s death has since been removed from the platform, but the incident was briefly listed in its trending tab. In the hours after Haro’s death, people created dozens of memecoins—a type of highly volatile crypto coin used as a vector for financial speculation—modeled after him. Sensing an opportunity to profit, traders piled into one of the coins in particular, driving its value to $2.1 million in aggregate. (The coin has since lost 96 percent of its value.) On X, some tried to argue that whoever was behind the MistaFuccYou coin had duly granted Haro’s final wish. But most denounced the impulse among traders to try to profit by his death. “If you’re trading this, you’re sick af,” wrote one user. Speculation ran rampant on X that Haro had ended his life because he had lost money to a memecoin rugpull—a maneuver whereby somebody creates a new coin, promotes it online, then sells off their holdings in one swoop, devaluing everyone else’s stake. WIRED was unable to confirm whether this had happened to Haro, but his friends have disputed the narrative. “It had nothing to do with crypto … It’s not what all these crypto nerds seem to think,” one of Haro’s friends, who goes by j nova on social media, told WIRED. Haro’s family, meanwhile, has described his death as the result of “his battle with depression.” The incident captures in microcosm the race to the bottom in memecoin trading circles, where only the most heinous and morally bankrupt ideas are now rewarded with attention, says Azeem Khan, cofounder of the Morph blockchain and venture partner at crypto VC firm Foresight Ventures. “We’ve reached the point where the most potentially exciting launch that people are looking at is Kanye trying to launch a swastika coin,” says Khan, in reference to now-deleted X posts made by an account associated with the artist Kanye West. “That’s how terrible this space is.” Until last year, launching a memecoin was relatively expensive and technically burdensome, which meant few came to market. Only Dogecoin—the original memecoin—and a handful of derivatives had any sort of longevity. That equation was reversed with the arrival of Pump.Fun, a platform that makes it simple for anyone to launch a memecoin at no cost. Since Pump.Fun launched in January 2024, many millions of memecoins have flooded the market, among them the coins modeled after Haro.

But the Post has struggled to hold on to the commercial success it had reached during the first Trump administration, falling far behind its biggest competitor, The New York Times. Its audience has shrunk and financial losses have ballooned. “He’s a highly rational person. I think he was comfortable losing $20 million a year. When it gets to $100 million a year, I don’t know what his appetite is for that,” says a former executive who worked closely with Bezos. “I don’t think for him it’s like a mission business in the same way it would be for [the Graham family] or for the Sulzbergers [who own the New York Times].” Credit: Robert Miller/The Washington Post/Getty Images The Washington Post chief executive Will Lewis answers employees’ questions last year. Bezos’ turn toward Trump has alarmed editorial staff, fueling an exodus of talent and subscribers. The Washington Post chief executive Will Lewis answers employees’ questions last year. Bezos’ turn toward Trump has alarmed editorial staff, fueling an exodus of talent and subscribers. Credit: Robert Miller/The Washington Post/Getty Images This commercial struggle occurred at the same time that Bezos cozied up to Trump and began intervening on the Opinion section, further sinking morale across the newspaper. At least a dozen seasoned journalists have left to work for competitors. Many saw a direct link between his muzzling of the newspaper and rewards for his other enterprises. Hours after Bezos spiked the paper’s endorsement of Harris, Blue Origin chief executive David Limp met Trump at a rally in Texas. Former Post editor-at-large Robert Kagan criticized the meeting as a “quid pro quo.” In the latest fracas, veteran journalist Ruth Marcus resigned last week, claiming that chief executive Sir Will Lewis had axed a column she had written criticizing Bezos’ overhaul of the Opinion section. “The Washington Post I joined, the one I came to love, is not The Washington Post I left,” Marcus said of the decision. Yet for all the drama in the newsroom, public grievances lodged by senior journalists, and relentless media coverage depicting the Post in crisis, there is little indication that Bezos is fazed. “He’s a really tough guy. He’s totally charming, but he didn’t get where he was by being kind and gentle. In that sense, I don’t know that he’s changed,” says the former Post executive. “I remember him talking about the people he gets for the Amazon board, and he specifically picks people who don’t give a shit what other people think. He doesn’t want people on the board like ‘oh my god, we don’t air condition the warehouses?’ I think he himself is like that, too.” In January, more than 400 Post journalists signed a letter imploring Bezos to meet them about their concerns. He did not respond. “He’s taken the organization in a certain direction and seems determined to do that, for whatever reason that he hasn’t fully explained,” says Baron, who adds that he has not tried to contact Bezos about his objections. “Quite simply, I do not think he cares what I think.” © 2025 The Financial Times Ltd. All rights reserved. Not to be redistributed, copied, or modified in any way.

The Nvidia Blackwell RTX Pro 6000 GPU was announced during the GTX 2025 keynote. These will use the same GB202 die that goes into Nvidia's RTX 5090 graphics card, but with some significant changes in some of the other aspects. There will be three variants of the RTX Pro 6000: the Blackwell Workstation Edition, Max-Q Workstation Edition, and Blackwell Server Edition. The core specifications for the RTX Pro 6000 are the same across all three models. You get 188 SMs enabled, out of a potential 192 maximum from GB202. That's 10.6% more SMs, shader cores, tensor cores, RT cores, etc., relative to the RTX 5090. Clock speeds weren't given, but Nvidia does list up to 125 TFLOPS of FP32 compute via the shaders, and 4000 AI TOPS from the tensor cores. That works out to a boost clock of around 2.6 GHz, but that won't be the same for all three variants. The RTX Pro 6000 features the full 128MB L2 cache of GB202, along with four NVENC and four NVDEC video blocks. RTX 5090 only has 96MB of L2 cache and three each for NVENC/NVDEC. It's very close to a fully enabled chip, with only 2% of the SMs disabled. The memory configuration is the same for all three variants. As discussed in the initial RTX Pro 6000 announcement, Nvidia uses 24Gb (3GB) GDDR7 chips rather than the 2GB chips used on the consumer GeForce RTX 50-series cards. That increases the memory capacity to 48GB per PCB side, and with chips on both sides of the PCB in 'clamshell' mode, there's 96GB total. The memory has the same 28 Gbps clocks as most of the 50-series parts, with 1792 GB/s of total bandwidth. Swipe to scroll horizontally Nvidia RTX Pro Specifications Graphics Card RTX Pro 6000 RTX Pro 5000 RTX Pro 4500 RTX Pro 4000 Architecture GB202 GB202 GB203 GB203 Process Technology TSMC 4N TSMC 4N TSMC 4N TSMC 4N Transistors (Billion) 92.2 92.2 45.6 45.6 Die size (mm^2) 750 750 378 378 SMs 188 110 82 70 GPU Shaders (ALUs) 24064 14080 10496 8960 Tensor Cores 752 440 328 280 Ray Tracing Cores 188 110 82 70 Boost Clock (MHz) 2600 2500? 2500? 2500? VRAM Speed (Gbps) 28 28 28 28? VRAM (GB) 96 48 32 24 VRAM Bus Width 512 384 256 192 L2 Cache 128 96? 64? 48? Render Output Units 192 144? 96? 80? Texture Mapping Units 752 440 328 280 TFLOPS FP32 (Boost) 125.1 70.4? 52.5? 44.8? TFLOPS FP16 (FP4/FP8 TFLOPS) 1001 (4004) 563 (2253) ? 420 (1679) ? 358 (1434) ? Bandwidth (GB/s) 1792 1344 896 672? TBP (watts) 600 300 200 140 Image 1 of 6 (Image credit: Tom's Hardware) (Image credit: Tom's Hardware) (Image credit: Tom's Hardware) (Image credit: Tom's Hardware) (Image credit: Tom's Hardware) (Image credit: Tom's Hardware) The Blackwell Workstation Edition looks basically the same as the RTX 5090, except with a glossy black finish in places rather than a matte black. TDP (TGP) for the card is 600W, 25W higher than the 5090, but otherwise, the two cards look about the same. You also get four DisplayPort 2.1b outputs, whereas the 5090 typically offers at least one HDMI 2.1b output. For the Max-Q Workstation Edition, the TGP gets capped at 300W. Half the power will naturally mean lower typical boost clocks for a lot of workloads, though there will undoubtedly be cases where it will still run nearly as fast as the 600W card. It also has a standard FHFL (full-height, full-length) dual-slot form factor with dual-blower fans at the back of the card. It also has four DP2.1b outputs. Finally, the Blackwell Server Edition has a similar form factor to the Max-Q card but ditches the fans, instead relying on the server fans to provide airflow and cooling. That's usually in ample supply for servers, and noise levels are usually less of a concern — you get high RPM fans moving lots of air in a regulated environment to make everything run sufficiently cool. The power on the Server Edition is configurable up to 600W, so some installations might opt for lower power to optimize the efficiency if they're power-limited. All three models use the same 16-pin connector found on desktop RTX cards. Servers and workstations tend to be built to much tighter specifications, and so far there haven't been any widespread reports of servers or workstations with melting connectors. That suggests perhaps that the biggest issues with 16-pin connectors are component quality and proper installation — companies are less likely to cheap out on the cables in a server or workstation, so there aren't impurities causing hot spots and melting. Pricing hasn't been discussed, but we typically see professional and server solutions like the RTX Pro 6000 selling for 4X~5X more than the equivalent consumer GPUs. It wouldn't be surprising if the various RTX Pro 6000 cards cost $10,000 or more. We'll find out exactly where they fall in the coming days.

Assassin’s Creed Shadows is out today, but if you happen to be a PS5 owner, you may wondering why your game looks so different than what you see from those on PlayStation 5 Pro. Digital Foundry found playing the game on “Performance” settings for regular PS5 results in what may be the worst way to experience Ubisoft’s beautiful but awkward stealth action title. If you’re one of those who ponied up $700 for a PlayStation 5 Pro, then you can get both performance and pretty graphics. For PS5 consoles, Assassin’s Creed Shadows offers three main graphics options: “Performance,” “Balanced,” and “Fidelity” modes. What setting you chose can have a drastic impact on how your game looks. A video breakdown of the game’s graphics by Digital Foundry shows that if you want to play at the max 60 FPS—which is considered the best baseline framerate for smooth gameplay modern gaming—playing it on PS5 results in a washed out, drab look compared to playing it on the console that costs $200 more. This is down to whether Performance mode on either console supports ray traced global illumination, or RTGI. This form of in-game lighting creates a simulation of light bouncing off multiple surfaces before it comes to the player, as if the screen were the lens of a camera. This setting is available in Shadows’ Performance mode on the PS5 Pro, but not Sony’s base $500 console. Without those lighting settings Shadows can look flat and drab. On PS5 Pro, the lines around shadows appear more crisp than PS5. There is more contrast in the shadows inside buildings, and light reflects through objects like bamboo far more realistically. The difference in graphics quality implies the game was built with that lighting simulation in mind. I can’t emphasize enough how important the visuals are to whether you’ll fully enjoy Assassin’s Creed Shadows. The game relies on its lighting for gameplay, such as the titular “shadows” players can hide in, and for its atmosphere. This is one of the first games to showcase such drastic difference in visual quality from PS5 to PS5 Pro. Invariably, those who spent more money on their PlayStation are getting a far better experience at 60 FPS. That above image spells out the difference in visual quality. It’s far more dramatic a change than on other games, like Final Fantasy VII: Rebirth. I’ve played games on both consoles, like Star Wars Jedi: Survivor, and while there are subtle difference in lighting, the game still maintains its overall look across both consoles at 4K resolution. The difference in Shadows is much less dramatic when the game is set to the Balanced or Fidelity modes. On the PS5 those modes include much more of the lighting goodies found on the PS5 Pro at the expense of lower frames per second. The distinct difference between the two versions of Sony’s latest console have never been more stark. We expect future games will make better PlayStation 5 Pro’s updated GPU architecture. Now, console gamers can feel the same FOMO pressure PC gamers have whenever they stare longingly at ray tracing toggles in a games’ options menu wondering “how much do I really need 60 FPS?”

When Apple launched the iPhone 16e, one of its new features was the use of Apple’s C1 modem. The company’s plan is to eventually replace Qualcomm’s modems with its own, so the question is, how does Apple’s modem hold up? Unfortunately, the answer isn’t what most Apple fans want to hear. According to Ookla’s benchmarks, the Apple C1 did not exactly beat out Qualcomm. How did the C1 perform? Ookla’s study measures performance in three tiers. The first tier, the 10th percentile, basically measures how the modems perform when connectivity is struggling. Apple’s C1 beat Qualcomm here. However, in the 90th percentile, where connectivity is optimal, Qualcomm beat out the Apple C1 modem in the benchmarks. As for the median tier, the C1 edged out Qualcomm ever so slightly. However, even then the results are mixed. This is because Speedtest data found that the C1 performed better on some networks than others. For instance, the iPhone 16e achieved faster speeds on AT&T and Verizon’s networks but was noticeably slower on T-Mobile’s network. Based on these results, it sounds like the Qualcomm is clearly the “better” choice, especially under ideal conditions, right? However, Ookla doesn’t see it that way. The company says, “Performance at the lowest 10th percentile often provides a more accurate reflection of overall quality of experience (QoE) than the fastest 90th percentile, which can be skewed by deployments in mmWave-covered locations and is subject to declining marginal returns.” A decent start for Apple To be fair, the C1 is Apple’s first attempt at creating its own modems. Apple bought Intel’s smartphone modem business years ago, so for its acquisition to finally bear fruit is good progress. Also, take into account the fact that the C1 doesn’t support mmWave connectivity right now. We’re hearing reports that Apple is working on an update to the C1, which will support mmWave. So maybe that version should have better benchmarks. Either way, it’s still kind of impressive that Apple got this close to Qualcomm’s results despite it being its first modem. Hopefully, we’ll see more improvements in future iterations. Apple will continue using Qualcomm’s modems until 2027, which is when its licensing deal runs out. This means that Apple will use a mix of its own modems and Qualcomm’s modems in its upcoming iPhones. The iPhone 17 Air is rumored to use the C1, while the rest of the iPhone 17 lineup will likely use Qualcomm’s modems.

After a year of regulatory blocking, Meta AI is now accessible in Europe on WhatsApp, Facebook, Instagram and Messenger. Meta's AI assistant must help user exchanges in its products. But its deployment raises strong regulatory concerns. All tech news for pros every day in our newsletter learn more about the use of personal data Officially, it is not based on European user data. This point was the explanation for the delay in its deployment in Europe. The Irish authority, which supervises Meta for the European Union, demanded the cessation of AI training on the publications of European users of Facebook and Instagram. Consequently, the European version of Meta AI has been adapted. It is currently limited to textual conversations and therefore does not offer certain features available in the United States. Best for customer service via the Messenger or Instagram canal? In addition, Meta assures that the training of its AI in Europe does not use the data of Europeans. But this declaration does not raise doubts about the fact that the model of AI Llama, at the heart of Meta Ai, is fueled "indirectly" by data collected on the Internet, the origin of which can be European. The fact remains that this type of AI functionality in messaging applications is eagerly awaited by companies that use Facebook or Instagram as their customer service. This deployment could allow, for example, an actor in e-commerce to better automate his first-level responses on Messenger, including the availability of products or order monitoring. AI already in great demand by customer service players According to an Opinionway survey for Esker in December 2024, more than half (56 %) of customer service managers sees AI as an opportunity on the French customer service market. Among the main perceived advantages of the AI, we will note: Significant time saving (68 %) Optimized management of orders or complaints (63%) A reduction in the processing deadlines for customer requests (62 %) The AI ​​is also considered to be a productivity lever for customer service teams (55 %), also strengthening customer satisfaction (54 %) and employees (46 %). The demand is therefore strong. But due to European regulations, Meta AI in Europe does not currently offer the generation and editing of images, and does not allow questions to be asked from photos.

The European Union is attacking a locking which it considers unfair. According to her, the iPhone will have to work better with the connected watches of other brands. In its sights of course, the privileged link that Apple maintains between its smartphones and the Apple Watch. All tech news for pros every day in our newsletter learn more about the use of personal data A prioritization that has been limiting smartwatches for other ecosystem actors with iPhones for years. And this has consequences for user habits. Imagine that you use Microsoft Teams on iPhone. If your connected watch is not an Apple Watch - a Galaxy Watch or a Fitbit Versa for example - it will be more difficult for you to receive and especially to respond to notifications and alerts from their iPhone. Impose better interoperability Under the Digital Markets Act (DMA), Apple is therefore now imposed by Brussels a better interoperability between its equipment and that of its competitors. And this first concerns accessory pairing and notification management. Concretely, Apple will have to document and share its technical protocols with its competitors with improving the interoperability of its systems. For manufacturers, guaranteeing compatibility with the iPhone has so far been a vow. For users, it is the promise to finally be able to fully use their mobile work tools. However, compliance may spread over time. As with the adoption of the USB-C, Apple could play the watch, invoking the protection of its "know-how".

The promise of fusion energy is cheap and abundant power for the entire planet. Scientists have made startling advances towards achieving it at scale, but there are still many problems holding it back. One of them is the production of fuel, which requires vast amounts of enriched lithium. Enriching lithium has been an environmental catastrophe, but researchers in Texas believe they’ve found a way to do it cheaply and at scale without poisoning the world. A team of researchers at Texas A&M University discovered the new process by accident while working on a method for cleaning groundwater contaminated during oil and gas extraction. The research has just been published in the scientific journal Chem under the title “Electrochemical 6-Lithium Isotope Enrichment Based on Selective Insertion in 1D Tunnel-Structured V2O5.” The effect the research has on nuclear fusion might be enormous. “Nuclear fusion is the primary source of energy emitted by stars such as the Sun,” Sarbajit Banerjee, a professor and researcher at ETH Zürich and Texas A&M and one of the authors of the paper, told Gizmodo. The simplest method of doing fusion on Earth instead of space involves tritium and deuterium isotopes. Tritium is rare and radioactive so reactors currently “breed” it on demand to generate energy. They breed the tritium by bombarding lithium isotopes with neutrons. Most lithium on the planet, more than 90% of it, is lithium-7. Breeding tritium works way more efficiently with the ultra-rare lithium-6. “When 7Li, the most commonly occurring lithium isotope, is used, tritium production is much less efficient as compared to 6Li,” Banerjee said. “As such, modern reactor designs are based on breeding blankets with enriched 6Li isotope that has to be specifically extracted from natural lithium.” You can turn naturally abundant mixtures of lithium isotopes into Lithium-6, “enriching” it, but the process is a toxic nightmare. “From 1955 to 1963, the United States produced 6Li at the Y12 plant at Oak Ridge National Laboratory in Tennessee for thermonuclear weapons applications, taking advantage of the slight difference in solubility of 6Li and 7Li isotopes in liquid mercury,” Banerjee said. “This did not go so well.” “About 330 tons of mercury were released to waterways and the process was shut down in 1963 because of environmental concerns,” he said. Mercury is a toxic nightmare substance that’s difficult to clean up. After 60 years, heavy metals from the process of extracting Lithium-6 from naturally abundant mixtures are still poisoning Tennessee today. Cleaning up the remnants of the environmental disaster is a major project for Oak Ridge National Lab’s current residents. During a different project, the team at Texas A&M developed a compound called zeta-V2O5 that it used to clean groundwater. As it ran water through this membrane it noticed something strange: it was really good at isolating Lithium-6. The team decided to see if it could harvest Lithium-6 from mixtures of Lithium isotopes without mercury. It worked. “Our approach uses the essential working principles of lithium-ion batteries and desalination technologies,” Banerjee said. “We insert Li-ions from flowing water streams within the one-dimensional tunnels of zeta-V2O5…our selective Li sponge has a subtle but important preference for 6Li over 7Li that affords a much safer process to extract lithium from water with isotopic selectivity.” Banerjee this could lead to a massive change in how fuel is developed for fusion generators. It also doesn’t require a massive re-design of the existing reactors. “Our work outlines a path to overcoming a key supply chain issue for fusion. However, to be clear we are not redesigning the actual reactors—tokamaks or stellarators—although there is tremendous recent excitement about new innovations and designs in plasma physics,” he said. A lot of people are banking on fusion being the path towards cheap and abundant energy. My entire life I’ve heard that the breakthrough that will make it real is “just around the corner.” It’s been a constant refrain that’s become a bit of a joke. Just last year the Bulletin of the Atomic Scientists asked if fusion might be “forever the energy of tomorrow.” But Banerjee was hopeful. “Despite the incredible challenges, fusion is too big of a prize to give up on,” he said. “The transformative potential has been clear but there have been critical gaps in engineering designs, materials science for extreme environments, and understanding of the complexity of plasma processes to enumerate just a few gaps. There is an intensifying global competition and billions of dollars in private and public investments—while still not imminent, there are promising signs of realistic fusion energy in about two or three decades.”

is editor-in-chief of The Verge, host of the Decoder podcast , and co-host of The Vergecast Today we’re talking about the protests against Tesla, which have been branded on social media under the hashtag Tesla Takedown. The protests are, of course, a reaction to Elon Musk, who has managed to install himself as basically a not-so-shadow president who is tearing the federal government apart, leaving confusion and destruction in his wake. A lot of people are deeply unhappy with this state of affairs. And because Musk is not an elected official but is, on some level, a car salesman, many of those frustrated and angry people are taking aim at the most visible and accessible symbol of his power and wealth: Tesla itself. There are protests at Tesla showrooms and charging stations every weekend now, and they’re only getting bigger — and attracting police presence and threats of prosecution from the government. President Donald Trump and Attorney General Pam Bondi have said that any violence aimed at Tesla vehicles or property would be prosecuted as domestic terrorism. This is very Trump, in that there is not actually a specific federal domestic terrorism law, but you know… Trump. In any case, these protests seem like they’re working; Tesla’s stock price has been sinking, new car registrations are down, and merely owning one has become uncomfortably political for a lot of people. We’re also seeing the hype around the company fading in ways that are frankly surprising — popular YouTubers are pointing out that vehicles like the Cybertruck simply aren’t very good and that Tesla’s missed ship dates and big claims about the capabilities of its self-driving tech don’t really hold up. And if the hype fades enough, it’s possible to break Tesla itself — it’s sky-high stock price is entirely based on hype, after all. To help pull this all apart, I asked Ed Niedermeyer to join me on the show. Niedermeyer published a book about Tesla in 2019 called Ludicrous: The Unvarnished Story of Tesla Motors, and he hosts a podcast called Autonocast, about autonomous vehicle technology. His sizable following on Bluesky has also made him one of the leading organizers of the Tesla protests in Portland, Oregon, where he lives. So I wanted to bring him on the show to talk me through what he sees with both the big picture of Tesla as a company and on the ground with the ongoing protest movement. There’s a lot happening and the story is changing fast, even since Niedermeyer and I talked at the beginning of the week. But that bigger picture — the disconnect between what he calls the “myth” of Tesla and the reality, and how that reality is now manifesting into a global protest movement — is shaping up to be the defining story of Tesla’s next chapter. If you’d like to read more about what we talked about in this episode, check out the links below: Questions or comments about this episode? Hit us up at decoder@theverge.com. We really do read every email!

Asus has unveiled a new BIOS feature called AI Cache Boost, which is designed to enhance AI performance on systems equipped with AMD’s Ryzen 9000 series processors. The new feature is part of a recent BIOS firmware update for Asus AMD 800 Series motherboards. It focuses on improving the efficiency of data transfers between the CPU cores, cache, and memory, which is crucial for handling extensive AI workloads. Upon enabling AI Cache Boost in the BIOS under the Extreme Tweaker tab, the system overclocks the Infinity Fabric clock (FCLK) to 2100MHz. This increase in FCLK enhances the bandwidth of data transfers within the CPU architecture, which Asus says facilitates more efficient processing of large datasets commonly associated with LLMs. As per Asus, enabling AI Cache Boost can boost performance by up to 12.75% when working with LLMs (Large Language Models). This feature is tailored for AI enthusiasts, researchers, and professionals who utilize AMD Ryzen 9000 Series processors, specifically those with AMD 3D V-Cache Technology, like the recently launched Ryzen 9 9950X3D. Asus says it is particularly beneficial for users engaged in AI workloads that involve processing massive datasets, such as training and deploying large language models. Asus has clarified that the new AI Cache Boost feature is specifically designed to enhance AI workloads, not gaming performance. While the feature can technically be enabled alongside other BIOS settings like GPU Boost, doing so may compromise overall system stability without delivering any noticeable benefit in gaming scenarios. For users focused solely on maximizing AI capabilities, Asus offers additional tuning options beyond AI Cache Boost. By activating both Game Turbo and AI Cache Boost modes, disabling simultaneous multi-threading, and deactivating one CCD (Core Complex Die) on Ryzen processors, Asus says AI performance can be pushed even further. In fact, Asus own tests on a Ryzen 9 9950X3D configured with a single eight-core chiplet showed up to a 19.4% improvement in AI tasks. As with all company-provided benchmarks, take that with a grain of salt. Asus conducted internal testing using a system equipped with ROG Crosshair X870E Hero motherboards, Nvidia RTX 5090 or 4090 GPUs, and Ryzen CPUs with AMD 3D V-Cache, showing performance uplifts in various AI benchmarks. In Geekbench AI tests, enabling AI Cache Boost resulted in improvements between 3.4% to 8%, with the Ryzen 7 9800X3D demonstrating the most significant gains. The uplift was similarly positive when tested with previous-gen RTX 4090 GPUs, showing a boost of up to 7.57%. Swipe to scroll horizontally Geekbench AI - (AMD Ryzen 9 9950X3D - RTX 5090 - ROG Crosshair X870E Hero) Row 0 - Cell 0 Default AI Cache Boost Improvement Single Precision Score 43932 46213 +5.19% Half Precision Score 63464 67009 +5.59% Quantized Score 32197 33347 +3.57% Swipe to scroll horizontally Geekbench AI - (AMD Ryzen 9 9900X3D - RTX 5090 - ROG Crosshair X870E Hero) Row 0 - Cell 0 Default AI Cache Boost Improvement Single Precision Score 43625 46074 +5.61% Half Precision Score 62772 66744 +6.33% Quantized Score 32169 33275 +3.44% Swipe to scroll horizontally Geekbench AI - (AMD Ryzen 7 9800X3D - RTX 5090 - ROG Crosshair X870E Hero) Row 0 - Cell 0 Default AI Cache Boost Improvement Single Precision Score 44099 46965 +6.50% Half Precision Score 63365 68393 +7.93% Quantized Score 32480 33774 +3.98% Swipe to scroll horizontally Geekbench AI - (AMD Ryzen 9 9950X3D - RTX 4090 - ROG Crosshair X870E Hero) Row 0 - Cell 0 Default AI Cache Boost Improvement Single Precision Score 37480 40181 +6.19% Half Precision Score 51786 54392 +5.03% Quantized Score 28097 29118 +3.63% Swipe to scroll horizontally Geekbench AI - (AMD Ryzen 9 9900X3D - RTX 4090 - ROG Crosshair X870E Hero) Row 0 - Cell 0 Default AI Cache Boost Improvement Single Precision Score 37866 39777 +5.05% Half Precision Score 51782 53642 +3.59% Quantized Score 28157 28798 +2.28% Swipe to scroll horizontally Geekbench AI - (AMD Ryzen 7 9800X3D - RTX 4090 - ROG Crosshair X870E Hero) Row 0 - Cell 0 Default AI Cache Boost Improvement Single Precision Score 38191 40788 +6.80% Half Precision Score 52531 56509 +7.57% Quantized Score 28247 29536 +4.56% Swipe to scroll horizontally Geekbench AI - (AMD Ryzen 9 9950X - RTX 4090 - ROG Crosshair X870E Hero) Row 0 - Cell 0 Default AI Cache Boost Improvement Single Precision Score 38277 40441 +5.65% Half Precision Score 52465 55823 +6.40% Quantized Score 28373 29284 +3.21% Asus testing using the UL Procyon AI Computer Vision Benchmark the Ryzen 7 9800X3D achieved double-digit improvements of up to 12.75%. As mentioned earlier, Asus says combining AI Cache Boost with Turbo Game Mode can yield even higher gains. For example, the company says the Ryzen 9 9950X3D saw up to a 19.35% performance increase in specific AI workloads when both features were enabled. This combination is particularly beneficial for AI users whose tasks don’t fully utilize the CPU's multithreading capabilities. Stay On the Cutting Edge: Get the Tom's Hardware Newsletter Get Tom's Hardware's best news and in-depth reviews, straight to your inbox. Contact me with news and offers from other Future brands Receive email from us on behalf of our trusted partners or sponsors Swipe to scroll horizontally UL Procyon AI Computer Vision Benchmark (RTX 4090 - ROG Crosshair X870E Hero) Row 0 - Cell 0 Default AI Cache Boost Improvement AMD Ryzen 9 9950X3D 1426 1485 +4.14% AMD Ryzen 9 9900X3D 1417 1473 +4.00% AMD Ryzen 7 9800X3D 1490 1680 +12.75% AMD Ryzen 9 9950X 1439 1475 +2.50% Swipe to scroll horizontally UL Procyon AI Computer Vision Benchmark (RTX 4090 - ROG Crosshair X870E Hero) Row 0 - Cell 0 Default AI Cache Boost Turbo Game Mode Improvement AMD Ryzen 9 9950X3D 1426 1485 1702 +19.35% AMD Ryzen 9 9900X3D 1417 1473 1570 +10.80% If it works as advertised, the AI Cache Boost features could be a straightforward way for Ryzen 9000 series users to extract more performance out of their systems when working with AI-related tasks.

