The Register

ZTE showcased at MWC Shanghai 2026 its AIR Core, designed for the era of AI agents. Building on deep expertise in core networks and AI, AIR Core drives a paradigm shift from traditional "connectivity pipes" to "intent-driven intelligent services". By integrating AI, computing, data, and security capabilities, it constructs an Agent Service Network (ASN) that spans across devices and ecosystems, comprehensively empowering diverse demands from individuals, smart homes, connected vehicles, and massive emerging AI agents. Powered by AIR Core, key services are elevated across three dimensions: Voice – AI New Calling: From "Basic Calling" to "Smart Services" Introducing the AI Assistant Calling, users can summon personalized assistants on demand. With native multimodal interactions, complex tasks can be seamlessly closed-looped during calls. Furthermore, an AI-powered noise-cancellation micro-model eliminates 99.9% of ambient noise, ensuring pristine call clarity. Bytes – Experience Monetization: From "Extensive" to "Precision" By integrating NWDAF and AI-UPF, the network achieves precise service identification, real-time QoE assessment, and sub-second dynamic policy adjustments, making premium experiences perceptible and quantifiable. For high-value scenarios like high-speed railways, NWDAF leverages spatiotemporal large models to build real-time user profiles and deliver dedicated network guarantees, truly achieving personalized experiences at scale. Token – New Business Model: From "Selling Connectivity" to "Selling AI Services" To accommodate new AI traffic patterns (multimodal, bursty real-time, and symmetric uplink/downlink), the AI-UPF integrates large models for precise AI traffic identification and experience measurement. It establishes dedicated AI PDU sessions with differentiated QoS. Device-network collaboration drives computing power into the network and enables intelligent scheduling, supporting precise token metering and empowering low-latency computing offloading and real-time responses for edge AI applications. AIR Core fundamentally breaks the boundaries of legacy networks, achieving three-dimensional upgrades in service scope, space, and content: Service Scope: From "Humans" to "All AI agents" Through ASN, service targets expand beyond humans and machines to include AI entities like home robots, digital humans, and virtual assistants. This provides efficient connectivity, trusted identity management, and integrated computing-network-intelligence empowerment for agentic communication. Service Space: From "Terrestrial" to "Space-Air-Ground Integration" With deep integration of Non-Terrestrial Networks (NTN), intelligent routing based on dynamic topology perception coordinates multi-orbit satellites and terrestrial networks. This provides highly reliable, ubiquitous connectivity for remote areas, maritime, aviation, and emergency scenarios. Service Content: From "Single Traffic" to a "Traffic + Token + Data" Trinity The AI-UPF evolves into a converged computing-network-intelligence gateway, ensuring precise AI service identification and on-demand resource scheduling. It also builds a Unified Data Plane (RTDS) for centralized data collection, storage, and analysis, supporting user profiling, experience monetization, and data value realization. Marking a fundamental architectural leap from Cloud Native to AI Native, AIR Core lays the foundation for agentic communication in 5G-Advanced and 6G networks. Currently in pilot and early commercial deployment, ZTE AIR Core is partnering with global operators to accelerate their transformation from "connectivity providers" to "intelligent service enablers", jointly exploring new monetization pathways for the agentic era. Contributed by ZTE.

A former Constellation Energy engineer has been accused of making $1.4 million from insider trading ahead of the company's announcement that it would restart a reactor at Three Mile Island to supply power to Microsoft. The Securities and Exchange Commission (SEC) and federal prosecutors allege in their lawsuits that Casey Muggleston purchased share options based on knowledge officially deemed material non-public information (MNPI). In September 2024, Microsoft and Constellation Energy – which owns the Three Mile Island facility – signed a 20-year power purchase agreement to bring TMI Unit 1 back online after it was retired in 2019. Earlier this month, the Nuclear Regulatory Commssion made a preliminary finding [PDF] that the proposed restart would not significantly affect the environment. The assessment remains subject to public comment and a final determination. Should the plant reopen, Microsoft has agreed to purchase energy from it at a predetermined price amid escalating demand for power-hungry cloud and AI services. According to a filing in the US District Court for the District of Delaware, Muggleston knew he was barred from trading on MNPI or buying options in Constellation stock. The criminal indictment claims that in May 2024, he became aware of the plan for the company to restart the Three Mile Island reactor through his work as a nuclear engineer at Constellation. He "exploited that MNPI in violation of the duties of trust and confidence owed to the energy company and its shareholders," the filing alleges, through a series "call options" on company shares between June and September for a total of around $66,595. In July 2024, directors authorized the $1.5 billion restart project. On September 20, the company issued a public statement about its plans to reopen the facility, along with Microsoft's involvement. The company's shares started the day trading at $181.80 and ended the day at $254.98. "After the announcement, Muggleston sold all his Constellation call options. In so doing, he made approximately $1.4 million," the SEC filing said. The SEC also said that the engineer "continued to express an interest in trading on material nonpublic information in violation of his duty to Constellation." Representatives of the parties are not responding to requests for comment. In the federal case, taking place in Wilmington, Delaware, Muggleston is set to make an appearance on July 2 before Judge Sherry R Fallon. ®

The California Science Center has unveiled its Space Shuttle Endeavour exhibit, featuring the retired spacecraft mounted in launch position with the familiar orange external tank and twin solid rocket boosters. The exhibit will be housed in the Samuel Oschin Air and Space Center, which is due to open on November 13, 2026, after a multi-year construction effort. Endeavour arrived at the California Science Center in 2012 after an overland journey from LAX. Following several years parked in what can best be described as a jumped-up hangar, it has now been moved into a vertical position with the tank and boosters attached. The museum isn't quite done. Although the Space Shuttle is now in place, months of artifact and exhibit installation work remain before visitors will be welcomed to view the vehicle. Prior to the installation, Atlantis at Kennedy Space Center in Florida was the de facto Space Shuttle exhibit, showing the vehicle as if on orbit with payload bay doors open. The California Science Center has upped the ante with a view of a Space Shuttle as if it were ready for launch. According to the museum, as well as admiring the stack, visitors will be able to ascend 140 feet alongside the vehicle in a gantry-style elevator, "evoking the experience of astronauts preparing for launch." The tank, ET-94, is the last remaining flight-qualified external tank and is mated to "real solid rocket boosters," the museum says. The earthquake risk posed by the 185-foot-tall stack has been addressed with a hefty concrete slab, approximately eight feet thick and weighing 1,800 tons, supported by six triple pendulum isolators. As construction got underway in 2023, Dennis Jenkins, a former NASA space shuttle engineer and director of the science center's Endeavour display project, told collectSPACE.com: "There is a 3-foot (0.9-meter) moat around the floating concrete slab so it can move three feet in any direction. "We don't dampen vertical motion, so whatever vertical motion the earthquake gives us is what we get, but horizontal motion we can move three feet in any direction." Jenkins also noted that the flight loads the vehicle was designed to take during launch meant that the hardware would likely survive "without permanent damage." Discovery, the most-flown surviving Space Shuttle orbiter, is currently at the Smithsonian's Steven F. Udvar-Hazy Center in Chantilly, Virginia. While Endeavour is now in launch position, and Atlantis is displayed as though on orbit, Discovery's ultimate fate remains unclear as lawmakers argue over whether and how it should be transported to Houston, Texas – a logistical challenge that could make the California Science Center's achievement seem like child's play. ®

