Schneider Electric announced on Monday that it will purchase PTC, an American industrial software firm, for $22.6 billion in cash. The transaction represents the largest acquisition in Schneider Electric's history and is expected to close in late 2025, pending regulatory approval. PTC, founded in 1985, develops software for computer-aided design and product lifecycle management. The deal aims to integrate PTC's design capabilities with Schneider's power and data infrastructure offerings, enabling software-defined automation in industrial systems. Schneider Electric's market value has more than doubled over the past four years, driven by rising demand for data center power and cooling solutions linked to artificial intelligence. The company previously acquired Cognite for $3.1 billion in June and invested $850 million for a 75 percent stake in Motivair in late 2024, a firm specializing in liquid cooling components. Schneider CEO Olivier Blum stated that combining PTC and Cognite would allow the company to enhance efficiency in data center design and operations by embedding automation early in development. The acquisition underscores Schneider's strategy to expand its role in the infrastructure supporting AI workloads, which require increasingly dense power delivery and advanced thermal management systems.

Etched, an artificial intelligence chip startup, is currently evaluating multiple financing proposals that would value the company at $40 billion or higher, according to individuals with knowledge of the discussions. The company has not confirmed these terms, and no agreement has been finalized.

These funding discussions follow Etched's previous $700 million raise in September at a $21 billion valuation. The new offers range from $40 billion to $50 billion, with terms still subject to change as negotiations progress. The company declined to comment on the specifics of the proposals.

Etched is focusing on developing proprietary hardware systems designed to accelerate AI inference workloads. A company representative indicated that a successful raise at the upper end of the proposed range could extend the company's operational runway to approximately 3.5 years.

The startup's technology centers on custom-designed components intended to improve the speed and efficiency of AI computation. According to Etched, its processors can handle more data tokens per second while reducing costs compared to competing solutions. This performance claim has attracted interest from clients including Jane Street, a quantitative trading firm that both invested in and deployed early versions of Etched's hardware.

Investor enthusiasm appears to stem from several factors: Etched's ability to attract talent from major semiconductor firms, its existing customer commitments totaling $1 billion, and its progress in manufacturing test chips through TSMC's facilities. Approximately 15% of Etched's 400-person engineering team previously worked at Nvidia, according to reports.

The company maintains production coordination capabilities in both Silicon Valley and Taiwan, where it operates a facility focused on semiconductor manufacturing oversight. Etched's leadership includes co-founders Gavin Uberti and Chris Zhu, who met during graduate studies at Harvard, and COO Robert Wachen, who previously worked alongside Uberti at another venture.

The startup has a history of rapid valuation increases, having previously raised $300 million at a $10.3 billion valuation in July before securing the larger $700 million round at $21 billion valuation in September. Such funding patterns reflect broader trends among high-growth technology companies seeking to scale specialized hardware development.

Marina Temkin, a technology reporter covering venture capital and startups, reports on this development for the publication. She previously worked analyzing venture investments for financial data firms and holds a professional investment certification. She can be reached via email or encrypted messaging for professional inquiries.

Danish authorities confirmed on October 5, 2026, that a cyber intrusion compromised the nation’s Central Person Register (CPR), a government database storing identity information for approximately 11 million individuals, including residents abroad and deceased persons. The breach, discovered on October 2, involved unauthorized access to records spanning decades, with data extending back to the 1960s. Minister of Digital Affairs Christina Egelund described the incident as a “serious incident” in an official statement, noting that names, addresses, and national identification numbers were among the stolen details. The CPR system, which holds tax and service-related identifiers for Denmark’s 6 million living citizens, was accessed through a vulnerability in a third-party vendor’s authorized data query process. No entity has claimed responsibility for the attack, though officials indicated the breach likely exploited a flaw in a contractor’s interaction with the database. The incident marks the largest data compromise in Danish history and follows prior breaches involving national identity systems in Turkey and India. Authorities have not disclosed whether affected individuals will be directly notified, citing ongoing investigations. The breach underscores growing risks to centralized public data repositories amid heightened global cyber threats targeting government infrastructure.

The iPhone 18 Pro, launched alongside Apple’s first foldable device, the iPhone Duo, represents the company’s most recent iteration of its premium smartphone line. Priced starting at £1,199 (€1,449/$1,199/A$2,099), the model carries a £100 increase over its predecessor, attributed by analysts to higher RAM configurations. The device weighs 211g and measures 150 x 71.9 x 8.75mm, making it slightly heavier than the iPhone 17 Pro while featuring a redesigned aluminium back that matches the body’s colour options.