What if you could listen to music or a podcast without headphones or earbuds and without disturbing anyone around you? Or have a private conversation in public without other people hearing you? Our newly published research introduces a way to create audible enclaves – localized pockets of sound that are isolated from their surroundings. In other words, we’ve developed a technology that could create sound exactly where it needs to be. The ability to send sound that becomes audible only at a specific location could transform entertainment, communication and spatial audio experiences. What is sound? Sound is a vibration that travels through air as a wave. These waves are created when an object moves back and forth, compressing and decompressing air molecules. The frequency of these vibrations is what determines pitch. Low frequencies correspond to deep sounds, like a bass drum; high frequencies correspond to sharp sounds, like a whistle. Controlling where sound goes is difficult because of a phenomenon called diffraction – the tendency of sound waves to spread out as they travel. This effect is particularly strong for low-frequency sounds because of their longer wavelengths, making it nearly impossible to keep sound confined to a specific area. Certain audio technologies, such as parametric array loudspeakers, can create focused sound beams aimed in a specific direction. However, these technologies will still emit sound that is audible along its entire path as it travels through space. The science of audible enclaves We found a new way to send sound to one specific listener: through self-bending ultrasound beams and a concept called nonlinear acoustics. Ultrasound refers to sound waves with frequencies above the human hearing range, or above 20 kHz. These waves travel through the air like normal sound waves but are inaudible to people. Because ultrasound can penetrate through many materials and interact with objects in unique ways, it’s widely used for medical imaging and many industrial applications. In our work, we used ultrasound as a carrier for audible sound. It can transport sound through space silently – becoming audible only when desired. How did we do this? Normally, sound waves combine linearly, meaning they just proportionally add up into a bigger wave. However, when sound waves are intense enough, they can interact nonlinearly, generating new frequencies that were not present before. This is the key to our technique: We use two ultrasound beams at different frequencies that are completely silent on their own. But when they intersect in space, nonlinear effects cause them to generate a new sound wave at an audible frequency that would be heard only in that specific region. Crucially, we designed ultrasonic beams that can bend on their own. Normally, sound waves travel in straight lines unless something blocks or reflects them. However, by using acoustic metasurfaces – specialized materials that manipulate sound waves – we can shape ultrasound beams to bend as they travel. Similar to how an optical lens bends light, acoustic metasurfaces change the shape of the path of sound waves. By precisely controlling the phase of the ultrasound waves, we create curved sound paths that can navigate around obstacles and meet at a specific target location. The key phenomenon at play is what’s called difference frequency generation. When two ultrasonic beams of slightly different frequencies, such as 40 kHz and 39.5 kHz, overlap, they create a new sound wave at the difference between their frequencies – in this case 0.5 kHz, or 500 Hz, which is well within the human hearing range. Sound can be heard only where the beams cross. Outside of that intersection, the ultrasound waves remain silent. This means you can deliver audio to a specific location or person without disturbing other people as the sound travels. Advancing sound control The ability to create audio enclaves has many potential applications. Audio enclaves could enable personalized audio in public spaces. For example, museums could provide different audio guides to visitors without headphones, and libraries could allow students to study with audio lessons without disturbing others. In a car, passengers could listen to music without distracting the driver from hearing navigation instructions. Offices and military settings could also benefit from localized speech zones for confidential conversations. Audio enclaves could also be adapted to cancel out noise in designated areas, creating quiet zones to improve focus in workplaces or reduce noise pollution in cities. This isn’t something that’s going to be on the shelf in the immediate future. For instance, challenges remain for our technology. Nonlinear distortion can affect sound quality. And power efficiency is another issue – converting ultrasound to audible sound requires high-intensity fields that can be energy intensive to generate. Despite these hurdles, audio enclaves present a fundamental shift in sound control. By redefining how sound interacts with space, we open up new possibilities for immersive, efficient and personalized audio experiences. Jiaxin Zhong, Postdoctoral Researcher in Acoustics, Penn State and Yun Jing, Professor of Acoustics, Penn State. This article is republished from The Conversation under a Creative Commons license. Read the original article.

It’s hard to picture the universe in its infancy. The middle-aged cosmos now stretches across 93 billion light-years, holding up to two trillion galaxies and more than 200 billion trillion stars. But in the beginning, things were simple. As a baby, the universe was essentially a hot, dense soup of particles, which has been expanding and cooling over billions of years. A telescope perched in northern Chile has peered back at the early universe, capturing the afterglow of the Big Bang that radiated throughout the cosmos. The Atacama Cosmology Telescope (ACT) released the sharpest images yet of the universe’s first light, capturing the material that would later form the earliest galaxies and stars during the cosmic infancy. The new images, to be presented at an upcoming meeting of the American Physical Society, date back to when the universe was only 380,000 years old. The cosmos is now a more mature 13.8 billion years old, which means the light had to travel more than 13 billion years to reach the telescope. “By looking back to that time when things were much simpler, we can piece together the story of how our universe evolved to the rich and complex place we find ourselves in today,” Jo Dunkley, physics and astrophysical sciences professor at Princeton University, and the ACT analysis leader, said in a statement. This is the earliest cosmic time accessible to our viewing. That’s because light would frequently scatter off of free electrons, making the universe opaque. It wasn’t until 380,000 years after the Big Bang when particles began to combine, allowing light to travel freely and ending the cosmic dark ages. The cooled remnant of the first light that permeated the universe is known as the cosmic microwave background—leftover radiation from the Big Bang that can still be detected in the distant universe. This ancient light carries with it clues to the universe’s past, as well as its future, allowing astronomers to get as close as possible to the Big Bang to be able to understand the birth and evolution of the cosmos. After staring at the cosmic skies from a mountaintop in Chile for 15 years, ACT was able to measure the intensity and polarization of the universe’s first light with extreme sensitivity. This allowed scientists to estimate the temperature, density, and velocity of the swirling material that occupied the baby universe, gauging just how much of it was there before it began forming galaxies and stars. The polarization of the material reveals the detailed movement of hydrogen and helium during cosmic infancy. “We are seeing the first steps towards making the earliest stars and galaxies,” Suzanne Staggs, director of ACT and professor of physics at Princeton University, said in a statement. “And we’re not just seeing light and dark, we’re seeing the polarization of light in high resolution…Like using tides to infer the presence of the moon, the movement tracked by the light’s polarization tells us how strong the pull of gravity was in different parts of space.” The images are helping scientists gather clues to the universe’s origin story. By looking over ACT’s measurements, the team behind the research was able to confirm the age of the universe at 13.8 billion years old, with an uncertainty of only 0.1%. “A younger universe would have had to expand more quickly to reach its current size, and the images we measure would appear to be reaching us from closer by,” Mark Devlin, professor of astronomy at the University of Pennsylvania, and ACT’s deputy director, said in a statement. The team was also able to measure more precisely that the universe extends out to about 50 billion light-years in all directions away from us, and contains as much mass as 1,900 zetta-suns, or the equivalent of almost two trillion trillion Suns. Rather than coming up with new theories, the measurements confirm that it’s business as usual for our surrounding cosmos. “Our standard model of cosmology has just undergone its most stringent set of tests. The results are in and it looks very healthy,” David Spergel, professor of astronomy at Princeton University, said in a statement. “We have tested it for new physics in many different ways and don’t see evidence for any novelties.”

ASML is the world’s undisputed leader in the design and provision of advanced chip-making equipment. The manufacture of the most state-of-the-art semiconductors by industry titans like TSMC, Intel, and Samsung is highly reliant on equipment from this Dutch firm, founded in 1984. However, like many household names in tech, ASML has quite a humble origin story. In ASML’s case, company lore states that it all began in a “leaky shed.” The current kings of the lithography machine world, ASML (Advanced Semiconductor Materials Lithography), started as a collaboration between Philips, a household name in the electronics business, and Advanced Semiconductor Materials International (ASMI). The vision was to set up a new company laser-focused on lithography systems, which the partners identified as a growing market. Standing on the shoulders of giants | ASML - YouTube Watch On ASML’s first base was the aforementioned leaky shed, and this isn’t just a humorous name; it really leaked. The building was located on the former Philips campus in Eindhoven, Strijp-T. A photo of the ramshackle single-story prefab-style construction shows it is joined to the first floor of the Philips TQ building, home of the firm’s ‘Industrial Electronics’ division. Building on R&D that had been in the pipeline since the early 1970s, ASML’s first system, called the PAS 2000 stepper, was quickly realized by the team. According to the company's post, the shipping container visible on this side of the leaky shed was put there “to keep everyone sane.” Apparently, the PAS 2000’s hydraulic oil pumps were so noisy the team was inspired to externalize them this way. Image 1 of 2 (Image credit: ASML on LinkedIn) (Image credit: ASML on LinkedIn) Thankfully for ASML and all the employees from the leaky shed, it wasn’t long until the operation’s business flourished, and the move to a purpose-built office and factory unit was possible in 1985. Another year later, the refined PAS 2500 came to market and would become the foundation for many future machines. Coincidentally, ASML’s partnership with lens manufacturer Carl Zeiss was inked at this time and continues to this day. In the 1990s, ASML says its “breakthrough platform,” dubbed the PAS 5500, was launched. Off the back of this success, credited to the classic litho system’s “industry-leading productivity and resolution,” ASML enjoyed a successful IPO in both the Amsterdam and New York stock exchanges. The first Twinscan machines with revolutionary dual-stage technology were launched in 2001. This same litho tool series stepped up a notch in 2010 with the introduction of the Twinscan NXE:3100, which leveraged EUV technology. Closer to the present day, ASML is again pushing the boundaries like no rival can with its EXE platform and High NA technology. Stay On the Cutting Edge: Get the Tom's Hardware Newsletter Get Tom's Hardware's best news and in-depth reviews, straight to your inbox. Contact me with news and offers from other Future brands Receive email from us on behalf of our trusted partners or sponsors The most recent annual report from ASML says it enjoyed nearly $31 billion in revenue, operates across more than 60 locations, and employs over 44,000 people. ASML is by no means alone in having such a humble origin story, however we don’t think the fabled U.S. garage-startups Apple, HP, or Google corporations had to suffer with a leaky roof. Dampness and electronics aren’t usually a desirable combination.

Nvidia acquired the Gretel synthetic data company for a figure in the order of billions of dollars, according to two people directly aware of the agreement. The price at which the transaction has been closed exceeds the most recent evaluation of Gretel, equal to 320 million dollars, underline the sources, even if the exact terms of the purchase are unknown. Gretel and its team of about 80 employees will become part of Nvidia, which will insert the company's technology in its growing suite of generative artificial intelligence services based on cloud and aimed at developers. Nvidia has already started developing and putting tools on the market for the generation of synthetic data, which allow developers to train their AI models and develop them for specific applications. In theory, synthetic data could create an almost infinite reserve of data for training and help to solve the problem of the scarcity of data that has looked in the sector since Chatgpt has become a commonly used system in 2022. But experts also argue that the use of this type of data entails risks. The spokesperson of Nvidia and Gretel did not want to provide comments for this article. What Gretel is and what is synthetic data for Gretel was founded in 2019 by Alex Watson, John Myers and Ali Goldhan, who still plays the role of CEO. The startup offers a synthetic data platform and a suite of bees to developers who want to build models of generative but do not have access to a sufficient number of training data or would meet privacy problems using information from real people. Gretel does not build or licensed border models, but perfects existing open source systems by adding privacy and safety functions. According to PitchBook, before the acquisition the company collected more than 67 million dollars in risk capital. Unlike those created by human or existing beings, synthetic data are generated by a computer by imitating real ones. According to its supporters, this approach makes the generation of the data necessary to build models of the most scalable, less demanding and more accessible to smaller developers or with less resources. Among the strengths there is also the protection of privacy, which makes synthetic data an interesting option for those who provide health services, banks and government agencies.

There's nothing like retro gaming on the Raspberry Pi but we haven't quite seen a gaming rig like this. Leave it to the Pi community to blow our minds and expectations out of the water. This project, created by maker and developer John Park is using our favorite SBC — the Raspberry Pi 5 — to drive a cool wall arcade featuring RGB LED matrix panels as the main display. Pico-8 Wall Arcade - YouTube Watch On According to Park, this setup doesn't just look the part. You can actually play games on the system like a real arcade using wired USB controllers. That said, you're limited by the display capabilities of the matrix panel display. It can run demos with cool retro-style animations but also play a few homebrew games that are created using the PICO-8 Fantasy console. The project is built around the latest Raspberry Pi 5 which not only drives multiple HUB75 RGB matrix panels but also runs the operating system used for gaming. The hardware is housed inside of an Ikea Sannahed picture frame which has been fitted with some custom 3D-printed components to mount everything together. The brackets are part of the design and can be seen around the perimeter adding a nice dash of color to its overall look. Image 1 of 2 (Image credit: John Park) (Image credit: John Park) The board Park selected to run this setup is the Raspberry Pi 5 4GB model. It's connected to four 64 x 64px HUB75 RGB LED matrix panels via an Adafruit RGB matrix bonnet that makes short work of proving power and data to the display. A USB external stereo speaker provides audio output. For user input, it relies on a couple of USB controllers. The software used to run the arcade games is known as PICO-8 Fantasy Console . This is a system that comes with a handful of custom games that are reminiscent of existing titles. It's not freeware, however, and costs $15 (USD). It's compatible not only with the Raspberry Pi but also Mac and Windows machines. PICO-8 can also be used to code, create music and even develop game assets like sprites and maps. If you want to get a closer look at this Raspberry Pi project , you're in luck. Park uploaded a full build guide for the project complete with links to everything you need to reconstruct it at home over at the Adafruit website.

While we have some time before the next major holiday sales event, it doesn’t mean you can’t still find a great deal. After searching diligently, we’ve found something very practical which you can treat yourself to so you can start the Spring off right. Whether for outings or emergencies, a backup portable battery is essential. This Anker Solix C1000 portable power station is now available at half off, offering a substantial $500 discount. This brings the price down from $999 to $499—the lowest price ever for this power station. See at Amazon The Anker Solix C1000 provides 2,400W of power, sufficient for 99% of appliances. Its 11 ports make it an excellent backup power source for emergencies or a convenient tool for camping trips and tailgate parties. Additionally, it features a built-in light so it’s great for either daytime or nighttime use. This thing is a beast when it comes to charging or powering your devices. With a fully charged Anker Solix C1000, you can recharge an iPhone 92 times over. A projector can operate for nearly 24 hours. In case of a blackout, you can power your Wi-Fi router for 52.1 hours, keep the lights on for 50 hours, or run your refrigerator for 14 hours. Don’t panic if you forget to charge your Anker Solix C1000 power station before a major trip. With its UltraFast recharging technology, you can reach 80% battery capacity from zero in just 43 minutes. Amazingly, it takes less than an hour to fully charge to 100%. The Anker Solix C1000 guarantees dependable power for ten years. Its battery is engineered to endure 3,000 charging cycles, translating to more than a decade of usage. Harness the Sun’s Energy If you’re living an coconscious lifestyle or are planning a multi-day camping trip, no sweat. When your in need of more than just what the power station can hold in a single charge, go ahead and pair your Anker Solix C1000 with a compatible solar panel. This way you can have practically unlimited power on any long outdoor camping escapade. The portable power station can be connected to the Anker app on your smartphone. Through it, you can customize charging speeds, view battery life, gain real-time insights, all over Wi-Fi or Bluetooth. You can even track power statistics in real-time. For a limited time, the Anker Solix C1000 portable power station has been slashed down to its lowest price ever. Right now, you can save over 50% on your power station, paying only $499. That’s a savings of $569. See at Amazon

An Asus exec at GTC 2025 has shown off what might be both the ultimate PC collector's item and the most valuable consumer graphics card on Earth. Asus Director of Marketing, Ernest Cheng, shared a picture of one of the extravagant ROG Astral GeForce RTX 5090 Dhahab OC Edition graphics cards. Lifting this to a higher level of geek Nirvana, this particular sample bears Nvidia CEO Jensen Huang's autograph. We think this lavish gold graphics card sample is probably worth over $16,000 after looking at previous auctions. As Cheng points out above, this 'Golden ROG RTX 5090 Astral' is now one of its kind, bearing Jensen's autograph and his "RTX ON!" inscription. However, it was already one of the rarest graphics cards among the rare-as-hen's-teeth RTX 5090 hardware that has reached retail. As a reminder of the calibre of product we have in the photo, you can check back through our extensive RTX 5090 review, and recap the ROG Astral GeForce RTX 5090 Dhahab OC Edition story we ran in February (linked top). The ROG Astral line is a new quad-fan flagship family from Asus, and an RTX 5090 OC model carries an official MSRP of $3,099, which is over 50% more expensive than a reference design. Now, add the golden shroud and design featuring detailed engravings of skyscrapers, camels, and Arabic calligraphy – and you have true opulence. This Middle Eastern exclusive will obviously carry a further price premium to the $3,099. Add in Jensen's autograph, and the sample graphics card in our picture will surely be worth a king's ransom. One thing collectors will be aware of, though, is an overabundance of signatures can devalue the addition of any autograph. Jensen seems to be pretty open to signing things from fans willy-nilly, which is nice but could limit future collector valuations. Already this year we have seen the official Nvidia GeForce social media channels offer up several historical GPUs accompanied by the CEO's scrawl. We also reported on Der8auer buying a Jensen-autographed Asus ROG Matrix GeForce RTX 4090 (with proceeds to charity) last year. That raised $16,000, which might help draw a baseline expectation on the sum the headlining 'Golden ROG RTX 5090 Astral' from the GTC 2025 event could achieve.

Screamboat‘s cast and crew talk bringing in Art the Clown himself to play Steamboat Willie. The Boys spinoff Vought Rising has found a few more mystery cast members. Plus, another look at what’s coming on Oh My God… Yes!. Spoilers away! Supergirl: Woman of Tomorrow In an interview with Screenrant, Jason Momoa briefly teased his appearance as Lobo in the new Supergirl movie. Well, this is the role I’ve always wanted to play. That’s the comic I loved, so I’m really nervous about it. It’s kind of a no-brainer to play this character. It’s pretty big. I don’t want to give away too much, but I mean, we look pretty dead on, exactly like the character, and he’s pretty rough and gruff and… I’ll say the bike’s really cool… It’s [Milly’s] movie, so it’s great. I just come in for a little bit. Invincible Speaking to Comicbook, Robert Kirkman is still holding out hopes that one day the comic series will also get a live-action adaptation, after the rights were picked up back in 2017. Unfortunately, there’s not much I can say. I will say that the television show makes it vastly easier and vastly more difficult in various different aspects of the development of the movie. That’s kind of where I have to leave it, unfortunately. It’s still early going. I would be shocked if it doesn’t happen one day. Screamboat The cast and crew of Steamboat discuss David Howard Thornton’s performance as Steamboat Willie in a new featurette. Fréwaka A home care worker’s new client is agoraphobic woman who fears the Na Sídhe–”sinister entities who she believes abducted her decades before” in the trailer for Fréwaka. Párvulos: Children of the Apocalypse We also have a trailer for Párvulos: Children of the Apocalypse, a dystopian zombie movie in which three kids must feed monster they keep in their basement. Vought Rising Deadline reports Elizabeth Posey has joined the cast of The Boys prequel series, Vought Rising, in an undisclosed role. Variety additionally reports Will Hochman has also joined the series in an undisclosed capacity. School Spirits School Spirits has been officially renewed for a third season at Paramount+. [Deadline] Oh, My God…Yes! Finally, Adult Swim has released a clip from this week’s episode of Oh, My God…Yes!

Hassan Youssef is the classic rags-to-riches story. Born in Quebec, Canada, to immigrant Lebanese parents, he climbed his way to vast wealth by building a digital empire — first in porn, and now in entertainment news. The founder and CEO of Valnet, based in Montreal, owns Screen Rant, Collider, MovieWeb and the comic book news site CBR, which according to the company, collectively drives 260 million page views every month. In the past decade, Valnet has scooped up most of the independently owned fan-facing news sites and is constantly seeking to acquire more, regularly sending pitch emails to independent content owners to buy them out and add traffic to its online juggernaut. “Valnet has one clear goal: become the greatest content and digital investment company the world has ever seen,” boasted a press release announcing the acquisition of tech site How-To Geek in 2023. The company does not disclose revenues. Lately, Valnet has been seeking credibility in Hollywood, such as with a March 11 screening Q&A and red carpet event at The Grove in Los Angeles, featuring Oscar-winning composer Hans Zimmer in honor of “the launch of music divisions on Collider and ScreenRant,” as a press release noted. Collider founder and editor-in-chief Steven Weintraub was the moderator. Hans Zimmer and Steven “Frosty” Weintraub at Valnet Celebrates “Hans Zimmer & Friends: Diamond in the Desert” at The Grove. (Screen Rant) But there’s a wrinkle in this story of a bootstrapping entrepreneur. When Valnet takes over a fan site, the playbook is well established: employees are replaced by contractors, compensation plummets and writers who complain land on a blacklist that blocks them from working for Valnet sites altogether. TheWrap spoke to 15 current and former contributors who said Valnet routinely exploits and discards writers while prioritizing mass quantity over quality to churn out mind-numbing SEO bait. “Everyone is underpaid, overworked and really pretty — extremely — exploited,” a former Collider contributor told TheWrap. “They are a company without clear objectives or morals who present themselves as a safe haven for young and experienced writers, but they end up paying bargain-basement rates and forcing contractors to create junky clickbait,” a former Valnet contributor told TheWrap on condition of anonymity out of fear of retaliation. Now one of those freelancers has filed a lawsuit claiming oppressive work conditions, and is seeking to establish a class action against Valnet. Daniel Quintiliano’s suit seeks $40,000 for violations including failing to pay minimum wage, overtime, provide meal or rest breaks and reimburse business expenses. The sum is unimpressive, but the prospect of extending this across a class of contractors has Valnet spooked — the company is now seeking signed releases from its contractors to not engage in any class action lawsuits against the company in exchange for $100 payment. Few have accepted the offer, and a new offer has gone out … for just $200. Valet release offer for $100. The Valnet Blacklist Valnet may find it has few allies in the writing pool, because another operating tactic of the company is blacklisting those who are perceived as difficult. TheWrap has obtained a spreadsheet of more than 400 “2025 Blacklisted Freelancers,” as the document is titled. When questioned about this, Valnet declined to comment. In an internal email accidentally forwarded to TheWrap, a Valnet executive said to another, “our documents are being leaked.” The list is a remarkable catalogue of criticism or perceived misconduct tracked meticulously. “Applied to Screen Rant and went to Twitter with the rates,” one entry says. Another freelancer was blacklisted for “creating drama” by “discouraging a past writer from returning,” while others faced a Valnet ban for calling rates “abysmal” or requesting transparency. One freelancer who suggested Valnet include payment information in job listings was permanently banned. (TheWrap has blacked out the writers’ names.) Other criticisms cited on the spreadsheet: “extremely rude and aggressive with hiring associate over rates and demanded a salary job”; “Combative, rude and argumentative with HR over pay”; and “Poor attitude, created drama via Writers Google Group email chain by discouraging a past writer from returning — speaking poorly about Screen Rant editorial team and company as a whole while being an active writer.” A spokesman for Valnet said he was unavailable for comment because he was on vacation. An M&A executive reached out to TheWrap saying the company would not be subject to “coercion” in responding, but did not respond to any specific questions. “I’d like to bluntly ask you: What is Umberto trying to achieve?” asked head of M&A Rony Arzoumanian in an email. “Valnet invests tens of $millions in content, annually. We support journalism and would gladly welcome your questions.” He did not respond to further emails or phone calls. “Valnet Blacklist 2025” “Valnet doesn’t care about exclusivity or quality,” said an insider familiar with digital publishing. “They arbitrage web traffic. They chase keywords in the same way day traders chase stocks. It’s quick gains over actual content. Their strategy treats digital content like a commodity. They flood the market with stories optimized for clicks rather than insight or originality.” From foosball to digital porn empire Hassan Youssef may be little-known in the clubby world of entertainment, and there’s a good reason for that. A civil engineer by training, he ventured into content creation after exiting an online porn business that he successfully built with his brother Sam and others, but that ended with a $6 million seizure of their funds by the Secret Service for alleged money laundering in 2009 before paying a $2.2 million fine. The origins of Valnet can be traced back to Montreal’s competitive foosball circuit, according to a 2011 New York Magazine investigative report titled, “The Geek-Kings of Smut.” In 2003, while still students, foosball players Matt Keezer, Stephane Manos, Sam Youssef and his brother Hassan decided to get into the world of online porn. Together, they started their first websites, with names like Jugg World, Ass Listing, KeezMovies and XXX Rated Chicks. “We started with $5,000, and we grew the company … Everything happened gradually. We had a little money, we reinvested it,” Sam Youssef told The Montreal Journal last year. Their venture proved hugely profitable. The group rapidly expanded, creating their own affiliate network (Jugg Cash) and their own paysite, Brazzers — named after a private joke, a “throaty immigrant-Arabonics version of ‘brothers,’” according to the report. The company went from 80 employees in 2007 to 250 by 2009. Sam Youssef emerged as the business visionary, while Manos handled sales and motivation. Hassan did analytics. Keezer became a wizard of algorithms and Search Engine Optimization (SEO), helping their websites dominate search for adult keywords. It was their mastery of SEO that would help them later, as they expanded to more outwardly legitimate fan sites, helping their sites consistently rank at the top of Google searches. Under the Brazzers brand, the company would develop a reputation for high quality productions, outsourcing the content creation for the websites to porn producers in Los Angeles, Las Vegas and Miami. The founders maintained a low profile and stayed under the radar amid their growing success. They operated through a complex web of corporate entities, with the main business housed under the name “Mansef” — a combination of the names Manos and Youssef. In January 2007, Keezer bought the domain name pornhub.com for $2,750 and went online through a separate company called Interhub. The Brazzers group were silent partners. But Interhub ran into trouble in 2009 when British teenager Rose Kalemba discovered her rape as a 14-year-old broadcast on Pornhub. Kalemba pleaded to have the content removed for over six months, but was ignored. Kalemba’s rape was viewed over 400,00 times. “The titles of the videos were ‘teen crying and getting slapped around’, ‘teen getting destroyed’, ‘passed out teen’. One had over 400,000 views,” Rose told the BBC. “The worst videos were the ones where I was passed out. Seeing myself being attacked where I wasn’t even conscious was the worst.” The video was finally removed after Kalemba set up a new email address posing as a lawyer and sent Pornhub an email threatening legal action. By 2009, before they drew the attention of federal investigators, the Montreal-based enterprise had grown into one of the largest adult entertainment operations online, with the founders maintaining offices in both Canada and the United States. Their success in adult entertainment would ultimately provide the capital and the SEO playbook that Hassan Youssef would eventually apply a few years later in his pivot into mainstream digital media publishing. Federal investigation and corporate transformation In October 2009, the U.S. Secret Service’s Organized Fraud Task Force seized $6.4 million from two Mansef bank accounts, according to New York magazine. Federal investigators uncovered over $9 million in wire transfers from Israel, and the feds said that had been wired into the two accounts over a three-month period from banks in Israel and other countries on financial-fraud watch lists. Of the $6.4 million that was seized, the government returned $4.15 million and $2.2 million of the sum was forfeited in a settlement filed with the court. In March 2010, the founders decided to get out of online porn and sold their adult entertainment assets to German businessman Fabian Thylmann, who paid $140 million for Pornhub.com and its related properties including Brazzers, according to Celebrity Net Worth. Thylmann changed the company’s name from Mansef to Manwin, but the Youssef brothers were out. From adult entertainment to entertainment media After selling Mansef in 2010, Hassan and Sam Youssef took some time off before launching new ventures. They founded Valnet in August 2012. By 2015, after losing money investing in gyms and real estate, Sam Youssef founded Valsoft Corporation, focusing on acquiring and improving existing software companies. Despite making their fortune in online porn, the Youssef brothers have scrubbed any mention of Mansef or their adult entertainment origins from their public personas of the Valnet site. The Valnet sites serve passionate fans of film, television, video games and comics and need major scale traffic to make money off programmatic advertising. Hassan Youssef speaks onstage at Valnet’s 10th anniversary in Montreal , Canada. (Linkedin) Remnants of the movie blog heydays of the late 2000s and 2010s when sites like Ain’t It Cool News made a mark, these Valnet-owned websites largely serve as fan-focused aggregation, commentary and listicle repositories focused on movies, TV, comic books and gaming. Many articles are designed to maximize consumer engagement (and fervent agreement or disagreement) with purposefully inflammatory headlines like “Arnold Schwarzenegger’s Perfect Replacement as the Terminator Is So Obvious, I Can’t Believe It Hasn’t Happened Yet.” The current state of these websites reflect a Valnet playbook to maximize clicks and revenue. Once acquired, rates for freelancers at some of these sites dropped from $150 for reviews to $40, according to multiple freelancers who spoke to TheWrap. The current rate for reviews is now even lower at $30. Access to health insurance and benefits evaporated as full-time employees were rehired as at-will contractors, they said. “I had written for Collider a couple of times prior to its Valnet acquisition, at a time when the freelance rate I was paid was $250 an article,” a former contributor said. “But soon after Valnet got involved, that rate dropped to $40.” Content creators allege exploitation According to another former contributor, Collider maintains “a content mill, borderline like almost sweatshop-level” where hundreds of writers are “constantly being pushed to write more, to do it quicker,” while editors face “pretty high level of quotas” despite being understaffed. The former contributor also alleged a pattern of firing employees who questioned company practices and recalled being called into a meeting where the individual was told “you should quit if you think it’s so bad here.” In an interview with TheWrap, former Valnet writer Stephanie Ann Grant detailed her experience working for Valnet’s “The Travel” vertical. After accepting a contract for $30 per published article on Oct. 3, 2024, Grant described being suddenly locked out of Valnet on Oct. 23, 2024. She received no response from editors for six hours before getting an email announcing the termination of her contract. “I then emailed the head of HR and then Valnet’s legal department. It appeared that Valnet had ghosted me,” Grant said. When her first paycheck failed to arrive on Nov. 1, Grant says her emails went unanswered for days until she received excuses that “it didn’t go through.” Only after posting about her experience on LinkedIn did she receive payment — though she said she was still shorted $65. “I checked my PayPal account settings to make sure it wasn’t a problem on my end, and as a freelance writer, this was the only time I’ve ever had someone having a ‘problem’ paying me,” Grant said. Valnet declined to respond to questions about Grant’s experience. According to the legal complaint filed by former MovieWeb writer Daniel Quintiliano, writers must submit to “Valnet’s complete and total discretion as to approval or revision of the work product” — a level of oversight that the lawsuit argues is inconsistent with independent contractor status under California law. The lawsuit details that Quintiliano regularly produced about three to four MovieWeb articles per day, five days or more per week, spending approximately two to three hours per article. Despite the intensive workload, he was paid only $15 per article with no additional compensation. Valnet continues to aggressively expand its portfolio of consumer-facing websites, sending acquisition inquiries and seeking legal releases from freelancers while CEO Hassan Youssef remains largely behind the scenes. The former Collider contributor said: “In journalism, there are really bad jobs. And then there is a place like Valnet,” adding it’s “one of the worst places that I’ve ever worked and is probably one of the worst journalism publications I’ve ever seen.”

