"The only reason we're still talking to these people is we need them and we need them now. The problem for these guys is they are that good." -- Defense official
Defense Secretary Pete Hegseth gave Anthropic CEO Dario Amodei until Friday to grant the military unfettered access to its AI model Claude or face penalties including being declared a "supply chain risk" or forced compliance under the Defense Production Act.
The Pentagon is escalating its dispute with Anthropic over AI safeguards, demanding the company remove restrictions on how the military uses its industry-leading Claude model. While the Pentagon wants to punish Anthropic for refusing to allow mass surveillance of Americans or development of autonomous weapons, officials are simultaneously concerned about losing access to Claude, which is currently the only AI model used in classified military systems. This standoff reflects a broader tension between national security demands and AI safety guardrails, with the Pentagon exploring alternatives like xAI's Grok, OpenAI, and Google's Gemini to reduce its dependence on Anthropic.
Summary not available
"Google says 3.1 Pro is ready for 'your hardest challenges.'" -- Google
The product: Gemini 3.1 Pro is Google's updated flagship AI model designed for improved problem-solving and reasoning capabilities. It is rolling out in preview for developers and consumers with enhanced abilities in generating graphics, simulations, and complex logical reasoning.
Cost: API cost for developers remains $2 input and $12 output per 1M tokens.
Availability: The updated model is available today in AI Studio and the Antigravity IDE in preview. Enterprise users will see 3.1 Pro in Vertex AI and Gemini Enterprise. Gemini 3.1 Pro is available for both the Gemini app and NotebookLM today.
"The whole supply chain is kind of strained. We're lucky, because we have our own TPUs, so we have our own chip designs. But it still, in the end, actually comes down to a few suppliers of a few key components." -- Demis Hassabis
Google DeepMind CEO Demis Hassabis said memory chip shortages are creating a "choke point" that constrains both AI deployment and research across the industry.
The global memory chip market is heavily supply-constrained, with only three suppliers—Samsung, Micron, and SK Hynix—dominating production. Even tech giants like Google, which manufactures its own TPUs (Tensor Processing Units), cannot escape the bottleneck because they still depend on external suppliers for critical memory components. This shortage is particularly acute for AI companies seeking high-bandwidth memory (HBM) chips, which differ from the memory chips PC manufacturers need, and the constraint is hampering both the deployment of existing AI models and the ability of researchers to experiment with new ones at scale.
"Starting in September 2026, Android will require all apps to be registered by verified developers in order to be installed on certified Android devices" -- Google (from official program description)
Google announced in August 2025 that starting in September 2026, all Android app developers will be required to register with Google, pay fees, provide government identification, and submit their private signing keys before their apps can be installed on certified Android devices.
This policy change represents a fundamental shift away from Android's foundational promise of being an "open" platform, according to the Keep Android Open campaign. The registration requirement threatens consumer choice by allowing Google to block apps from unverified developers, undermines independent developers' ability to distribute software directly to their communities, and concentrates digital sovereignty in the hands of a single corporation. The campaign argues that without evidence Google will allow users to bypass verification without friction, the policy effectively gives the company unilateral control over what software Android users can install.
OpenAI is developing its first hardware product—a smart speaker with a camera priced between $200 and $300—as the company expands into physical devices following its acquisition of Jony Ive's hardware company.
OpenAI's move into hardware represents a significant expansion of the AI company beyond software, joining competitors like Apple in developing AI-powered physical devices. The smart speaker, expected no earlier than March 2027, will feature object and conversation recognition along with facial recognition for purchases. OpenAI is also exploring smart glasses and a smart lamp, though these devices remain in early development stages with unclear release timelines, reflecting the broader industry trend of integrating AI capabilities into consumer hardware products.
Summary not available
Amazon has officially dethroned Walmart as the world's biggest company by revenue, with 2025 sales of $717 billion compared to Walmart's $713.2 billion for the 12 months ending January 31.
This milestone marks a significant shift in retail dominance, with Amazon's revenue growing at nearly 10 times the pace of Walmart's over the past decade. The e-commerce and cloud-computing giant, which began as an online bookseller in a Seattle garage in 1994, has surpassed the company that held the largest-by-revenue position for more than a decade. Amazon's growth has been fueled by changing consumer shopping patterns—a shift from physical stores to online retail—and the rapid expansion of its cloud-computing division, Amazon Web Services, demonstrating how diversified business models can outpace traditional retail operations.
"In other reports I've seen on this, the driver did actually call their supervisor and asked it they should follow their GPS and we're told in no uncertain terms that if the GPS said to follow that route they had to do it." -- nangua
An Amazon delivery van became stuck in tidal mudflats near Southend-on-Sea, Britain, after the driver followed GPS directions onto The Broomway, a dangerous route to Foulness Island known locally as the "deadliest footpath in Britain."
