RSS 생중계

Nvidia is facing backlash for allegedly manipulating the review process of its GeForce RTX 5060 GPU by withholding drivers, selectively granting early access to favorable reviewers, and pressuring media to present the card in a positive light. As The Verge's Sean Hollister writes, the debacle "should be a wake-up call for gamers and reviewers." Here's an excerpt from the report: Nvidia has gone too far. This week, the company reportedly attempted to delay, derail, and manipulate reviews of its $299 GeForce RTX 5060 graphics card, which would normally be its bestselling GPU of the generation. Nvidia has repeatedly and publicly said the budget 60-series cards are its most popular, and this year it reportedly tried to ensure it by withholding access and pressuring reviewers to paint them in the best light possible. Nvidia might have wanted to prevent a repeat of 2022, when it launched this card's predecessor. Those reviews were harsh. The 4060 was called a "slap in the face to gamers" and a "wet fart of a GPU." I had guessed the 5060 was headed for the same fate after seeing how reviewers handled the 5080, which similarly showcased how little Nvidia's hardware has improved year over year and relies on software to make up the gaps. But Nvidia had other plans. Here are the tactics that Nvidia reportedly just used to throw us off the 5060's true scent, as individually described by GamersNexus, VideoCardz, Hardware Unboxed, GameStar.de, Digital Foundry, and more: - Nvidia decided to launch its RTX 5060 on May 19th, when most reviewers would be at Computex in Taipei, Taiwan, rather than at their test beds at home. - Even if reviewers already had a GPU in hand before then, Nvidia cut off most reviewers' ability to test the RTX 5060 before May 19th by refusing to provide drivers until the card went on sale. (Gaming GPUs don't really work without them.) - And yet Nvidia allowed specific, cherry-picked reviewers to have early drivers anyhow if they agreed to a borderline unethical deal: they could only test five specific games, at 1080p resolution, with fixed graphics settings, against two weaker GPUs (the 3060 and 2060 Super) where the new card would be sure to win. - In some cases, Nvidia threatened to withhold future access unless reviewers published apples-to-oranges benchmark charts showing how the RTX 5060's "fake frames" MFG tech can produce more frames than earlier GPUs without it. Some reviewers apparently took Nvidia up on that proposition, leading to day-one "previews" where the charts looked positively stacked in the 5060's favor [...]. But the reality, according to reviews that have since hit the web, is that the RTX 5060 often fails to beat a four-year-old RTX 3060 Ti, frequently fails to beat a four-year-old 3070, and can sometimes get upstaged by Intel's cheaper $250 B580. And yet, the 5060's lackluster improvements are overshadowed by a juicier story: inexplicably, Nvidia decided to threaten GamersNexus' future access over its GPU coverage. Yes, the same GamersNexus that's developed a staunch reputation for defending consumers from predatory behavior, and just last month published a report on "GPU shrinkflation" that accused Nvidia of misleading marketing. Bad move! [...] Nvidia is within its rights to withhold access, of course. Nvidia doesn't have to send out graphics cards or grant interviews. It'll only do it if it's good for business. But the unspoken covenant of product reviews is that the press, as a whole, gets a chance to warn the public if a movie, video game, or GPU is not worth their money. It works both ways: the media also gets the chance to warn that a product is so good you might want to line up in advance. That unspoken rule is what Nvidia is trampling here.

Read more of this story at Slashdot.

An anonymous reader quotes a report from Ars Technica: Researchers have found malicious software that received more than 6,000 downloads from the NPM repository over a two-year span, in yet another discovery showing the hidden threats users of such open source archives face. Eight packages using names that closely mimicked those of widely used legitimate packages contained destructive payloads designed to corrupt or delete important data and crash systems, Kush Pandya, a researcher at security firm Socket, reported Thursday. The packages have been available for download for more than two years and accrued roughly 6,200 downloads over that time. "What makes this campaign particularly concerning is the diversity of attack vectors -- from subtle data corruption to aggressive system shutdowns and file deletion," Pandya wrote. "The packages were designed to target different parts of the JavaScript ecosystem with varied tactics." [...] Some of the payloads were limited to detonate only on specific dates in 2023, but in some cases a phase that was scheduled to begin in July of that year was given no termination date. Pandya said that means the threat remains persistent, although in an email he also wrote: "Since all activation dates have passed (June 2023-August 2024), any developer following normal package usage today would immediately trigger destructive payloads including system shutdowns, file deletion, and JavaScript prototype corruption." The list of malicious packages included js-bomb, js-hood, vite-plugin-bomb-extend, vite-plugin-bomb, vite-plugin-react-extend, vite-plugin-vue-extend, vue-plugin-bomb, and quill-image-downloader.