As the mercury climbs to disgusting heights, The Reg FOSS desk has picked out a few FOSS highlights from its overflowing mailbox, in case you fancy some Super Productivity while monitoring your AI habit. Prism Carbon Tracker The Prism carbon tracker is a neat idea for helping to instill a tiny bit of self control in the more dedicated botlickers on your development team. It's a plugin that integrates into Visual Studio Code. As the developer uses various "online coding assistants," Prism shows an estimate of how much CO₂ they're releasing in the process of outsourcing their mental activity to a datacenter somewhere. As the plugin's homepage says: "It captures token usage from GitHub Copilot Chat, Claude Code, and runtime LLM API calls, then calculates CO₂ equivalents and surfaces them in the sidebar, status bar, and a live dashboard." Prism comes out of a collaborative effort between the University of Bristol and ustwo, a majority employee-owned "global digital product studio." Ustwo's Nick Hegarty said: "Developers are not the problem, but a fundamental part of the solution. By making AI emissions visible during development, we hope to create greater awareness and support better decision-making. We believe even directional estimates can be valuable if they help start conversations and encourage more thoughtful AI usage." The emissions estimate uses "a transparent token-to-energy-to-carbon methodology informed by Green Software Foundation guidance and recent academic research." Super Productivity Maintainer Johannes Milan wrote to tell us about "Super Productivity." Normally a name like that would set our mental alarm bells ringing, but he did it just right. As a handy lesson for all the other breathlessly excited organizations out there, this is how to do it: with no hyperbole, and a terse text-only email with no formatting. He told us: "I maintain Super Productivity, a mature MIT-licensed local-first task manager and time tracker. It runs on Linux, Windows, macOS, web, and mobile; has Linux packages for Flatpak, Snap, AppImage, deb, and rpm; and works without accounts or telemetry." This is a good intro, so what does it do? It is "a local-first productivity app that is not another cloud suite: tasks, timeboxing, Pomodoro, time tracking, review, and issue tracker integrations in one offline-capable app." The phrase "local-first" twice makes us think he has a better idea of our interests than the PR people who tell us that they liked a random article from three years ago and so should write about their new AI-based product. If you want to see the source, it's on GitHub, and if you are more used to a bit of hyperbole, the product's homepage TAMOSS TAMOSS is a far more specialized tool, but may be very valuable to the people who need it. The name is short for Time-Addressable Media Open Source Store, and the product's homepage says "TAMOSS is an open source, Kubernetes-native implementation of the BBC TAMS API." Which of course merely left us wondering what the BBC TAMS API was, but happily, the Beeb has a bit over 100 years of experience at explaining stuff. It has some pretty good educational resources of its own. We suggest starting with the BBC R&D division's TAMS in 2025 – we found the introductory video there packed a decent summary into three and a half minutes. TAMS, short for Time-Addressable Media Store, is an API for a server that "can be used to store, query and access segmented media over HTTP." Most audio-visual media online is in the form of single, big, fat media files, which contain video, soundtrack, subtitles and more, all mixed together. This is very unhelpful if you want to edit the material, split off a soundtrack, or slow part of it down or whatever. It's a lowest common denominator sort of thing, but that's what all the tools understand and what existing servers know how to send. The TAMS spec lets a client request and receive separate parts of this (the sound, or the video, or both, or other data) by timeline using AMWA NMOS-compatible UUIDs over HTTP. NMOS here is the Networked Media Open Specifications from the Advanced Media Workflow Association. Did you know the BBC has its own GitHub? Us neither, but on it you'll find the open source TAMS API spec. The TAMS standard has its own site as well, and the Beeb hosts the BBC R&D white paper that the spec is based on. TAMS is a spec, not a product. There's nothing you can deploy there, which is where LiveWyer, a London-based Kubernetes and cloud-native consultancy, came in. It developed TAMOSS as open source, and it's publishing it for free under the Apache 2.0 license. The company told us "it lets any organisation run a fully functional, TAMS-compatible media store on its own infrastructure, whether that's from a laptop or a production cluster. No proprietary licensing, no vendor lock-in." ®

EPISODE 12 "What's everybody here for?" the Boss asks, peering worriedly at the long line of people extending out of the wedged-open door of Mission Control, continuing on down the corridor. "It's the amnesty," the PFY replies. "The amnesty?" "Oh, I forget!" the PFY says. "You've not been here that long. Every year or so we run an amnesty, where users can fess up to damaging, destroying, or losing our equipment, no questions asked." "No questions?" "Well, typically they'll explain what they did, because it's amnesty day – for one day only – and there's no retribution." "Retribution?!" "Yeah. Well, take this bloke. This is his six-month-old laptop. It's a laptop that we expressly told him – several times – not to carry around with the cables still plugged in. And what did he do?" "I... uh..." "He caught it in the lift door, damaging the power socket so that it only works if you put a book on the plug to hold it in a particular position." "Well, surely you can repla-" "The rest of the USB-C functionality – i.e. the ability to use a dock, plug in an external mouse, etc., no longer works from that port." "Can't he use another port?" "Remember how I said several times? Yeah... There are no ports left. So anyway, retribution would normally see him 'upgraded' to a 19-inch i3 series 3 laptop, retrofitted with lead-acid batteries, along with an OHP and some slides – for his presentations." "But not today," I say, entering the conversation and handing our user a new-ish replacement machine. "Today we'll take his laptop and replace it with a similar model from a location where it's not needed." "So... you're just replacing people's equipment?" the Boss asks, as our user wanders off happily. "Sometimes. Other times, people just want to admit to, say, plugging an HDMI cable directly into a meeting room display instead of using the wall socket provided – then tripping over the cable and munging up the only spare HDMI socket in the screen. Then sneaking out of the room and pretending it was like that when they came in. Isn't that what happened?" The next user in front of us looks rather sheepish. "But no matter!" I continue. "We'll move that screen to another location with a lower use requirement, and it'll be good as new!" Our next user departs, relieved. "How's all this being paid for?" "As he said," the PFY responds, nodding at me, "we often just move items to an area that has a lower use requirement – somewhere the damage is unlikely to be noticed, or to someone that everybody hates. The rest of the damaged items were – tragically – destroyed in that fire." "What fire?" "The big fire we had," I reply. "When was that?" the Boss gasps. "Tomorrow morning," I say. "Probably." "Yes," the PFY says. "The big fire in the storage room. Like the one about four years ago – after which the insurance company told us we couldn't store large quantities of isopropyl in the building." "Or the one about two years ago, when the insurance company told us that storing piles of small pieces of timber – kindling, they called it, though the PFY maintained that he'd bought it to make a three-dimensional model of a CPU – on top of a piece of powered-up equipment with a dodgy heatsink during stress testing." "No," the PFY says, "we're very cautious about that sort of stuff now. Anyway, changing the topic a little – I'm thinking about taking up a hobby." "Really?" I cry. "I collect matchboxes and matchbooks – I've got boxes of them. Would you like me to bring in my collection to show you, tomorrow maybe?" "I was more thinking about collecting antique firearms," the PFY admits. "And it turns out there's a local company who supplies bulk black powder – with next-day delivery!!" "Wait a minute," the Boss interrupts. "That's not how insurance works." "Sure it is. All the stuff's damaged. We're just... consolidating... our claim. Think of the admin fees that a single large claim will save the insurer. They should be thanking us." "No," the Boss insists. "So what do we tell this user?" I say, pointing to the next in line. "By the look of it she has a laptop that her children have been using... as... an ice cream container?" "I… uh..." she says. "I'll let you decide," the Boss says smugly. "But we're not spending money on your little side hustle – and we're not indulging in insurance fraud." ... The next day ... "There's something wrong with my laptop," the Boss says. "The keys are sticking." "Have you been using this as an ice cream container?" the PFY asks, looking at the keyboard carefully. "No." "It certainly looks like it." "Wait. That's not my laptop!" "Sure it is. It was on your desk, wasn't it?" "Yes, but it's not my laptop!" "Please. Next you'll be telling me that your wall monitor didn't always have a broken HDMI cable." "What?! Look, I want my old laptop back!" "Ah, you mean the one with the broken USB-C ports." "NO. I want MY old laptop." "But that IS your old laptop. I could check the serial number in the fixed asset register, but I'm pretty sure it'll match that machine." "I bet it won't." "Oh it does. I just checked it this morning, coincidentally. In fact, I checked all your asset numbers and all of them match. The iPad with the cracked screen, the multifunction printer that's had vinyl run through it..." "I don't have a multifunction printer." "The asset register says you do. A real monster, with four A3 trays. I'm surprised you didn't smell the burned plastic when you opened the door." "Was that what that smell…" "Then there's the secure shelf unit with the electronic lock that doesn't have a way of supplying power when the internal battery runs out. Apparently that's where you keep all those spare laptops with the broken hinges. That said, not many of our users have a portable diesel backup generator that you accidentally filled with petrol after taking it camping." "I've never BEEN camping," the Boss sighs. "OK, you win. We can have an amnesty day." "Yeah. We did. It was yesterday. One day only. TODAY, you just have a lot of broken crap." "So, nothing in my office works?" the Boss snaps. "There is a gas-powered fire starter in your top drawer," the PFY suggests. "Oh, and a 5 kg carton of black powder." "They're on top of that generator..." I add. ® BOFH: Previous episodes on The Register The Compleat BOFH Archives 95-99