A key visual change involves the Dynamic Island, which now accommodates three simultaneous on-screen activities—such as music playback, timers, and navigation—up from two on earlier models. The cutout for the Face ID components has been reduced, enabling this expansion without compromising functionality. The 6.3-inch Super Retina XDR display retains its 120Hz refresh rate and 460ppi resolution, with an added anti-glare coating intended to improve outdoor visibility.

Under the hood, the phone incorporates Apple’s A20 Pro chip, which analysts note offers approximately 27% faster central processing performance and 45% improved graphics capabilities compared to the previous generation. A revised cooling system supports sustained performance during intensive tasks like gaming. Battery endurance has been extended by about six hours relative to the iPhone 17 Pro, with real-world testing indicating up to two days of mixed use before requiring recharge; under heavy usage, approximately 40% battery remained after a full day.

The device runs iOS 27, which introduces refinements to the interface and faster app launching. A notable addition is an updated version of Siri, described as more task-oriented and less conversational than competing AI assistants. The new functionality includes contextual suggestions—such as calendar integration from messages—and limited interoperability with Apple’s ecosystem, including Spotlight search and camera-based insights. However, its capabilities are restricted to Apple’s native applications, limiting utility for users of third-party messaging or email platforms.

The camera system features a variable aperture on the main 48MP sensor, adjustable between f/1.78 and f/2.8, intended to enhance low-light performance and enable manual control over depth of field. While the physical impact on bokeh is modest due to the compact lens array, the feature supports creative effects like starburst rendering and limited foreground-background focus adjustments. The ultrawide and 4x telephoto lenses remain among the most capable in the segment, with the latter offering up to 8x hybrid zoom before digital enhancement. Video capabilities include post-recording cinematic effects, and a new AI-powered focus tracking system aims to maintain subject clarity during motion.

Sustainability disclosures indicate the device contains over 40% recycled materials and is rated highly for repairability by independent evaluators. Battery longevity is projected to exceed 1,000 charge cycles while retaining 80% capacity, with out-of-warranty repairs priced at £349 for the screen and £119 for battery replacement. Trade-in and recycling programs are available for non-Apple devices.

At launch, the iPhone 18 Pro was positioned against competitors such as the Google Pixel 11 Pro (£1,079) and Samsung Galaxy S26 Ultra (£1,359). Reviewers characterized the update as incremental rather than transformative, noting meaningful improvements in battery life, processing speed, and camera versatility, but limited perceptual differences in daily use. The device is widely regarded as the most refined iPhone to date, though its impact may be most noticeable to users transitioning from older generations.

Proponents highlight its compact form factor relative to larger flagship models, robust build quality, and extended software support. Critics point to the elevated price point, marginal gains over the previous model, and AI features that currently trail behind leading alternatives in responsiveness and contextual awareness.

Glasgow has become Europe's leading hub for small satellite manufacturing, but rising investment by other nations threatens its dominance.

In 2005, engineer Craig Clark returned to Glasgow after working abroad and founded Clyde Space, a company focused on CubeSats — compact, modular satellites measuring roughly 10cm per side. These devices are cheaper and faster to develop than traditional satellites and can be stacked or customized with sensors, cameras, or communication tools. Their data supports everyday functions like navigation, weather forecasting, and disaster monitoring.

Over the past two decades, Glasgow has built a full ecosystem around small satellite production, including manufacturing, launch services, and ground stations. The sector now employs more than 2,800 people across Scotland, up 81% since 2018, and generated £335 million in revenue in 2022–23. The University of Strathclyde has played a key role, partnering with companies like Clyde Space and AAC Clyde Space to train engineers and conduct research.

American firm Spire opened a European base in Glasgow in 2015, and other firms such as Craft Prospect and Alba Orbital have followed. According to Space Scotland, satellite manufacturing alone supported over 1,000 jobs in 2022–23. The industry’s growth was largely organic, driven by private initiative rather than state planning.

However, European competitors are increasing investment. Germany has pledged €35 billion for military space capabilities, including over 100 encrypted communication satellites. Finland’s ICEYE, which specializes in radar imaging satellites used in conflict zones like Ukraine, recently raised over €1 billion and received state backing. These efforts reflect a broader trend: governments are not only funding research but also becoming direct customers, seeking strategic autonomy in critical technologies.

This shift poses challenges for Glasgow. While the city still produces more small satellites than any other European location, companies like Alba Orbital have relocated headquarters to the U.S., and Spire has expanded manufacturing to Munich. Experts warn that without coordinated support, Glasgow risks losing ground as other regions scale up faster.