Read more of this story at Slashdot.

Microsoft is testing AI-powered text generation in Notepad for Windows 11 Insiders with Copilot Plus PCs, allowing users to draft content from prompts or build on existing text through a right-click menu. The update also introduces Paint's AI sticker generator and Snipping Tool enhancements including automatic screenshot cropping. The Write feature requires Microsoft account sign-in and uses the same credit system as other Windows 11 AI features.

Read more of this story at Slashdot.

Broadcom has upped VMware licensing costs by between eight to 15 times since it took over the organization, and a lack of alternatives in the tech industry means trade and end customers have no choice but to play ball. From a report: This is the according to the European Cloud Competition Observatory (ECCO), an independent body formed by customer organizations, and CISPE -- a trade association of 37 cloud providers in the region -- to monitor the behavior of software vendors accused of abusing their monopoly position. The report also calls for regulatory intervention. The current subscription model "creates a material risk for the company and their shareholders should Regulators investigate and challenge the legality of such model," the report adds.

Read more of this story at Slashdot.

China has achieved substantial technological advances across robotics, AI, and semiconductor manufacturing as part of a seven-year self-reliance campaign that has tripled the country's research and development spending to $500 billion annually. Chinese robot manufacturers captured nearly half of their domestic market by 2023, up from a quarter of installations just years earlier, while AI startups now rival OpenAI and Google in capabilities. The progress extends to semiconductors, where Huawei released a high-end smartphone powered by what industry analysts believe was a locally-produced advanced processor, despite U.S. export controls targeting China's chip access. Morgan Stanley projects China's self-sufficiency in graphics processing units will jump from 11% in 2021 to 82% by 2027. Chinese companies have been purchasing as many industrial robots as the rest of the world combined, enabling highly automated factories that can operate in darkness. In space technology, Chinese firms won five of 11 gold medals when U.S. think tanks ranked the world's best commercial satellite systems last year, compared to four for American companies.

Read more of this story at Slashdot.

Mozilla has announced that it is shutting down Pocket, a bookmarking service acquired by Mozilla in 2017, this coming July. "Pocket has helped millions save articles and discover stories worth reading. But the way people use the web has evolved, so we're channeling our resources into projects that better match their browsing habits and online needs."

Apple is aiming to release smart glasses at the end of next year as part of a push into AI-enhanced gadgets, Bloomberg News reported Thursday, but it has shelved plans for a smartwatch that can analyze its surroundings with a built-in camera. From the report: Company engineers are ramping up work on the glasses -- a rival to Meta Platforms's popular Ray-Bans -- in a bid to meet the year-end 2026 target, according to people with knowledge of the matter. Apple will start producing large quantities of prototypes at the end of this year with overseas suppliers, said the people, who asked not to be identified because the products haven't been announced.

Read more of this story at Slashdot.

BrianFagioli writes: In a surprising move that will frustrate longtime fans, Mozilla has announced it will shut down Pocket on July 8, 2025. The once-popular read-it-later service, which helped users save and organize web content for later reading, will no longer function as normal after that date. While existing users can continue saving and reading articles until July, the service will switch to export-only mode afterward, with all user data permanently deleted on October 8. The Firefox-maker will also shut down Fakespot, a service that allows users to identify unreliable reviews, on July 1.

Read more of this story at Slashdot.

An anonymous reader writes: The US Treasury is phasing out production of the penny and will stop putting new one-cent coins into circulation. The US Treasury has made its final order of penny blanks this month, and the mint will continue to manufacture pennies as long as its supply of penny blanks exist. President Donald Trump stated that production of pennies are wasteful, as the coins cost more to produce than their one-cent value.

Read more of this story at Slashdot.