The iPhone 16e's 5G modem has outperformed the iPhone 16's in some tests Its lack of mmWave support might be the only thing holding it back This bodes well for future iPhones and future versions of this modem One interesting aspect of the iPhone 16e is that it uses Apple’s C1 modem. This is Apple’s first in-house 5G modem, and so far the iPhone 16e is the only device you’ll find it in. With this being a cheaper phone than the rest of the iPhone 16 series you might expect the C1 to trail the Qualcomm modem used in those, but in fact tests suggest the C1 might offer better performance in most situations. Ookla (via 9to5Mac) has collected data from iPhone 16 and iPhone 16e users in the US, and found that the iPhone 16e offers 10th percentile download speeds of 27.35Mbps, compared to 16.66Mbps on the iPhone 16. Those are essentially the speeds that the users experiencing the lowest performance on either phone can expect. (Image credit: Ookla) For median download speeds the iPhone 16e also comes out on top, at 217.64Mbps compared to 210.55Mbps on the iPhone 16. However, at the top end the iPhone 16 takes the win, with 90th percentile download speeds of 756.13Mbps, compared to 560.4Mbps on the iPhone 16e. Notably though those upper speeds could be skewed by the fact that the iPhone 16e doesn’t support faster mmWave 5G, while Ookla suspects future versions of the C1 modem will support this. Just as good for uploads (Image credit: Ookla) The iPhone 16e and its C1 modem are similarly impressive when looking at median upload speeds, having the iPhone 16 comfortably beat across all three tested carriers (T-Mobile, Verizon, and AT&T). This all is good news for Apple, since it suggests the company could switch to its in-house modem across all new iPhones without sacrificing speeds, and it’s all the more impressive as Apple claims the C1 is also more efficient than other iPhone modems, meaning it shouldn’t impact battery life as much. Get daily insight, inspiration and deals in your inbox Sign up for breaking news, reviews, opinion, top tech deals, and more. Contact me with news and offers from other Future brands Receive email from us on behalf of our trusted partners or sponsors It does put the iPhone 16e in the unusual position of actually having an advantage over Apple’s top models – other than its lack of mmWave support – but that’s one more reason to buy this phone, and looking ahead, it’s a promising sign for future iPhones, especially since the next generation of this modem might be even better.

is a news writer who covers the streaming wars, consumer tech, crypto, social media, and much more. Previously, she was a writer and editor at MUO. YouTube is looking into an issue causing “lower than usual video quality” on iOS, desktop, and smart TVs. The platform says the apparent bug could make YouTube videos and shorts stream at 144p or 360p even with a strong internet connection. Over the past week or so, many users on Reddit have reported having YouTube automatically set their video quality to low resolution. One user notes, “Every video I click on YouTube the quality automatically sets to 144p even if I try to manually change it higher, it just reverts back to 144p.” Others found that when trying to switch to a higher video quality, the stream will start buffering. Trying to reset the device, changing the connection, or reinstalling the YouTube app doesn’t seem to help, either. Though YouTube currently doesn’t have a fix, the company says it’s “actively looking into” the issue and will update its support thread with new information.

Demand for Nvidia's latest GeForce RTX 5080 and RTX 5090 — some of the best graphics cards money can buy these days — is so high that it is nearly impossible to get them at their recommended prices in the U.S. However, Nvidia seems to have a solution for attendees at its GPU Technology Conference: buy them on site. Nvidia has limited quantities of its enthusiast-grade GeForce RTX 5080 and flagship GeForce RTX 5090 add-in-boards (AIBs) available for purchase at their $999 and $1,999 recommended prices from its Nvidia gear van at GTC park. The graphics cards must be purchased from the van from 7AM to 12PM on Thursday or Friday and then picked up at South Hall main entrance the same day. There are several catches. The graphics boards are available to conference pass ($1,145 for one day, or $2,295 for five days) and exhibit pass holders only, with a limit of one card per person. Also, Nvidia will only make 2,000 boards available in total: 1,000 GeForce RTX 5080 cards, and 1,000 of the GeForce RTX 5090 models. Finally, even Nvidia itself says that the boards are available spontaneously, in a kind of flash sale manner. For example, "90 GeForce RTX 5090 units are available for the next 30 minutes." 🚨 GeForce RTX 5090s are available at the Gear Store in the park right now at #GTC2590 units are available for the next 30 minutes, with more coming tomorrow. Come say hi! pic.twitter.com/oHdwMlp87rMarch 19, 2025 As it turns out, Nvidia's prioritization of datacenter GPU production over client GPU production has both positive and negative sides for the company. On the one hand, Nvidia can sell more ultra-expensive B100/B200 GPUs for AI at circa $50,000 a piece. On the other hand, there are not enough graphics cards for enthusiasts (who generate positive buzz around the brand) and software developers (who work on next-generation software that will drive demand for Nvidia's products in the future). Selling 2,000 advanced GeForce RTX 50-series boards at GTC to some degree solves this problem and lays the cards in the hands to software developers. However, this is certainly not enough. Of course, software makers with deep pockets are certainly looking forward to get Nvidia's DGX Station with GB300 Blackwell Ultra inside or the miniature GB10-based DGX Spark. The majority of developers will still use something more down-to-earth, which are GeForce RTX 50-series graphics cards for desktops.

You often find your various devices competing for the only available outlet? It’s frustrating having to make the tough call to which you’ll plug in, your phone or your laptop. You can connect one and then go plug in the other into a different outlet I suppose, but then that’s sometimes all the way on the other side of the room. Completely impractical. You’re better off investing in a reliable wall charger block that can power all your devices simultaneously. The Anker Nano 65W USB-C charger has multiple ports to plug into and can recharge your phone, tablet, or laptop fast. It’s normally $55, but right now you can secure one for 29% off at just $39. That’s a savings of $16. See at Amazon Fast-Charging in a Small Package The Anker Nano charger delivers impressive power despite its compact size. Comparable in size to an AirPods Pro Case, it is 53% smaller than a Mac charger, yet it can charge a MacBook Pro 13″ at full speed. This charger block is built for resilience. Its plug folds neatly, making it convenient for travel, whether it’s thrown into your bag or tucked into your pocket. There’s no need to worry about potential damage. Additionally, the exterior is both smudge-proof and scratch-resistant, ensuring it always looks pristine. Anker prioritizes the safety of your devices in all its designs. The Anker Nano 735 65W charger incorporates Anker’s exclusive MultiProtect technology and adheres to the latest safety certifications. This technology safeguards against short circuits and high-voltage outputs, regulates current, automatically matches current, controls temperature, and ensures the charger is resistant to static. This charging block is compatible with a wide range of devices. Whether you have a Samsung phone, JBL earbuds, or an Apple watch, you can be assured that this charging block will connect and charge your devices efficiently. The Anker Nano 735 features three charging ports: two USB-C ports and one USB-A port. When using one USB-C port by itself, it delivers up to 65W of power. If both a USB-C and a USB-A port are used simultaneously, the USB-C port provides 40W. And when all three ports are in use, one of the USB-C ports will still supply 40W of power. See at Amazon

During World War I, navies painted their ships in “dazzle” camouflage, also known as “razzle dazzle.” Unlike traditional camouflage, which helps objects blend into their surroundings, dazzle camouflage used stark geometric patterns to try to confuse German U-boat captains’ perception of a ship’s direction and speed, making it harder to target. But did the dazzle actually dazzle, or did it simply look ridiculous? Aston University researchers have investigated the efficacy of dazzle camouflage on WWI battleships by re-analyzing a 106-year-old study. According to the new paper, the unintentional “horizon effect”—when a boat seems to be traveling along the horizon, even if it’s not—was a far greater source of deception than the dazzle paint itself, exposing a key oversight in the 1919 analysis. Their findings are detailed in a study published on March 14 in Sage Journals. In 1919, MIT naval architecture and marine engineering student Leo Blodgett conducted a study on dazzle camouflage for his thesis. The study involved painting dazzle patterns onto a model battleship and observing how the patterns affected an onlooker’s perception of the ship’s direction of travel when viewed through a periscope. Blodgett ultimately concluded that dazzle camouflage achieved its aim. More than a century later, however, researchers Timothy Meese and Samantha Strong expressed significant concerns about Blodgett’s methods. Specifically, they suspected that the onlookers’ warped perception was not totally due to the dazzle paint. “It’s necessary to have a control condition to draw firm conclusions, and Blodgett’s report of his own control was too vague to be useful,” Strong, a senior lecturer in optometry, said in a university statement. “We ran our own version of the experiment using photographs from his thesis and compared the results across the original dazzle camouflage versions and versions with the camouflage edited out. Our experiment worked well. Both types of ships produced the horizon effect, but the dazzle imposed an additional twist.” The horizon effect dictates that viewers will perceive ships as traveling along the horizon even if they’re traveling at an angle of up to 25 degrees relative to the horizon. More broadly, viewers underestimate this angle, even when it’s greater than 25 degrees. If dazzle camouflage were single-handedly responsible for the visual deception noted in Blodgett’s study, viewers should have consistently seen the front of the ship, called the bow, “twist” away from the direction of travel, according to the researchers. Meese and Strong, however, pointed out that in certain instances—specifically, when the model boat was moving away from the viewer—the onlooker saw the bow “twisting” toward them. This indicates that another factor, beyond dazzle camouflage, was influencing the illusion. They identified the horizon effect and concluded it played a greater role in deceiving viewers than dazzle camouflage. The researchers “knew already about the twist and horizon effects” from a previous study that Meese, a professor of vision science, co-authored in 2024. However, “the remarkable finding here is that these same two effects, in similar proportions, are clearly evident in participants [from the 1919 study] familiar with the art of camouflage deception, including a lieutenant in a European navy,” Meese explained. “This adds considerable credibility to our earlier conclusions by showing that the horizon effect—which has nothing to do with dazzle—was not overcome by those best placed to know better.” In other words, the horizon effect hoodwinked even experienced individuals. Back then, however, “the horizon effect was not identified at all,” Meese added, so the effect was truly “deceiving the deceivers.”

The space science community has long prided itself on its ability to inspire and move people of all backgrounds, but President Donald Trump’s recent executive orders demanding the end of diversity programs have thrown that optimism into chaos. In response, NASA has suspended funding for diversity and outreach programs, paused the meetings of community groups that interface with space scientists, and banned the activities of internal employee resource groups for women, queer people, and others. The White House additionally moved to terminate thousands of probationary NASA employees before suddenly reversing the decision at the last minute, though the threat of deep layoffs and budget cuts of up to 50 percent continues to hang over the heads of agency workers. NASA has also closed three of its offices this week, including the diversity, equity, inclusion, and accessibility (DEIA) branch of its equal opportunity office, and laid off its workers. Those who remain at NASA are hiding symbols of LGBTQIA pride and removing pronouns from their email signatures. Talented scientists who receive federal funding are in fear not only for their jobs but also for their children’s futures, with many looking for work outside the federal funding structure or even considering leaving the US altogether. “Executive orders are not optional,” said Charles Webb, acting director of NASA’s planetary science division, at a conference on Monday, describing NASA as “racing to comply” with the orders. These changes demanded by the Trump administration are rolling back decades of evidence-supported work in diversity and outreach. Experts warn that these actions will impede scientific discovery; create a smaller, less creative, and more generic agency; and could even lead to more accidents and loss of life as people working on space missions don’t feel that they can speak out about problems that they see. Taking an axe to diversity programs isn’t making NASA a more efficient agency — it’s undermining the values of science. The chaos is the point Those working within NASA are bound by the government’s whims, but the chilling impact of these changes is being felt far beyond civil service. The legality of Trump’s executive orders is up for debate, leaving individual federal grant recipients grappling with how to handle them. The Verge spoke to nine people working in space science (others declined to speak to the media out of concern for their positions), several of whom described receiving conflicting emails from their university employers on what work is or is not allowed, and on what initiatives are being supported, with guidance changing daily. A general atmosphere of fear and worry is leading many to keep their heads down in hopes that their research may remain unaffected. Taking an axe to diversity programs isn’t making NASA a more efficient agency — it’s undermining the values of science. Scientists typically look to professional groups for guidance, though most have failed to step up to the plate. The American Geophysical Union (AGU) has been widely seen as mishandling its response to the crisis by preemptively deleting content from its website related to diversity and then restoring it after criticism. Other groups like the Space Science Institute (SSI) have also removed pages about a commitment to diversity from its website. Even NASA itself briefly removed mentions of inclusion as one of its core values before reinstating them. “The removal of a statement (which is just words on a page) does not diminish the commitment of individual researchers and educators in the organization to the idea that diversity in the space science workforce is crucial, and that educational access for all Americans regardless of gender, creed, national origin or other identity breeds innovation and progress in the next generation of scientists,” the SSI said in a statement to The Verge, citing the executive orders as necessitating the change. With NASA under threat of deep cuts, universities in damage control mode, and most professional organizations floundering, scientists are having to look outside their typical organizational structures to resist Trump’s executive meddling. “We’ve not seen the kind of leadership that I think we need, and that tells me that the way to push back on this is largely going to be kind of grassroots. It won’t be coming from the top down. It’ll have to come from the bottom up,” says Paul Byrne, a planetary scientist at Washington University. NASA’s Artemis II crew in December 2024. AFP /AFP via Getty Images No basis in scientific evidence A theme that experts repeatedly drew attention to was that the promotion of diversity in the sciences was not a matter of window dressing or checkboxes, but an important pillar of critical thought. While they supported wider access to science for ethical and human reasons, they also emphasized that, in purely pragmatic terms, greater diversity among scientists leads to better science. Trump’s crusade against diversity in sciences “has no basis in scientific evidence,” according to Julie Rathbun, who works in Cornell University’s astronomy department and has worked as a leader in DEIA programs in the past. “Scientific evidence for years has said diverse groups do better science,” she says. “Social science tells us diverse and inclusive groups do better science and better technology, and have better outcomes.” NASA previously embraced diversity because it was following the evidence, Rathbun says, and it adopted policies that enabled a range of voices in the room to question and challenge each other. That attitude has been largely embraced by space scientists of all backgrounds, who see the practical and ethical values in diversity. “We’ve not seen the kind of leadership that I think we need, and that tells me that the way to push back on this is largely going to be kind of grassroots.“ The 1986 Challenger disaster — in which seven crew members were killed when their Space Shuttle broke apart shortly after launch — was directly linked to a homogeneity of thought among NASA personnel. The agency’s lack of diverse perspectives fed into the tendency toward groupthink that contributed to the disaster, while research has shown that more cultural and ethnic diversity in groups leads to more creative and higher quality ideas — and lower risks for space missions. Diversity initiatives are necessary if the field is to accurately represent the US population, supporters say, noting that they have been making progress toward that goal. In planetary science, for example, a 2011 demographic survey found that only 25 percent of US researchers were women and just 1 percent were Black or Latinx. There were slight improvements made in the following years, with an increase in a 2020 survey to 35 percent women and 4 percent Latinx researchers, though Black researchers remained distinctly underrepresented with no improvements made in their numbers. The stark lack of Black voices in the field is exactly the kind of issue that DEIA programs hoped to address through outreach and support. “DEIA is not just lip service. It’s actually trying to make a level playing field,” Byrne says. “And there is a very kind of cold statistical argument behind DEIA that transcends just the moral, which is that we know statistically and scientifically that the more varied a set of perspectives we have, the better the outcome.” This view was echoed by Mark Sykes, CEO of the Planetary Science Institute (PSI), one of the few leaders of a professional group who did send an unambiguous statement of support for diversity. The role of imagination is key for scientific advances, he says, and a more varied group of people can imagine a broader range of possibilities. “It’s not rocket science,” Sykes says. “A lot of times, it’s just being considerate of everybody.” WASHINGTON D.C., UNITED STATES - SEPTEMBER 15: The National Aeronautics and Space Administration (NASA) Headquarters is seen in Washington D.C., United States on September 15, 2023. (Photo by Celal Gunes/Anadolu Agency via Getty Images) Anadolu Agency via Getty Images An ahistorical departure This recent slash-and-burn approach by the Trump administration is striking because, traditionally, space science has been beloved on both the left and right. Conservatives have generally been pro-NASA spending, seeing it as a way to burnish the US’ image and to make it a leader in space development as a source of national pride. That right-wing support for NASA has extended to support for its diversity initiatives, too. Inclusion was added as a core NASA value by then-administrator Jim Bridenstine, a Republican and first-term Trump appointee. Much of the Artemis program’s promotion under Bridenstine’s tenure revolved around the desire to put the first woman and person of color on the moon. Bridenstine also issued a policy statement affirming commitment to equal opportunities within NASA and its partner institutions. That policy statement has now been deleted. NASA’s interest in diversity is neither new nor a purely left-wing phenomenon. Threatening to gut the agency’s staff and cudgeling its outreach programs is not a return to a traditional conservative approach but rather veering into a crudely “anti-woke” ethos that has no interest in evidence, reality, or history. An ahistorical departure In such circumstances, space scientists are finding their own ways to band together and resist. Sykes says he is committed to promoting diversity as a key value of the PSI. If that means federal funding dries up then they will look for other sources of money, perhaps even crowdfunding. He sees outreach and providing input on DEIA topics to be crucial work that they will keep doing, saying, “If NASA wants, or the government wants to go after us for doing that, then, well, the hell with them.” For Rathbun, who was active in programs like NASA’s now-suspended Here to Observe initiative that provided outreach for underrepresented students, she says she and her fellow scientists will be continuing their mentoring and support of these students, even if the official program is closed, and inviting them to meetings and workshops to let them know that there are still avenues open to them. Scientists are even taking it upon themselves to support their colleagues through mutual aid funding. The largely early career researchers who make up the Choir Collaboration, a group dedicated to studying galaxy evolution and promoting intersectionality in science, have started a mutual aid initiative and are collecting financial donations to be distributed to space scientists who have been affected by the cuts, especially those working in DEIA. For those who are working to diversify science, these efforts are a fundamental duty that scientists have to the broader community. “We view caring and enabling and fostering and uplifting DEIA initiatives, equity, diversity in science, as well as the people who are doing the science, as a part of our job description,” says Choir member Erini Lambrides of NASA Goddard. NASA’s interest in diversity is neither new nor a purely left-wing phenomenon. Financial pressures are a real concern for early career scientists, for whom a missed paycheck can be an enormous hardship. They described embarking on postdoc positions and having just a few hundred dollars in the bank, or hearing from colleagues who were told with less than a week’s notice that they might not be getting paid that month. No one becomes a space scientist for the money, but neither scientists can’t be expected to do good work when they can’t afford to pay their rent. That’s particularly true of those who come from marginalized backgrounds or who work in DEIA initiatives, which are often unfunded and are frequently considered by employers as not as worthwhile work for a scientist as, say, producing a highly cited paper. That devaluing of outreach work by prospective employers overlooks the human aspect that powers all of science. “Science is never done in isolation, right?” says Choir member Taylor Hutchison, also of NASA Goddard. “We’re people first. And we’re part of communities.” That sense of community has also powered the mutual aid initiative, which has been taken up and supported by scientists at all career stages. It has spread even beyond the US, reflecting the global nature of the field and an increasing push against astronomy’s imperialist roots and its centering of the Global North. “The response has been fantastic,” said choir member Gourav Khullar of the University of Washington. “And we’re engaging with the broader astronomy community even outside North America, which is very representative of what the astronomy field is.” A SpaceX Falcon 9 rocket lifts off from launch pad 30A at the Kennedy Space Center, carrying four astronauts to the Space Station. (Photo by Manuel Mazzanti/NurPhoto via Getty Images) NurPhoto via Getty Images An attack on science Researchers working outside of federal agencies expressed sympathy for those working within NASA. Federal workers have little freedom to challenge executive orders or to stand up to the government, even in self defense. “Government employees are not the enemy,” Rathbun said. “They do really good work, especially people at NASA.” There is a sense that workers at NASA and other federal agencies are being thrown under the bus in the name of a culture war, with their work being denigrated in a way that they don’t have the ability to refute. “Every time you talk to someone who actually works with career civil servants, you realize that overwhelmingly, they’re super motivated, well-meaning, passionate people,” said Byrne. “These are not bad people. There’s no deep state. These are people who care about their country and the planet and their species, and want to help.” “Government employees are not the enemy.” While few would argue that there are some areas in which NASA could curtail its spending and reduce costs, such as the ever-ballooning pork barrel buffet that is the agency’s Space Launch System rocket, the nixing of DEIA values is neither efficient nor supported by the evidence. It’s the partisan nature of these orders that have people worried, as they undermine the structures that are especially necessary at the cutting edge of scientific thought. “Many of us work in fields where we’re trying to challenge and break existing paradigms within our science,” Lambrides says. “And in order to really test and break those paradigms, you do need diversity of thought. You do need people coming from different frameworks of life, and how they solve and how they think about problems. Because these are brand new problems we’re working on.” This isn’t the first time that scientists have seen their work disregarded, and researchers see a clear through-line in right wing attacks on scientific principles. “If you look at these executive orders in total, it’s an attack on science as a whole,” Rathbun says. “An attack on DEIA is the same as an attack on vaccines, the same as an attack on climate change. It’s an attack on the process. It’s not letting us fulfill that scientific goal of getting our best understanding of the universe.”

Ten years ago, the mythical 'Nintendo PlayStation' became a solid reality when a prototype was discovered. This would eventually change hands for $360,000 in 2020, and was believed to be the only surviving example of the unreleased system—until now, that is. Japan-based videographer and photographer Julian Domanski recently posted that he got to "fondle" a second prototype during a meeting with former Sony Computer Entertainment CEO Ken Kutaragi himself—also known as the Father of PlayStation. Kutaragi spearheaded the PlayStation project when the deal with Nintendo went south. The arrangement had been for Sony to create the CD-ROM add-on for the SNES as well as the stand-alone, all-in-one PlayStation, but the story goes that Nintendo was unhappy with Sony's insistence that it took a cut of every SNES CD game sold. At CES 1991, Nintendo threw a curve ball by announcing it was teaming up with Philips instead—embarrassingly for Sony; it had only revealed it was working with Nintendo the previous day. The last one in existence was believed to have sold at auction for $300,000. Turns out the ex CEO of SonyCEA has one in his closet. Ken Kutaragi, top bloke. Signed my PS1 too! I never thought I'd see something so rare, but today I actually got to fondle a Nintendo PlayStation!The last one in existence was believed to have sold at auction for $300,000. Turns out the ex CEO of SonyCEA has one in his closet. Ken Kutaragi, top bloke. Signed my PS1 too! pic.twitter.com/EGRos3IVT8 March 1, 2025 "I never thought I'd see something so rare, but today I actually got to fondle a Nintendo PlayStation," says Domanski on social media. "The last one in existence was believed to have sold at auction for $300,000. Turns out the ex CEO of SonyCEA has one in his closet. Ken Kutaragi, top bloke. Signed my PS1 too!" Given how close he was to the project, it's not a massive shock to learn that Kutaragi has a Nintendo PlayStation in his house; what is surprising is that it's taken this long for it to become common knowledge.