Still, signs of resilience remain. Simera Sense, a Glasgow-based satellite camera manufacturer, recently announced plans to open a new facility in the city, creating 20 skilled jobs. The city’s deep network of universities, engineers, and suppliers — developed over 20 years — remains difficult to replicate elsewhere.

Experts caution against simply matching foreign spending. Instead, they suggest focusing on filling market gaps and securing public contracts that provide stable revenue without stifling innovation. The UK Space Agency estimates that every pound invested in space generates nearly £8 in economic returns, supporting broader sector growth.

The central question now is not whether Glasgow can build satellites — it already does — but whether its emerging industry can sustain long-term growth at home. As one engineer noted, complacency could allow competitors to overtake a once-unique advantage. The legacy of the River Clyde, once known for shipbuilding, may yet extend into space — but only if the region adapts to a rapidly changing global landscape.

Google is testing an expansion of its Gemini AI assistant to manage outbound phone calls for personal communication, according to findings from Android Authority. The development follows a teardown of the Gemini app's code that revealed a new interface labeled "Gemini Calling," featuring example prompts such as "Call Mom and tell her I will be 15 minutes late" and "Call John and ask if he is coming for dinner."

The feature appears designed to deliver brief, text-based messages via phone calls rather than facilitate full conversations. Potential use cases include notifying contacts about delays, checking availability for plans, or reaching individuals who may not regularly check messages. Android Authority also identified granular permission settings within the Gemini app that could allow users to control which app functions Gemini accesses, including the ability to opt out of incoming automated calls.

Google has not confirmed a public rollout timeline for the expanded functionality. The company previously limited Gemini's calling capabilities to business-related interactions, but the discovery suggests broader ambitions for the AI to handle routine personal communications. This move aligns with Google's broader strategy to integrate Gemini across its ecosystem, though the scope of the feature and its eventual availability remain under development.

A newly identified class of vulnerabilities in the Model Context Protocol (MCP) is allowing attackers to move malicious instructions between AI agents within corporate networks. Security researcher Syed Anas Mohiuddin demonstrated that the protocol, designed to facilitate communication between AI tools, creates trust pathways that bypass traditional security checks.

Mohiuddin's testing revealed that five organizations—including Google, JP Morgan Chase, Weviate, Rapid7, the French government's digital directorate, and a U.S. federal agency—had MCP implementations vulnerable to prompt injection attacks. Rather than targeting the underlying language model directly, these exploits leverage the agent-to-agent communication channel.

The core issue lies in MCP's architecture: internal agents automatically trust instructions received through the protocol, and many implementations lack adequate guardrails. When an agent receives a crafted prompt, it may execute the request and forward it to other agents, creating a chain reaction. This "protocol pivoting" allows attackers to escalate privileges across systems.

One specific vulnerability, CVE-2026-97228 in Rapid7's network, received a severity rating of 2.7. Google's more critical flaw (rated 8) involved its MCP database toolbox initializing HTTP clients without proper redirect handling. Attackers could manipulate URL parameters to bypass security checks and make unauthorized network requests.

Google addressed the issue by implementing IP allow-lists and stricter validation at startup, requiring explicit authorization before processing certain requests. Other organizations have since applied similar fixes, including enhanced credential management and stricter trust boundaries between agents.

Experts note that these vulnerabilities stem from longstanding security patterns like server-side request forgery (SSRF), but the MCP context introduces new attack vectors. The incidents highlight a broader trend where organizations deploying AI agents have overlooked zero-trust principles, assuming internal communication channels are inherently secure.

Markus Vervier of X41 D-Sec described the technique as a subset of prompt injection rather than a novel attack vector, emphasizing that the underlying risks are not new but become more complex in agent-based architectures. Douglas McKee of Rapid7 stressed that any input from an LLM should be treated as potentially untrusted, regardless of its source within an organization.

The findings underscore the need for stronger security frameworks as AI agent ecosystems expand. While fixes are being implemented, the episode reveals gaps in how new communication standards are evaluated for safety before deployment.

A former Anthropic employee who publicly resigned last month told a New York City Council hearing that the pace of artificial intelligence development had outstripped safeguards. Jacob Coxon, who left Anthropic in December, said the industry’s culture of “move fast, break things” was unsuitable for technology with unprecedented power.

Coxon appeared before the council alongside former staff from Google, OpenAI and Meta, where they faced questions about potential risks. Their testimony coincided with growing concern among researchers about the long-term implications of advanced AI systems, a concern that has recently entered broader policy debates.

OpenAI policy representative Morgan Dwyer was asked to assess the level of risk posed by AI. She responded that she could not provide a precise measurement but emphasized that any risk was unacceptable. Council members pressed for greater clarity, with Public Advocate Jumaane D. Williams describing the answer as “God-awful, fantastically terrible.”