Anthropic launched Claude Opus 4 and Claude Sonnet 4 today, positioning Opus 4 as the world's leading coding model with 72.5% performance on SWE-bench and 43.2% on Terminal-bench. Both models feature hybrid architecture supporting near-instant responses and extended thinking modes for complex reasoning tasks. The models introduce parallel tool execution and memory capabilities that allow Claude to extract and save key facts when given local file access. Claude Code, previously in research preview, is now generally available with new VS Code and JetBrains integrations that display edits directly in developers' files. GitHub integration enables Claude to respond to pull request feedback and fix CI errors through a new beta SDK. Pricing remains consistent with previous generations at $15/$75 per million tokens for Opus 4 and $3/$15 for Sonnet 4. Both models are available through Claude's web interface, the Anthropic API, Amazon Bedrock, and Google Cloud's Vertex AI. Extended thinking capabilities are included in Pro, Max, Team, and Enterprise plans, with Sonnet 4 also available to free users. The startup, which counts Amazon and Google among its investors, said Claude Opus 4 could autonomously work for nearly a full corporate workday -- seven hours. CNBC adds: "I do a lot of writing with Claude, and I think prior to Opus 4 and Sonnet 4, I was mostly using the models as a thinking partner, but still doing most of the writing myself," Mike Krieger, Anthropic's chief product officer, said in an interview. "And they've crossed this threshold where now most of my writing is actually ... Opus mostly, and it now is unrecognizable from my writing." Krieger added, "I love that we're kind of pushing the frontier on two sides. Like one is the coding piece and agentic behavior overall, and that's powering a lot of these coding startups. ... But then also, we're pushing the frontier on how these models can actually learn from and then be a really useful writing partner, too."

Read more of this story at Slashdot.

Our recent article on Home Assistant observed that the project emphasizes installations using its own Linux distribution or within containers. The project has now made that emphasis rather stronger with this announcement of the deprecation of the "core" and "supervised" installation modes, which allowed Home Assistant to be installed as an ordinary application on a Linux system.

These are advanced installation methods, with only a small percentage of the community opting to use them. If you are using these methods, you can continue to do so (you can even continue to update your system), but in six months time, you will no longer be supported, which I'll explain the impacts of in the next section. References to these installation methods will be removed from our documentation after our next release (2025.6).

Support for 32-bit Arm and x86 architectures has also been deprecated.

The Russian government has introduced a new law that makes installing a tracking app mandatory for all foreign nationals in the Moscow region. From a report: The new proposal was announced by the chairman of the State Duma, Vyacheslav Volodin, who presented it as a measure to tackle migrant crimes. "The adopted mechanism will allow, using modern technologies, to strengthen control in the field of migration and will also contribute to reducing the number of violations and crimes in this area," stated Volodin. Using a mobile application that all foreigners will have to install on their smartphones, the Russian state will receive the following information: Residence location, fingerprint, face photograph, real-time geo-location monitoring.

Read more of this story at Slashdot.

Verizon is officially asking for a waiver of the FCC's phone unlocking requirements. From a report: "Given the substantial and growing harms to consumers, competition and Verizon from this obligation -- and the lack of offsetting benefits -- the commission should waive this rule," the operator wrote. Verizon faces phone unlocking requirements stemming from its acquisition of 700MHz spectrum in 2008, and also from conditions the FCC placed on the operator's acquisition of prepaid provider TracFone in 2021. The requirements mean that when a customer buys a phone from Verizon it's locked to Verizon's network for 60 days, so that they can only use it with a Verizon SIM card. After 60 days, Verizon automatically unlocks the phone, allowing that customer to use their phone on another carrier's network.

Read more of this story at Slashdot.

The Fedora Council has ruled on the Fedora Engineering Steering Council's (FESCo) decision last year to revoke Peter Robinson's provenpackager status. In a statement published to the fedora-devel-announce mailing list, the council has announced that it has overturned FESCo's decision:

FESCo didn't have a specific policy for dealing with a request to remove Proven Packager rights. In addition, the FESCo process was handled entirely in private. The contributor didn't receive a formal notification or warning from FESCo, and felt blindsided by the official decision when and how it was announced. The Fedora Council would like to extend our sincerest apology on behalf of the Fedora Project to them.