ON CALL Supporting IT and keeping it secure is a serious endeavor. Which is why The Register lightens up Friday mornings with a fresh installment of On Call, the reader-contributed column that shares your tales of tech support trauma. This week, meet a reader we'll Regomize as "Colin" who told us about a recent gig at a customer that decided to improve the security of its Microsoft 365 implementation – chasing the Secure Score that Redmond uses to rate resilience. "We spent a good amount of time working with the customer and agreed a rollout plan to ensure multi-factor authentication (MFA) was enabled across the board in accordance with a security baseline." Colin and his crew knew what to do, so when they flicked the switch on various upgrades, all went smoothly. Until it didn't. "The following morning, one of the senior directors of the company – who was allegedly the COO of a cybersecurity company – called our service desk and started yelling." Amid the yelling and accusations, Colin and his colleagues picked out an allegation that the company had been brought to its knees by the need to register for MFA, which had crippled an invoicing system and would surely result in ruin within a disastrously short time frame. "Once she allowed us to speak, it turned out that the problem only impacted three or four phones," Colin wrote. The support team investigated and quickly learned the real problem was with the invoicing software, which promised MFA support but relied on buggy software to make it happen. The director didn't care for that explanation and ordered an instant rollback that we understand remains in place. Colin found it stunning that the former COO of a security company wasn't willing to wait for a workaround, so delivered the desired result: no MFA, and worse security. He told us this client often made nonsensical requests, such as demanding that a particular engineer – who cannot drive – visit a remote site ASAP to fix a printer. On another occasion, the same person claimed Colin's work on M365 caused a power outage! Have you ever been told to make IT worse? If so, click here to send On Call an email so we can make the column better on a future Friday. ®

Microsoft has oh-so-quietly extended its offer of extended security updates for consumer users of Windows 10. The Windows giant ended support for Windows 10 on October 14, 2025, when it stopped shipping fixes and updates – unless users coughed up for an extended patch package. Redmond’s offer for business users includes an option for an extra three years of support, albeit with ratcheting costs that demonstrate Microsoft has an enormous preference for customers to adopt Windows 11. Consumer users of Windows 10 could acquire an extra year of patches any time up to October 13, 2026. In the last couple of days, Microsoft’s page explaining extended security support changed to mention users can sign up until October 12, 2027. Microsoft hasn’t explained why it’s extended the offer but it’s not hard to guess why: millions of users continue to use the OS – HP says 30 percent of its customers haven’t moved. StatCounter found that 26 percent of all Windows users it can see online continue to use Windows 10. The global population of active Windows boxes exceeds one billion, so hundreds of millions of PCs running Windows 10 remain in service. Microsoft has already committed to providing business customers with patches until 2028, so isn’t giving itself a massive amount of extra work by extending support for ordinary punters. Redmond will charge $30 for a year of extra patches, a far smaller sum than the price of the new PC that many users will need to get decent performance out of Windows 11 – or to run it at all given the newer OS requires TPM 2.0 and at least 4GB of RAM plus 64GB of storage. PC prices are also on the rise thanks to the AI boom meaning deep-pocketed hyperscalers are paying a premium for memory and solid-state storage devices. No tech manufacturer is immune: Apple has hiked Mac and iPad prices to cover the rising cost of RAM. Microsoft knows that’s a problem for the PC market because it recently said its developers are “lowering the baseline memory footprint for Windows.” That effort hasn’t yet delivered a change to Microsoft’s requirement for PCs running Windows 11 to employ 4GB of memory – or 16GB on AI PCs. Keeping Windows 10 secure for a little while longer might earn Microsoft some grudging appreciation – and a few bucks it might not be able to access from fresh Windows 11 license sales at a time punters are keeping their wallets shut. ®