The hearing marked one of several recent efforts by city officials to examine the societal impact of AI, as companies and lawmakers grapple with how to balance innovation with oversight.

Meta is introducing two new audio-focused features for its smart glasses, beginning today. The updates, revealed during the company's Connect conference, include Audio Balance, which lets users adjust left and right channel volume independently, and Mono Audio, designed to combine stereo output into a single channel for improved clarity in certain listening scenarios. Both tools are now being rolled out to existing devices.

The company is also broadening access to navigation functions within its Meta AI app. Turn-by-turn pedestrian directions are now available more widely, offering step-by-step guidance for walking routes. Additionally, users can share their real-time location from their smartphone directly to their glasses, aiming to simplify location sharing during outdoor activities.

These changes reflect Meta's ongoing effort to refine the functionality of its wearable technology, focusing on practical adjustments rather than transformative features. The updates arrive amid broader industry interest in augmented reality interfaces, though they represent incremental improvements rather than a fundamental shift in the platform's capabilities.

OpenAI announced plans to embed an invisible watermark in text produced by ChatGPT and Codex within the European Union, aligning with the EU AI Act’s transparency mandates. The regulation, effective August 2, obliges AI developers to label synthetic content in a manner detectable by downstream systems.

The watermarking mechanism subtly modifies word selection patterns during generation, creating statistical signatures imperceptible to readers but identifiable through specialized tools. This approach ensures the marker persists through copying or pasting, though OpenAI confirmed it does not reveal user identities. Performance metrics indicated no significant degradation in model output quality when the feature operated.

Implementation will unfold gradually over upcoming weeks, initially targeting eligible EU users across all subscription tiers. API-accessible developers globally may activate the feature for select models immediately, though it remains disabled by default. OpenAI emphasized it will not enable watermarking as a universal default upon release.

Technical documentation detailing the methodology, dubbed textGrain, appeared alongside the announcement. Co-authored with researchers from the University of Pennsylvania and Yale, the framework explains how predetermined keys guide sequential word predictions to encode detectable patterns. Aggregated across hundreds of such adjustments, detectors can identify AI authorship using only the text and key.

OpenAI’s internal testing revealed watermark resilience challenges. Reducing word substitutions to 10% of content diminished detection rates from approximately 92% to 66%. Short-form texts, mathematical responses, and translated materials also presented greater detection hurdles. The company noted missing watermarks cannot definitively indicate human authorship, as brief passages or heavily edited content might lack detectable markers regardless of origin.

The practice differs from Anthropic’s recently adopted approach for Claude-generated content, which applies watermarking globally. Some Claude users criticized the move, arguing the tool merely executes instructions rather than creates original work. OpenAI previously developed internal watermarking capabilities but deferred public release due to concerns users might migrate to non-compliant alternatives, according to industry reporting.

The initiative forms part of broader industry adherence to the EU’s voluntary code of practice for AI-generated content, which includes commitments from Anthropic, Google, Meta, Microsoft, and OpenAI. Aditya Mehta, TechCrunch reporter covering artificial intelligence, holds support from the Tarbell Center for AI Journalism and earned his degree from UC Berkeley. He may be reached at aditya.mehta@techcrunch.com or via encrypted Signal communication at adymehta.74.

Jagex revealed a fourth RuneScape massively multiplayer online game built on Unreal Engine during RuneFest 2026, alongside an animated series adaptation. The project, internally titled RS4, remains in early development with no release date announced. Concurrently, the studio partnered with YouTuber Charlie White of MoistCr1TiKaL and Lyrical Media to produce the animated adaptation. Meanwhile, RuneScape Reignited, a new mode for RuneScape 3, launches on December 2 with a revised economy, Remastered Journey, and free access through December 12. A plugin API will release later this month following an early access period. In 2027, RuneScape 3 will receive three story arcs beginning with Hoard of Dragons in January, followed by City of Gnomes in spring and Secrets of Karamja in autumn, alongside an Agility rework. Old School RuneScape's roadmap includes Fractured Archive on October 20, The Chosen Commander and Dungeons of Gielinor in winter, Golden Gates to Menaphos in spring, Fairytale questline and Mining and Smithing rework in summer, and Tower of Xorphoros in autumn. RuneScape: Dragonwilds will receive the Luminance update in December, adding Prayer skill and Elves, with monthly and quarterly updates, Hunter and Slayer skills, and console-compatible CurseForge modding support planned for 2027. KitGuru notes the breadth of announcements at RuneFest 2026.