LWN covered the story in December 2024.

Google's expansion of Gemini's data access through "personal context" represents a fundamental shift in how AI assistants operate. Unlike competitors that start from scratch with each new user, Gemini can immediately tap into years of accumulated user data across Google's ecosystem. The Verge adds: Google first started letting users opt in to its "Gemini with personalization" feature earlier this year, which lets the AI model tap into your search history "to provide responses that are uniquely insightful and directly address your needs." But now, Google is taking things a step further by unlocking access to even more of your information -- all in the name of providing you with more personalized, AI-generated responses. During Google I/O on Tuesday, Google introduced something called "personal context," which will allow Gemini models to pull relevant information from across Google's apps, as long as it has your permission. One way Google is doing this is through Gmail's personalized smart replies -- the AI-generated messages that you can use to quickly reply to emails. To make these AI responses sound "authentically like you," Gemini will pore over your previous emails and even your Google Drive files to craft a reply tailored to your conversation. The response will even incorporate your tone, the greeting you use the most, and even "favorite word choices," according to Google.

Read more of this story at Slashdot.

BrianFagioli writes: Signal has officially had enough, folks. You see, the privacy-first messaging app is going on the offensive, declaring war on Microsoft's invasive Recall feature by enabling a new "Screen security" setting by default on Windows 11. This move is designed to block Microsoft's AI-powered screenshot tool from capturing your private chats. If you aren't aware, Recall was first unveiled a year ago as part of Microsoft's Copilot+ PC push. The feature quietly took screenshots of everything happening on your computer, every few seconds, storing them in a searchable timeline. Microsoft claimed it would help users "remember" what they've done. Critics called it creepy. Security experts called it dangerous. The backlash was so fierce that Microsoft pulled the feature before launch. But now, in a move nobody asked for, Recall is sadly back. And thankfully, Signal isn't waiting around this time. The team has activated a Windows 11-specific DRM flag that completely blacks out Signal's chat window when a screenshot is attempted. If you've ever tried to screen grab a streaming movie, you'll know the result: nothing but black.

Read more of this story at Slashdot.

Testing filesystems is a frequent topic at the Linux Storage, Filesystem, Memory Management, and BPF Summit (LSFMM+BPF); the 2025 edition was no exception. Boris Burkov led a filesystem-track session to discuss stress-testing filesystems—and running those tests for lengthy periods. He reviewed what he has been doing when testing filesystems and wanted to gather ideas for what could be done to catch more bugs before the filesystems hit production.

OpenAI CEO Sam Altman told employees Wednesday the company plans to ship 100 million AI "companion" devices as part of what he called "the chance to do the biggest thing we've ever done as a company here," according to WSJ. Speaking at an internal meeting, Altman said the $6.5 billion acquisition of former Apple designer Jony Ive's startup has the potential to add $1 trillion in value to OpenAI. The pocket-sized, screen-free devices will be contextually aware of users' surroundings and designed to help wean people from traditional screens. Altman said the device will not be a phone or glasses, but rather a third core device that would sit on a desk alongside a MacBook Pro and iPhone. The company aims to ship the devices "faster than any company has ever shipped 100 million of something new before," with a target release of late next year.

Read more of this story at Slashdot.

Greg Kroah-Hartman has announced the release of the 6.14.8, 6.12.30, 6.6.92, 6.1.140, and 5.15.184 stable kernels. As usual, each contains a long list of important fixes throughout the kernel tree.

Security updates have been issued by AlmaLinux (kernel, kernel-rt, and webkit2gtk3), Fedora (mozilla-ublock-origin and sudo-rs), Oracle (.NET 8.0, compat-openssl10, grafana, osbuild-composer, redis:6, ruby:2.5, and webkit2gtk3), SUSE (dante, firefox-esr, gnuplot, govulncheck-vulndb, grype, postgresql13, postgresql14, postgresql15, postgresql16, postgresql17, python-tornado6, python314, thunderbird, ucode-intel, and xen), and Ubuntu (bind9, libfcgi-perl, linux-ibm-5.4, linux-oracle-5.4, postgresql-17, and Tomcat).