With all of the great smartphones, wearables, and AV devices that Samsung has been releasing, sometimes we forget that they’ve also been the world’s #1 flash memory brand for over 20 years. But since 2003 the performance, reliability, and durability of Samsung’s portable solid-state drives and other flash memory products has earned them that lofty title. A great example of Samsung’s mastery of portable hard drives is the T7 Portable SSD, and the 2TB version of that highly rated drive is on sale at Amazon right now for just $140 down from $269, a 48% price cut. The 2TB T7 has more than enough capacity for students and professionals alike to store their files or use it as an automatic backup, and it’s fast enough for gamers and creative professionals who deal in massive files every day. Everyone across the board will love how it’s amazingly compact and easy to carry, yet tough enough to be taken on outdoor trips or photo shoots. See at Amazon Made for Creators and Influencers Influencers and creative professionals will love the T7’s built-in PCIe NVMe technology, delivering impressive read and write speeds of up to 1,050/1,000 MB/s. This translates to nearly double the performance of the previous T5 model, drastically cutting down file transfer times and ensuring a seamless experience when handling large media projects. Built with a durable aluminum unibody, the T7 is designed to withstand drops from up to six feet, offering excellent shock resistance. This makes it a dependable choice for photographers, videographers, and digital nomads who need reliable storage in unpredictable conditions. For videographers, the T7 is a game-changer—it allows direct recording of Ultra High-Resolution (4K 60fps) videos from compatible cameras and mobile devices. It even supports ProRes 4K at 60fps on the iPhone 15 Pro and 16 Pro, which eliminates extra steps in the workflow and making content creation more efficient. See at Amazon Universal Storage Whether you’re working on a PC, Mac, Android device or the latest iPhone, the T7 offers seamless cross-platform compatibility including gaming consoles. With its amazingly small profile and up to 2TB of storage, it’s the perfect solution for everything from work projects to gaming libraries, all without compromising on speed or reliability. Getting started with the T7 is effortless, thanks to its plug-and-play functionality—no complex setup required. Weighing barely 2 ounces, it’s incredibly portable, slipping easily into a pocket or bag. For those prioritizing security, the T7 offers optional password protection and AES 256-bit hardware encryption and keeps your sensitive data safe from unauthorized access. To see the deal: See at Amazon

The French are addicted to screens. Two out of three even declares that they cannot be able to do without your more day smartphone! The 2025 edition of the digital barometer, carried out at the request of Arcep, Arcom, the General Council of the Economy and the National Agency for Territorial Cohesion (ANCT) draws up an implacable observation of our addiction to digital. All tech news for pros every day in our newsletter learn more about the use of personal data This survey, carried out with 4,066 people representative of the French population aged 12 and over, even reveals increased use compared to the previous edition. On average, the French spend about 4 hours a day in front of their screens, a quarter of the waking time. For one in four people, exposure to screens even exceeds 5 hours. If the youngest have - not surprisingly - a more intensive use of digital than the rest of the population, they are also more aware of this overexposure. 61 % of 18-24 year olds spending more than 3 hours a day ahead of the screens estimate this excessive time, compared to 31 % of the over 70s. This awareness is beneficial while the government is trying to reduce the hyperconnection of children. 96 % of teens have a smartphone The number one screen remains, of course, mobile. In 2024, 91 % of the population was equipped with a smartphone, an increase of four points in one year. The adoption is almost universal among young people (96 % of 12-17 year olds) and adults under the age of 60 (98 % of 18-39 year olds and 95 % of 40-59 year olds). Notable fact, the elderly gradually fill their delay with a strong increase, ten points in one year in the 60-69 year olds (90 %) and hui points among 70s and over (70 %). The computer equipment rate is also increasing slightly (89 %, + 2 points), as well as that of the tablet (54 %, + 1 point). Connected speakers finally find their audience, with a third of the equipped population, a gain of 4 points in one year. More broadly, 40% of French people have at least one connected object linked to health, safety, home automation or household appliances. The disparities linked to socio-demographic profiles tend to be reduced. If many homes, executives and young people remain the most equipped, the proportion of non-graduates with a smartphone increased by 6 points in 2024 to 63 %. The underestimated environmental impact This multiplication of digital uses leads to abuse. On average, each household has around ten equipment with screen, two of which are unused and could be recycled or sold. If the lifespan of smartphones is lighter - more than a quarter of respondents keep their mobile more than 3 years against 16 % in 2020 -, the purchase of new phones remains the norm. Only 22 % of smartphones are bought second -hand or reconditioned. Likewise, we observe an inflation of the mobile packages swollen with data. The share of package holders of more than 100 GB has doubled in four years, from 15 to 32 %. "However, two thirds of individuals almost never consume all of the volumes available," notes the barometer. And if the average consumption of data on mobile networks (16.5 GB per month and by SIM card) continues its progression, it is less sustained. Growth exceeded 10 % in 2024, against around 25 % per year between 2020 and 2023. The generalization of fiber optics accompanies this movement. In 2024, three-quarters of people surveyed with a fixed connection to the Internet use a fiber or cable subscription (+ 8 points in one year), while copper (ADSL) was still in the majority in 2020. The proportion of fiber subscribers in municipalities of less than 20,000 inhabitants, which reaches 69 % in 2024, approaches agglomerations. A digital life outside social networks What are the French doing of all this time spent in front of the screens? Social networks monopolize their attention and relate to 75 % of Internet users, three points in one year. Their consultation is generally passive. 62 % of users read the content published by others every day, but only 29 % share or comment daily, and 19 % publish personal content. Surprise: Seniors publish more content than adolescents. Beyond social media, uses are diversifying. 75 % of French people go on the internet to search for routes, 67 % to make a medical appointment, 31 % defend causes in the form of a petition, and 29 % look for accommodation or a job. Online administrative procedures are generalized. They concern 73 % of French people (+ 2 points), with a significant increase among those over 70 (+ 7 points). After a decline in 2023, online purchases of non -food goods increased again in 2024 and concern 77 % of the population (+ 4 points). The clothing remains the most purchased category online (56% of French people), ahead of books or music (40%). The second hand is a significant success: 53 % of French people have made at least one purchase or a sale on specialized sites, such as Le Bon Coin or Vinted, during the past year. The use of AI progresses, distrust persists Chatgpt phenomenon obliges, the use of artificial intelligence is progressing strongly. 33 % of French people used AI tools in 2024, compared to 20 % a year earlier. Uses in privacy dominate (26 %, +10 points in one year) and slightly exceed those related to professional life (22 %, also +10 points). Despite this adoption, distrust of AI remains the majority within the population (56 %). The main fears concern the destruction of jobs (62 %), its negative impacts on artistic creation (53 %), the environment (48 %), education and training (47 %). This distrust falls with the use: only 26 % of people who have already used AI is wary of it, against 71 % of those who have never experienced it. The digital divide is reduced Is this hyperconnection judged positively? Overall, two thirds of French people believe that digital technology facilitates their daily lives (+ 4 points). However, this proportion is reversed for non-graduates. Almost 60 % of them consider that digital complicates or has no effect on their daily lives. Despite the massive use of digital technology, almost two thirds of individuals feel fears of its use. The main concerns relate to the inappropriate use of personal data (28 %), followed by the risks of scam, fraud or online harassment (19 %). The digital fracture tends to be reduced. The proportion of the population encountering brakes on a complete use of digital tools decreases significantly, going from 48 % in 2022 to 36 % in 2024. 20 % of French people have the feeling of not sufficiently controlling IT tools. Almost one in two (44 %) declares to encounter difficulties in carrying out administrative procedures online. Visual generated by AI, Microsoft Bing

SoftBank Group today said it would acquire Ampere Computing, a developer of Arm-based datacenter CPUs, for $6.5 billion in cash. Ampere will continue operating under its current name as a wholly owned subsidiary with 1,400 employees and will continue to offer its processors. SoftBank, which also controls Arm Holdings, is expected to use Ampere CPUs for its own datacenters, including proposed Stargate datacenters. Ampere was established in 2018 in Silicon Valley with the backing from Carlyle Group and Oracle with an aim to develop CPUs tailored for the needs of cloud service providers (CSPs), which the company calls 'cloud-native' processors. Since then, the company released two product families — the Ampere Altra with up to 128 cores made on TSMC's N7 node and the AmpereOne with up to 192 cores made on TSMC's N5 manufacturing technology — that gained limited traction in the CSP. The company's roadmap includes launch of AmpereOne CPUs with up to 256 cores produced on TSMC's N3 fabrication process in 2025 and then moves to something more advanced in 2026. With Arm's plans to offer its own datacenter-grade processors targeting hyperscalers and CSPs, Ampere's chances to gain a significant market share got substantially lower. Also, Ampere Computing has to compete with Intel's Xeon CPUs — such as Sierra Forest with up to 144 cores and its successor Clearwater Forest in 2026 — and this is something particularly hard given the ample ecosystem available for x86 CPUs. (Image credit: Ampere) However, as a statement from Ampere reads, the company "is expected to play a key role in future Softbank growth markets, including recently announced AI infrastructure investments like Stargate." SoftBank's AI strategy involves building massive cloud datacenters. Next-generation cloud datacenters for AI are expected to house tens or hundreds of thousands of nodes, which will represent a huge market. If Ampere supplies millions of CPUs to its owner, then its business will expand beyond where it is now. Also, if other Stargate partners — OpenAI and Oracle — adopt Ampere's CPUs as a default choice for their datacenters, this could be a breakthrough for the company. This may never happen, though. "The future of Artificial Super Intelligence requires breakthrough computing power," said Masayoshi Son, Chairman and CEO of SoftBank Group Corp. "Ampere's expertise in semiconductors and high-performance computing will help accelerate this vision, and deepens our commitment to AI innovation in the United States." As part of the agreement, SoftBank will purchase Ampere for $6.5 billion in cash. Ampere’s major investors — Carlyle and Oracle — are divesting their stakes in the company. The deal is contingent on standard closing requirements, including regulatory approvals, and is anticipated to be finalized in the latter half of 2025. "With a shared vision for advancing AI, we are excited to join SoftBank Group and partner with its portfolio of leading technology companies," said Renée J. James, Founder and CEO of Ampere. "This is a fantastic outcome for our team, and we are excited to drive forward our AmpereOne roadmap for high performance Arm processors and AI."

Summit of members of the Linux Foundation - (NAPA, California) - The Open Source code has become a vector of malicious software. For example, a free software developer discovered that Jia Tan, chief programmer and manager of the Linux XZ data compression library, was also a computer pirate. He had introduced a stolen door into the code to allow attackers to take control of Linux systems. All tech news for pros every day in our newsletter learn more about the use of personal data One of the main problems is that no one knew who Jia Tan was. And we still don't know it. Jim Zemlin, executive director of the Linux Foundation, addressed this fundamental problem of the identity of the developers at the top of the members of the Linux Foundation this week in Napa, California. Mr. Zemlin began his intervention by declaring: "Open Source is now a fundamental element of modern IT, and IT pirates pay attention. In addition, we have seen a whole series of new regulations around open source appear, such as the Cyber-resilience law (CRA) of the European Union. The time when open source software was the subject of a minimal exam. Develop a decentralized system of confidence To further complicate things, we live in an era of "geopolitical tensions and the rise of techno-nationalism". The Linux Foundation, however, wants to preserve anyone's ability to participate in the open source from any place on the planet. " This goal is easier to say than to do. Mr. Zemlin stressed that if security is crucial, confidence is just as much. Confidence not only means ensuring that software is secure, but also to verify the identity and intentions of contributors. To meet these challenges, Mr. Zemlin proposes to develop a decentralized confidence system, including a "trusted score card" (Trust Scorecard) similar to those of the Open Source Software Foundation (OpenSSF). This system would help users to assess the reliability of open-source projects on the basis of factors such as the verification of contributors and project history. The Linux Foundation explores the first person project In such a system, developers should gain levels of confidence before they can take care of an important project. The factors that could be taken into account in this evaluation are: Proof of identity The history of contributions to the project Code quality Security history The reputation of the community Thus, for example, using such a system of confidence, by comparing Jia Tan to Greg Kroah-Hartman, the famous stable nucleus of Linux, we would see that Jia Tan would barely reach level 1. Kroah-Hartman would obtain a perfect score. To build such a system, the Linux Foundation explores the First Person project, which aims to establish a decentralized accreditation system using blockchain technology. This system would allow contributors to verify their identity while preserving their anonymity. An attempt not so operating as that with PGP Project managers would establish trusted policies. For example, they could ensure that their contributors meet specific criteria, such as verified employment or participation in industrial events. A group, those responsible for the Linux kernel, already use this system. After a safety flaw in the Linux code filing in 2011, the Linux team uses a Pretty Good Privacy (PGP) signature system to ensure that any new code comes from a known programmer. However, this system does not work so well that it told me Kroah-Hartman. "We hate this system, but we do with it." Kroah-Hartman and Linus Torvalds, the creator of Linux, therefore plan to use another system. However, any system of this type must have a "low friction" so that it is easy to integrate new people. To achieve this, Mr. Zemlin called for the collaboration of large companies and free software communities to develop and implement these confidence mechanisms. This effort involves working with organizations such as the Trust Over IP Foundation and the Openwallet Foundation to create evolving and respectful approaches to privacy.

For at least two decades, the software as a (SaaS) service has been at the center of the technological universe, with fully accessible applications via the Internet. Today, artificial intelligence upsets this situation. In fact, applications work more and more independently to provide services. This is called "services as software" (AS Software services). All tech news for pros every day in our newsletter learn more about the use of personal data AI creates this new world for IT pros. "Historically, companies bought software and hired professionals to operate it," describes Frank Diana, a prospectivist at Tata Consultancy Services, in a recent post. "AI is overthrowing this model. More and more, companies will design their processes according to AI, from the start. This change will have an impact on almost all functions. An HFS study reveals that 60 % of the 1,000 companies questioned are already looking to acquire services as a technological offer. The same number "plans to replace part or all of their professional services with a form of AI in the next three to five years". Imminent changes in the roles of IT pros Similar trends are at work because routine computer maintenance, HR, supply chain, accounting and customer service "become much easier to reproduce in latest generation AI and agentity software, supported by public and private cloud capacities to secure and increase transactions volumes" say the authors of the study. What once required teams of specialists "is more and more supported by systems fueled by AI and capable of performing tasks independently," says Diana. "This transformation is not only a question of automation - it is a question of redefining the way in which companies consume and provide services. The emerging model marks a deep break with the past. The software is no longer a tool for professionals. In many cases, it is the professional." This announces imminent changes in the roles of IT pros. "The SaaS models were mainly articulated around applications hosted in the cloud," said Diana to Zdnet. Services as software "integrate technologies such as AI, digital twins, predictive analysis and autonomous systems, which obliges professionals to go from infrastructure management to dynamic and adaptive service orchestration." Realign skills This transition "underlines the critical need for professionals to realign their skills. And this by giving priority to mastery of AI, strategic reflection and man-machine synergy", continues Diana. "In addition, it will be essential to adapt to new professional careers focused on the orchestration of integrated service ecosystems and on the improvement of interfunctional collaboration." Despite all automation, human expertise remains essential, underlines Franck Diana. If AI "can process data, humans will provide context, vision and leadership."

When Samsung dropped the Galaxy Watch Ultra last July, it was instantly compared to the Apple Watch Ultra 2 which had come out several months earlier. In fact, we compared them directly, and basically ruled this clash of flagship smartwatches to be a draw in terms of quality and performance. But when Samsung decides to make the Galaxy Watch Ultra available for as little as $325 with trade-in credit or $420 without (down from $649), suddenly the Apple Watch Ultra 2 is looking very pricey indeed. See at Samsung.com If you’re able to max out the trade-in credit from Samsung and get the full $325 they’re offering, the Galaxy Watch Ultra comes in at less than half the price of the Apple Watch Ultra 2, basically making the choice for you. Even without a trade-in, the $230 in instant savings currently offered by Samsung makes the Galaxy Watch Ultra more affordable than the Apple Watch Ultra 2 by over $350. Either way, Samsung’s further sweetening the pot with great discounts on accessories, like 61% off a pair of Galaxy Buds3 or 41% off the Galaxy Buds3 Pro. It Loves a Workout Samsung has packed the Galaxy Watch Ultra with advanced fitness-tracking technology so it can go toe-to-toe with the vaunted Health app of the Apple Watch lineup. The Galaxy AI-powered heart-rate monitor eliminates inaccuracies caused by motion, providing exceptionally precise readings. The Galaxy Watch Ultra also delivers VO2 Max measurements, estimated sweat loss, and performance tracking for runners and cyclists, allowing you to compare previous workouts and gauge improvements over time. This is the watch for you if you’re a hardcore adventurer and/or watersports enthusiast, because it can take anything dished out by your active lifestyle. The titanium case of the Galaxy Watch Ultra is ultra-durable, and engineered to keep up with your toughest workouts. With 10 atmospheres of water resistance, it’s ideal for swimmers, divers, and kayakers, while hikers and climbers will appreciate enhanced GPS tracking and turn-by-turn navigation that ensures you never lose your way. See at Samsung.com The Numbers Don’t Lie You won’t have to worry about the Galaxy Watch Ultra’s battery running out at inopportune times — it has the longest battery life of any Samsung smartwatch to date. That means after it tracks your toughest workout, you’ll be ready for bed and the Galaxy Watch Ultra will still be up and running to monitor your sleep quality. Unless you have extreme brand loyalty to the Apple Watch or other fitness-oriented wearable, the option to buy Samsung’s newest flagship smartwatch for well under half the price of the Apple Watch Ultra 2 when you max out the $325 trade-in credit offer is hard to pass up on, especially when you consider they are basically neck-and-neck in terms of performance. Even with no trade in, the $420 price with instant credit makes the Galaxy Watch Ultra the clear choice. See at Samsung.com

Is a senior reviewer who has worked at the virgin since of its founding in 2011. His coverage areas including audio (Apple, Bose, Sony, etc.), Home Theater, Smartphones, Photography, and More. FUJIFILM JUST ANNOUNCED ITS LATEST MEDIUM FORMAT CAMERA, THE GFX100RF. This one differs from the Company's Other GFX Models in Two Key Ways: It's the Lightest Body in the GFX Series So Far, and It's The First to Featu a Feature a Fixed-Lens Design. UP Front You're Getting at 35mm Lens (equivalent to 28mm in Full Frame), on Focal Length Fujifilm Chose For Its Versatility. With the GFX100RF's Styling and Compact Design, Fujifilm is Trying to Bring Some of That X100VI LOOK AND FEEL TO A CAMA CAPABLE OF PRODUCTING AN INCREDIBLE LEVEL OF DETAIL THANKS TO ITS 102-MEGAPIXEL SENSOR. YES, FUJIFILM ALSO HAVES THE LEICA Q3 SQUARELY IN ITS SIGHTS. The GFX100RF is Certainly a Looker, with a Top Plate That's Precision-Craft from Single Block of Aluminum. IT'LL AVILABLE IN APRIL FOR $ 4.899.95. AT 1.62 POUNDS (735 grams), It's an Easy Carry by Medium Format Standards and in the Same Ballpark As Some Full-Frame Cameras Like The Sony A1 II. The Camera's Design Isn't Far Off from the X100VI. Image: Fujifilm FUJIFILM Says the permanently-upfixed f/4 lens offers a minimum focusing distance of 20 centimeters, which should allow for some creative close-ups, and "the focal length can be digitailly extended to 45 mm, 63 mm, and 80 mm." And Believe We Call That a Crop in Layman's Terms, But When You're Working with a Massive Sensor, It's Better Than Nothing. Speaking of which, there's no in-body image stabilization on the gfx100rf, which is dyspointing from an extremly expensive camera that owners will be hand-holding much of the time. Ditching Ibis Likely Helped Fujifilm Make the gfx100rf the size that is, But Smart Design Alsoed a role. "By adopting to Leaf Shutter System Insteats of Focal Plane Shutter System, The Back Focus Has Been Shortened," The Company Wrot in Its Press Rethase. "By use the space created by this and optimizing the arrangement of the lens and camera components, a significant reduction in size has been good as." That Leaf Shutter Also Synccs Flash at Year Shutter Speed. The 35mm (28mm equivalent) lens uses ten elements in eight groups, including Two Ascherical Lenses. Image: Fujifilm This is a camera brand That's never afraid to introduce a new dial, and in the case of the gfx100rf, there's an aspect ratio dial that lets you immediately switch between nine different aspect rats included 3: 2, 16: 1, 1: 1, and 17: 6. The Camera Alo Features in Built-in Four-Stop ND Filter-Another First for the GFX Series. Where back of the camera is 3.15-Inch, 2.1-Million Dot Touchscren LCD, and IF (Like Me) You prefer to peer through the electronic Viewfinder More Ophten Than Not, Fujifilm included in High-Resolution 5.76-Million Dot Evf. As an x-h2 Owner, and Can Attest To How Sharp That Is. FUJIFILM's OWN GFX 100S II IS HEAVIER AND PERHAPS LESSS ELEGA, But That Camera Gives You Ibis and A Selection of Interchangeable Lenses. Still, Deep-Pocketed Photography Out There are tempted by Fujifilm's Answer to the Leica Q3. Stay Tuned for a Review in the Near Future.

Artificial intelligence has to date been enlisted as a bogeyman in cultural circles: Software will take the jobs of writers and translators, and AI-generated images ring the death toll for illustrators and graphic designers. Yet there’s a corner of high culture where AI is taking on a starring role as hero, not displacing the traditional protagonists—art experts and conservators—but adding a powerful, compelling weapon to their arsenal when it comes to fighting forgeries and misattributions. AI is already exceptionally good at recognizing and authenticating an artist’s work, based on the analysis of a digital image of a painting alone. AI’s objective analysis has thrown a wrench into this traditional hierarchy. If an algorithm can determine the authorship of an artwork with statistical probability, where does that leave the old-guard art historians whose reputations have been built on their subjective expertise? In truth, AI will never replace connoisseurs, just as the use of x-rays and carbon dating decades ago did not. It is simply the latest in a line of high-tech tools to assist with authentication. A good AI must be “fed” a curated dataset by human art historians to build up its knowledge of an artist’s style, and human art historians must interpret the results. Such was the case in November 2024, when a leading AI firm, Art Recognition, published its analysis of Rembrandt’s The Polish Rider—a painting that famously confounded scholars and led to many arguments as to how much, if any of it, had actually been painted by Rembrandt himself. The AI precisely matched what most connoisseurs had posited about which parts of the painting were by the master, which were by students of his, and which involved the hand of over-enthusiastic restorers. It is particularly compelling when the scientific approach confirms the expert opinion. We humans find hard scientific data more compelling than personal opinion, even when that opinion comes from someone who seems to be an expert. The so-called “CSI effect” describes how jurors perceive DNA evidence as more persuasive than even eyewitness testimony. But when expert opinion (the eyewitnesses), provenance, and scientific tests (the CSI) all agree on the same conclusion? That’s as close to a definitive answer as one can get. But what happens when the owner of a work that, at first glance, looks totally inauthentic to the point of being laughable, recruits a slick firm with the task of gathering forensic evidence to support a preferable attribution? Lost and Found Back in 2016, an oil painting surfaced at a flea market in Minnesota and was bought for less than $50. Now its owners are suggesting that it could be a lost Van Gogh, and therefore would be worth millions. (One estimate suggests $15 million.) The answer—at least to anyone with functioning eyeballs and a passing familiarity with art history—was a resounding “nah.” The painting is stiff, clumsy, utterly lacking the feverish impasto and rhythmic brushwork that define the Dutch artist’s oeuvre. Worse still, it bore a signature: Elimar. And yet, this dubious painting has become the center of a high-stakes battle for authenticity, one in which scientific analysis, market forces, and wishful thinking collide. The owners of the “Elimar Van Gogh,” as it has come to be derisively known in art circles, are now an art consultancy group called LMI International. They are investing heavily in getting experts to say what they want to hear: that it is, in fact, a genuine Van Gogh. This is where things get murky. The world of art authentication is not a straightforward affair. Unlike the hard sciences, art history deals in probabilities, connoisseurship, and competing expert opinions. It is also, crucially, an industry driven by financial incentives. If the painting is deemed real, its value skyrockets. If it’s deemed a fake, or rather in this case a derivative work by someone named Elimar who daubed a bit on canvas, distantly inspired by Van Gogh perhaps, but with none of his talents, it's virtually worthless—about as valuable as you might expect to find at a flea market in Minnesota for under 50 bucks. This imbalance in stakes has led to a dangerous trend: hiring experts not to determine authenticity, but to affirm it.

The paragon case does not stop. The report published by The Citizen Lab, the team of the University of Toronto which carried out forensic analysis on the smartphones of the activist Luca Casarini, of the journalist and director of Fanpage, Francesco Cancellato and on others spied on by the Graphite software produced by the Israeli society, throws new shadows on what could prove to be a real case of espionage for political purposes. Wiredleaks, how to send us an anonymous report do you have confidential information on the use of spyware in Italy? You can communicate safely with Wired through our platform for anonymous reports The spied on the paragon software could be many more The researchers speak openly of "a worrying model that targets groups for human rights, government critics and journalists", but the alarming detection is that the victims of these irregular operations could be many more than 90 (of which 7 Italians) ascertained to date. According to The Citizen Lab, in fact, the cases that emerged would be only those notified by WhatsApp, but it is not excluded that the spyware has also used other vectors to enter the phones of other subjects who may not be aware of the infection of their terminals. "Even if the mercenary spyware has been acquired for a primary purpose, such as the investigation on organized criminal groups - reads the report - the experience shows that, over time, the temptation to use these powerful technologies for political purposes is strong". Mediterranea: "We will send the report to the Prosecutors and the International Criminal Court" The document published by the Toronto team was immediately relaunched by the Mediterranean NGO, who in addition to Luca Casarini has at least two other activists spied on with the Graphite software, the owner Beppe Caccia and the parish priest Don Mattia Ferrari, the on -board chaplain of the Humanitarian ship Mar Ionio. The report by The Citizen Lab, in the part in which it deals with the cases of Italians canceled, Casarini and Caccia, explains that BigPretzel, a tool was used in the study "that we believe uniquely identifies the infections with the Graphite spyware of Paragon. We have analyzed the devices of the three. Based on the notification received by WhatsApp, we believe that everyone has been targeted by the SPYWARE". There is also a priest among the spy of the paragon case still widens the spy story that involves refugees, activists, the director of Fanpage and now also Don Mattia Ferrari, the on -board chaplain of Mediterranea Saving Humans Mediterranean will also be the relationship to those who are investigating the case. "There are the tests - reads a note from the organization - of the presence of Paragon Graphite, used by the secret services on the phones of our activists. We will deliver everything to the 5 Prosecutors (Palermo, Naples, Venice and Rome) who are investigating, and to the President of the European Parliament Roberta Metsola. We will also send the report to the international criminal court, because behind this case there is the Libyan situation and the relationships between secret services. Oil and weapons, and militias that manage power in that country, could have played a role. The paragon case is a state secret. Fratoianni to Wired: "I also spied on and other parliamentarians?" The spy story that agitates politics widens. The Mediterranean NGO publishes an analysis of the Citizenlab of Toronto in which Luca Casarini was said to have been spied on from February 2024. The leader of the Italian left: "I met him several times and I was there for dinner, was there anyone who recorded everything?" The company in the eye of the cyclone diminishes the story with a skimpy press release signed by John Fleming, executive president of the company, sent to the Team of The Citizen Lab. According to the Manager of Paragon Solution in the report published by the University of Toronto there would be "numerous inaccuracies", but, adds the manager, "without more details we cannot be more specific or to provide comments". And also Amnesty International announces that spy people could be many more: "An independent investigation conducted by the AMNESTY INTERNATIONAL Security Lab - reads a press release - in the last six months has identified further cases of apparent use of spyware against other activists engaged in the rescue in the sea in Italy. The research underway of civil society will almost certainly bring to light new cases. of the iceberg ". Copasir moves: listened to the top management of the services and two representatives of Meta In the meantime, Copasir, the Committee for information on the security of the Republic, has listened to two Meta representatives. It was the Alerts sent by Mark Zuckerberg's company via WhatsApp who made people involved discover that they have a "zero click" spy software in their phones. It should be remembered that the Graphite software, which the Company Paragon Solution provides government apparatuses of Western countries, can hiddenly access the sensitive data on the smartphones of spied people.

Nvidia’s investment over the next four years is expected to reach nearly $500bn. (Photo: Shutterstock) Nvidia is reportedly planning a significant investment in semiconductor production and electronic components within the US. Over the next four years, total spending is expected to reach approximately $500bn, according to reporting by the Financial Times. Huang, in an interview with the publication, indicated that a substantial portion of this investment could be directed towards domestic chip production. He suggested that the Trump administration may play a role in accelerating the growth of the AI industry in the US by supporting manufacturing initiatives. Nvidia has been addressing investor concerns about continued demand for its high-cost AI chips, particularly amid increased competition. This follows the launch of a chatbot by China’s DeepSeek, which reportedly operates with fewer AI chips, raising questions about potential shifts in the AI hardware market. Huang has confirmed that Nvidia’s latest AI systems can now be manufactured in the US through partnerships with major suppliers, including Taiwan Semiconductor Manufacturing Company (TSMC) and Foxconn. He noted that TSMC’s investment in US-based manufacturing represents a significant step towards strengthening supply chain resilience, reducing dependency on overseas production, and ensuring steady access to critical semiconductor components. The company’s expansion strategy comes at a time of heightened global competition in the AI sector, particularly from Chinese technology firms. Nvidia has acknowledged Huawei’s growing influence in the market, underscoring the need for continuous investment in advanced chip production and infrastructure to maintain its leadership position. At the company’s developer conference in California on Wednesday, Huang told analysts that demand for Nvidia’s Blackwell AI chips could be significantly higher than current projections. He noted that orders for 3.6 million units from four major cloud firms underestimated overall market demand, as they did not account for purchases from Meta Platforms, smaller cloud providers, or AI-focused startups. Expanding AI infrastructure through partnerships Beyond manufacturing, Nvidia is expanding its role in AI infrastructure development through its participation in the AI Infrastructure Partnership (AIP). The consortium, which includes BlackRock, Global Infrastructure Partners, Microsoft, and MGX, aims to drive investment in AI-ready data centres and energy infrastructure. Initially, AIP plans to unlock $30bn in capital, with total investment potential reaching $100bn when including debt financing. Nvidia will continue to serve as a technical advisor to AIP, leveraging its expertise in AI computing and data centre infrastructure. The initiative is expected to foster collaboration across a diverse network of investors, corporations, and technology firms, focusing on scalable AI solutions. AIP’s investments will be primarily directed toward the US and partner countries within the Organisation for Economic Co-operation and Development (OECD). Last month, Nvidia forecasted revenue of approximately $43bn for the first quarter of fiscal 2026, surpassing market expectations. The company expects gross margins to be around 70.6%, with operating expenses estimated at $5.2bn on a GAAP basis. It also anticipates additional income of approximately $400m, excluding gains and losses from equity securities.

Average attacks last 253 days. In Germany, 69 percent of companies are affected by at least one cyber security incident in the year. In 2024, 35 percent of the cyber attacks lasted more than a month. On average, the duration of the attack was 253 days. This emerges from the global Incident Response 2024 Report by the security provider Kaspersky. In Germany, 69 percent of companies were affected by at least one cyber security incident last year - 31 percent even several times. The attacks in data encryption and loss often resulted. This meant that complex measures were necessary to restore. The middle duration of the Incident Response measures in 2024 was 50 hours. The most common attack vectors included exploits against publicly accessible applications (39 percent), access to valid access data (31 percent) and exploiting trustworthy relationships (13 percent). "It is crucial for companies to understand the developing cyber threat landscape in order to be able to protect themselves comprehensively," explains Konstantin Sapronov, Head of Global Emergency Response Team at Kaspersky. "Our analysis shows that cybercriminals are becoming more and more adaptable and further develop their methods with technological progress. Companies are not only allowed to react to attacks, but must continuously develop their security strategies."

About eight million French people concerned by an exposure of their personal information after a data leak in 2024, compared to 5 million in 2023. All tech news for pros every day in our newsletter learn more about the use of personal data Here is the substantial estimate of the Incyber forum, the new name of the FIC professional show - its next edition will take place in early April in Lille -, and the Hexatrust association, a group of French cybersecurity companies, in their last barometer of data leaks. A figure from data published in Open Data by the CNIL. In all, the French guardian of personal data was notified on average of 16 incidents per day, against 13 in 2023, continue the Incyber and Hexatrust forum. "This significant increase is mainly explained by the resurgence of cyber attacks (+ 25 %) and the enlargement of the exposure surface to risks," they say. Year 2024 Black So many leaks first accidental - the share of those of malicious origin is around 17% - who exhibit "individuals at increased risks of fraud and identity theft", they add. And to emphasize that CNIL data mentions "the presence of sensitive personal data (financial data, health data, etc.) in 20% of leaks, which adds a particularly disturbing dimension to this phenomenon". After a year 2024 marked by numerous data leaks, such as that of the operator Free for example, the number of French people concerned by an exhibition of their worried personal information. But the exact volume of people concerned is visibly debated. At the beginning of March, before the Senate, Hervé Pétry, the boss of the gendarmerie national unit cyber, had indeed reported another statistics, even more dramatic, without however specifying his sources. "It is estimated today that 80% of the population had their personal data stolen in 2024," he said. 54 million people, France with 68 million inhabitants in 2024. CNIL checks Anyway, the CNIL had already set the tone at the start of the year. After the bad vintage 2024, she had announced upcoming checks around cybersecurity. Whether at the end of data violation, "to verify the implementation of suitable corrective measures". Or by encouraging the sector to become more virtuous. For example by promoting security culture in the field, notably by disseminating "adapted educational resources", pointed out the guardian of personal data.