The incident sparked MetaFilter community discussion about over-reliance on GPS navigation and workplace pressure to follow automated directions without question. According to reports shared in the comments, the driver called his supervisor to question whether he should follow the GPS route onto the mudflats, but was allegedly told to comply with the GPS directions. The discussion highlights broader concerns about how GPS technology can override human judgment and how employment pressures may prevent workers from making safe decisions when technology directs them into dangerous situations.
"This acquisition represents a major step forward in our vision to deliver a safer, smarter home that understands context, recognizes risk, and provides peace of mind, all while protecting customer privacy. By integrating AI sensing into the home, we plan to add a new intelligence layer, elevating traditional security and smart home capabilities by allowing ADT to verify human presence, classify and analyze motion, and add context to alarm events and other conditions within the home." -- Omar Khan, ADT's executive vice president and chief business officer
ADT has acquired Origin Wireless, the company that invented Wi-Fi motion sensing technology, in a $170 million deal to expand its smart home security capabilities.
Origin Wireless developed algorithms that analyze how wireless RF signals bounce around a space to detect people and objects without additional sensors. ADT plans to integrate this AI sensing technology into its security systems to add context to alarm events, reduce false alarms, and potentially require fewer physical sensors during installation. The acquisition allows ADT to leverage existing Wi-Fi infrastructure while continuing to support existing partners like Philips Hue and WiZ that already use Origin Wireless' technology.
"Under the agreement the new partner will lead sales, marketing, and logistics across the region, while Panasonic provide expertise and quality assurance to uphold its renowned audiovisual standards with full joint development on top-end OLED models." -- Panasonic representative
Panasonic announced that Chinese company Skyworth will take over manufacturing, marketing, and selling Panasonic-branded TVs, marking the end of the Japanese electronics giant's direct TV production.
This deal represents a significant milestone in the decline of Japanese TV manufacturing. Panasonic, once the dominant plasma TV market leader with 40.7 percent market share in 2010, has been gradually exiting the TV business for over a decade—stopping plasma production in 2014, leaving the US market by 2016, and considering a complete exit as recently as February 2025. The Skyworth partnership means virtually no TV production remains in Japan, as other major Japanese manufacturers like Sharp, Toshiba, Hitachi, and Pioneer have already exited the industry, leaving South Korean and Chinese manufacturers to control the majority of global TV sales.
"For the foreseeable future, there will be people who play a role in the vehicles' behavior, and therefore have a safety role to play. One of the hardest safety problems associated with self-driving is building software that knows when to ask for human help." -- Philip Koopman, autonomous-vehicle software and safety researcher at Carnegie Mellon University
Waymo and Tesla have revealed new details about their "remote assistance" programs in government documents, disclosing how human workers help guide their self-driving vehicles when the technology encounters problems on public roads.
These remote assistance programs are critical to safety because self-driving cars regularly encounter situations their software cannot handle alone, such as traffic light outages or interactions with law enforcement. Industry experts emphasize that the humans supporting these robots play a safety-critical role, and if they make mistakes, it can result in crashes. The companies' previous reluctance to discuss these programs fueled conspiracy theories that robotaxis are simply remote-controlled cars, making the newly disclosed details important for public understanding of how autonomous vehicles actually operate.
"To use a medical phrase — we have stopped the bleeding. And while we know much more now than we did 24 hours ago, the extent and the scope of the intrusion is still not fully understood." -- LouAnn Woodward, UMMC vice chancellor for health affairs
The University of Mississippi Medical Center has closed all 35 of its clinics statewide after a ransomware attack launched Thursday compromised its phone and electronic systems, including its electronic health records platform Epic.
The cyberattack has disrupted patient care across the state, forcing the cancellation of appointments including chemotherapy and elective procedures. While UMMC hospitals and emergency departments remain operational, the medical center has taken all systems offline as a precaution and is working with law enforcement, including the FBI, to determine the full extent of the intrusion and whether patient information was compromised. Staff are currently using paper documentation to continue time-sensitive care while cybersecurity specialists work to restore the systems.
"There's a human foible, human frailty involved in this — in terms of building human confidence based upon this machine's ability to inform that confidence, so the human is willing to push the Big Red Button." -- Chris Inglis, former U.S. national cyber director
Former senior U.S. cybersecurity officials warn that artificial intelligence is accelerating the prospect of catastrophic cyberattacks powerful enough to shut down hospitals, black out cities, and disrupt core government systems.
As AI capabilities advance, experts believe adversaries could use these tools to scale existing cyberwarfare tactics into large-scale crises, with particular vulnerabilities in utilities, healthcare, communications, and logistics. Rather than a single massive attack, officials worry about either accidental escalation through AI systems spinning out of control or a series of smaller coordinated strikes that could be equally devastating. The concern is compounded by the risk that defenders may come to over-trust AI tools while attackers exploit them more effectively than ever before.