Chinese cybersecurity vendor Qihoo 360 claims it’s built an AI bug-finder that’s better than Anthropic’s Mythos model. CEO Zhou Hongyi revealed the model in a speech at the 14th Beijing Cybersecurity Conference, which Qihoo 360 organizes. Chinese media outlets have transcribed the talk, in which Zhou described Mythos as “equivalent to a ‘cyber nuclear weapon’,” because the USA’s ban on foreign nationals accessing the model gives America a tool with which to find flaws in software upon which other nations rely. Zhou thinks China needs equivalent capabilities as a deterrent, but suggested replicating Mythos is not a viable approach. “Mythos follows a typical large-scale model approach: the strongest model, the strongest computing power, and the strongest chips – a strategy of sheer brute force,” he said. “However, this path has an implicit prerequisite: your model capabilities must be sufficiently strong. Objectively speaking, domestically developed models still lag behind by 20 percent to 30 percent in underlying capabilities.” The CEO therefore thinks China can’t wait for its own models to catch up and needs to find another way to build Mythos-grade bug-finders. Helpfully, Qihoo 360 has found those alternative methods by distilling its 20 years of experience fighting cyber-threats and colossal malware library into security-specific models and agents. The company has put that to work in what Zhou described as a “multi-agent swarm.” “If the American approach is about cultivating a genius hacker, the 360 approach is about organizing a professional attack and defense team,” he said. “When faced with a target, the swarm doesn't perform single-point analysis, but rather collaborates: first, it models the threat and filters high-risk attack surfaces; then, it follows the data flow across files to discover potential vulnerabilities.” The company’s agents apparently “automatically build sandbox environments, automatically generate exploit code, and conduct real-world testing. The result is that every vulnerability is ‘confirmed’ rather than just suspected. After completing a task, the swarm also summarizes and reviews its performance, becoming smarter with each use. This is something a single large model can hardly do.” Qihoo calls this approach “Tulongfeng” and says it’s already finding flaws in open-source and commercial software. “We automatically discovered a Windows kernel privilege escalation vulnerability that had been dormant for five years, an Office remote code execution vulnerability that had been dormant for eight years, and an Excel vulnerability that had been dormant for 10 years, earning official recognition from Microsoft,” Zhou boasted. The CEO said the tool found plenty of flaws in OpenClaw – a feat that human researchers have also achieved. Zhou said Qihoo 360 has created another AI-powered security tool called “Yitianzhen” that automatically simulates potential attacks against an organization’s cyber-defenses, then suggests and/or implements remediations. The company has created an alliance of local cybersecurity companies to use it and create a bulwark against Project Glasswing – the group of entities Anthropic allows to use Mythos under controlled conditions. US authorities have sanctioned Qihoo 360 on grounds that it probably supplies China’s military. China's National Computer Virus Emergency Response Center (CVERC) often cites and publicizes the company’s research, sometimes in its documents that allege the US hacks itself to make China look bad. ®

AI-POCALYPSE Several leading US AI companies joined with former US government officials on Thursday to form a bipartisan coalition focused on helping to prepare workers for the AI tsunami, even as many companies seem to be in a race to see who can replace human workers the fastest. Founded by former US Commerce Secretary Gina Raimondo and former Indiana Governor Eric Holcomb, RAISE US aims to "design and pilot new corporate incentives to retrain and redeploy workers, new approaches to support people through job transitions, and new training models tied to changing employer demand." The announcement talks up employer incentives but makes no mention of worker rights. It does, however, tease the possibility of wage insurance and career navigation "so changing jobs no longer means financial ruin." Winning political support for such benefits seems unlikely in a country where the federal minimum wage, frozen at $7.25 since 2009, "is at its lowest real value in 77 years," and healthcare costs could erase the solvency protection promised by wage insurance. Amazon, Anthropic, Microsoft, and the OpenAI Foundation – vendors of the AI models inspiring recent layoffs – are anchor partners in RAISE US initiative. But its corporate backing is broader still, with financial sponsorship from Bank of America and the support of more than two dozen major companies and foundations. Amazon, having sacrificed 14,000 jobs on the altar of AI transformation last fall, said, "For Amazon, joining RAISE US is a natural extension of what we have been doing for years: investing in people, partnering with governments, and building programs that help workers and communities succeed." Microsoft, which ousted about 9,000 employees last July and recently dangled buyouts before a similar number of staff, offered a similar endorsement from VP and president Brad Smith. RAISE US intends to focus on: state workforce policy and programs; employer-driven worker transition plans; worker education and re-training programs; and policy recommendations. "America has a technology strategy for leading the global AI competition," said Raimondo, who will serve as CEO of RAISE US, in a statement. "It does not yet have a people strategy – and we cannot lead without one. If we build the best AI systems in the world and leave millions of Americans behind, we won’t have won anything; we’ll have automated our own decline." We haven't yet automated away software developer roles, but there are signs that's happening for designers and marketers. A recent employment report from venture capital firm SignalFire notes that tech hiring is at 25 percent less than what it was before the 2019 COVID pandemic, on a 12-month trailing basis. But some jobs have fared better than others. Engineering hiring is down only 11 percent during this period at major tech companies and is up seven percent at startups. Designers have fared worse, with hiring down 48 percent at large tech companies and down 22 percent at startups. Marketing roles have declined 36 percent at large organizations and are down 18 percent at startups. And among product managers, hiring is down 39 percent at tech giants but up two percent at startups. Amid this overall retrenchment, software engineers are proportionally more in demand, accounting for 55 percent of tech hiring, compared to 46 percent seven years ago. The AI job apocalypse is most evident among new hires with less than a year of experience, and college graduates. SignalFire's data shows that in 2016, about 22 percent of new hires at major tech companies had a year or less of experience. Today, that figure is about eight percent. Among startups, where about 15 percent of new hires were inexperienced a decade ago, today that number is just three percent. "The critical systemic risk arises when an entire industry stops investing in early-career talent," SignalFire says in its report. "By eliminating its new grad pipeline to optimize current balance sheets, the tech industry could face a severe leadership vacuum over the next decade." RAISE US certainly has its work cut out for it as its members keep cutting jobs. ®

A new self-destructing backdoor called Mistic used in intrusions since April appears to be linked to a criminal gang that compromises corporate networks and then sells that access to ransomware groups, according to security researchers. This backdoor, also tracked as MLTBackdoor, was first documented by Zscaler earlier this month, with the security shop suggesting the novel malware is “likely used in ransomware attacks to establish a foothold for lateral movement.” In a Wednesday threat brief, Symantec and Carbon Black threat hunters say the backdoor has been used to access multiple organizations' networks over the past few months, including those in insurance, education, IT, and professional services. Additionally, the security sleuths reported, “Mistic may be linked to the financially motivated initial access broker (IAB) tracked publicly as KongTuke (which we track as Woodgnat) and it was used in one intrusion that also involved the group's ModeloRAT remote access trojan.” KongTuke and other IABs don’t deliver the final payload – such as ransomware – to compromised companies. Rather, they break into company systems, and then sell that foothold to other criminals, like ransomware gangs. Symantec and Carbon Black arrived at their low-confidence attribution after at least one case where Mistic was deployed in close proximity to ModeloRAT, the Python-based remote access trojan KongTuke also developed. KongTuke has previously been linked to attacks from various ransomware crews including Qilin, Interlock, Rhysida, Akira, 8Base, and Black Basta. “Our Threat Hunter Team has separately observed ModeloRAT used in attacks that deployed Qilin ransomware, linking this tool to ransomware deployment,” Symantec and Carbon Black noted. Plus, Zscaler reported Mistic being delivered in a multi-stage ClickFix infection chain, which is another pointer to KongTuke, as the group is known to use that initial access technique. In one case that Symantec and Carbon Black responded to, Mistic was side-loaded through a legitimate file, MpExtMs.exe, and then loaded from a DLL named EndpointDlp.dll, which likely helped the backdoor blend in with legitimate software. Mistic has all the usual backdoor functionality: It can upload, download, move, rename, and delete files. It can also create new folders, and check for additional commands from the attacker-controlled command-and-control (C2) server. But here’s the stealthy part: it can run remote payloads from C2 directly in memory – so it doesn’t write malicious files to the hard drive – which helps it dodge file-based detection in antivirus and endpoint detection products. When the mission is accomplished, it then terminates and deletes itself. “The fact that Mistic executes in memory and also has a kill switch built in means that it is very stealthy, potentially allowing for long-term, stealthy access for attackers,” the threat hunters wrote. ®