Google Pixel 9a, Ph. Kerry Wan/Zdnet All tech news for pros every day in our newsletter learn more about the use of personal data We have reached a point in the mobile industry where each new update in a range amounts to playing "find the differences". Whether it is the iPhone 15 and the iPhone 16, the Samsung Galaxy S24 and the Samsung Galaxy S25 or the Google Pixel 8 and the Google Pixel 9, customers must dig a little more to find out what is really new. This is particularly true with the new Google Pixel 9a. It is sold at the same launch price of € 550 as the Pixel 8A from last year. The screen size is now different, but its quality is the same. The battery capacity is different, but the load speeds are the same. A bigger screen than the 8a pixel It is not necessarily a bad thing. The Pixel 9a is as pleasant to hold as its predecessors, with a 6.3 -inch screen easy to manage and sweep. This 120 Hz OLED screen is slightly larger than the 6.1 inch frame of Pixel 8A, but it aligns with the rest of the Pixel 9 family, including the Pixel 9 Pro and the Pro Fold. It's all about consistency here. Kerry Wan/Zdnet The screen is also much brighter this time, with a peak of 2,700 nits (against 2,000), which makes the Pixel 9a a little more readable under direct light. Do not look at the screen too close, because the borders are thicker than ever, with a black around one that clearly indicates in which price category is the smartphone. High -end processor ... and again AI By keeping the smartphone online with the rest of the Pixel 9 series, Google upgraded the Pixel 9A processor to the G4 Tensor. Thanks to the new chip and just enough RAM (8 GB), the mid -range pixel is now supporting imaging tools such as Add Me (add of his person in a photo), astronomical photography mode), Pixel Studio (image generation by AI) and more. From left to right: peony, iris, porcelain, obsidian. Kerry Wan/Zdnet Google specifies that users who want more RAMs must opt ​​for its Pixel 9 Pro models, higher -end. Same main sensor as the Pixel 9 Pro Fold Photo touch -up tools should be able to play with the 48 megapixel main camera, even if it displays a lower definition as its Pixel 8A predecessor (64MP sensor). It is actually a main main camera similar to that of the much more expensive Pixel 9 PRO Fold, and it has the same ultra-large target of 13 Mpx as the Pixel 8A. It will be necessary to see what it is during tests in real conditions, but the sensor is promising. To complete this year's changes, the device has an IP68 index to withstand deeper water and dust infiltrations, a larger 5,000 mAh battery (compared to 4,492 mAh) and a more flush camera module. The rest of the Pixel 9a is almost the same as the Pixel 8A, including load speeds with and without wireless, the promise of seven years of software and security updates, storage options and price. Source: "Zdnet.com"

Commercial quadcopters have been on the mainstream gadget scene for 15 years, proliferating across industries and among hobbyists. There's a swanky DJI store on New York City's Fifth Avenue, and you probably have a neighbor, not to mention a roofer, who owns a drone. So when researchers at the embedded-device security firm Red Balloon started seeing surprising quadcopter accessories on Chinese shopping platforms like Temu and AliExpress, they didn't think much of it at first. As with any popular gadget type, there's a whole ecosystem of niche, wacky, and comical add-ons available for drones. But the more Red Balloon CEO Ang Cui thought about it, the more unsettled he and his colleagues became about how cheap and easy it would be for anyone to buy seemingly disparate add-ons that could easily turn a mainstream quadcopter into a war machine. The accessories the researchers found include AI drone guidance modules—essentially small mounted cameras that use object recognition to identify humans and road vehicles at long range—and miles-long fiber optic tethers. Like plugging an ethernet cable directly into your laptop, miles-long tethers allow drones to fly around a large area without being vulnerable to disruption by signal jammers. The researchers recognized them from battlefield footage and other reports that such tethers—not to mention AI guidance modules—are being used by both sides in the Russia-Ukraine war to drop explosives or autopilot crash entire drones themselves into tracked objects without requiring operator control. The fact that battlefield technology is widely available in the United States and around the world at low cost augurs a climate in which any actor—from criminal syndicates to paramilitary groups, from disgruntled employees to ostracized teens—can quickly and cheaply gather the needed equipment to remotely go on a destructive and violent rampage. “The more we looked at this and started to see the big picture, the more my heart sank, because these are toys that are really amazing at killing people,” Cui tells WIRED. “This tech didn't exist in the commercial space two years ago, it wasn't in demand from hobbyists, but now manufacturers are making it for war and it's spilling over.” The Red Balloon researchers found that long-range quadcopters are readily available on marketplaces like AliExpress for under $300, with many priced at around $200 each. And the unit price of a long-range drone can drop even lower when purchased in bulk. The researchers noticed sellers who were claiming that they ship 60,000 of the long-range drones per month. They purchased a 1-mile-long tether from a seller on AliExpress for about $260 and another that is 7.5 miles long for about $700. Meanwhile, the researchers also purchased an AI guidance module from a different AliExpress seller for $325. Cui and his colleagues also purchased cargo holders that strap to drones and can be used to transport beer cans or water bottles, but could also be loaded with mortar bombs. These were $106 each. In the two months since the researchers made the purchases, they say the prices of most of the products they bought have even dropped slightly. “This equipment is mainstream, but it does not just cost a few dollars to make, so I think this stuff is being sold either at or below cost,” Cui says.

After years of simultaneously competing with Qualcomm and slugging it out in court, Apple has launched its long-awaited C1 modem in its new iPhone 16e handset. On Wednesday, a benchmark test released by Ookla revealed that the C1 beat Qualcomm’s 5G modem in some areas, complete with surprising results. The Ookla tests tracked performance on the top three US carriers across three different tiers: 10th percentile (lowest performance experience) Median performance 90th percentile (highest performance experience) The full data offers a nuanced picture of how each device fares against the other, but there were a number of key takeaways. First, the iPhone 16e with C1 was the clear winner in the 10th percentile. This means that when analyzing those situations when connectivity is a struggle, and download speeds suffer, the C1 offered a better experience than Qualcomm’s modem. Qualcomm’s modem fared better in the 90th percentile, wherein the iPhone 16 with Qualcomm had better top download speeds across AT&T, T-Mobile, and Verizon. However, Ookla says this is unlikely to matter as much to users: “Performance at the lowest 10th percentile often provides a more accurate reflection of overall quality of experience (QoE) than the fastest 90th percentile, which can be skewed by deployments in mmWave-covered locations and is subject to declining marginal returns.” Where median performance was concerned, the iPhone 16e with Apple’s C1 modem offered faster median speeds across both AT&T and Verizon. Still, T-Mobile users got faster median speeds with Qualcomm’s hardware as opposed to than C1. Finally, upload speeds were consistently faster for iPhone 16e—across all major carriers. With AT&T and Verizon the C1 offered in the ballpark of double the upload speeds of Qualcomm. For T-Mobile customers, the C1 was able to edge out Qualcomm’s modem, but just barely. As such, Apple has assembled a good modem, especially for its first-generation hardware, and is said to be readily developing future versions. Stay tuned for additional details as they become available. Via 9to5Mac and Ookla

When we talk about AI chatbots on e-commerce sites, you often imagine intrusive contextual windows that disrupt the shopping experience. However, a new trend emerges. Customers are now turning to generative AI chatbots to help them make their choices. According to a study carried out by Adobe Analytics, traffic towards merchant sites from these artificial intelligence tools continues to increase. All tech news for pros every day in our newsletter learn more about the use of personal data Adobe's analysis is based on more than a thousand billion visits to sales sites in the United States, and on the responses of more than 5,000 American customers. The conclusions are clear: AI becomes a real purchasing assistant for many users. "Our data show that consumers have great confidence in the recommendations of generative AI platforms," ​​said Vivek Pandya, head of Adobe Digital Insights and the study, in an interview with ZDNET. "They are looking for speed and efficiency above all. »» Use of generative AI for online purchases Among those questioned, 39 % have already used the generative AI for their online purchases. The most common uses include: Product search Personalized recommendations The hunt for good deals Inspiration for gift ideas This trend is only accelerating: 53 % of respondents plan to use AI for their online purchases this year. These uses are already reflected in concrete benefits for traders. During the holiday season, Adobe observed an unprecedented increase in traffic generated by generative AI to American merchant sites. Between November 1 and December 31, 2024, this traffic jumped 1,300 % compared to the previous year, with a spectacular increase of 1,950 % just for Cyber ​​Monday. This trend did not stop at the end of year celebrations. In February 2025, traffic from the generative AI had increased by 1,200 % compared to July 2024. Although its volume is still modest compared to more established traffic motors, its growth is dazzling. Growth that receives several sectors The rise of traffic from generative AI is not limited to online commerce. The travel sector has recorded the strongest increase, with an increase of 1,700 % between July 2024 and February 2025. This increase in particular concerns visits to sites dedicated to travel, leisure and hotel. Other industries follow this trend: retail and bank sectors have also experienced a significant increase, with growth of 1,200 % over the same period. © Adobe In the banking field, the generative AI plays a key role by supporting consumers in their financial decisions. According to the report, the main uses include: The search for recommendations for checks and savings accounts The request for explanations on financial terms or strategies The development of personalized budgets Buyers' behavior in the AI ​​era Consumers who use AI chatbots before buying show signs of reinforced engagement once on merchant sites. According to the study, these visitors: Display an 8 % increase in their commitment Consult 12 % more pages per visit Record a 23 % lower rebound rate However, if these users are more active, their conversion rate remains slightly behind. They have a 9 % chance of finalizing a purchase compared to other traffic sources. This figure is nevertheless clearly improving compared to July 2024, where the gap reached 43 %. This rapid progression testifies to an increasing interest of consumers for AI as a tool for buying purchasing. "This is quality trafficking that resellers and brands are looking to hire, because it generates effective conversions," explains Vivek Pandya. According to him, the challenge for merchants is not limited to capturing these new visitors, but to integrate them durably into their ecosystem. "When these customers arrive via an AI assistant, an opportunity arises: how to maintain this relationship and promote long -term recurring purchases? He adds.

ジャーナリズムの世界には昔から、「愛していると母親に言われても、裏を取れ」という格言がある。つまり、誰よりも信頼できる相手にさえ懐疑的にならなければいけないということだ。では、話している相手が母親ではなく、OpenAIの「ChatGPT」のような生成人工知能（AI）モデルだったらどうだろう。コンピューターは、信頼できるのだろうか。 先頃開催された、映画、音楽、インタラクティブメディアのカンファレンスである「South by Southwest」で、カーネギーメロン大学（CMU）の2人のコンピューター科学者が語った話については、どうだろう。これも同じことだ。自分で確かめよう。 テキサス州オースティンで開かれたこのカンファレンスでも、話題が集中したのはAIだった。信頼性の問題、人間の仕事の変化といったテーマについて、AIの将来の展望と全体像が専門家の間で論じられた。CMUの助教を務めるSherry Wu氏とMaarten Sap氏は、とりわけAIの現在の状況に焦点を当て、大規模言語モデル（LLM）でトレーニングされたAIチャットボットなど、現在出回っている一般的な生成AIツールを、誤ることなく適切に使うヒントを示した。 「生成AIは、実のところ完璧にはほど遠く、人々が使いたいと考えている用途のすべてに適しているとは言いがたい」。Sap氏はこう話している。 AIより賢くあるにはどうすればいいのか、そのための5つのヒントをご紹介しよう。 指示は明確に、「察して」はNG 「Twitter」（現「X」）や「Bluesky」などのSNSでジョークを投稿して不発に終わった経験がある人なら、文字で皮肉を伝えることの難しさを十分に承知しているだろう。そうしたSNSのユーザーは（少なくとも人間であれば）、文字どおりの発言でないことを対話上の手がかりから理解する。LLMには、それができない。 現在のLLMは、文字どおりではない発言でも文字どおりに解釈してしまうことが多く、社会的な状況や関係性からの推論が不得手だとSap氏は説明する。 Wu氏によると、プロンプトはできるだけ具体的に、かつ体系化するとよいという。重要なのは、自分がどんなことを聞き出そうとしているのかをモデルに理解させることだ。求めている内容を厳密に絞り込むべきであり、LLMが質問の真意を察してくれるなどと期待してはならない。 チャットボットは堂々とウソをつく 生成AIツールで最大の問題は、おそらくハルシネーションを起こす、つまり虚偽の情報を作り上げることだろう。Sap氏によれば、ハルシネーションは多ければ4回に1回は起こるもので、法律や医学など専門性が上がるとその確率は上がるという。 問題は、間違った内容を答えることだけではない。チャットボットは、完全に間違っていても自信たっぷりにそうに回答する、とSap氏は指摘する。 「間違った答えに、さも確実そうな表現を使うので、人はついそれを信じてしまう」（Sap氏） これについての対策は単純で、LLMの回答を確認すればよい。同じ質問を何度か繰り返したり、同じ質問でも聞き方を変えたりすれば、AIの一貫性を確かめることができる、とWu氏は説明する。違う答えが返ってくる場合もある。「ときには、自分が答えている内容を実際には分かっていないことさえある」（Wu氏） 最も重要なのは、別の情報源を使って検証することだ。自分が答えを知らない疑問を尋ねるときは、要注意ということでもある。生成AIの回答が最も有用なのは、質問者がよく知っているテーマを尋ねたときだとWu氏は言う。何が正しく、何が正しくないかを質問者自身が判断できるからだ。 「AIモデルを信頼するときと信頼しないときを意識的に峻別し、自信たっぷりな回答は疑うようにすべきだ」（Wu氏） AIは秘密を守れない LLMについては、プライバシーの問題も多い。LLMにサプライズパーティーの企画を手伝わせるというテストでは、秘密にしておきたい当のサプライズ相手への連絡にパーティーのことを含めてしまう結果になった、とSap氏は話している。 「LLMは、誰がいつ何を知るべきか、どんな情報を隠しておくべきかといった類推が不得意だ」、とSap氏。 機密データや個人情報はLLMに共有しないように、とWu氏も警告する。 「自分が作ったものをAIモデルと共有する場合は、LLMに公開したくない情報が含まれていないかどうか常に再確認すべきだ」 相手は機械だと肝に銘じる AIチャットボットが人気を博している理由の1つは、人間の話し方をまねるのがうまいことだ。しかし、それはあくまでも物まねであってリアルな人間ではない、とSap氏は釘を刺す。AIモデルが「こうではないかと思います」とか「こう想像します」といった言葉を使うのは、そういう語句を含む言葉づかいを学習したからであって、実際に思ったり想像したりするわけではない。「われわれが言語を使うとき、こうした語句には必ず暗黙的に認知が伴っている。そのために、LLMが何か想像するかのように、内面世界を持っているかのように感じてしまうのだ」、とSap氏は解説している。 AIモデルを人間のようにとらえると、誤った信頼感を生む危険性がある。LLMは人間と同じように動くわけではなく、人間のように扱っていると、社会的な固定観念を強めてしまうおそれがある、とSap氏は言う。 「人は、AIシステムが人間に似ている、意識を持っていると実際以上に考えてしまう傾向がある」 LLMの利用が妥当ではない場面もある LLMは高度な調査や類推の機能を備えているとうたわれてはいるものの、実際にはまだそこまでではないとSap氏は話している。AIモデルは博士号を取得した人間並みの能力を発揮できるというベンチマークはあるが、それはあくまでもベンチマークにすぎない。その分析の裏でどんなテストが実施されていようと、個々人が使いたい用途と同じ水準でその性能が発揮されるとは限らないのである。 「AIは何でもできるという幻想が先行しすぎたために、自分のタスクに使うべきだと人々が早計に判断するようになっている」、とSap氏は語る。 生成AIモデルを、あるタスクに使うかどうかを判断するときには、使った場合のメリットと潜在的なリスクは何か、使わない場合のメリットと潜在的なリスクは何かをそれぞれ考慮すべきだ、とWu氏も指摘している。 単行本「生成AI 社会を激変させるAIの創造力」（価格をAmazonでチェック）

Googleが新型スマートフォン「Pixel 9a」を発表した。米国や英国、豪州で4月に発売予定。日本でも時期は不明だが発売予定だ。 米国での価格は499ドル（税別約7万4000円）で、iPhone 16eの599ドル（税別約8万9000円）より約1万5000円も安い。それでいて120HzディスプレイやPixel 9 Proと同じプロセッサ、そしてミドルレンジ最強のカメラを備える。 Amazonで現在開催中のセールを見る ●Pixel 9aとは 新たに発表されたPixel 9aは、大きな期待を背負っている。なぜなら、昨年のPixel 8aが機能やバッテリー駆動時間、長期的なソフトウェアサポートなど、総合的なバランスに優れ、筆者の2024年のお気に入りの1台だったからだ。幸いにも、スペック上では本機は8aをほぼすべての面で上回っているように見受けられる。 ただ、Pixel 9aのリリース時期は、Appleが「手頃な価格」とうたう599ドルの「iPhone 16e」を発表した直後にあたり、比較されやすい状況でもある。一方で、iPhone 16eはAppleの中では安価とはいえ、500ドル以下という「お手頃スマホ」の価格帯からは外れているのが現実だ。 その点、GoogleはAppleのように価格を上げず、Pixel 9aを499ドル（約7万4000円）に据え置いている。予算を抑えたい人には有力な選択肢になりそうだ。 筆者は実機に約30分ほど触れたが、その堅牢な作りや新デザインのいくつかの要素に好印象を受けた。特に背面がほぼフラットになった点は気に入っている。一方で、ほかのPixel 9シリーズにある特徴的なカメラバーが省かれたため、見た目の面では「Pixelらしさ」がやや薄れた印象だ。とはいえ、ディスプレイや内部仕様については、まさに「Pixelらしさ」が詰まっている。 Pixel 9aはGoogleストアや米国、英国、豪州では各小売店で4月に発売される予定だ。 Pixel 9aは、従来のPixel aシリーズとは外観が大きく異なる。Pixel 9や9 Proと同様、エッジがフラット化され、8a以前のように丸みを帯びた形状ではなくなった。Googleは極限までミニマリズムを追求しているようで、背面はほぼ完全にフラットだ。カメラの突起もわずかに出っ張っている程度で、指でなぞると少し引っかかるぐらいである。 2025年現在、Pixel 9aのように机の上に置いてもほとんどガタつかず、平らに安定して置けるスマホは珍しい。カメラバーを備えるPixel 9や9 Proも実はそれほどグラつかないが、9aはさらにフラットさが際立っている印象だ。個人的には、この平らなデザインがとても気に入っている。 ●Pixel 9aのデザイン カメラバーがなくなったことを少し残念に思う人もいるかもしれない。以前のGoogleスマホでは、あのカメラバーがデザイン上のアクセントになっていたためだ。実際にはケースを付ける人が多いとはいえ、個性がやや薄れたと感じる可能性はある。 フォームと機能のせめぎ合いの中で、Googleは機能を優先したようだ。その結果、Pixel 9aはaシリーズの中で最も耐久性が高い機種だと同社は主張している。背面には81％再生プラスチックが使われており、人によっては「プラスチック製か」と気にする向きもあるだろうが、見た目は悪くなく、持った感触もしっかりしていた。コストダウンにも貢献しているはずだ。さらにIP68の防水防塵性能を備え（先代Pixel 8aはIP67）、水深1.5mで30分の耐久性があるとされる。 Pixel 9aは4色展開で、オブシディアン、ポーセリン、ピオニー、そしてアイリスが用意される。実物を見た限りでは、個人的にアイリスが最も好みだった。 本体サイズはPixel 8aとほぼ同じ（わずかに縦長）だが、ディスプレイは6.3インチ（Pixel 8aは6.1インチ）に拡大された。最大輝度は2700ニトへと大幅に向上（8aは2000ニット）しており、カリフォルニアの強い日差しの下でもどの程度見やすいのか試してみたいところだ。スペック上は、ディスプレイの品質が無印のPixel 9と同等という点も注目に値する。 ●Pixel 9aのカメラ カメラはメインが4800万画素、超広角が1300万画素という2眼構成だ。メインカメラではGoogleの「Macro Focus」が使え、接写撮影が可能になっている。多くのスマホは超広角カメラをマクロ代わりに利用するが、本機はメインカメラでマクロを実現している点が興味深い。 画素数だけを見ると、8aの6400万画素から減少しているようにも見えるが、Googleは長年ソフトウェアの力で高画素相当の写真を実現してきた。単純に数字が下がったからといって、一概に性能ダウンとは言えないだろう。 さらにPixel 9や9 Proと同様、AIを活用したさまざまな機能が搭載されている。たとえば、より良いポートレートを作成するBest Take、Photo Unblur、不要な要素を消せるMagic Editor、夜景撮影に強いNight Sightなど。また、Add MeというAR/AI機能を使えば、集団写真にあとから撮影者本人を合成できる。筆者はPixel 9 ProでAdd Meを何度か試したが、その仕上がりには毎回驚かされる。 これまでのPixel Aシリーズ同様に、9aも「価格以上」のカメラ性能を期待させる。本機が新たに採用した4800万画素センサーでどこまで撮れるのか、早く試してみたい。 ●Pixel 9aの性能とバッテリー Pixel 9aには、Pixel 9や9 Proと同じTensor G4チップが搭載されており、GoogleのAI機能やGeminiに対応している。RAMはPixel 9の12GB、9 Proの16GBより少ない8GBだが、発売時点でAndroid 15を搭載し、以後7年間の主要OSアップデートとセキュリティ更新が保証されるのは大きな強みだ。 Tensor G4はベンチマーク上で特別高い数値を叩き出すわけではないが、Pixel 9や9 Proの使用感から推測するに、日常使い（ゲーム、SNS、写真編集など）では十分に快適なパフォーマンスを期待できるだろう。 バッテリー容量は5100mAhと、Pixel aシリーズ史上最大だという。初めてスペックを見たときは、まるでSamsungのGalaxy Aシリーズかと思うほどで驚いた。実際、1300ドルのGalaxy S25 Ultraよりも大容量を備えている点は興味深い。 ●Pixel 9aの現時点でのまとめ その他にも、盗難防止機能や家族向けに利用制限・共有が管理できるFamily Link、GmailやGoogleマップなどの定番サービスまで、いわゆる「Pixelらしさ」をしっかり備えている。ただし、Pixel 9シリーズにある衛星を利用したSOS機能は9aには非搭載だ。コストを抑えるためには何らかの機能を省略せざるを得なかったのだろう。 ストレージは128GBと256GBの2種類から選択可能。購入時には3カ月間のGoogle One、3カ月間のYouTube Premium、そして6カ月間のFitbit Premiumが付いてくる。今後、実際にPixel 9aを入手し次第、カメラ性能やバッテリーの持ちなどを詳しく検証する予定だ。実際に手元に届くのが待ち遠しい。 Amazonで現在開催中のセールを見る Amazonで現在開催中のセールを見る

Googleが新型スマートフォン「Pixel 9a」を発表した。米国や英国、豪州で4月に発売予定。日本でも時期は不明だが発売予定だ。 米国での価格は499ドル（税別約7万4000円）で、iPhone 16eの599ドル（税別約8万9000円）より約1万5000円も安い。それでいて120HzディスプレイやPixel 9 Proと同じプロセッサ、そしてミドルレンジ最強のカメラを備える。 Amazonで現在開催中のセールを見る ●Pixel 9aとは 新たに発表されたPixel 9aは、大きな期待を背負っている。なぜなら、昨年のPixel 8aが機能やバッテリー駆動時間、長期的なソフトウェアサポートなど、総合的なバランスに優れ、筆者の2024年のお気に入りの1台だったからだ。幸いにも、スペック上では本機は8aをほぼすべての面で上回っているように見受けられる。 ただ、Pixel 9aのリリース時期は、Appleが「手頃な価格」とうたう599ドルの「iPhone 16e」を発表した直後にあたり、比較されやすい状況でもある。一方で、iPhone 16eはAppleの中では安価とはいえ、500ドル以下という「お手頃スマホ」の価格帯からは外れているのが現実だ。 その点、GoogleはAppleのように価格を上げず、Pixel 9aを499ドル（約7万4000円）に据え置いている。予算を抑えたい人には有力な選択肢になりそうだ。 筆者は実機に約30分ほど触れたが、その堅牢な作りや新デザインのいくつかの要素に好印象を受けた。特に背面がほぼフラットになった点は気に入っている。一方で、ほかのPixel 9シリーズにある特徴的なカメラバーが省かれたため、見た目の面では「Pixelらしさ」がやや薄れた印象だ。とはいえ、ディスプレイや内部仕様については、まさに「Pixelらしさ」が詰まっている。 Pixel 9aはGoogleストアや米国、英国、豪州では各小売店で4月に発売される予定だ。 Pixel 9aは、従来のPixel aシリーズとは外観が大きく異なる。Pixel 9や9 Proと同様、エッジがフラット化され、8a以前のように丸みを帯びた形状ではなくなった。Googleは極限までミニマリズムを追求しているようで、背面はほぼ完全にフラットだ。カメラの突起もわずかに出っ張っている程度で、指でなぞると少し引っかかるぐらいである。 2025年現在、Pixel 9aのように机の上に置いてもほとんどガタつかず、平らに安定して置けるスマホは珍しい。カメラバーを備えるPixel 9や9 Proも実はそれほどグラつかないが、9aはさらにフラットさが際立っている印象だ。個人的には、この平らなデザインがとても気に入っている。 ●Pixel 9aのデザイン カメラバーがなくなったことを少し残念に思う人もいるかもしれない。以前のGoogleスマホでは、あのカメラバーがデザイン上のアクセントになっていたためだ。実際にはケースを付ける人が多いとはいえ、個性がやや薄れたと感じる可能性はある。 フォームと機能のせめぎ合いの中で、Googleは機能を優先したようだ。その結果、Pixel 9aはaシリーズの中で最も耐久性が高い機種だと同社は主張している。背面には81％再生プラスチックが使われており、人によっては「プラスチック製か」と気にする向きもあるだろうが、見た目は悪くなく、持った感触もしっかりしていた。コストダウンにも貢献しているはずだ。さらにIP68の防水防塵性能を備え（先代Pixel 8aはIP67）、水深1.5mで30分の耐久性があるとされる。 Pixel 9aは4色展開で、オブシディアン、ポーセリン、ピオニー、そしてアイリスが用意される。実物を見た限りでは、個人的にアイリスが最も好みだった。 本体サイズはPixel 8aとほぼ同じ（わずかに縦長）だが、ディスプレイは6.3インチ（Pixel 8aは6.1インチ）に拡大された。最大輝度は2700ニトへと大幅に向上（8aは2000ニット）しており、カリフォルニアの強い日差しの下でもどの程度見やすいのか試してみたいところだ。スペック上は、ディスプレイの品質が無印のPixel 9と同等という点も注目に値する。 ●Pixel 9aのカメラ カメラはメインが4800万画素、超広角が1300万画素という2眼構成だ。メインカメラではGoogleの「Macro Focus」が使え、接写撮影が可能になっている。多くのスマホは超広角カメラをマクロ代わりに利用するが、本機はメインカメラでマクロを実現している点が興味深い。 画素数だけを見ると、8aの6400万画素から減少しているようにも見えるが、Googleは長年ソフトウェアの力で高画素相当の写真を実現してきた。単純に数字が下がったからといって、一概に性能ダウンとは言えないだろう。 さらにPixel 9や9 Proと同様、AIを活用したさまざまな機能が搭載されている。たとえば、より良いポートレートを作成するBest Take、Photo Unblur、不要な要素を消せるMagic Editor、夜景撮影に強いNight Sightなど。また、Add MeというAR/AI機能を使えば、集団写真にあとから撮影者本人を合成できる。筆者はPixel 9 ProでAdd Meを何度か試したが、その仕上がりには毎回驚かされる。 これまでのPixel Aシリーズ同様に、9aも「価格以上」のカメラ性能を期待させる。本機が新たに採用した4800万画素センサーでどこまで撮れるのか、早く試してみたい。 ●Pixel 9aの性能とバッテリー Pixel 9aには、Pixel 9や9 Proと同じTensor G4チップが搭載されており、GoogleのAI機能やGeminiに対応している。RAMはPixel 9の12GB、9 Proの16GBより少ない8GBだが、発売時点でAndroid 15を搭載し、以後7年間の主要OSアップデートとセキュリティ更新が保証されるのは大きな強みだ。 Tensor G4はベンチマーク上で特別高い数値を叩き出すわけではないが、Pixel 9や9 Proの使用感から推測するに、日常使い（ゲーム、SNS、写真編集など）では十分に快適なパフォーマンスを期待できるだろう。 バッテリー容量は5100mAhと、Pixel aシリーズ史上最大だという。初めてスペックを見たときは、まるでSamsungのGalaxy Aシリーズかと思うほどで驚いた。実際、1300ドルのGalaxy S25 Ultraよりも大容量を備えている点は興味深い。 ●Pixel 9aの現時点でのまとめ その他にも、盗難防止機能や家族向けに利用制限・共有が管理できるFamily Link、GmailやGoogleマップなどの定番サービスまで、いわゆる「Pixelらしさ」をしっかり備えている。ただし、Pixel 9シリーズにある衛星を利用したSOS機能は9aには非搭載だ。コストを抑えるためには何らかの機能を省略せざるを得なかったのだろう。 ストレージは128GBと256GBの2種類から選択可能。購入時には3カ月間のGoogle One、3カ月間のYouTube Premium、そして6カ月間のFitbit Premiumが付いてくる。今後、実際にPixel 9aを入手し次第、カメラ性能やバッテリーの持ちなどを詳しく検証する予定だ。実際に手元に届くのが待ち遠しい。 Amazonで現在開催中のセールを見る Amazonで現在開催中のセールを見る

"The era of robotics which performs current tasks has arrived," said Jensen Huang, CEO of Nvidia. All tech news for pros every day in our newsletter learn more about the use of personal data The manager spoke during the GTC 2025 conference, which is held this week in San Jose, California. The company has unveiled a training model for humanoid robotics. "With GR00T N1, robotic developers will be able to open new paths in the AI ​​era," said Jensen Huang. A two -headed architecture GR00T N1 adopts an architecture with two systems inspired by the principles of human cognition. System 1 is a rapid reflection model that imitates the reflexes and the intuition of man System 2 is a prudent and systematic decision -making model The robot can perform daily tasks such as enter and move objects, as well as tasks requiring complex steps that require a combination of general skills. Developers can post-entracting the GR00T N1 using real or synthetic data to adapt it to humanoid robots and specific tasks. The main developers of humanoid robots, including Boston Dynamics, Agility Robotics and Menti Robotics, had early access to GR00T N1. Partnership with Google Deepmind and Disney Research Nvidia also announced the development of Newton, an open-source physical engine for robotic development, in partnership with Google Deepmind and Disney Research. Newton is based on the Warp framework. It is compatible with simulation frameworks such as Mujoco by Google Deepmind and Isaac Labs of Nvidia. With Deepmind, the company co-develops Mujoco-Warp, which should speed up the automatic robotics learning tasks over 70 times. Disney Research will use Newton to feed the next generation of entertainment robots, including the Star Wars BDX droid presented at the GTC.