Microsoft disclosed earlier this year that it would separate several of its development studios, among them Undead Labs, Double Fine, Compulsion Games and Ninja Theory. While some of those studios have secured new publishing arrangements, two prospective transactions involving Ninja Theory collapsed, leaving the studio without a clear path forward. Recent reports indicate that Ninja Theory has initiated a round of workforce reductions that encompassed design, art and technical positions. The cuts follow a consultation period that may culminate in the permanent shutdown of the studio. At the time of the announcement, Ninja Theory was said to be developing a title tentatively referred to as Senua, a sequel set within the Hellblade universe that would emphasize combat mechanics over narrative depth. The project was unveiled during the Xbox Games Showcase earlier this summer, though internal documents suggest the reveal was intended to generate interest from potential acquirers rather than to signal an imminent launch. Industry observers note that the lack of a confirmed publishing partner has cast doubt on the game's completion, and that the studio's current restructuring efforts suggest a wind‑down of operations. Ninja Theory has not yet been formally dissolved, but the commencement of layoffs signals that the organization is moving toward closure. KitGuru reported that the situation raises questions about the future of the Senua project and its potential impact on the broader market.

Zain Javaid, 19, announced Monday that Ghost has emerged from stealth after securing $11 million in seed funding led by Andreessen Horowitz. The round also included participation from Abstract, Audacious Ventures, SV Angel and Nova. Javaid, who previously worked as a quant investor, shifted focus after the 2022 release of ChatGPT, which he described as the catalyst for his current work.

Ghost's first product, Core, is a $3,499 computer designed to run AI agents that perform tasks and manage personal data. The device includes an Nvidia RTX Pro 4000 SFF Blackwell GPU, bundled in the price, and runs software built specifically for AI functionality. Core operates without a screen, accessed through a phone or app, and features voice control via the Core app. Three models are preinstalled: Qwen-3.8-Next, Qwen-3.8-27B and Gemma-4-31B, with options to install additional models from Hugging Face or upload custom-trained versions.

The hardware and software are tailored for AI agents rather than general computing, unlike devices such as the Mac Mini. Javaid emphasized that Core receives over-the-air updates for new models, allowing users to adopt improvements without manual intervention. Updates can be configured to install automatically or require user approval.

Manufacturing occurs at a San Francisco facility, with preorders opening Monday and shipments scheduled for the last week of October. The system stores and processes all data locally, encrypting personal information to prevent access by cloud providers or third parties, including Ghost. A firewall monitors outgoing network requests to restrict unauthorized use of credentials, and the device continues functioning even if Ghost ceases operations. Javaid stated that all source code, logic and model weights reside on the device, ensuring user ownership of data and functionality.

An example use case involves Core detecting elevated resting heart rate, late-night work activity and seasonal flu trends to proactively notify a user of potential illness. Javaid noted that while Apple produces capable hardware, Ghost aims to address software limitations in how AI systems remember, reason and act across a user's digital life.

The founding team consists of Nicholas Chua, 19; Yifei Chen, 24; Gautam Sharda, 23; and Javaid. The company describes Core as a "brain in a box" focused on creating a product users love, independent of hardware competitors. All content is stored and processed on-device, with no reliance on external service maintenance.

Two sisters have launched Hot Girl Hotline, a startup that uses artificial intelligence to provide relationship advice tailored to young women. The company, founded by Sumrin Mudgil, a Stanford computer science graduate, and her sister Baila Mudgil, a former social media consultant, aims to offer a safer alternative to anonymous forums and general-purpose chatbots.

The startup addresses growing use of large language models for personal advice, a practice that has raised concerns after several lawsuits alleged manipulative interactions by general AI systems like ChatGPT, including cases linked to adverse mental health outcomes. Hot Girl Hotline emphasizes safety by incorporating safeguards such as referrals to crisis resources like 988 when conversations indicate risk.

Rather than positioning itself as an AI companion, the company describes its role as facilitating self-reflection through Socratic questioning, encouraging users to act in the real world rather than remain in cycles of rumination. The app avoids clinical claims and does not replace human interaction, according to Baila Mudgil.

User privacy is a stated priority: conversations are encrypted in transit and at rest, the platform does not sell or train on user data, and third-party AI models used adhere to zero-data-retention policies. Advice is personalized based on individual values, goals, and identity.

Hot Girl Hotline operates on a $9.99 monthly subscription model, with early adopters already paying for access. The company plans to expand into beauty and lifestyle domains. It is currently bootstrapped and will present at TechCrunch Disrupt 2026 in San Francisco, scheduled for October 13–15. The app is available on iOS and via web.