A sealed scroll from the Roman town of Herculaneum, which was destroyed by Mount Vesuvius' eruption nearly 2,000 years ago, has finally given up its secrets, thanks to a combination of machine learning and high-resolution CT scans. In 2023, researchers managed to decipher a few words from among the char and ash that make up the bulk of the scrolls. Some of those same prize-winning researchers recovered more passages from one of the scrolls, PHerc.Paris.4, netting them the $700,000 grand prize from the Vesuvius Challenge contest in early 2024. Fast forward two more years, and those grand prize winners are now part of the Vesuvius Challenge team that managed to read the surviving portion of a rolled scroll end-to-end, as the VC team shared in a Thursday announcement and detailed in an accompanying paper [PDF]. According to the research paper, the ability to make out the entirety of the scroll was thanks to high-resolution phase-contrast X-ray microtomography performed at the European Synchrotron Radiation Facility in France - an improved imaging technique over prior methods used to capture prior images that were analyzed in the prize competition. That wasn’t all, though: They say that much of their work succeeded because of a new "workflow" they developed to scan scrolls, detect ink on charred papyrus, virtually "unroll" the scrolls by modeling their deformed surfaces, and preserve those surfaces digitally, allowing machine learning models to identify letters across an entire scroll rather than just isolated patches. "The key transition marked by the present work is therefore from exceptional local recovery to systematic scroll-scale recovery," the team wrote. In other words, provided they can account for the particularities of the hundreds of sealed scrolls recovered from Herculaneum's Villa of the Papyri, the world's only surviving intact library from antiquity, this could mark the beginning of an explosion in new material for historians. So, what did it say? PHerc.Paris.4 wasn’t at the center of this breakthrough either, though they did have some exciting news to share on that front that we’ll get to. Instead, the breakthrough centered on PHerc. 1667, a previously unread rolled scroll whose preserved text was read continuously from end to end for the first time. The work appears to be a treatise on Stoic philosophy focused on ethics - a favorite subject of Zeno, Marcus Aurelius, Epictetus, and their intellectual fellows. “Having certainly strained ourselves to the utmost through research and learning, we will no longer be inferior to them in any respect,” a passage from the latter part of the scroll reads, “accomplishing in like manner the things that befit them and possessing the same practical wisdom as they.” Quite a fitting bit of ancient wisdom to be the first to see the light of the modern world. While the team digitally unrolled the scroll, detected its ink, and transcribed the preserved text from end to end, portions of the original PHerc. 1667 were lost long ago during earlier attempts to physically open the scroll, before archaeologists had access to sophisticated X-ray imaging and AI-assisted analysis. “Earlier attempts to open it by hand — in the nineteenth century, and again in 1969 and the 1980s — destroyed its outer layers,” the Vesuvius Challenge team said, noting that only an 8 cm-high core remains of the original scroll, which originally measured between 19 and 24 cm in height when standing upright. Nonetheless, “it is the first time the preserved text of a rolled Herculaneum scroll has been read continuously, end to end, rather than in isolated words or patches,” the team said. In addition to the full reveal of what’s left of PHerc.1667, the team also managed to pick out some information from a couple of other scrolls using their new workflow. One, PHerc.139, was determined to be a copy of book eight of epicurean philosopher Philodemus’ treatise On Gods, meaning scholars can expect to know what they’re looking at once the scroll is fully digitally unrolled. The second concerns, as mentioned above, PHerc.Paris.4. The new higher-resolution images taken for this latest experiment make the words on the scroll directly visible for the first time, meaning that there’s no need to rely on algorithmic detection of individual words and phrases from CT scans. Most crucially, the new scans of Paris.4 perfectly matched what the grand prize team made out several years ago, providing independent confirmation that the prize went to the right team. There are still challenges to meet in unwrapping and deciphering the rest of the ancient library, with the team calling out geometric challenges in surface prediction that can render an unrolled scan unreadable, and radiometric challenges that make ink identification difficult, as ancient recipes were inconsistent. Still, it’s a massive leap forward and the team believes the X-ray and machine learning workflow they’ve developed is ready to scale. “The thoughts of the ancient world, sealed in darkness for two millennia, are coming back into the light — a whole scroll at a time,” the Vesuvius Challenge team said. I, for one, can’t wait to see what ancient secrets they discover next. ®

Security firm Huntress allegedly has a turncoat insider leaking info to a ransomware operation, according to an ex-employee who took his grievances to social media after claiming the security shop tried to “silence” him with legal threats. And it all started with a Pinocchio GIF and clown emoji. Late last week, Huntress disclosed that it is among the “hundreds of Klue customers” compromised in the supply-chain attack, stating that “Huntress believes in radical transparency about security incidents, including when it affects our company.” Ben Folland, a former security operations analyst at Huntress who left the company in February, responded with a Pinocchio GIF and clown emoji - although, to be clear, his complaints about his former employer have nothing to do with the Klue incident. These stem from an earlier incident that Folland also detailed in a series of posts. According to Folland’s resignation letter, which he also shared on LinkedIn, he left the security firm for “personal reasons, and due to a conflict of interest,” with his last day of work being February 19. This conflict, Folland alleges, arose from his December discovery that “another Huntress employee passed communications from US law enforcement to a cybercriminal, DevMan, who is actively and publicly targeting my family and me.” DevMan is a ransomware operation that first emerged in April 2025 and uses modified DragonForce code. “Since December 2025, I believe Huntress has been actively trying to conceal a serious security incident from its partners, customers, and employees involving an insider who is still employed at the company,” Folland said in a LinkedIn post. The alleged insider was “caught by the FBI,” according to Folland, and continues to work as a Huntress employee. “The incident in question would cause significant reputational damage to Huntress and, in my view, continues to put clients at risk,” his LinkedIn post continued. “With an IPO on the horizon, it appears their priority was not transparency, but keeping this away from the press.” Folland also promised to publish, over the next two weeks, “evidence supporting the claims made in my resignation email,” such as communications with the FBI and those between the Huntress employee and DevMan, recorded phone calls, internal Huntress memos, and threats targeting Folland and his family. The Register reached out to Folland for more information and did not receive a response. “If you are an employee at a cybersecurity company, you should not be helping cybercriminals,” he wrote on LinkedIn. “You should not be informing them of active investigations. You should not be engaging in cybercriminal activity yourself.” We also contacted Huntress about Folland’s accusations, and CEO Kyle Hanslovan responded via a spokesperson. "A former employee raised concerns that a teammate exercised poor judgment in communicating with a cybercriminal,” Hanslovan said. “By nature of our work as security researchers, teammates occasionally need to communicate with possible cybercriminals to gather intel that ultimately supports our partners and customers,” he continued. “I appreciate the hell out of that former employee's concerns and we've taken them seriously every step of the way. I also have to make sure Huntress upholds its responsibility to protect the confidentiality of our teammates involved and the investigation underway.” Hanslovan also assured Huntress’ partners, customers, and employees that if he learns “new information that changes our assessment of the current situation, I will take quick and appropriate action.” In a more direct response on Reddit, Hanslovan said he “firmly disagree[s]” and doesn’t “understand Ben's accusations.” His company “strongly disagree[s] with this ‘insider’ narrative,” he wrote. “We sure af didn’t prioritize an IPO over the safety of our partners, customers, or team.” And about the FBI allegations: “Some aspects of this matter involve ongoing active coordination with law enforcement and legal proceedings that prevent us from providing a complete public account,” Hanslovan wrote. “We're not gonna litigate this on LinkedIn with Ben but will likely publish some form of official comms to make our stance clear for those needing something more than my reddit reply.”®