Tile window managers are one of the most efficient ways to operate your OS. Whether this is suitable or not depends on the user's preferred method of operation. If you prefer keyboard-based operations whenever possible and are unwilling to master a number of new keyboard shortcuts, a tiled window manager is the perfect choice. Conversely, for people who like to use a mouse, I don't recommend tiled window managers. For those who have been curious after reading this far, this article introduces Linux distributions that will help you learn how to use the tile window manager, i3. "Regolith Linux" (Regolith Desktop) is a Linux distribution (desktop environment) that emphasizes productivity. It combines the i3 with "Sway" and "GNOME" to not only create a fast workflow for a tiled window manager that operates with a keyboard, but also incorporates a bit of functionality from a traditional desktop environment. Regolith includes support for both X11 (i3) and Wayland (Sway) as well as "gnome-session" for system management, "regolith-look" for adjusting themes and settings. Choose between i3 and Sway Yes, Regolith allows you to choose from i3 and Sway as a tiled window manager. I would like Sway to be considered an alternative to the i3 that works with Wayland. Regolith comes with both i3 and Sway, so you can enjoy a similar experience regardless of whether you have the window server you need (X11 or Wayland). Sway and i3 are very similar (they are not exactly the same), so if you master one, you will be able to use the other as well. There are the following differences between i3 and Sway: Sway uses Wayland, whereas i3 uses X11 Sway comes with a function to adjust the gap as standard. It also supports multiple unqualified keys when assigning key bindings Sway allows you to create containers from floated windows to move and resize groups of windows Sway handles input, output, and wallpaper settings without relying on another application Using Sway will slightly longer battery life on your laptop Some X11 apps may not work with Wayland. This means that i3 offers more widespread application support Compared to the i3, Sway is more active in development and has more GUI features. There is one thing to be careful about. In the past, ISO images of Regolith were provided. Unfortunately, its ISO image is now an older version. To get the latest version of Regolith, install the required packages on an instance running "Ubuntu 22.04" or "Ubuntu 24.04". To do this, run the following command: wget -qO - https://archive.regolith-desktop.com/regolith.key | gpg --dearmor | sudo tee /usr/share/keyrings/regolith-archive-keyring.gpg > /dev/null echo deb "[arch=amd64 signed-by=/usr/share/keyrings/regolith-archive-keyring.gpg] https://archive.regolith-desktop.com/ubuntu/stable noble v3.2" | sudo tee /etc/apt/sources.list.d/regolith.list sudo apt update sudo apt install regolith-desktop regolith-session-flashback regolith-look-lascaille -y To install the Sway version, run the following command: sudo apt-get install regolith-session-sway -y If you are installing the Sway version (the author's recommendation), select "Regolith on Wayland" on the login screen. Once these packages are installed, restart your Linux instance and select either i3 or Sway at the login prompt. I would be very grateful if the developer of Regolith released the latest version of the ISO image, but if you already have an instance of running Ubuntu (there is no problem with "Debian"), it's not difficult to install. The installation should be completed within 5 minutes. Installing the latest version of Regolith should allow you to switch between i3 and Sway (older versions do not support Wayland/Sway, so only i3 is provided).

Tile window managers are one of the most efficient ways to operate your OS. Whether this is suitable or not depends on the user's preferred method of operation. If you prefer keyboard-based operations whenever possible and are unwilling to master a number of new keyboard shortcuts, a tiled window manager is the perfect choice. Conversely, for people who like to use a mouse, I don't recommend tiled window managers. For those who have been curious after reading this far, this article introduces Linux distributions that will help you learn how to use the tile window manager, i3. "Regolith Linux" (Regolith Desktop) is a Linux distribution (desktop environment) that emphasizes productivity. It combines the i3 with "Sway" and "GNOME" to not only create a fast workflow for a tiled window manager that operates with a keyboard, but also incorporates a bit of functionality from a traditional desktop environment. Regolith includes support for both X11 (i3) and Wayland (Sway) as well as "gnome-session" for system management, "regolith-look" for adjusting themes and settings. Choose between i3 and Sway Yes, Regolith allows you to choose from i3 and Sway as a tiled window manager. I would like Sway to be considered an alternative to the i3 that works with Wayland. Regolith comes with both i3 and Sway, so you can enjoy a similar experience regardless of whether you have the window server you need (X11 or Wayland). Sway and i3 are very similar (they are not exactly the same), so if you master one, you will be able to use the other as well. There are the following differences between i3 and Sway: Sway uses Wayland, whereas i3 uses X11 Sway comes with a function to adjust the gap as standard. It also supports multiple unqualified keys when assigning key bindings Sway allows you to create containers from floated windows to move and resize groups of windows Sway handles input, output, and wallpaper settings without relying on another application Using Sway will slightly longer battery life on your laptop Some X11 apps may not work with Wayland. This means that i3 offers more widespread application support Compared to the i3, Sway is more active in development and has more GUI features. There is one thing to be careful about. In the past, ISO images of Regolith were provided. Unfortunately, its ISO image is now an older version. To get the latest version of Regolith, install the required packages on an instance running "Ubuntu 22.04" or "Ubuntu 24.04". To do this, run the following command: wget -qO - https://archive.regolith-desktop.com/regolith.key | gpg --dearmor | sudo tee /usr/share/keyrings/regolith-archive-keyring.gpg > /dev/null echo deb "[arch=amd64 signed-by=/usr/share/keyrings/regolith-archive-keyring.gpg] https://archive.regolith-desktop.com/ubuntu/stable noble v3.2" | sudo tee /etc/apt/sources.list.d/regolith.list sudo apt update sudo apt install regolith-desktop regolith-session-flashback regolith-look-lascaille -y To install the Sway version, run the following command: sudo apt-get install regolith-session-sway -y If you are installing the Sway version (the author's recommendation), select "Regolith on Wayland" on the login screen. Once these packages are installed, restart your Linux instance and select either i3 or Sway at the login prompt. I would be very grateful if the developer of Regolith released the latest version of the ISO image, but if you already have an instance of running Ubuntu (there is no problem with "Debian"), it's not difficult to install. The installation should be completed within 5 minutes. Installing the latest version of Regolith should allow you to switch between i3 and Sway (older versions do not support Wayland/Sway, so only i3 is provided).

Today, we are talking about a colossal contract of nearly $ 12 billion signed between Openai and Coreweave. A contract that could clearly shake up the hand of Microsoft on Openai. All tech news for pros every day in our newsletter learn more about the use of personal data But before going further, I am sure you have already heard of Openai. He is the publisher of the generative artificial intelligence solution Chatgpt. But this is surely the first time that you hear the name of Coreweave. So who is Coreweave? This American startup, supported by Nvidia, is in full ascent. It provides cloud infrastructure optimized for AI. And so it competitions directly Azure from Microsoft and AWS from Amazon. Its turnover exploded in one year, from 229 million to almost $ 2 billion in 2024. And it is preparing a much awaited IPO. A small pebble in the Microsoft shoe Then, why is this contract between Openai and Coreweave is a real pebble in the Microsoft shoe? So far, Openai was mainly based on Microsoft Azure to run its AI models. But with this 5 -year investment in Coreweave, Openai affirms its independence. A strong signal while relations between Microsoft and Openai seem to be stretching, especially on the AI ​​market intended for businesses. The entry of Coreweave into the Openai ecosystem therefore reduces its dependence on Microsoft. The phenomenal race for computing power But more broadly, this deal reflects a broader trend, that of the race for computing power. The generative AI explosion boosts GPU demand and cloud infrastructure, and AI players seek to secure their resources by finding alternatives to hyperscal infrastructures like Microsoft, Amazon or Google. By betting on Coreweave, which is a close partner of Nvidia, Openai ensures privileged access to these new infrastructures, and killing two strokes with a potential competitor of Microsoft. The ZD Tech is on all Podcast platforms! Subscribe!

Michal Černý The author is a journalist on the loose leg. He studied journalism at FSV UK, Poté worked as an editor of IT magazine Chip. Víc than the Deset Flight Psal on Technologies in the RESPPT, published such in the Němec magazine Bild der Wissenschaft and MIT Technology Review Deutschland. It will create graphics, like a cameraman and a scorch of equalized documentary films. It is justified by information technology and the topic of the round of ecology and the maintenance of the consumption. He is a member of the West Bohemian writer in Pilsen, his literary patterns belong to the writer Stanislaw Lem and Andrzej Sapkowski. He lives in Sokolov.

Each year, at the Adobe summit, the company shows its latest experimental innovations, 40 % of which are officially deployed. All tech news for pros every day in our newsletter learn more about the use of personal data Adobe shared in preview with ZDNET an experience that changes the game for marketers as well as brands. What is the Panorama project The Panorama project helps brands understand user interactions on mobile applications. This task is complex due to the multitude of elements present in applications, such as screens, offers and messages. The Panorama project offers a card that brands can use to view the user route. According to Adobe, the card will update in real time, following the performance indicators and user behavior to present each page and screen of the user experience. Marketing specialists could use this information to deploy new offers and send messages from the Project Panorama interface without creating code. What facilitate optimization to increase user engagement.

A Norwegian man said he was horrified to discover that ChatGPT outputs had falsely accused him of murdering his own children. According to a complaint filed Thursday by European Union digital rights advocates Noyb, Arve Hjalmar Holmen decided to see what information ChatGPT might provide if a user searched his name. He was shocked when ChatGPT responded with outputs falsely claiming that he was sentenced to 21 years in prison as "a convicted criminal who murdered two of his children and attempted to murder his third son," a Noyb press release said. ChatGPT's "made-up horror story" not only hallucinated events that never happened, but it also mixed "clearly identifiable personal data"—such as the actual number and gender of Holmen's children and the name of his hometown—with the "fake information," Noyb's press release said. ChatGPT hallucinating a "fake murderer and imprisonment" while including "real elements" of the Norwegian man's "personal life" allegedly violated "data accuracy" requirements of the General Data Protection Regulation (GDPR), because Holmen allegedly could not easily correct the information, as the GDPR requires. As Holmen saw it, his reputation remained on the line the longer the information was there, and—despite "tiny" disclaimers reminding ChatGPT users to verify outputs—there was no way to know how many people might have been exposed to the fake story and believed the information was accurate. "Some think that ‘there is no smoke without fire,’" Holmen said in the press release. "The fact that someone could read this output and believe it is true, is what scares me the most."

is transportation editor with 10+ years of experience who covers EVs, public transportation, and aviation. His work has appeared in The New York Daily News and City & State. Tesla protesters are planning their “biggest day of action” yet, aiming for 500 demonstrations at Tesla showrooms across the world on March 29th, organizers said during a mobilizing call Wednesday. There have also been an uptick in incidents of arson, vandalism, and violence against Tesla showrooms that, while unrelated to the protests, have led to Musk and President Donald Trump labeling them “domestic terrorism.” Other members of the Trump administration have signaled the protesters could come under scrutiny as well. Attorney General Pam Bondi promised “severe consequences on those involved in these attacks, including those operating behind the scenes to coordinate and fund these crimes.” AFP via Getty Images The hour-long call, which included actors, filmmakers, members of Congress, federal workers, academics, and journalists, tried to steer clear of Trump’s talk of “terrorism,” keeping the focus on Musk and the effort to tank the company’s stock price. Tesla’s stock has lost nearly 40 percent of its value since the beginning of the year, as poor sales and rising competition in the US and overseas have fueled growing pessimism about the company’s future. “There is no conspiracy, there is no well-funded cabal,” said actor and filmmaker Alex Winter, who helped promote the protests early on on BlueSky. “It’s just Elon Musk who has taken Tesla down.” Alice Hu, executive director of Planet Over Profit, said that protesters were aiming for 500 events across the world on March 29th, with demonstrations at all 277 Tesla showrooms in the US, as well as hundreds more overseas. Protesters should even feel free to demonstrate at Supercharger stations, she said. “There is no conspiracy, there is no well-funded cabal.” “We need to show Elon that he can throw a tantrum online because his stocks are tanking,” Hu said. “He can get Trump to put on a humiliating used car show in front of the White House. These wannabe authoritarians can try to intimidate us from exercising our First Amendment rights, but they can’t stop us from fighting back.” Organizers were adamant that their movement was peaceful, often going out of their way to stress the nonviolent nature of the demonstrations. “The things that we’re fighting for, we are fighting for our country,” Rep. Jasmine Crockett (D-Tex.) said. “We’re fighting for democracy. We’re fighting for our freedoms. And when I say fighting, I’m saying that figuratively. Obviously, everything that I am promoting is nonviolent.” With Trump promising domestic terrorism charges for violence against Tesla, organizers advised that protesters consult attorneys to better understand the laws in their states. Some states have statutes that could be used to intimidate protesters, so its worthwhile to know what you’re up against, said Lauren Regan, executive director and senior staff attorney at the Civil Liberties Defense Center. She said in her experience, states are often hesitant to prosecute activists because there’s a strong likelihood those statutes will be found unconstitutional. “Their goal is to just pluck a few individuals out and scare the rest of us into submission and apathy,” Regan said. “There are going to be some areas of the country that are very conservative and are gonna be hard on dissidents or activists, no matter what the timing.” The sharp decrease in Tesla’s stock in recent weeks has clearly invigorated the protests. Several speakers spoke of Tesla’s collapse as not only possible, but likely. Micah Lee, an investigative journalist who was among a group of Twitter users to have their accounts banned by Musk shortly after his acquisition of the social platform, said that going after Tesla’s value was a “solid strategy.” “If we kill the Tesla brand, if we drive down the stock price low enough, we can force him to sell his stock to pay back the billions of dollars of debt that he took on to buy Twitter,” Lee said. “This will drive Tesla’s stock into a death spiral.” Musk’s status as the richest man in the world is largely thanks to Tesla’s stock price. He owns 13 percent of the company, making him the single largest shareholder. As of today, the company is worth $739 billion — down from 1.08 trillion earlier this year, meaning Musk’s stake is worth about $96 billion. And Tesla’s board of directors is composed of close friends and relatives, raising concerns about its independence from the controversial CEO. Several board members, including chair Robyn Denholm and James Murdoch, have sold over $100 million in Tesla stock in recent weeks. “This will drive Tesla’s stock into a death spiral.” But it’s not clear that hurting Tesla will actually matter much to Musk. He remains in Trump’s good graces, and is wielding vast amounts of control within the federal government. Even if these protests can seriously affect Tesla, Musk has consolidated so much political power that, after a certain point, it’s not clear whether market forces still apply as strongly. Musk’s love of memes — he recently quipped “I am become meme” at CPAC — is a sign that the world’s richest man is living in a different reality than most people, which could be an advantage, said Joan Donovan, an assistant professor of journalism and emerging media studies at Boston University who studies media manipulation, disinformation, and online political movements. “He thinks of himself as this black hat hacker that’s broken into the government and socially engineered his way into the Treasury and he’s gonna abscond with all the data, it’s an obvious data heist,” Donovan said. “But he believes he’s living in a meme, and so we need to be very clear about what our demands are, about what our bright lines are, and that we’re not gonna stop until Tesla is done with Musk.” In simplest terms, the protesters said they were not backing down.

An electric truck with a battery replaceable battery begins to drive through the streets of Tokyo. A multi-billion yen project has begun to take action to operate vehicles such as battery-replaceable electric trucks and other delivery vehicles by Mitsubishi Fuso Trucks and Buses and Mitsubishi Motors at fully automatic battery replacement stations installed in Tokyo. Ample, a US company that handles battery replacement systems for electric vehicles (EVs), will provide the technology and install multiple exchange stations in various locations in Tokyo. This project has been selected as a support project for the "Technology Development Support Project for Promotion of New Energy" by the Tokyo Metropolitan Government and the Tokyo Environmental Corporation, and it is expected that the total investment amount will be around billions of yen in subsidies from the Tokyo Metropolitan Government. Mitsubishi Fuso Trucks and Bus have already begun demonstration experiments in Kyoto City in September 2024 using the eCanter, a small battery-replaceable electric truck. This electric truck is equipped with Ample's replaceable battery module, and was developed in 2023 in partnership between the two companies. Amplé's fully automatic battery replacement station has already been installed in Kyoto City, and Yamato Transport uses the eCanter battery replacement model for collection and delivery services. Ample's fully automatic battery replacement station was installed in Kyoto city. A demonstration experiment is currently underway using a battery-replaceable model of the small electric truck "eCanter." ENEOS Holdings also participates in the operation. The same applies to the demonstration experiment in Tokyo, with Ample battery replacement stations being installed in various locations in Tokyo, and delivery companies plan to operate eCanter battery replacement models and other models starting from the first half of this year. When a vehicle with a low battery level arrives at the exchange station, the robot automatically replaces the battery. The battery replacement time is approximately five minutes, and one station can accommodate more than 100 vehicles. "Overall costs are about 10-20% cheaper than traditional gasoline or diesel. With the amount of energy consumed and the installation costs lower, we expect it to be 3-1 times cheaper than fast charging over time," explained John de Souza, Ample co-founder and president. Strength in replacing batteries in 5 minutes and reducing vehicle costs Ample is planning to use this project as an opportunity to accelerate its business expansion in urban areas in Japan and Asia, and will begin expanding its target vehicles, starting with the logistics field. "We were able to demonstrate a lot in our demonstration experiments in Kyoto," says De Souza of Ample. "We were able to demonstrate that Ample stations can be installed in a few days, that it consumes significantly less power than fast charging, that it does not require expensive charging infrastructure or vehicle downtime, and that it can efficiently support vehicles without changing the way it operates." With Ample's system, the time required to replace the battery is approximately 5 minutes. In contrast, regular small electric trucks take around 10 hours to charge in Japan, and 1-2 hours to charge them on a fast charge. For this reason, replacing the battery can minimize the time the truck will be stopped, which increases the operating rate.

Humans spend about a third of their lives sleeping. It's a shame, but we humans continue to sweat and remove our skin and hair while we sleep. Some reviewers of WIRED eat snacks in bed. You may not be that far, but it's a good idea to know effective ways to keep your bed clean. This time we received advice from sleep and bedding experts. The three people spoke to the story: Angela Holiday-Belle (a doctor and sleep expert at the University of Chicago), Parima Ears (textile expert and founder of bedding brand Pure Prima), and Tandy Avery (vice president of product development at interior fabric maker Luzome). Keep your bedding clean all year round and get a comfortable sleep! Photograph: Nena Farrell Basics of care Some people may want to make their bed first thing in the morning. However, first, flip through the futon and sheets to let the air flow through the bed. Even just a few minutes can make a big difference. By removing moisture, you can also prevent dust mites. Use cold water when washing bedding. Tandy Avery, vice president of product development at US bedding manufacturer Luxome, says it's best to choose a cold water course and run the washing machine to avoid damage to the fibers. Be sure to follow the manufacturer's recommended care instructions. Avery also says that it is recommended to wash items with zippers and hooks separately from bedding to prevent them from getting caught. Textile expert and founder Pure Parima Iyaz says bedding should be dried at low temperatures to protect delicate fabrics. Also, when storing bedding, it is best to store it in a cool, dry place to prevent mold. If your home washing machine or dryer has a small capacity, consider using a coin laundry. The large front-loading washing machine allows you to quickly wash and dry bulky bedding. A home dryer takes longer, but a coin laundry can dry in about 20 minutes. If the time and weather allow, it's fine to dry naturally with a clothesline. Photograph: Kinga Krzeminska/Getty Images About detergents and washing machines How to choose detergent: There are many types of detergents, each one with a good variety. For those who are sensitive to skin, those who use a particular washing machine, and those who value the care of delicate fabrics, unscented detergents may be suitable. "A small amount of detergent is enough," says Angela Holiday-Bell, a doctor and sleep expert at the University of Chicago. This is because excessive use will quickly wear out the fibers and will actually damage your bedding.

The US Army is using CamoGPT, a test-stage generation AI tool, for new uses. The purpose is to identify and remove statements relating to diversity, equity, inclusiveness and accessibility (DEIA) from training materials in accordance with the latest executive order issued by President Donald Trump. According to an internal document confirmed by WIRED, the US Army Training Doctrine Command (TRADOC), an important command division responsible for training soldiers, training executives, and formulating military activity guidelines, strategies and initiatives, uses AI tools called "CamoGPT" on a daily basis, with the aim of "reviewing the various policies, plans, public relations and initiatives relating to DEIA and reporting on the facts we have learned." The internal document was prepared after President Trump signed an executive order on January 27, 2025, entitled "Restoring America's Fighting Force." The executive order that ordered Secretary of Defense Pete Hegses to eliminate any references of the Pentagon's policies on race and gender that could encourage "non-American, division-inducing discriminatory, radical, extremist and irrational logic," extends to past social media posts sent from official US military accounts. In an email to WIRED, Army Major Chris Robinson, who serves as TRADOC's spokesman, acknowledged that the CamoGPT's purpose was a review of materials that referenced the DEIA. "[TRADOC] will carry out all the instructions listed in the President's executive orders. We will use all the professionalism and efficiency to implement these directives in line with national security policies," Robinson said. "We will not disclose any details about our internal policies or methods, but TRADOC aims to increase productivity in all respects by utilizing all of our hand-held tools, including CamoGPT. We also believe that this is possible." Military AI with approximately 4,000 users CamoGPT was developed in the summer of 2024 with the aim of improving productivity and enhancing immediate service across the US Army. The number of users now reaches around 4,000, and users are routinely "interacting information" with the tool, says Army Captain Aidan Doyle, a data engineer at CamoGPT. CamoGPT has a wide range of uses, from developing teaching materials for various training programs to creating translated materials in other languages. According to Robinson, TRADOC presented a "proof-of-concept demonstration (PoC: an experiment showing the feasibility of new ideas and technologies) for the tool at the annual meeting of the United States Army Association (AUSA) held in Washington, DC in October 2014. Doyle avoids specific mentions about how TRADOC officials use CamoGPT when checking DEIA-related statements, but states that they are not using any particularly unusual techniques for document searches. "We'll get all the documents that need to be investigated, then load them into CamoGPT in order, then ask them about what they need," explains Doyle. "The Search Extension Generation (RAG) feature allows CamoGPT to return more detailed information, the more specific the questions about the document's purpose." The actual procedure is that TRADOC staff will load a large number of documents into CamoGPT and order the Large-Scale Language Model (LLM) to search for keywords such as "dignity" and "respect" to be searched for. These are also used as keywords for searching for past digital content. This is how TRADOC is trying to find the material to be revised and change it to something that matches Trump's executive order.

Table of contents We won't let Gaza become the "Middle Eastern Riviera" Everything is on a pile of rubble Destruction and reconstruction What we Palestinians want I was in New York when I first heard President Donald Trump talk about what will happen in Gaza. This is to attend a special preview to celebrate the broadcast of the new season of the Netflix drama "Mo", produced and starring by his friend Mo Amar. I received a message from another friend. "Damn Trump's press conference, did you see it? It says America will take over Gaza. Let's talk tomorrow." I was in a hurry. Who on earth will the US take over Gaza? The Israeli army retreated after destroying this area [Editor's note: at the time of writing. On March 18, 2025, Israeli forces resumed their attacks. Ahmad, a friend of mine who lives in Beit Hanoon, a town in northern Gaza, said residents are beginning to return to town, not to regain their previous lives, but to "live on the rubble of their homes." But even those rubbles have meaning to us. It is the place where our precious people lived and died. When the battle is over, we are the only ones to clear this rubble. We have no choice but to recover from this. Don't let Gaza become the "Middle Eastern Riviera" Trump said: "The Gaza people will have peace. They will no longer be killed in gunfire or destroyed the town. They will no longer be forced to the great people who built this land. The Palestinians want to return to Gaza because they have no other options. It seems like the demolition site is here." He did not mention until the end that it was the Israeli army supported by the US government that committed the shooting, slaughter and destruction. Instead, Trump's mouth was to make the Gaza Strip a "Middle Eastern Riviera." It was like there was no one living there. Later, when asked how many people could drive their homeland, Trump replied: "All of them are. 1.7 million or 1.8 million... We should be able to live happily at our destination. We will no longer be afraid of death, day after day." He also said that when rebuilding the Gaza Strip, both the King of Jordan and the Egyptian president felt that they would "show understanding" the Palestinians. It's like someone other than Gaza residents will be working on this difficult project, which takes time. The numbers Trump said at this time don't matter. The question lies elsewhere. Who said that Gaza residents were scared of death? Many people are scared of death all over the world. This is probably true for Americans who do not have health insurance and Americans who live in areas where wildfires are on the verge of. But we are not afraid of death. What Palestinians are scared of is the death of Israeli soldiers and migrants by bombs and gunshots. How do you stop people from being killed? The way to do this should not be to move people who have been shot or bombed, but to stop the actions of those who are shooting or bombing. Everything is on a pile of rubble The ceasefire agreement that came into effect in Gaza in January was virtually powerless, with more than 100 Palestinians still being killed and many more injured. People close to me can hear that they are about to return to where they lived before October 7th, 2013. His wife's family also returned to his hometown to see that his three-story home was still built. This is an exception in Gaza, where most of it has been bombed. However, when I stepped in, the interior was completely burned down. There were almost no traces of the closet, mattress, blankets, etc. I watched the video I had sent to, and every time someone walked I could hear the tiles on the floor cracking. The walls and ceiling were charred. My friend, Sabre, has two children, but is not in a hurry to return to the Northern District. After October 7th, he left northern Gaza and lived in a tent in Haan Eunice, but in November his apartment was bombed. On January 27th, 2015, the day he was allowed to return to the North, my friend returned to his hometown and found out there was nothing there. We left the next day. "I walked through for five hours," Sabre said. There was only two days in the north. "Then I walked another five hours back to Hahn Eunice's tent."

Apple is making major changes to its AI leadership team as the company struggles to keep up with competitors in artificial intelligence. According to Bloomberg’s Mark Gurman, CEO Tim Cook has lost confidence in John Giannandrea, Apple’s longtime AI chief, and is bringing in a new leader to fix Siri. Mike Rockwell, the exec behind Apple’s Vision Pro headset, is now taking over Siri development and will report directly to software chief Craig Federighi. This move effectively removes Siri from Giannandrea’s control. Sources familiar with the situation told Bloomberg that Apple’s AI efforts have been falling behind, and internal discussions at a recent secretive company summit focused on how to turn things around. The biggest concern is Siri, which has failed to deliver promised AI features for iPhone 16 users. Apple has advertised a more powerful Siri with deeper access to user data, but so far these updates aren’t even available yet six months later. Internally, Apple executives have acknowledged the problems. Siri’s outgoing manager, Robby Walker, recently called the delays “ugly” and “embarrassing” in a team meeting. He also admitted that Apple doesn’t have a firm timeline for when the AI upgrades will actually be released, which is pretty shocking. The decision to put Rockwell in charge of Siri is a sign that Apple wants someone with a track record of shipping complex products. Rockwell helped lead the development of the Vision Pro, a high-tech but expensive niche device. Now, Apple is betting that he can bring the same level of innovation and urgency to Siri. Giannandrea is staying at Apple and will continue working on AI research and long-term projects. Siri has been a disaster when compared to other smart assistants and generative AI chatbots–there’s no way to sugar coat it. Apple needs to act quickly here to bring forth its promised super smart Siri it advertised for the iPhone 16 series.