Sarah, a TechCrunch reporter since 2011 with prior experience at ReadWriteWeb and in industrial IT, covers the startup. She can be reached at sarahp@techcrunch.com or via encrypted Signal message at sarahperez.01.

HackerRank, a platform that helps companies assess technical talent, has made its AI interview system Chakra generally available to customers. The tool, which entered beta about six months ago, has already conducted over 500,000 interviews, according to the company. Early adopters include Snowflake, Snorkel, and Capgemini, while HackerRank also tested the system internally.

Chakra is designed to simulate real work rather than administer a standard coding test. Candidates receive a task within a live code repository and work in a shared canvas that includes an AI assistant. As they write code, Chakra observes their process and can ask follow-up questions about their approach, such as how they would adjust their solution under new constraints.

The system generates a detailed report for hiring managers after each interview. HackerRank says Chakra consolidates what was previously a three-stage process — recruiter screening, take-home assessment, and engineer follow-up — into a single interview session.

CEO Vivek Ravisankar told TechCrunch that the shift reflects a broader change in how technical ability should be measured. He noted that AI can now produce correct answers with minimal effort, making the traditional focus on output less meaningful. Instead, he said, employers need to evaluate the thinking behind a solution, which he described as "AI fluency" — the ability to frame problems for AI, assess its responses, and guide it toward a desired outcome.

Ravisankar argued that giving candidates access to AI during interviews reduces the temptation to use hidden tools, citing internal data showing fewer suspicious activity flags in Chakra interviews compared to conventional HackerRank assessments. The rate of flagged behavior varied by region and role level, but overall declined by 70 to 80 percent.

The company, founded in 2012 and backed by Y Combinator, has grown to serve over 3,000 enterprise clients, including Amazon, Nvidia, and Replit, and supports a global community of more than 30 million developers. Chakra represents a strategic pivot from HackerRank’s original model, which emphasized solving coding challenges correctly. Ravisankar compared the transition to Apple’s shift from iPod to iPhone, noting that while the legacy product still has value, the future lies in AI-integrated evaluation.

He emphasized that Chakra is intended to score candidates, not replace human decision-making. Final hiring decisions, he said, remain with people, while AI handles structured evaluation using consistent criteria set by employers. Ravisankar maintained that properly tuned AI can apply rubrics uniformly, reducing the influence of subjective factors such as educational background.

However, experts have raised concerns that even standardized AI evaluation can perpetuate or amplify bias if the training data or design reflects existing inequities. Automated hiring tools have already attracted regulatory attention; New York City, for example, mandates bias audits and candidate notification for certain AI-driven recruitment systems. HackerRank stated that compliance with such regulations is integrated into Chakra’s design and rollout.

The broader use of AI in hiring continues to provoke debate over transparency, accountability, and fairness. While some argue that algorithmic assessment can reduce human inconsistency, others warn that without careful oversight, it may entrench systemic disparities under a veneer of objectivity.

AMD's latest X3D processor lineup is receiving notable price reductions through bundled offers at Newegg. The Ryzen 7 9800X3D, typically priced higher, is currently listed at $469 but drops to $429 with a $40 promotional code. This bundle includes an MSI MAG CoreLiquid A13 240mm all-in-one liquid cooler valued at $70 and a digital copy of Onimusha: Way of the Sword, typically priced at $70. The effective cost reduction brings the processor's price down to approximately $290 when factoring in the included components. A similar bundle for the Ryzen 7 9850X3D is available at $469, offering the same cooler and game, though it remains priced above the 9800X3D's discounted rate. The Ryzen 9 9950X3D and 9950X3D2 models are also listed with bundled cooling solutions and game copies, priced at $699 and $899 respectively, though these do not reflect direct discounts on the processors themselves. Additional bundles include the Ryzen 7 7800X3D at $348 with a free 850W power supply, down from $416, and the Ryzen 7 7700X3D at $309 with a comparable power supply offer. These promotions appear to be part of a broader pricing strategy by AMD to clear inventory ahead of potential next-generation releases, with discounts observed across multiple retailers. The bundles are available for a limited time, though specific end dates have not been disclosed by the retailer.