A company can learn a lot about the market by looking at its own employees. OpenAI says that its team members are switching from chatbots to agents as their primary form of AI interaction, a trend also detected (though less pronounced) among external organizations and users. Instead of one-off ChatGPT prompts, workers are asking Codex agents to tackle multi-step tasks that take long periods of time. And those doing so are increasingly non-developers. OpenAI insists that its findings have implications for other companies, labor researchers, and policymakers, not the least of which would be a brighter revenue picture for OpenAI. Longer running tasks consume more tokens, and to the extent those can be billed, that should help diminish hundreds of billions in debt obligations. "We find that agentic AI usage is growing rapidly: the number of active users has grown more than fivefold in the first half of 2026, with the most rapid increase occurring outside the initial audience of software developers," said company researchers and academics in a paper [PDF] titled, "The Shift to Agentic AI: Evidence from Codex." OpenAI did not immediately respond to a request to clarify whether it incentivizes or encourages employees to use its AI tools – through internal communiques, token allocations, token use leaderboards, or tying tool usage to performance metrics. But we'll take it on faith that when there's enough Kool-Aid on-premises, employees may just develop a taste for it regardless of whether their jobs depend on Kool-Aid consumption. "Through August 2025, the average OpenAI worker spent less than 10 percent of their tokens on Codex," the biz explained in a blog post accompanying its paper (that suggests employee token allocations). "Now, every department, including non-technical departments such as Legal and Recruiting, uses Codex as their primary AI tool for work." Within OpenAI, 97.9 percent of employees are now using Codex, up from around 40 percent in August 2025. External organizations have also seen a usage uptick, to 17.3 percent presently. With individuals, Codex isn't much to speak of – about 0.7 percent. The thing about Codex is that, as an agent, it can operate for long periods of time. "Since the start of the year, the share of individual Codex users who submit at least one request for a task estimated to require more than eight hours for an experienced human to complete has increased nearly tenfold," the paper says. We note that comparing the time a human might take for a task (as estimated by an LLM-as-judge) to the time an AI model takes is only part of the picture if the workflow isn't entirely automated. Generating code at, say, 10x the rate a person might manage may expand the time required for code verification and deployment. OpenAI also points out that, since August 2025, non-developer usage of Codex has risen 137x for individuals, 189x for organizational users, and 12x within OpenAI. The company concedes that technical usage remains the dominant mode, but insists that adoption by non-devs shows how a broader set of knowledge workers can take on coding or technical tasks, such as automation, data transformation, and data analysis. "In June 2026, the median OpenAI employee in a legal role generated 13 times more monthly output tokens across Codex and ChatGPT than they did in November 2025," the paper says. Given that the number of US federal lawsuits filed against OpenAI and associated entities only grew about 11 percent (35 to 39) between the last six months of 2025 and the first six months of 2026, it looks like OpenAI's legal team, with its 13x token surge, is making the company's case for the productivity benefits of AI tools. ®

Yet another Martian rock formation has revealed what may be signs of ancient life on the Red Planet. While these findings aren't as scientist-wowing as those reported last year, they add to growing evidence that ancient Mars contained organic carbon and may once have been habitable. An international team of scientists working on examining Martian rock samples collected and analyzed by the Perseverance rover reported in Science Advances Wednesday that they’d found signs of macromolecular carbon (MMC) in the Martian rock dubbed Cheyava Falls. Like most of the other life-affirming samples collected from Mars, this one comes from Jezero Crater, where Perseverance has spent the entirety of its five-year mission. Jezero Crater, and particularly the long-dried riverbed known as Neretva Vallis that fed what’s believed to be a large lake that once filled Jezero, has been the site of several potential signs of ancient life on Mars, and it’s no different in this case. Two instruments at the end of Perseverance’s arm, the Scanning Habitable Environments with Raman and Luminescence for Organics and Chemicals (SHERLOC) and the Wide Angle Topographic Sensor for Operations and eNgineering (WATSON), were collectively responsible for the discovery of MMCs in Cheyava Falls. MMCs found on Earth are often associated with organic compounds essential to life, including carbohydrates, lipids, proteins, and nucleic acids, and there's a distinct possibility that the MMCs found on Mars could have come from similar sources. The Cheyava Falls team was also excited to link their findings to a discovery made by Curiosity in 2018, which found evidence of similar materials more than 3,500 kilometers away in Gale crater. The potentially organic matter discovered by Curiosity was dubbed “kerogen” by that team. From a strictly scientific standpoint, kerogen can be used interchangeably with MMC, but the Cheyava Falls team opted to avoid that terminology in order to “avoid implications of biogenicity.” That said, it’s still exciting to the team, and the widespread discovery of MMCs across Mars could indicate that “habitability … and the availability of organics, may have been widespread across the planet billions of years ago.” This discovery hasn’t led the researchers to jump to conclusions, however. As they note in the paper, there are a number of lifeless sources of MMCs that could be the cause of their detections, with meteorite impacts and abiotic synthesis of organic matter that could both be responsible. “There are multiple potential pathways to form abiotic organics on Mars, including condensation from igneous systems and associated volatiles, serpentinization, carbonation, and the electrochemical reduction of carbon dioxide in aqueous/hydrothermal fluids,” the team noted. In other words, this is nowhere near the confounding mystery that NASA admitted it had on its hands in September of last year. Perseverance drilled a core sample called Sapphire Canyon from the Cheyava Falls rock, which NASA said last year contained potential signs of ancient life that it couldn't readily explain through abiotic processes, though non-biological explanations remain possible. "A year ago we found what we believe to be signs of microbial life on Mars' surface," then-acting NASA administrator and transportation secretary Sean Duffy said at a press conference last year. "After a year of review [the scientific community] has come back and said it can't find another explanation." Jezero Crater has given up other signs that it may have been a vibrant, living ecosystem too. In March, researchers reported that the same region also turned up signs of nickel compounds that indicate the formation of enzymes related to bacterial biological processes, but again those signs are also explainable by non-biological action. Proof of life requires samples brought to Earth Both the report released Wednesday and the research leaders behind it are unequivocal on what it’ll take to confirm the growing body of evidence that there was once life on Mars: Samples need to be returned to Earth for inspection. “The instruments on the Perseverance rover were not designed to distinguish between rocks that formed through biologic or abiotic processes, but they are able to identify compelling rocks to sample for possible return to Earth,” SHERLOC principal investigator and NASA Jet Propulsion Laboratory scientist Kyle Uckert told us in an email. Uckert reiterated to us that humanity’s best chance at fully understanding whether the various organics detected in Jezero Crater is some sort of sample-return mission, and he explained that rocks they sampled for the study are packaged, sealed, and waiting for the possibility that something like that still happens. That’s far from a sure thing at this point, as NASA’s Mars Sample Return plans are essentially scuttled at this point - something Uckert and his co-lead author, Planetary Science Institute researcher Ashley Murphy, declined to weigh in on when asked. Meanwhile, China is already pressing ahead with its own Mars sample-return mission, with launches planned for 2028 and a sample return targeted for around 2031. China already has a leg up on the US in the sample return space, with the China National Space Administration returning samples from the far side of the Moon to Earth in 2024. NASA has collected samples from an asteroid and returned them to Earth, to be fair, but Mars is still a long way off, and life or no life, getting samples back to prove it is going to be a race that China may win if its current plans work out. ®