Is Apple experiencing a "Vista-like drift into systemically poor execution?" That was a question posed by well-known technology analyst Benedict Evans, in a recent blog post covering Apple's innovation and execution, or seemingly lack thereof as of late. He is referring to Microsoft's Windows Vista operating system, which was widely criticized when it launched in 2007 due to software bugs, performance issues, frequent warning dialogs, polarizing design changes, and several other problems. Evans said Apple stumbled with the Vision Pro headset, which he believes was not ready to launch. Then, it previewed personalized Siri features at WWDC last year that were merely conceptual, and are now delayed. "This is a concern," said Evans. His thoughts on the Vision Pro: The Vision Pro is a concept, or a demo, and Apple doesn't ship demos. Why did it ship the Vision Pro? What did it achieve? It didn't sell in meaningful volume, because it couldn't, and it didn't lead to much developer activity ether, because no-one bought it. A lot of people even at Apple are puzzled. He said the personalized Siri delay is a "mirror image" of the Vision Pro situation: Apple showed a demo, and it only does demos when things are nearly done, and it said it would ship 'later this year' and it never misses deadlines like that. So we should be using it, today. And now we find that we didn't really see a demo, only a mock-up of a great concept, and that this product won't ship until maybe late 2025, and possibly (going by the rumour-mill) 2026, or even 2027. All of this led to his Vista-like comment: The failure of Siri 2 is by far the most dramatic instance of a growing trend for Apple to launch stuff late. The software release cycle used to be a metronome: announcement at WWDC in the summer, OS release in September with everything you'd seen. There were plenty of delays and failed projects under the hood […] but public promise were always kept. Now that seems to be slipping. Is this a symptom of a Vista-like drift into systemically poor execution? Nevertheless, Evans acknowledged that critics have been claiming that Apple is no longer innovative since at least the 1980s, and that the company has historically continued to deliver more innovative and category-defining products over the years. Still, he said he is left wondering if that Apple still exists today. Evans' full blog post, highlighted by Techmeme today, is worth a read.

AI: Apple Intelligence or Apple Ineptness? Another dinobaby blog post. No AI involved which could be good or bad depending on one’s point of view. I read a very polite essay with some almost unreadable graphs. “Apple Innovation and Execution” says: People have been claiming that Apple has forgotten how to innovate since the early 1980s, or longer – it’s a standing joke in talking about the company. But it’s also a question. Yes, it is a question. Slap on your Apple goggles and look at the world from the fan boy perspective. AI is not a thing. Siri is a bit wonky. The endless requests to log in to use Facetime and other Apple services are from an objective point of view a bit stupid. The annual iPhone refresh. Er, yeah, now what are the functional differences again? The Apple car? Er, yeah. Is that an innovation worm? Is that a bad apple? One possibility is that innovation worm is quite happy making an exit and looking for a better orchard. Thanks, You.com “Creative.” Good enough. The write up says: And ‘Apple Intelligence’ certainly isn’t going to drive a ‘super-cycle’ of iPhone upgrades any time soon. Indeed, a better iPhone feature by itself was never going to drive fundamentally different growth for Apple So why do something which makes the company look stupid? And what about this passage? And the failure of Siri 2 is by far the most dramatic instance of a growing trend for Apple to launch stuff late. The software release cycle used to be a metronome: announcement at WWDC in the summer, OS release in September with everything you’d seen. There were plenty of delays and failed projects under the hood, and centres of notorious dysfunction (Apple Music, say), and Apple has always had a tendency to appear to forget about products for years (most Apple Watch faces don’t support the key new feature in the new Apple Watch) but public promise were always kept. Now that seems to be slipping. Is this a symptom of a Vista-like drift into systemically poor execution? Some innovation worms are probably gnawing away inside the Apple. Apple’s AI. Easy to talk about. Tough to convert marketing baloney crafted by art history majors into software of value to users in my opinion. Stephen E Arnold, March 20, 2025 Share this: Facebook Twitter LinkedIn More Email Reddit Tumblr Pinterest Print Comments

Despite the iPhone 18 launch being at least 18 months away, the rumour mill has already started churning up stories about the devices. Once again, a new report has suggested that Apple's iPhone 18 series will feature the C2 modem, Apple's custom cellular chip. According to a 9to5Mac report, Jeff Pu of GF Securities has mentioned in his latest note to investors that he expects Apple to pack the “C2” modem in the iPhone 18 Pro and iPhone 18 Pro Max in 2026. Related Articles Apple debuted the C1 chip in the recently launched iPhone 16e, after years of featuring Qualcomm's cellular modems in its iPhones. However, the C1 chip is unlikely to feature in the iPhone 17 series expected to launch later this year. This is because the C1 lacks support for ultra-fast mmWave 5G, meaning users miss out on the extreme speeds available in locations like stadiums, airports, and dense urban areas. Apple has acknowledged this limitation, stating that the C1 is only the beginning and that future iterations will improve. This most likely means that Apple isn't going to put this modem chip inside their flagship handsets. Instead, the iPhone 17 series will most likely feature Qualcomm's modems. Apple's current agreement with Qualcomm runs till March 2027, allowing Apple time to refine its modem technology while still using Qualcomm’s hardware in premium iPhone models. Notably, supply chain analyst Ming-Chi Kuo had earlier suggested that the iPhone 17 Air will also use the C1 modem. Earlier, Qualcomm CEO Cristiano Amon had mentioned how Qualcomm's newly launched X85 modem was miles ahead of Apple's C1 chip, and it would create a "huge delta" in performance between premium Android devices and Apple’s iPhones.

Apple is shaking up its executive team in light of ongoing struggles to roll out advanced Apple Intelligence features for Siri. Bloomberg reports that Apple CEO Tim Cook “has lost confidence in the ability of AI head John Giannandrea to execute on product development.” As such, Mike Rockwell, current vice president of the Vision Products Group and creator of Apple Vision Pro, is taking the reigns of the Siri team. Giannanndrea is not leaving Apple, but Siri is being completely removed from his responsibilities. He will continue to oversee research, testing, and other AI-related technologies. Rockwell will report to Apple’s software boss, Craig Fderighi. According to the report, Giannanndrea “long considered Rockwell a potential successor.” The report also says another “senior manager” from Rockwell’s team is moving to the Siri group: In the past several days, Apple started moving over another senior manager from Rockwell’s team — Aimee Nugent — to the Siri group. Like Vorrath, she has a reputation for fixing challenging projects. The changes allowed two of Rockwell’s trusted executives to evaluate the organization before he became heavily involved. Inside Apple, Rockwell hasn’t been shy about criticizing Siri, according to people familiar with the matter. For years, he has pitched senior vice presidents on ideas for overhauling the voice assistant to make it more personalized. He has also been advising the AI group in recent weeks. With Rockwell shifting to the Siri team, Paul Meade will take over the Apple Vision Pro team. Meade previously led hardware engineering for Vision Pro and worked under Rockwell. Read the full report at Bloomberg.

Mark Gurman, with a blockbuster scoop for Bloomberg: Apple Inc. is undergoing a rare shake-up of its executive ranks, aiming to get its artificial intelligence efforts back on track after months of delays and stumbles, according to people familiar with the situation. Chief Executive Officer Tim Cook has lost confidence in the ability of AI head John Giannandrea to execute on product development, so he’s moving over another top executive to help: Vision Pro creator Mike Rockwell. In a new role, Rockwell will be in charge of the Siri virtual assistant, according to the people, who asked not to be identified because the moves haven’t been announced. Rockwell will report to software chief Craig Federighi, removing Siri completely from Giannandrea’s command. Apple is poised to announce the changes to employees this week. The iPhone maker’s senior leaders — a group known as the Top 100 — just met at a secretive, annual offsite gathering to discuss the future of the company. Its AI efforts were a key talking point at the summit, Bloomberg News has reported. [...] My quick take on this is that it’s a turf battle that Craig Federighi just won. It’s not just putting a new executive in charge of Siri, it’s moving Siri under Federighi’s group. How Gurman got this scoop before Apple had announced the changes — even internally — is rather unbelievable. It’s not “Bloomberg” that got this scoop. It’s Mark Gurman. And trust me, Apple PR did not leak this to him deliberately. I’m sure they’re now accelerating an announcement, at least internally, framing it on their own terms. I can only guess that Gurman hinted at his sourcing in the passage above: Tim Cook must have announced these changes at the Top 100 retreat this week, and at least two of those attendees leaked the news to Gurman. Unprecedented. Also: Rockwell is currently the vice president in charge of the Vision Products Group, or VPG, the division that developed Apple’s headset. As part of the changes, he’ll be leaving that team and handing the reins to Paul Meade, an executive who has run hardware engineering for the Vision Pro under Rockwell. I don’t find it surprising at all that Rockwell was given this task. Giannandrea will remain at the company, even with Rockwell taking over Siri. An abrupt departure would signal publicly that the AI efforts have been tumultuous — something Apple is reluctant to acknowledge. Giannandrea’s other responsibilities include oversight of research, testing and technologies related to AI. The company also has a team reporting to Giannandrea investigating robotics. This I find a little surprising. ★ Thursday, 20 March 2025

Venture capital investment in European startups passed $52 billion last year, reflecting the market’s long-term growth trajectory and gradual stabilization after the out-sized peaks of 2021-2022 (driven largely by the COVID-19 pandemic), and the comparative slump of 2023, according to a new report. Although 2024 has seen political and regulatory turmoil, Europe’s talent pool of startups continues to increase, even if funding shortages kicked in last year, per global law firm Orrick’s new “Deal Flow” report, which covers 2024. An analysis of more than 375 VC and growth equity investments in Europe last year reveals a handful of key takeaways. Compared to previous years, Europe’s startup market stabilized, with a modest rebalancing of investment terms compared to the extreme highs and lows of the pandemic hype and post-pandemic slowdown. There was also a lot more adoption of the British Venture Capital Association’s new model form documents in European deals, which tend to more closely align with U.S. practices. With this de facto standard emerging, this trend is likely to accelerate future deal-making because it’s a lot easier to push through deals where everyone is familiar with the structure. European companies also appeared to expand option pools, with over 70% of equity financings including a top-up, highlighting a stronger European talent pool and focus on scaling companies rather than selling early. There were signs of an improvement to deal volume and size, too, with the average size of deals Orrick did with investor clients growing by 66%, while deals initiated by startups saw a slight decline, even though company-side deals still represented the majority. However, the report reflected the fact that Europe remains constrained in the number and amount of growth-stage funding deals. While Europe is well-served for early-stage, later-stage and growth-stage funding is scarcer. Equity-based deals were stronger than debt-based deals, with companies preferring extension rounds over debt rounds. The two most common types of equity-based deals emerging in this instance are ASAs (Advanced Subscription Agreement) and SAFE (Simple Agreement for Future Equity). Some 30% of rounds were either a standalone secondary financing or rounds that included a secondary component. Founders tended to access secondary transactions earlier in the funding stage, with some occurring as early as Series A. Startups with some kind of SaaS or platform-based business model represented 21% of financings, deep tech increased to 23%, deals with an AI and ML (machine learning) component maintained a 33% share, and fintech rose to 16% of European deals.

The boss of the world’s biggest computer chipmaker, Nvidia, has promised that the company will shell out “several hundred billion” dollars to make semiconductors and other electronics in the US over the next four years. The comments from Jensen Huang illustrate how the California-headquartered AI chipmaker is remodelling its supply chain away from Asia amid unpredictable tariff threats by Donald Trump. The chief executive, who also co-founded Nvidia, told the Financial Times: “Overall, we will procure, over the course of the next four years, probably half a trillion dollars worth of electronics in total. And I think we can easily see ourselves manufacturing several hundred billion of it here in the US.” The remarks are the latest example of how Trump’s “America First” policy is having an impact on business investment, even forcing groups such as Nvidia – which is among the world’s most valuable companies – to rethink their global footprint. The Silicon Valley company, founded in 1993, has been the engine of the AI market boom, with hype around the technology propelling its valuation to a stratospheric $2.9tn (£2.2tn). But in recent years it and other big US tech companies, such as Apple, have become increasing dependent on chipmaking facilities in Taiwan run by TSMC and Foxconn – a supply chain now under threat from trade wars and from China’s territorial designs on its neighbour. Huang said he was confident that Nvidia was well-positioned to deal with any worsening situation in Taiwan, which is also vulnerable to earthquakes. “At this point, we know that we can manufacture in the US, we have a sufficiently diversified supply chain,” he added. He added that the Trump administration could be in a position to bolster the US AI industry at a time of growing competition from China. Huawei, for example, had become the “single most formidable technology company in China”, having “conquered every market they’ve engaged”. Huang argued US efforts to constrain the Chinese company had been “done poorly”, as evidenced by Huawei’s continued success. skip past newsletter promotion Sign up to Business Today Free daily newsletter Get set for the working day – we'll point you to all the business news and analysis you need every morning Enter your email address Sign up Privacy Notice: Newsletters may contain info about charities, online ads, and content funded by outside parties. For more information see our Newsletters may contain info about charities, online ads, and content funded by outside parties. For more information see our Privacy Policy . We use Google reCaptcha to protect our website and the Google Privacy Policy and Terms of Service apply. after newsletter promotion For now, “having the support of an administration who cares about the success of this industry and not allowing energy to be an obstacle is a phenomenal result for AI in the US”, Huang said, referring to the large amount of electricity needed to power its chips in datacentres. Efforts to onshore manufacturing have been helped by a recent $100bn US investment by TSMC. The Taiwanese semiconductor firm’s investment means that Nvidia’s Blackwell chips – its top-of-the-line graphics processing unit – are being produced in the US, providing for “a substantial step up in our supply chain resilience”.

Nvidia is making a major move to fortify its supply chain, committing billions of dollars over the next four years to expand U.S.-based semiconductor and electronics manufacturing. The decision comes amid escalating geopolitical tensions, surging demand for AI infrastructure, and increased competition from rival chipmakers. With this investment, Nvidia aims to reduce its reliance on overseas suppliers, particularly Taiwan Semiconductor Manufacturing Company (TSMC), while reinforcing its dominance in AI hardware. The move is part of a broader industry shift, as companies like Apple and Intel also increase their semiconductor investments in the United States. TSMC itself is investing $100 billion in Arizona, though it has acknowledged production delays that could impact supply chains. Building a Domestic AI Chip Supply Chain For years, Nvidia has depended on TSMC and Samsung to fabricate its GPUs. However, rising geopolitical risks and new U.S. regulations—such as the CHIPS and Science Act—are reshaping semiconductor manufacturing. The CHIPS Act provides billions in incentives for domestic production, aligning with Nvidia’s new investment strategy. CEO Jensen Huang has emphasized the importance of securing a stable, U.S.-based supply chain. Despite these investments, challenges remain. TSMC’s Arizona plants have already faced significant delays, and building new facilities requires extensive time and resources. Additionally, U.S.-made chips may carry a higher price tag compared to those produced in Taiwan or South Korea. Blackwell Ultra AI Factory: The Next Generation of AI Computing At GTC 2025, Nvidia just unveiled its new Blackwell Ultra AI Factory Platform, which is designed to significantly improve AI reasoning and inference. This infrastructure includes the GB300 NVL72, which integrates 72 Blackwell Ultra GPUs and 36 Grace CPUs, delivering a 1.5× increase in performance over previous models. Major cloud providers, including Microsoft, Google, and Oracle, have already committed to integrating Blackwell Ultra-powered instances into their services. Additionally, Nebius has announced it will be an early adopter, offering Nvidia-powered AI compute capabilities. As part of a broader push to enhance AI infrastructure, Nvidia has also joined forces with xAI (Elon Musk’s AI venture), BlackRock, and Microsoft in a $30 billion AI infrastructure fund. The initiative aims to develop energy-efficient data centers optimized for generative AI workloads, addressing the growing power consumption concerns of AI models. Surging Orders and Rising Competition The demand for AI hardware has never been higher. Nvidia has received orders for 3.6 million Blackwell GPUs from cloud providers, though notably, Meta has yet to place an order. This aligns with Meta’s ongoing strategy to develop in-house AI chips, reducing its dependence on Nvidia. Meanwhile, Nvidia might face growing competition from China-based firms in the future, which are ramping up their AI chip efforts. With the U.S. imposing new restrictions on high-performance chip exports to China, Nvidia is focusing on strengthening its presence in Western markets. Will Nvidia’s U.S. Investment Pay Off? If successful, Nvidia’s expansion could set a precedent for other semiconductor firms considering similar moves. However, the challenges of domestic chip production—including supply chain disruptions, workforce shortages, and high production costs—must be addressed for long-term sustainability. Nvidia’s approach indicates that it is not willing to wait for external factors to dictate its future. By securing a U.S.-based manufacturing ecosystem, the company is positioning itself to meet the increasing demands of AI computing while reducing supply chain vulnerabilities.

What's particularly noteworthy is the existence of "QDriver" that corresponds to the technical differences between different types of quantum computers and processors. QDriver acts as a hardware abstraction layer and is responsible for converting a general-purpose quantum instruction set (NetQASM) into instructions that rely on specific quantum hardware. The research team has developed two completely different types of quantum processors, the ion trap method (a method in which ions trapped by an electromagnetic field are manipulated to use as qubits) and the diamond spin method (a method in which lattice defects that occur in the crystal structure of a diamond are used as qubits), and demonstrated its versatility. "The system is like software on a home computer. It can be used without knowing how the hardware works," explains Maria Grazia Iuriano, a doctoral researcher at Delft Institute of Technology. Successfully implementing multitasking function Another feature of QNodeOS is its multitasking function that allows multiple quantum applications to be run simultaneously. The experiments successfully ran two different quantum applications simultaneously: distributed quantum calculation (DQC) and single-node local gate tomography (LGT). This multitasking function greatly improves the efficiency of quantum hardware utilization. For example, while one program is waiting for a response from the network, another program may make use of the quantum processor. Experiments showed that increasing the number of applications significantly improved device utilization. While QNodeOS makes practical use of quantum networks more realistic, challenges still remain. For example, there may be improved efficiency in generating entanglement between quantum processors, improved reliability of quantum memory, and improved communication speed and synchronization of the network. Furthermore, in order to further develop QNodeOS, it is necessary to create a scalable system by integrating different quantum devices together. The research team will continue to integrate hardware and software in the future with the aim of achieving secure quantum computing and large-scale quantum networks on the cloud. (Edited by Daisuke Takimoto) *Click here for related articles on quantum computers by WIRED. Related Articles Magazine "WIRED" Japanese version VOL.56 "Quantumpedia: Quantum Computers Beyond" If traditional classical computers are "defining calculations" through "the logic and circuits designed by humans," then quantum computers can be seen as an approach to "derive calculations" by taking the "the rhythm of information processing that nature itself possesses" - the flow in which multiple possibilities coexist and intervene in a way that interferes and entangles together, converges to an optimal solution. In other words, the countless possibilities engraved in the deep layers of nature and the existence that could become the "knot" with us humanity. That's a quantum computer. How will quantum computers bring about changes in society, culture, and industry in the future? A must-read "Quantum Technology Encyclopedia" for the upcoming 2030s (Quantum Age)! Learn more here.

An analyst who is one of the biggest boosters of Tesla on Wall Street says Elon Musk is facing a “moment of truth” at his EV company because of a “crisis” he’s created by spending so much in the Trump administration. The analyst, Wedbush’s Dan Ives, pleaded with Musk in a note published Thursday to reduce his role working with the Department of Government Efficiency (DOGE). “If you agree or disagree with DOGE it misses the point that by Musk spending 110% of his time with DOGE (and not as Tesla CEO) since President Trump got back into the White House this has essentially turned Tesla into a political symbol…. and this is a bad thing,” Ives wrote. This has led to what Ives described as a “brand tornado crisis moment for Musk and Tesla.” Ives’ plea comes as Tesla’s stock price has fallen to around $230 — less than half its all-time peak in late December of around $480, and lower than it was before Trump got elected. Protests against the company are growing worldwide, an the level of vandalism against Tesla’s vehicles and stores has increased so much that it prompted the new administration to label the actions as “domestic terrorism.” Ives used similar language in a note published earlier in March, but couched it in a more optimistic tone, saying he believed the company was at the “start of the biggest innovation and technology cycle in Tesla’s history.” Thursday’s note was more somber. While he wrote on X that Wedbush has “never been more bullish on Tesla’s future,” he specifically called on Musk and Tesla’s board of directors to “step up, stop being silent, and help resolve this crisis.” Ives didn’t call on Musk to step down — although another high-profile Tesla investor who once tried to get on Tesla’s board in 2023, Ross Gerber, is calling for just that. Instead, Ives urged the CEO to announce he will “balance” his roles at DOGE and Tesla. He also wants Musk to offer a more concrete roadmap for the mysterious “more affordable models” that are supposed to debut at some point this year, as well as more detail on its attempt to launch a robotaxi service. If the Tesla CEO does those things, Ives wrote, the “heat from Musk around DOGE will start to dissipate among most of the critics and this will leave a scar,” but won’t lead to “permanent brand damage.”

Microsoft is adding another 475 megawatts to its already considerable renewable-powered portfolio to feed the growing energy appetite of its data centers. The company recently signed a deal with energy provider AES for three solar projects across the Midwest, one each in Illinois, Michigan, and Missouri. The ramp up reflects the immediacy of Microsoft’s needs. When it comes to powering data centers, it’s hard to argue with solar. Quick to install, inexpensive, and modular, it’s a perfect fit for tech companies that need electricity now. Microsoft has been tapping solar with some regularity. In February, it contracted 389 megawatts from three solar projects across Illinois and Texas. And late last year, the company announced it was anchoring a $9 billion renewable power coalition that’s organized by Acadia. The Redmond-based company’s own renewable portfolio already includes over 34 GW of capacity. While tech companies have shown increasing interest in nuclear power in recent months, the cost and speed advantages of renewables have kept solar deals flowing. Though renewable power on its own doesn’t have the same consistency as nuclear or natural gas, developers are increasingly pairing it with battery storage to provide around-the-clock electricity. The combination is more expensive than solar or wind on its own, but given the rapid declines in cost for both solar and batteries, so-called hybrid power plants are beginning to encroach on prices for a new natural gas generating capacity. So far, new nuclear prices have remained significantly higher than either renewables or natural gas power plants. For tech companies and data center developers, time is of the essence. Demand for new computing power has risen at such a rate that up to half of all new AI servers could be underpowered by 2027. Most new natural gas and nuclear power plants aren’t scheduled to come online until several years after that. But renewables can start supplying power quickly, with utility-scale solar projects starting to produce electrons in about 18 months. That speed has proven attractive, leading to some massive deals: Microsoft, for example, signed a deal with Brookfield Asset Management last summer for 10.5 gigawatts of renewable capacity in the U.S. and Europe, all of which will be delivered by 2030.

In Brief Elon Musk-owned social media platform X is suing the Indian government, accusing the country’s IT ministry of unlawfully expanding its powers to remove online content, Reuters reported Thursday. X has long chaffed at government orders to remove or block content in the country, such as during the farmers’ protests last year. Though it ended up complying with those executive orders, citing the risk of substantial penalties, it did also file a legal challenge. Now it’s further stepping up opposition to state censorship efforts via fresh litigation. According to a court filing X made earlier this month, the company has accused the Indian government of creating an unlawful mechanism through which “countless” public officials can execute content takedown or blocking orders via a website that social media companies should regularly check. Per Reuters, X’s lawsuit alleges that this mechanism fails to meet Indian legal safeguards on content removal that require such orders “to be issued in cases such as harm to sovereignty or public order, and … with strict oversight of top officials”. X claims the website creates “an impermissible parallel mechanism” that causes “unrestrained censorship of information in India”, according to the news agency’s report of its filing. The company is hoping to use a legal route to quash the directive, it said. A brief hearing on the case was held earlier this month, Reuters also reported, but no ruling has been made. Another hearing is expected to take place in the High Court of southern Karnataka state next week.

Tesla is issuing a recall for around 46,000 Cybertrucks sold to date because of an exterior steel trim panel on the side of the windshield that can peel off. The National Highway Traffic Safety Administration (NHTSA) said in a filing that the piece can come off while in motion, which can create a hazard for the driver and other people on the road. The filing states that Tesla became aware of the problem in early January and has identified 151 warranty claims related to this panel coming off since then. The company is not aware of any resulting crashes or injuries, according to the filing. The problem with the panel occurs because the “structural adhesive” that holds it in place on the vehicle is “susceptible to environmental embrittlement,” according to the filing. Tesla will fix Cybertrucks covered by the recall at no cost to owners by using a “different structural adhesive not prone to environmental embrittlement.” It will also use a nut to “clamp” the steel panel to the vehicle structure. The new recall is the eighth for Tesla’s Cybertruck — a vehicle which is outselling other electric trucks but has underperformed Tesla CEO Elon Musk’s hopes for its success. Many of its past problems have been fixed using an over-the-air software update, but some have required owners to bring their EVs in for service. Two of the recalls that required service have been related to pieces of the Cybertruck coming loose. The first, in April 2024, happened when owners noticed the cover of the throttle pedal was coming loose and sliding up, potentially causing a situation where it could get stuck at full throttle. The second, in June 2024, had to do with a trim piece on the truck’s bed that was prone to coming loose because of insufficient adhesive. That history with adhesive failures did not seem to raise a ton of red flags at Tesla this time around, though, according to the filing. After Tesla was alerted to the first complaint about the windshield panel, it began an “engineering study” to investigate the problem. On February 6, the company closed the study after an “inspection and pull test” found “no detections of separation.” It was only after NHTSA reached out to Tesla on February 21 with a customer complaint about the problem that Tesla dug deeper into social media posts and service records, according to the filing. It agreed to recall the Cybertrucks on March 11.

Hackers are ramping up their attempts to exploit a trio of year-old ServiceNow vulnerabilities to break into unpatched company instances, security researchers warned this week. Threat intelligence startup GreyNoise said in a blog post on Tuesday that it had observed a “notable resurgence of in-the-wild activity” targeting the three ServiceNow vulnerabilities, tracked as CVE-2024-4879, CVE-2024-5178, and CVE-2024-5217. The vulnerabilities were first disclosed by researchers at Assetnote in May 2024 and patched by ServiceNow months later in July 2024. GreyNoise said that all three flaws have seen a resurgence in targeted exploitation attempts in the past week. It’s not known exactly who is behind this latest wave of targeting, but GreyNoise said that 70% of the malicious activity it observed in the past week targeted systems based in Israel, with activity also seen in Germany, Japan, and Lithuania. As first noted by Assetnote last year, GreyNoise also confirms that the vulnerabilities can be chained together for “full database access” of affected ServiceNow instances. Organizations often use the ServiceNow platform to host sensitive data about their employees, including their personally identifiable information and HR records related to their employment. ServiceNow spokesperson Erica Faltous told TechCrunch that the company first learned of the vulnerabilities “nearly a year ago”, and, “to date, we have not observed any customer impact from an attack campaign.” Following Assetnote’s disclosure of the flaws last year, U.S. security firm Resecurity warned that foreign threat actors had attempted to exploit the three ServiceNow vulnerabilities to target both private sector companies and government agencies around the world. Resecurity said it saw targeted attempts at an energy company, a data center organization, a Middle Eastern government agency, and a software developer. Cybersecurity company Imperva released another report in July 2024 warning that it had also observed exploitation attempts across 6,000 sites across various industries, with a focus on the financial services sector.

Nvidia announced Thursday it’s partnering with EPRI, a power industry R&D organization, to use AI to solve problems facing the the electrical grid. Perhaps ironically, the issues are largely caused by rising power demand from AI itself. The Open Power AI Consortium, which includes a number of electrical utilities and tech companies, says it will use what are known as domain-specific AI models to devise new ways to tackle problems that the power industry is predicted to face in the coming years. The models will be open sourced and available to researchers across academia and industry. The power industry is facing surging demand from data centers in the United States and elsewhere as AI ramps up the need for computing power. Electricity demand is expected to grow by 4% annually in the coming years, according to the International Energy Agency, nearly double over 2023 figures. In addition to Nvidia and EPRI, the consortium includes PG&E, Con Edison, Constellation Energy, Duke Energy, the Tennessee Valley Authority and ENOWA, NEOM’s energy and water company. On the tech side, Microsoft and Oracle are both members. In an attempt to stay ahead of the trend, tech companies have been racing to secure generating capacity as power has transformed from a simple line item to a competitive advantage. Over the last year or so, tech companies have been consistently inking new contracts. They’ve largely been spread across renewable energy projects, spurred mostly by solar’s low cost, modularity, and the speed at which it can be deployed. Microsoft, for example, recently added 475 megawatts of solar power to its sizable renewable portfolio. Last year, it became an anchor investor in a $9 billion renewable development project run by Acadia, and earlier in the year it said it was working with Brookfield asset management to deploy 10.5 gigawatts of renewable power in the U.S. and Europe, all of which is expected to come online by 2030. But even though new power sources may be the most obvious answer to losing power shortages, they aren’t the only one. One recent study found that by curtailing use when demand on the grid peaks, including shifting tasks that aren’t time sensitive to periods when demand is low, an additional 76 GB of capacity could be unlocked in the U.S. It’s a not insignificant amount, making up approximately 10% of peak demand in the U.S. It’s likely those are the sorts of solutions, among others, that this new consortium will be exploring.