Summary not available
The product: WorldWideWeb is a rebuilt version of the original NeXT web browser developed at CERN in December 1990. The 2019 recreation allows users to experience the original browser within a contemporary browser, enabling worldwide access to this foundational web technology.
Availability: Launched in February 2019 at CERN in celebration of the thirtieth anniversary of the original WorldWideWeb development. The project was supported by the US Mission in Geneva through the CERN & Society Foundation.
Platforms: Runs within a contemporary web browser (compatible with modern browser environments).
Summary not available
"This is not a debate about rejecting technology. It is a question of aligning educational tools with how human learning actually works. Evidence indicates that indiscriminate digital expansion has weakened learning environments rather than strengthened them." -- Jared Cooney Horvath
The U.S. spent over $30 billion since 2002 putting laptops and tablets in schools, but neuroscientist Jared Cooney Horvath testified before the U.S. Senate that Gen Z has become the first generation in modern history to score lower on standardized tests than the previous one, with research linking increased screen time to declining cognitive capabilities.
Despite the promise that widespread access to technology would empower students with knowledge, the opposite has occurred. Research shows a stark correlation between more classroom screen time and worse academic performance, with students engaging in off-task activities on computers nearly two-thirds of the time. Experts argue that technology's addictive design and interruption of sustained attention has weakened rather than strengthened learning environments, leaving Gen Z facing not only eroding cognitive skills but also disproportionate impacts from generative AI on the job market and reduced capacity to tackle complex societal challenges.
Colorado lawmakers have introduced Senate Bill 26-051, which shifts online age verification responsibility from individual websites to operating system providers and app stores rather than regulating the open web directly.
The bill represents a strategic pivot in Colorado's ongoing efforts to protect children from harmful online content while navigating constitutional, privacy, and technical constraints. By embedding age attestation within operating system account layers instead of requiring individual websites to verify age, the proposal targets the segments of the digital ecosystem where centralized control already exists. However, the measure faces familiar legal and practical challenges: critics question whether operating system-level restrictions can comprehensively prevent access when the same content remains available through web browsers and other pathways, and courts in other states have raised concerns about whether age verification mandates are narrowly tailored and constitutional.
Summary not available
Summary not available
"The core cost structure facing the auto industry hasn't fundamentally changed overnight." -- Jessica Caldwell, Edmunds' head of insights
Despite a Supreme Court ruling that struck down some of President Trump's tariffs, car prices are unlikely to decrease because most auto industry tariffs remain in effect under a different law.
The Supreme Court's decision blocked tariffs imposed under the International Emergency Economic Power Act, but the majority of tariffs affecting automobiles—including 15 percent duties on cars from Europe, Japan, and South Korea—were levied under section 232 of the Trade Expansion Act and remain in place. According to Edmunds data, new car prices have risen to a historic average of $48,576, and while automakers have absorbed some tariff costs so far, Jessica Caldwell, Edmunds' head of insights, warns that "if cost pressures continue to build, automakers may have less room to shield shoppers from higher prices."
"the fighter's 'computer brain,' including 'its cloud-based components', could be cracked to accept third-party software updates, just like 'jailbreaking' a cellphone, according to the Dutch State Secretary for Defense." -- Military news site TWZ
The Dutch State Secretary for Defense has suggested that the F-35 combat aircraft's software systems could potentially be jailbroken to accept third-party updates, similar to jailbreaking an iPhone.
This statement raises significant concerns about the F-35 program's security architecture and operational independence for foreign operators. The F-35's reliance on the cloud-based ALIS/ODIN network—which handles sensitive mission data, intelligence uploads, and logistics—underscores larger vulnerabilities in the aircraft's design. While jailbreaking may be technically feasible, such action would risk legal consequences from Lockheed Martin and U.S. government friction, and would likely render jets inoperable without U.S.-controlled maintenance support and spare parts.
"What I see is the result of my own choices rather than a system trying to capture and monetise my attention." -- Susam Pal
Author Susam Pal argues that social networks have fundamentally transformed from genuine platforms for connecting with known contacts into "attention media" designed to capture and monetize user attention through algorithmic feeds and manipulative notifications.
Pal traces the decline of social networking platforms from their origins in the mid-2000s, when they facilitated meaningful connections between people users actually followed, to their current state dominated by infinite scroll, bogus notifications, and algorithmic feeds that prioritize engagement over relevance. He contrasts this evolution with Mastodon, a decentralized platform he contends preserves the original social networking model by displaying only updates from users one deliberately chooses to follow, free from algorithmic manipulation or arbitrary notifications.