Neither LLVM nor GCC directly support the BASIC programming language – but a former Microware boffin proposes fixing that. An interesting new proposal on the Discourse forum of the LLVM compiler suite has turned into a new standalone BASIC compiler. The original RFC was Adding BASIC09 frontend tool to LLVM. Author Boisy Petre proposed adding a front-end to enable LLVM to compile BASIC source code, and this has now turned into a standalone compiler called basic09c, which uses LLVM as a library. We are irresistibly reminded of the recent addition of ALGOL-68 to GCC. As we said in 2023, BASIC is anything but dead, and, for BASIC’s 60th birthday, the following year we covered new versions of three modern FOSS dialects. The late Dr Thomas Kurtz would have been proud. It’s not just any old home-computer BASIC, either. Dr Petre is working on a compiler for Microware BASIC09. This was one of the compilers that Microware offered for its multitasking, multiuser Unix-like OS, called OS-9. Way back in 1999, many users of Apple’s then-new MacOS 9 – often just called “OS 9” – confused it with the already 20-year-old Microware OS-9, and they pestered OS-9 newsgroups and communities with Mac questions and chatter. As The Register reported back then, Microware even sued Apple over the trademark. BASIC09 is a structured BASIC: it has named procedures with local variables, supports constructs such as IF…THEN…ELSE, user-defined variable types – and no, it did not need line numbers. The best reference we can find to BASIC09 today is its Wikipedia article, but you can also read the manual [PDF]. Microware is still around and still supports OS-9, which is marketed as an RTOS these days. Home microcomputer users, though, might have known OS-9: it was an optional OS for the British Dragon 32 micro and was on the list for the first British laptop too. For Stateside readers, the Dragon 32 was a relative of the American Tandy Color Computer, which had the same Motorola 6809 CPU and could also run OS-9. Aside from its technical merits, there are other good reasons he chose this particular BASIC: Although Boisy Gene Petre became Dr Petre last year, he has been in the industry for quite a while. He worked at Microware early in his career, and even quarter of a century ago he was writing articles about the Tandy CoCo. This is not the first time that the Reg FOSS desk has been bamboozled by boffin Boisy’s brilliance. That was way back in 2012, when he created one of our favorite ever retrocomputing projects: the astonishing Liber809. This performed a total brain transplant on the original Atari 8-bit machines, installing a 6809 CPU and supporting firmware. As he explained in a post called The Beginning, the goal was to marry the most advanced eight-bit CPU with the most capable eight-bit computer of its time. Retrocomputing blog The Byte Cellar described it well. This would of course render the machine incompatible with all existing Atari software, but the plan was to make it able to run NitrOS-9 – a community distribution of the original 6809 version of the Microware OS. ®

AI agents now write code and run tasks at machine speed, and sometimes those same agents delete the wrong thing. AI-enabled attackers can also use AI to find zero-day vulnerabilities more easily. Most recovery systems still run at human speed, which widens the gap between how fast data can be lost and how fast it can be restored. Eon was built to close that gap. In our latest Hot Seat, Tim Phillips talks to Gonen Stein, president and co-founder of Eon, about what recovery should look like in the agentic AI era. Stein and his co-founders previously built CloudEndure, a cloud migration and disaster recovery outfit that AWS acquired. They found that backup had never been redesigned for cloud-native infrastructure: the old model assumed static servers and scheduled maintenance windows, while cloud environments change constantly. There's also a gap between the C-suite's faith in recovery and today's reality. Eon's research found that 98 percent of executives were confident in recovery, while most had suffered three or more failures last year. As Stein puts it, a completed backup is not the same as a tested restore. Configurations drift, new services appear, and backup policies fall behind, so the plan looks fine until you need it. The risk is not hypothetical. In April, an AI agent tasked with resolving a credential mismatch in PocketOS's staging environment deleted the production database and took the attached backups with it. The event took nine seconds. The agent used valid credentials and a legitimate API, so no alert fired. Eon's answer is to hold backups in immutable, logically air-gapped vaults with separate credentials, so that when something goes wrong the restore can target a single table or record rather than rebuild the whole environment. Watch the Hot Seat to hear Stein on AI-augmented attacks, the case for isolated recovery, and the checks to run first. You will learn about: Why confidence outruns evidence: a green tick on a backup dashboard does not confirm that a full recovery has been tested, and most organizations cannot pass that test. AI on both sides: coding agents ship faster than humans can review their work, while AI-augmented attackers shrink the window between finding a vulnerability and exploiting it. Recovery outside the blast radius: backups that share credentials or storage with production are vulnerable to the same attacks and faults, which is why air-gapped vaults matter. Granular restore: why you should be able to recover a specific table or record at a precise timestamp, rather than rehydrating an entire environment. If you are responsible for cloud data, backup policy, or recovery planning, watch this one. Sponsored by Eon.io + Inc.

The online Apple Store went offline Thursday morning, but instead of such a move marking the launch of upgraded products, it came back online with nothing new to offer but higher prices. The budget-priced MacBook Neo wasn’t even spared from the spike, with its entry-level price climbing from $599 to $699. Along with the Neo, entry-level MacBook Air prices rose by $200 and base MacBook Pro machines increased by $300, putting the lowest-spec machines in the lines at $1,299 and $1,999, respectively. iPads also went up in price, with the base-model iPad Air (11” w/128GB of storage) jumping from $599 to $749 and the basic iPad Pro (11”, 256GB of storage) now priced at $1,199 up from $999. iPhones were spared from the price increase, for now at least, but that may not last either if the current trend in hardware pricing is any indicator. As we’ve been covering of late, and pretty much everyone is well aware, memory prices are skyrocketing thanks to the AI and datacenter boom, leading to shortages, longer lead times, and the need to increase prices on everything that contains even the barest minimum of anything that looks slightly like RAM. With entry-level MacBooks now starting at a minimum of 16 GB of sweet, sweet integrated memory, it’s no wonder the prices are increasing alongside everything else. Terminal Apple CEO Tim Cook even warned that price hikes were coming in an interview with the Wall Street Journal last week, telling the publication that increases were “unavoidable.” “We’re doing our best to mitigate the huge increases that are being passed to us, and we’ve been trying to shield our customers from the increases, but the situation has become unsustainable,” Cook told the WSJ in an interview. Cook didn’t give any details on the what or when of Apple price increases, but there’s no reason to guess any longer. Microsoft has been affected by memory crunches too, and quietly rolled out a 12" Surface Pro and 13" Surface Laptop with a paltry 8 GB of memory this week. Given they have half of the minimum RAM spec needed to run Copilot+, the two machines are devoid of the AI marketing suffusing everything else that Microsoft now sells. As for when iPhone price increases may rear their heads, that’s still unclear, but don’t expect it to take too much longer. Micron, one of Apple’s RAM suppliers, said in its Q3 earnings call that it had signed agreements with 16 of its commercial customers to lock in historically high memory prices through 2030 on the back of continued uncertainty. “Our customers are recognizing that supply shortages in memory and storage will take considerable time to improve,” Micron CEO, president, and chairman Sanjay Mehrotra said during this week’s earnings call. “Even as we expect industry supply to improve gradually in 2028, we currently do not have line of sight as to when memory supply will be able to catch up with increasing demand.” In other words, if that iPhone is looking long in the teeth, it might be time to upgrade before those prices, too, begin to climb. We reached out to Apple for comment, but didn’t hear back. ®