Google is rolling out a new Gmail update that is designed to help you find the email you’re looking for more quickly. The company announced on Thursday that it will now use AI to consider factors like recency, most-clicked emails, and frequent contacts when surfacing emails based on your search query. Up until now, Gmail has simply displayed emails in chronological order based on keywords. “With this update, the emails you’re looking for are far more likely to be at the top of your search results — saving you valuable time and helping you find important information more easily,” the company wrote in a blog post. Google is also introducing a new toggle so people can switch between “Most relevant” or “Most recent” emails on a search results page. The toggle is aimed at users who prefer seeing search results displayed in chronological order, rather than the new “Most relevant” default option. Image Credits:Google The update is rolling out globally for users with personal Google accounts and is available on the web and in the Gmail app for Android and iOS. Google plans to expand the feature to business users in the future. The launch of the new search functionality comes as Google has been building out its email offering to better compete with Apple’s Mail app, which got a slew of Gmail-like features with iOS 18 last year. For instance, Gmail recently gained a Gemini-powered feature that lets users add events to a Google Calendar directly from an email. A few months ago, Gmail rolled out “Summary cards” that allow users to take actions in their inbox, like tracking packages, checking in for flights, setting reminders, marking bills as paid, and more. Gmail also introduced the ability for users to chat with Gemini about their inbox directly within the app on both iOS and Android.

In Brief CoreWeave said on Wednesday that it hopes to sell its initial public shares between $47 and $55 per share, putting the money it may raise between about $2.2 billion and $2.6 billion. Word on the street is that CoreWeave wanted to raise at least $3 billion and possibly over $4 billion. But price range announcements are always a bit of a game. The announced range is often lower than the bankers think the company can really get. Then the company officially sets a higher price. Doing that helps build excitement for the first day of trading, saying to retail investors that big institutions were so interested that they were willing to pay more. So we’ll have to wait and see if CoreWeave can get over $3 billion. Since publicly filing to go public, the company signed a $12 billion customer agreement with OpenAI and added Meg Whitman to its board of directors.

Anthropic’s AI-powered chatbot, Claude, can now search the web — a capability that had long eluded it. Web search is available now in preview for paid Claude users in the U.S., Anthropic said on its blog, with support for free users and additional countries coming soon. Users can toggle on web search in their profile settings from the Claude web app, and Claude will automatically search across sites to inform certain responses. For now, web search only works with the latest Anthropic model powering Claude, Claude 3.7 Sonnet, Anthropic said. “When Claude incorporates information from the web into its responses, it provides direct citations so you can easily fact-check sources,” the company wrote in its blog post. “Instead of finding search results yourself, Claude processes and delivers relevant sources in a conversational format. This enhancement expands Claude’s extensive knowledge base with real-time insights, providing answers based on more current information.” In my brief testing of the feature, web search didn’t consistently trigger for current events-related questions. But when it did, Claude indeed delivered an answer with inline citations, pulling from sources including social media (e.g. X) and new sources like NPR and Reuters. Searching through web sources with Anthropic’s Claude chatbot. Image Credits:Anthropic Claude’s ability to search the web brings it to feature parity with most rival AI-powered chatbots, including OpenAI’s ChatGPT, Google’s Gemini, and Mistral’s Le Chat. Anthropic’s argument against it, previously, was that Claude was “designed to be self-contained.” No doubt competitive pressure had something to do with the reversal in course. Of course, the risk is that Claude hallucinates or miscites web sources. Other chatbots suffer from this. According to a recent study from the Tow Center for Digital Journalism, popular chatbots including ChatGPT and Gemini provide incorrect answers to more than 60% of questions. A separate report from The Guardian found that ChatGPT’s search-focused experience, ChatGPT Search, can be fooled into generating completely misleading summaries.

OpenAI is bringing new transcription and voice-generating AI models to its API that the company claims improve upon its previous releases. For OpenAI, the models fit into its broader “agentic” vision: building automated systems that can independently accomplish tasks on behalf of users. The definition of “agent” might be in dispute, but OpenAI Head of Product Olivier Godemont described one interpretation as a chatbot that can speak with a businesses’ customers. “We’re going to see more and more agents pop up in the coming months” Godemont told TechCrunch during a briefing. “And so the general theme is helping customers and developers leverage agents that are useful, available, and accurate.” OpenAI claims that its new text-to-speech model, “gpt-4o-mini-tts,” not only delivers more nuanced and realistic-sounding speech but is more “steerable” than its previous-gen speech-synthesizing models. Developers can instruct gpt-4o-mini-tts on how to say things in natural language — for example, “speak like a mad scientist” or “use a serene voice, like a mindfulness teacher.” Here’s a “true crime-style,” weathered voice: And here’s a sample of a female “professional” voice: Jeff Haris, a member of the product staff at OpenAI, told TechCrunch that the goal is to let developers tailor both the voice “experience” and “context.” “In different contexts, you don’t just want a flat, monotonous voice,” Harris continued. “If you’re in a customer support experience and you want the voice to be apologetic because it’s made a mistake, you can actually have the voice have that emotion in it […] Our big belief, here, is that developers and users want to really control not just what is spoken, but how things are spoken.” As for OpenAI’s new speech-to-text models, “gpt-4o-transcribe” and “gpt-4o-mini-transcribe,” they effectively replace the company’s long-in-the-tooth Whisper transcription model. Trained on “diverse, high-quality audio datasets,” the new models can better capture accented and varied speech, OpenAI claims, even in chaotic environments. They’re also less likely to hallucinate, Harris added. Whisper notoriously tended to fabricate words — and even whole passages — in conversations, introducing everything from racial commentary to imagined medical treatments into transcripts. “[T]hese models are much improved versus Whisper on that front,” Harris said. “Making sure the models are accurate is completely essential to getting a reliable voice experience, and accurate [in this context] means that the models are hearing the words precisely [and] aren’t filling in details that they didn’t hear.” Your mileage may vary depending on the language being transcribed, however. According to OpenAI’s internal benchmarks, gpt-4o-transcribe, the more accurate of the two transcription models, has a “word error rate” approaching 30% for Indic and Dravidian languages like Tamil, Telugu, Malayalam, and Kannada. That means that the model misses around three out of every 10 words in those languages. The results from OpenAI’s internal speech recognition benchmarks. Image Credits:OpenAI In a break from tradition, OpenAI doesn’t plan to make its new transcription models openly available. The company historically released new versions of Whisper for commercial use under an MIT license. Harris said that gpt-4o-transcribe and gpt-4o-mini-transcribe are “much bigger than Whisper” and thus not good candidates for an open release. “[T]hey’re not the kind of model that you can just run locally on your laptop, like Whisper,” he continued. “[W]e want to make sure that if we’re releasing things in open source, we’re doing it thoughtfully, and we have a model that’s really honed for that specific need. And we think that end-user devices are one of the most interesting cases for open-source models.”

Meta’s social network Threads, an X competitor with more than 320 million monthly active users, is now rolling out new features related to topics, reply controls, and video playback in the app, among other things. It’s also making political content suggestions more personalized to the user, the company says. Combined, the features aim to make the app more relevant to users and keep them engaged with the network’s content, at a time when X is no longer the only competitor Threads faces. Today, Meta faces the rise of alternative networks like Bluesky, which now has over 33 million users and a platform that’s inspiring rivals to Instagram, as well. To help Threads users better connect, users will now be able to add up to 10 topics they’re interested in to their profile on the social network. Tapping on those topics will then show you related posts, which makes it easier to engage with others on the social network who share your same interests. The company began testing the feature to add interest to profiles earlier this month, TechCrunch reported. The feature is still being described as a “test” with its broader launch today. Image Credit: Threads When composing a new post, users will now see prompts to add a suggested trending topic to your post, if relevant, or a topic related to other posts you’ve previously shared. These topic tag suggestions aim to help posts reach an audience who’s also interested in that subject, Meta explains. Threads is also introducing a redesigned topic tag, which looks like a pill and is found next to the posts on the For You feed. Users will be able to quickly tap on the tag to see all the posts tagged with that particular topic. Meta claims posts with tagged topics “generally receive more views” than those without, but didn’t share any specific data to back up that statement. Image Credits: Threads Another new feature allows you to manage how people interact with your posts and the posts you see. Now, you’ll be able to allow only your followers to reply to or quote a particular post. The addition builds on the quote control Threads launched back in November 2024 that lets you limit quote posts to anyone, profiles you follow, or those you’ve mentioned only. Image Credit: Threads Also in November, Threads began to experiment with allowing users to pick a default feed that’s not the algorithmic feed after rolling out custom feeds to compete with the social network Bluesky. Now, Threads will allow users to shuffle the order of their custom feeds. That means you can even make a custom feed the default that displays when you first open the app — a feature users have been asking for. Another update improves video playback on Threads. Along with the other changes, Meta is launching a redesigned video player that has easy-to-access pause/play and skip controls along with a pinned progress bar to easily scrub through the video. Image Credits: Threads Notably, Meta also said that it’s started phasing back civic and political content into the Threads in a more personalized way. The company controversially decided in 2024 to stop proactively recommending political content to users on Instagram and Threads, ahead of a critical U.S. presidential election. After the election wrapped, with President Trump in power, Meta announced it would then begin to phase back in political content on Facebook, Instagram and Threads. Meta CEO Mark Zuckerberg also dismantled the company’s fact-checking program after the election, choosing to rely only on community moderation efforts going forward. As a part of those efforts, the company said it would begin to re-add political content to users’ feeds, based on what Meta determines they want to see. In other words, it aims to re-create the echo chamber where the user’s worldview is reinforced without exposure to the other side. Meta says it will continue to listen to user feedback as it further develops Threads. Additional reporting: Sarah Perez

As Generative AI moves increasingly into the physical world through robotics, vehicles and drones, the need to map cloud-based AIs to real-world environments can lead to a huge mismatch. Rerun, a Stockholm-based dev tools startup, which builds a data stack for this emergent kind of “Physical AI”, has secured $17 million in seed funding to help AI developers smooth out the bumps from code to road, as it were. The seed round was led by Point Nine capital out of Germany, bringing the 2022-founded startup’s total funding to date to $20.2 million. Nikolaus West, Co-Founder and CEO of Rerun, told TechCrunch that AI which operates in the physical world is already booming. Per West, Rerun’s database and cloud data platform is built from the ground-up for Physical AI data and workflows, managing multimodal data such as video streams, 3D scenes, and tensors. The startup’s multimodal data stack supports visual debugging, helping developers to track back through their bots’ movements to better understand where things might be going wrong — importing data from real world sensors to power spatial visualization and analytics. “The data infrastructure that exists for AI does not really understand physical AI data. And old-school robotics tools are not built for the machine learning world, so the data breaks, and trying to use both these things together creates a huge amount of friction for the teams,” West explained. He suggests there is a potential “data flywheel” effect in creating AI geared to the physical world. The smarter the system is, the more robotics units can be deployed, which in turn means more data to train the AI model on, ensuring better outcomes, per West. But to get there the startup’s conviction is that AI developers need dedicated tools to help their models move smoothly through the real world. Visualizations of Physical AI data created by Rerun’s open source platform have been incorporated into other open source projects by companies such as Meta, Google, Hugging Face, and Unitree. The Rerun team is drawn from veterans from the likes of Apple, AWS, Meta, Unity, Zenly and Zipline. CTO, Emil Ernerfeldt, is also the creator of egui, the biggest open source GUI framework in Rust. Commenting in a statement, Ricardo Sequerra Amram, a partner at Point Nine, said Rerun’s strength in open source led to the investment — which is a notably large seed round for a European startup. “Rerun’s work on open source has allowed them to gain the trust of some of the most ambitious companies in the world of Physical AI,” he suggested. Sunflower Capital also participated in Rerun’s seed round, along with existing investors Costanoa Ventures and Seedcamp.

Social network Bluesky, an open source X rival, has made a change that will allow publishers to better track the referral traffic originating from posts on its platform. According to a comment from Bluesky employee Emily Liu, the company is now sending referral traffic through Bluesky’s “go” subdomain to make it easier for publishers to track when visitors to their site came by way of a link shared on the social network. When you click on a link shared on Bluesky, you’ll notice that it briefly displays as a “go.bsky.app” URL before landing you on the publisher’s own domain. This allows publishers’ analytics systems to determine that the visitor came from Bluesky. Liu notes in a post on Bluesky that the change came about because newsrooms were seeing engagement from likes and reposts in the app but had a harder time determining how that was translating into “actual website traffic.” Publishers are an important demographic for Bluesky to attract, especially if it aims to become a source of real-time news, similar to X. A number of news outlets already have been reporting increases in Bluesky referral traffic, with some even finding that Bluesky surpassed referrals from other platforms, like Meta’s Threads and Elon Musk’s X. For instance, The Boston Globe reported in November 2024 that it was seeing 3x more traffic from Bluesky, compared with Threads, and 4.5x higher conversions to paid subscriptions. Also in November, Guardian Australia editor Dave Early shared that Bluesky traffic to The Guardian’s website was already 2x that of Threads, and Bluesky traffic to its posts was higher than X’s referral traffic in 2024. The New York Times as well as other smaller publishers have also seen higher user engagement with Bluesky. Bluesky COO Rose Wang explained at the time that referrals from Bluesky were up because “unlike other platforms, we don’t de-promote your links.” Her message was a dig at both Threads and X, the latter of which changed the way that links appeared in its app to encourage more people to post to X directly, instead of off-site. (X owner Musk confirmed that X’s algorithm was designed to optimize the time spent on X, which is why links didn’t get “as much attention.”) Market intelligence provider Similarweb noted that Bluesky generated 38.6 million outgoing visits in November 2024, spread across a variety of news websites, while Threads generated 24.5 million referrals. However, the firm pointed out that 42% of Threads’ traffic was directing users to its sister app, Instagram, not publishers’ websites. TechCrunch asked Bluesky when exactly it made the change to shift links through its “go” subdomain, as it’s not clear based on Liu’s post. The company has not responded to our request for comment at this time. However, the company last November decided to not move forward with another solution to the referral problem that included an optional opt-out and Liu said earlier this month that the referer issue had been fixed with the latest Bluesky update rolled out on March 10, 2025. Liu also replied to a post on Wednesday to add that the referral feature just got another bug fix on Monday night to address other issues.

Open-source data movement company Airbyte is launching additional connectors to help enterprises better utilize their data in the age of AI without compromising data sovereignty. The San Francisco-based startup announced on Thursday that it’s releasing a host of new capabilities designed to enable customers to securely move corporate data without tapping SaaS applications. The new features include support for transferring unstructured data out of applications like Google Drive and SharePoint and compatibility with Apache Iceberg, an open-source format for large analytic tables. Airbyte is also launching a connector bundle for enterprise customers that includes data connector pipelines for applications such as NetSuite, SAP, ServiceNow, and Workday. Michel Tricot, the co-founder and CEO of Airbyte, said that the beauty of these new features is that they give enterprises additional ways to extract and utilize their internal data, including for AI applications on-premise. They also provide an alternative to AI services that might put customers at risk of having their data exposed to third parties, he continued. Screenshot of Airbyte. “Fortune 500s like Airbyte because they can have the flexibility and the efficiency of a tool, a piece of infrastructure, but also, they have full control over where the data is moving and from what system,” Tricot said. “They are the only ones actually looking at the pipes. No one else can look at the pipes beside[s] them.” Tricot added that because Airbyte customers have full visibility into these data pipelines, they can go in and remove sensitive knowledge, like employee compensation information, from the data before it reaches its final destination. “One thing I like to say is, don’t give away your first-party data for intelligence,” Tricot said. “That is a bad trade. I think a lot of execs and companies are aware of that. There is so much noise about, oh, I don’t want this [AI model] to have access to my data when I’m chatting with the [model], and it makes sense, because now people are just putting their most sensitive data into this system. So we want to make sure that it’s kept protected within […] their infrastructure.” Tricot thinks support for Iceberg is one of the highlights of the company’s new feature set. Iceberg gives companies the ability to move their data into a data lakehouse, he said, creating a single “source of truth” that works with numerous applications. “[Iceberg is] compatible with Databricks, it’s compatible with Bitquery, it’s compatible with Snowflake, and it’s compatible with new AI apps,” Tricot said. “So it’s really being compatible with this portable schema.” Airbyte was founded in 2020 by Tricot and Jean Lafleur, the company’s COO. Airbyte has more than 7,000 enterprise customers, including Monday.com, Invesco, and Calendly, and around 250,000 installations. The startup has raised more than $181 million in venture capital from firms such as Coatue, Accel, and Benchmark, among many others.

TechCrunch Sessions: AI kicks off on June 5 at Zellerbach Hall at UC Berkeley — and you have the power to decide who you want to see lead an impactful breakout session. We have narrowed down the impressive number of speaker applications to six incredible finalists. Now, it’s up to you to choose who you want to see share their wisdom with 1,200 fellow AI leaders and enthusiasts! You can take part in Audience Choice voting through March 21 at 11:59 p.m. PT. You can only choose one speaker, so make your vote count. Don’t forget to lock in your ticket now to save up to $210 and guarantee your seat in the most-voted session and top AI discussions. Explore cutting-edge innovations, connect with industry leaders, and dive into the AI revolution! And your finalists are… Image Credits:TechCrunch Guide to Monetizing GenAI in Enterprise Applications – Mahesh Chayel, Product Management Lead, Meta Discover how to monetize generative AI in enterprise apps — from API licensing to premium features and more — during this session from Meta’s Mahesh Chayel. This talk will help you better understand ethical issues, pinpoint valuable use cases, and create lasting AI revenue streams. Who Is Coding Our Future? – Cristina Mancini, CEO, Black Girls Code How can we create a more equitable, innovative future for all? Women make up just 27% of tech roles, and Black women make up less than 2%. Cristina Mancini of Black Girls Code will lead a discussion about how greater inclusion can reshape tech. Behind Your Firewall: Secure Generative AI for Regulated Enterprises – Yann Stoneman, Staff Solutions Architect, Cohere AI Learn how to deploy secure, private AI in healthcare or finance without the cloud during this session from Cohere AI’s Yann Stoneman. You’ll see real-world use cases and demos, get expert tips, and learn how to manage data privacy while building AI solutions via a Q&A. The AI Policy Playbook: What Startups Need to Know – Hua Wang, Executive Director, Global Innovation Forum While AI is transforming startups, policy and global expansion remain challenging. Explore AI-driven tools for digital trade, compliance, and scaling during this session from Global Innovation Forum Executive Director Hua Wang — and gain insights on policies, data regulations, and international growth. Implementing Secure Generative AI with Guardrails – Hardik Vasa, Senior Solutions Architect, AWS Learn how to safeguard your generative AI with Amazon Bedrock Guardrails. Choose this session from AWS’ Hardik Vasa and discover how to block harmful content, filter responses, and protect sensitive information to ensure secure, compliant AI interactions for your business. A Big Tech Banker-turned-Investor’s Unique Lens on Identifying Exceptional AI Founders – Marcie Vu, Partner, Greycroft How do you find standout AI founders and actionable strategies for building enduring AI companies that go beyond the hype? This session from Greycroft partner Marcie Vu will help you cut through the noise and find founders that can have true market success. Which session do you want to see the most? Vote today for their chance to take the stage at TechCrunch Sessions: AI!

If you’re reading this, you already know: It’s time to amplify your brand. Get in front of thousands of TechCrunch event attendees, readers, and decision-makers who can take your business further this year by exhibiting at a TechCrunch event. The perks of exhibiting: What’s in it for you Exhibiting doesn’t just elevate your brand — it gives you and your team full access to all the perks of an attendee at whichever event you exhibit: TechCrunch Sessions: AI, TechCrunch All Stage, or TechCrunch Disrupt 2025. Full-day exhibit space (plus every day at TechCrunch Disrupt). Your brand is featured on-site, online, and in the event app. A bundle of tickets for your team. Access to high-value leads. And more! Your time to shine begins now. With distinct audiences and benefits at each event, find the one that’s right for you. Space is very limited at each event, so don’t miss your chance! TechCrunch Sessions: AI Building the next big AI breakthrough? Showcase your innovation to 1,200+ hungry AI pioneers, VCs, and enthusiasts at TC Sessions: AI on June 5 at UC Berkeley’s Zellerbach Hall. Learn more and reserve your exhibit table for TC Sessions: AI today — space is extremely limited! TechCrunch All Stage Does your brand provide the support and solutions startups need to thrive? Help take a startup to the next level by showcasing at TC All Stage on July 15 at SoWa Power Station in Boston. This is where 1,200+ founders and VCs gather to exchange insights and discover the next big opportunity. Secure your exhibit table for TC All Stage now before space runs out. With only eight tables available, this event offers the most limited exhibit opportunities. Get started today! TechCrunch Disrupt 2025 Disrupt is the global epicenter of the tech industry. If you’re looking to make a massive impact on a wide-ranging tech audience, this is the place to be. From October 27-29 at Moscone West in San Francisco, 10,000+ attendees will gather to explore innovations, gain invaluable insight, and create meaningful connections in AI, fintech, startups, space, and beyond. Exhibit front and center for all three days in front of thousands of industry leaders and TechCrunch’s global audience. Plus, enjoy exclusive founder benefits like access to the Deal Flow Cafe, a networking hotspot for investors and founders, and get the VC list. Discover more about exhibiting and all the perks here. Book your table early — this is a first come, first served opportunity. Amplify your brand before space runs out Seize the opportunity to showcase your innovation before all tables are taken. Each event offers unique exhibitor perks tailored to maximize value for your startup. Take the first step — select the event you’d like to exhibit at:

Pruna AI, a European startup that has been working on compression algorithms for AI models, is making its optimization framework open source on Thursday. Pruna AI has been creating a framework that applies several efficiency methods, such as caching, pruning, quantization, and distillation, to a given AI model. “We also standardize saving and loading the compressed models, applying combinations of these compression methods, and also evaluating your compressed model after you compress it,” Pruna AI co-fonder and CTO John Rachwan told TechCrunch. In particular, Pruna AI’s framework can evaluate if there’s significant quality loss after compressing a model and the performance gains that you get. “If I were to use a metaphor, we are similar to how Hugging Face standardized transformers and diffusers — how to call them, how to save them, load them, etc. We are doing the same, but for efficiency methods,” he added. Big AI labs have already been using various compression methods already. For instance, OpenAI has been relying on distillation to create faster versions of its flagship models. This is likely how OpenAI developed GPT-4 Turbo, a faster version of GPT-4. Similarly, the Flux.1-schnell image generation model is a distilled version of the Flux.1 model from Black Forest Labs. Distillation is a technique used to extract knowledge from a large AI model with a “teacher-student” model. Developers send requests to a teacher model and record the outputs. Answers are sometimes compared with a dataset to see how accurate they are. These outputs are then used to train the student model, which is trained to approximate the teacher’s behavior. “For big companies, what they usually do is that they build this stuff in-house. And what you can find in the open source world is usually based on single methods. For example, let’s say one quantization method for LLMs, or one caching method for diffusion models,” Rachwan said. “But you cannot find a tool that aggregates all of them, makes them all easy to use and combine together. And this is the big value that Pruna is bringing right now.” Left to right: Rayan Nait Mazi, Bertrand Charpentier, John Rachwan, Stephan Günnemann Image Credits:Pruna AI While Pruna AI supports any kind of models, from large language models to diffusion models, speech-to-text models and computer vision models, the company is focusing more specifically on image and video generation models right now. Some of Pruna AI’s existing users include Scenario and PhotoRoom. In addition to the open source edition, Pruna AI has an enterprise offering with advanced optimization features, including an optimization agent. “The most exciting feature that we are releasing soon will be a compression agent,” Rachwan said. “Basically, you give it your model, you say: ‘I want more speed but don’t drop my accuracy by more than 2%.’ And then, the agent will just do its magic. It will find the best combination for you, return it for you. You don’t have to do anything as a developer.” Pruna AI charges by the hour for its pro version. “It’s similar to how you would think of a GPU when you rent a GPU on AWS or any cloud service,” Rachwan said. And if your model is a critical part of your AI infrastructure, you’ll end up saving a lot of money on inference with the optimized model. For example, Pruna AI has made a Llama model eight times smaller without too much loss using its compression framework. Pruna AI hopes its customers will think about its compression framework as an investment that pays for itself. Pruna AI raised a $6.5 million seed funding round a few months ago. Investors in the startup include EQT Ventures, Daphni, Motier Ventures, and Kima Ventures.

Apple is reportedly shaking up its executive ranks, shortly after it delayed the personalized Siri features that it previewed at WWDC last year. Bloomberg's Mark Gurman today said that Vision Pro chief Mike Rockwell will be taking over the Siri team at Apple, which until now was led by Apple's artificial intelligence chief John Giannandrea. Apple CEO Tim Cook has apparently "lost confidence" in Giannandrea's ability to "execute on product development," but he will remain at the company for now to oversee artificial intelligence research and development in general. Vision Pro hardware engineer Paul Meade will be succeeding Rockwell as the head of Apple's spatial computing efforts, allowing Rockwell to focus entirely on Siri, according to the report. In his new role overseeing the Siri team, Rockwell will report to Apple's software engineering chief Craig Federighi, the report added. Apple plans to inform employees about the executive shakeup this week, according to the report. Apple has been facing a lot of criticism after delaying its more personalized version of Siri, with some people upset that the company chose to preview new features that were merely conceptual, rather than fully functional. It is a widely held belief that Apple is lagging behind competitors like OpenAI in the generative artificial intelligence space, and it looks like Apple is attempting to right the ship with this executive shakeup. Apple appointing Rockwell as its new Siri head is an interesting choice, as the Vision Pro has been described as a "commercial flop" and has faced criticism of its own. Nevertheless, the Vision Pro is certainly a technological feat, so Rockwell has proven experience and could be instrumental in improving Siri's underlying technologies. Apple said it anticipates rolling out the more personalized Siri "in the coming year." Even the current version of Siri has attracted a lot of criticism, including from some of the most devoted Apple fans, as it sometimes struggles to answer even the most basic questions. Rockwell certainly has his work cut out for him in his new role.

On Wednesday, Spotify launched a new personalized playlist designed to help listeners discover upcoming concerts in their area. The playlist, titled “Concerts Near You,” is updated every Wednesday with 30 songs from artists performing nearby, along with full event details and ticket links. Spotify partners with ticketing platforms like AXS, DICE, Eventbrite, and Ticketmaster. Like other personalized playlists on Spotify, such as “Daylist” and various “Mixes,” “Concerts Near You” is tailored to users’ listening habits, allowing them to discover artists they love. To receive the latest show updates, listeners should ensure they’re sharing their location through the “Live Events” section, found under the Search tab in the Spotify app. Spotify also recommends enabling push notifications and following favorite artists to receive personalized reminders. The launch of the new playlist signals the company’s ambition to make event discovery a more significant aspect of the overall Spotify listening experience while also providing artists more visibility. The announcement follows Spotify’s publication of its yearly Loud & Clear report, which provides detailed figures regarding its payments to counter claims that the platform doesn’t adequately compensate artists. For the first time, an artist who received one in every million streams on the platform earned an average of over $10,000 in 2024.

Over the last few years, developers have increasingly complained about paying fees to Apple or Google for App Store or Play Store distributions. App makers have often said that both companies make it difficult for them to earn money outside of their systems. New York-based app revenue management platform Adapty wants to change that with its new offering called FunnelFox, which helps acquire customers through websites or web apps. The idea behind FunnelFox is simple: It helps app makers create quizzes or experiences on the web to acquire or onboard customers. And once the customer pays, they get a link to download the app. This way, the app makers don’t have to pay up to 30% of the app store tax. “A year ago, I had a chat with our top customers of the revenue tracking product and realized that they acquire a big chunk of customers through the web by building acquisition funnels internally. We thought that this is a big opportunity and decided to build FunnelFox to make web funnels easier for everyone,” Adapty CEO Vitaly Davydov told TechCrunch over a call. Adapty has been working with a closed set of more than 100 customers for over a year to try out FunnelFox. The company said it has tracked $20 million in revenue during this phase, and customers have seen their revenue double as compared to direct revenue from mobile in-app purchases. FunnelFox has a drag and drop tool to build quizzes to acquire customers. It also has an AI component, which can suggest different questions based on the App Store or Play Store link of the app. The startup has two core parts of the business: an SDK that helps app makers with subscriptions and a platform for marketers to target audiences and understand the lifetime value of a customer. FunnelFox is the latest edition of its suite. Customers can use FunnelFox separately from Adapty while using another app revenue tracking solution, such as RevenueCat. FunnelFox works on a revenue sharing basis, with the company taking a cut on user acquisitions. Adapty plans tiered pricing for FunnelFox, but didn’t specify exact details. Adapty has raised $2.5 million in capital. Davydov said that the company is profitable. While the startup is not actively looking for funding, it is open to exploring fundraising. The company serves 10,000 applications and has tracked $1.2 billion in revenue per year. This is up from 2,500 apps it tracked in 2022. Davydov said the company is consistently growing both in terms of number of apps and revenue tracked. The startup wants to add more AI to its products. For Adapty, it wants to use AI to help app makers create payment screens that will give them better conversion rate by suggesting design and rates.

Apple Inc. is undergoing a rare shake-up of its executive ranks, aiming to get its artificial intelligence efforts back on track after months of delays and stumbles, according to people familiar with the situation. Chief Executive Officer Tim Cook has lost confidence in the ability of AI head John Giannandrea to execute on product development, so he’s moving over another top executive to help: Vision Pro creator Mike Rockwell . In a new role, Rockwell will be in charge of the Siri virtual assistant, according to the people, who asked not to be identified because the moves haven’t been announced.