Elon Musk confirmed plans to rename SpaceXAI, the entity formed from the merger of SpaceX and xAI, to SpaceXSI after President Donald Trump issued an executive order directing federal agencies to replace 'Artificial Intelligence' with 'Super Intelligence' in official communications. The directive, signed on September 29, applies to non-statutory documents and requires agencies to adopt 'SI' as the preferred term. Musk endorsed the change publicly, stating 'No more AI. SI. It's better' and affirming that SpaceXAI would adopt the new designation. He did not specify a timeline for implementation. At the time of reporting, the domain spacexsi.com was listed for sale. The rebrand appears to align with Musk's broader pattern of corporate renaming, including Twitter's transition to X in 2023. Trump's order does not mandate private companies to alter their names, though it encourages the terminology shift across federal operations. Musk has previously used 'superintelligence' to describe SpaceX's ambitions, though the company has not clarified how the new label will impact its technical or operational framework. OpenAI has not indicated plans to rebrand as 'OpenSI,' and no official updates have been released regarding SpaceXSI's structure or objectives. The move reflects a semantic pivot rather than a substantive change in SpaceX's activities, which remain focused on aerospace development and space exploration.

Contactless switch designs are gaining traction in the keyboard market, offering alternatives to traditional mechanical switches that rely on metal leaf contact. These newer technologies—optical, Hall-effect, tunnel magnetoresistance (TMR), and electrostatic capacitive (EC) switches—use light, magnetic fields, or capacitance to register keypresses without physical metal-to-metal contact.

The shift toward contactless switches stems from claims of longer durability and reduced susceptibility to oxidation, which can degrade traditional mechanical switches over time. While manufacturers often highlight these benefits, the practical advantages vary by design. Some switches, such as optical and Hall-effect types, eliminate debounce delays by avoiding mechanical bounce, potentially reducing input latency. However, actual performance gains depend on multiple factors, including polling rate and switch travel, and may not be noticeable in everyday use.

Optical switches, introduced in consumer keyboards around 2016, use an infrared light beam interrupted by a moving stem to register input. Companies like Razer have expanded their capabilities with features such as adjustable actuation points and analog input, which allows keys to register pressure-sensitive commands. Hall-effect switches, revived from 1990s controller technology, use magnets and magnetic sensors to detect keypresses. TMR switches, a newer development, employ tunnel magnetoresistance sensors, offering higher sensitivity and potentially finer control over actuation. Inductive switches, which use electromagnetic induction, are also emerging, though they remain less common.

Customization is a key feature of many contactless switches. Some allow users to adjust actuation points or assign multiple functions to a single key press, such as triggering different characters based on how far a key is depressed. Analog input support, which captures the depth of a keypress, is another differentiator, enabling gaming experiences similar to joysticks. However, software-based actuation settings do not always align precisely with physical behavior, as testing has shown discrepancies between intended and actual actuation distances.

Compatibility and availability present challenges. Hot-swappable keyboards with contactless switches are limited, and replacing switches can be difficult due to proprietary designs or alignment requirements. Hall-effect and TMR switches often require specific PCB configurations and magnetic polarity, complicating interchangeability. Additionally, while some companies market TMR keyboards under the broader “HE” label for recognition, the underlying technologies differ in power consumption, sensitivity, and noise resistance.

Price and accessibility vary widely. TMR switches currently carry higher costs due to complex manufacturing, while optical and Hall-effect options are increasingly available in prebuilt gaming keyboards. Despite marketing claims about speed and precision, the real-world impact on typing or gaming performance depends on system-wide factors like hardware and firmware.

For most users, the choice of switch technology may come down to feel, sound, and compatibility rather than measurable performance gains. While contactless designs offer theoretical benefits, their practical superiority remains context-dependent, and traditional mechanical switches with high durability ratings can offer comparable longevity.

Lucebox has unveiled the Zero 495, a high-end AI workstation featuring AMD's Ryzen AI MAX+ 495 processor and Radeon AI PRO R9700 graphics accelerator. The system is offered in two memory configurations: 128 GB and 192 GB of unified memory, with the latter expandable to 224 GB when paired with the dedicated GPU. Both versions support up to 4 TB of NVMe storage and are housed in an extruded aluminum chassis measuring 340x110x320 mm and weighing 6 kg. The workstation includes dual USB4 ports, dual USB Type-A, HDMI 2.1, dual DisplayPort 2.1, 5GbE Ethernet, Wi-Fi 7, and Bluetooth connectivity. A 1000W 80 Plus Platinum SFX power supply provides power, with peak draw around 500W under load.

The Zero 495 is designed for local AI workloads, targeting enterprises and developers shifting away from cloud-based inference services. It integrates with Lucebox Engine and ships with preinstalled software supporting OpenAI and Anthropic API compatibility, along with root access via SSH. An optional cluster kit includes an Intel E810 100 GbE network card to interconnect multiple units, priced at $799.