Users with hopelessly borked Windows devices have a new avenue of recovery in the form of point-in-time restore for Windows 11. The service is designed to restore a Windows 11 PC to its exact state at an earlier point in time and is accessed through the Troubleshoot menu in the Windows Recovery Environment. According to Microsoft, each restore point covers the operating system, apps, settings, and local files. It is stored locally and automatically deleted after 72 hours or when free disk space falls below 20 GB. New points are created approximately every 24 hours by default, although the interval can be configured. Like the existing System Restore functionality, point-in-time restore uses the Volume Shadow Copy Service (VSS) to get a device back to a previous state. However, there are differences. Point-in-time restores are scheduled, whereas System Restore requires an event or manual intervention to trigger a restore point. Point-in-time restore mitigates storage impact by using reserved storage and also has a maximum retention period of 72 hours for each restore point. Point-in-time restore is on by default on devices with an OS volume size of 200 GB or greater and not under enterprise management, such as Windows Home and Pro. It'll be off by default on managed systems until Windows 11 26H2 arrives. There are some downsides. A restore can only be triggered locally, although Microsoft plans to add remote initiation via Intune in the future. The upshot is that it has limited utility for administrators. There is also the potential for problems with Outlook, where a data file (.ost) mismatch might occur after a restore, necessitating the deletion or renaming of files, and Windows Recall may be disabled after a restore. Some might argue the latter is more of a feature than a known issue. "Point‑in‑time restore is an important foundation for the future of Windows recovery," Microsoft said, labeling it part of the company's Windows resiliency initiative. It's been in public preview for some time and is enabled on more than two million devices. However, nagging questions remain. Why has it taken so long to add this capability to Windows, and why does the operating system allow updates that can leave users needing to restore their PCs in the first place? ®

The Swift Package Index (SPI), a search engine for open source packages for the Swift programming language, is now part of Apple, though it will remain open source. Dave Verwer, who created SPI over six years ago along with Sven A. Schmidt, said on Mastodon that "I'll be joining Apple to continue working on everything related to Swift packages." Apple senior product manager Dave Lester said that SPI has become an "essential part of the Swift ecosystem" and referenced the intent to build a comprehensive package registry for Swift. The news follows Apple’s sponsorship of SPI back in March 2023. SPI is open source on GitHub under the Apache 2.0 license and Verwer, Schmidt, and Swift Core Team member Ted Kremenek said that it will remain open source and continue to operate as it does today, though they also promised accelerated development. New capabilities are promised, including package signing and identity to improve "robustness and security." Currently anyone can add a package to SPI, and developers are reliant on metadata to judge the reliability and trustworthiness of a package. Each package in the index shows information, including the number of contributors; how long the project has been in development; the number of open issues; how many dependencies it has; the project README from GitHub; release notes; and a Use this package button, which shows how to use it from Xcode (the official Apple IDE) or using the Swift Package Manager. On its launch in 2020, SPI contained around 2,500 packages, and now has over 11,000: significant growth, but miniscule in comparison to the likes of PyPi (Python Package Index), which has over 8 million, the main reason being the niche status of Swift outside development for Apple platforms. That said, SPI has some strong features including that it runs compatibility-testing builds against each version of every package on a range of operating systems, including macOS, iOS, WatchOS, VisionOS, Linux, Wasm, and Android. A snag is that many packages report no compatibility information along with the message "we are currently processing a large build job backlog," making the feature useless for recent releases. A common complaint about SPI is that it only supports packages hosted on GitHub. Soon after the project launched, a user asked for GitLab support to be added. Verwer said "I would definitely like to get to it one day," but two years later admitted that "the situation has only got worse and we are more tightly coupled to GitHub than we were at the start." In May this year, another user remarked that "being attached only to GitHub is not good for the wider Swift community," but Verwer said that "this isn’t a current priority for us." The Apple move appears to signal a change of direction though. On Hacker News, Verwer said that "the great thing about a registry is that it doesn't care where the original source is hosted. We will be moving away from that model completely as we work towards this." Verwer’s remark suggests that, despite the news announcement emphasizing that SPI will remain as-is for now, big changes are planned. Apple’s resources may fix issues such as long waits for build results, though developers will also be wary of the implications of a well-liked independent package index coming under the company’s full control.®

Amazon has promised a $13 billion chunk of its mammoth capex plans to build AI and cloud infrastructure in India. The cloud services and software giant has said it expects $200 billion capex for 2026 as, like other vendors in the market, it struggles to build compute capacity to support the expected explosion in demand for AI services. Having announced a $35 billion investment in India in 2025, Amazon's commitments in the country between 2010 and 2030 total more than $88 billion, some of which is for new retail fulfillment centers and delivery stations. The new $13 billion is earmarked for AI and cloud infrastructure, including additional AWS datacenter capacity in Mumbai and Hyderabad. AWS promises startups, enterprises, and government organizations access to custom AI chips, managed AI services, cloud technologies, and developer tools. Amazon CEO Andy Jassy announced the news following a meeting with Prime Minister Shri Narendra Modi. "As we grow Amazon in India, our business priorities align with India's priorities of democratizing access to AI, digitizing small businesses, creating jobs, and enabling exports, and we are investing over $48 billion in the coming five years to meet the strong demand across our business in India and to help India achieve these priorities. We are inspired by Prime Minister Modi's vision of a Viksit and Atmanirbhar Bharat [Developed and Self-Reliant India], we are committed to being a long-term partner in India's growth story," Jassy said. Speaking to CNBC, Jassy said the capital investment plans were designed to capture an inflection point in cloud and AI markets. "When you have shifts that are this momentous, you want to make sure that you invest in such a way that you can pursue the opportunity as broadly for your customers as possible," he said. In January last year, Amazon proposed an $11 billion investment in the US state of Georgia. AWS indicated the funds would be directed toward datacenters in Butts and Douglas counties. In July last year US president Trump described India as a "dead economy," claiming that India's purchasing of energy from Russia had fueled Putin's war on Ukraine and instituting retaliatory tariffs among other trade restrictions. The tariffs were imposed under emergency-power (IEEPA) laws and were to be paid by US importers, of course, not Indian exporters. US importers and shippers last month began receiving tariff refunds after a Supreme court ruling that Trump overstepped his authority in enacting them. US Customs and Border Protection has not released specific data detailing exactly how many individual importers claimed refunds on goods shipped from India, but many have applied. In November last year, Amazon said it would invest up to $50 billion to expand AI and supercomputing capabilities for US government customers of AWS. This would add nearly 1.3 gigawatts of additional compute capacity across AWS Top Secret, AWS Secret, and AWS GovCloud (US) Regions, it said. ®