Pricing begins at $5,999 for the initial 200 units, reflecting a $1,000 discount from the standard $6,999 retail price. This base configuration includes 128 GB of memory (expandable by 32 GB via the R9700), 2 TB of storage, and the Radeon AI PRO R9700 GPU. Additional upgrades include a 192 GB memory option at $2,000 extra and 4 TB storage for $250. The 192 GB base model therefore starts at $7,999, with the top configuration reaching $9,249.

AMD announced the Ryzen AI MAX 400 series SoCs recently, positioning them as part of a broader push toward localized AI computing. The Zero 495 is positioned as a distinct alternative to systems like NVIDIA's DGX Spark, which starts at $7,000 for a 128 GB variant. Lucebox claims performance exceeding the DGX Spark in specific benchmarks, reporting over 2x speed-up in token generation for models like Qwen3.8-27B and DeepSeek V4 Flash using ROCm software.

The company has published performance comparisons suggesting superior efficiency in certain workloads, particularly when leveraging ROCm over Vulkan-based rendering. These claims are based on internal testing and may vary by application. The Zero 495 is scheduled to begin shipping in January 2027, targeting users requiring high-capacity memory and local AI processing without reliance on external cloud infrastructure.

Apple’s design leadership structure has shifted following the departure of longtime executive Jeff Williams and the recent transition of John Ternus into the role of CEO, according to reporting by Bloomberg’s Mark Gurman. Since taking the helm, Ternus has taken an active role in overseeing both industrial design and human interface teams, engaging directly with design studios on a frequent basis. Individuals familiar with the departments noted that he visits the design studios multiple times per week, and on some occasions, more than once within a single day. This level of involvement contrasts with the approach of his predecessor, Tim Cook, who reportedly made studio visits approximately once a month during his tenure as CEO.

Ternus’ background in hardware engineering appears to inform his hands-on approach, though he does not hold a formal design education. His increased presence in design operations has coincided with efforts to replenish talent within the teams, which have experienced notable departures, including Alan Dye, who joined Meta, and several engineers recruited by external firms such as OpenAI. In response, Apple has promoted several junior designers and brought in new hires from outside the company to strengthen design capacity.

While the organization has not appointed a successor to former Chief Design Officer Jony Ive, who left in 2019, Apple has identified Molly Anderson and Steve Lemay as leads for industrial design and human interface, respectively. Ternus is described as the central decision-maker in design-related matters, ensuring continuity in strategic oversight despite the absence of a singular design executive.

The shift reflects a broader reorganization of leadership responsibilities following Ternus’ ascension to CEO. Though his direct involvement in design processes marks a departure from prior practices, Apple has not indicated plans to establish a formal Chief Design Officer role. Instead, the current model emphasizes distributed leadership with Ternus at the center of design governance.

This evolution in leadership style underscores a continued emphasis on design within Apple’s operational framework, even as the company navigates transitions in executive structure and talent retention.

Google and its partners have unveiled the initial lineup of devices branded as Googlebooks, marking the first wave of laptops to carry the new designation. Five models from Acer, Asus, Dell, HP, and Lenovo are now available for preorder, with retail availability scheduled for October 4th. Starting prices range from $899 to $1,299, placing them in the same segment as the MacBook Air and comparable Windows laptops. Each device includes a 12-month subscription to Google AI Pro, 5TB of cloud storage, and limited-time trials of YouTube Premium, Adobe Photoshop, and CapCut. Google has also committed to providing up to 10 years of software updates for these models.

The Acer Googlebook 14 is the only convertible model in the initial release, featuring a design reminiscent of the Acer Chromebook Plus Spin 714. It offers up to 16 hours of battery life and launches at $899, though Acer has indicated this price is temporary and subject to adjustment based on supply conditions and component availability. The model includes an Intel Core Ultra 5 or Ultra 7 processor, 16GB of RAM, 256GB or 512GB of storage, a 14-inch 2.8K OLED touchscreen, and a thickness of 13.9mm with a weight of 2.5 pounds. It provides one Thunderbolt 4 port, two USB-C ports, one USB-A port, and a headphone jack, along with a fingerprint sensor.

Asus introduces the Googlebook 14 with a metallic gray-blue chassis inspired by its ExpertBook Ultra line. Weighing 2.2 pounds, it is the lightest of the initial models but not the thinnest. It includes an HDMI port alongside two USB-C and two USB-A ports, and is offered starting at $1,299. Specifications include an Intel Core Ultra 5 or Ultra 7 Series 3 processor, 16GB or 32GB of RAM, 256GB or 512GB of storage, a 14-inch 3K OLED touchscreen, and a thickness of 14.0mm. It also features a fingerprint sensor.

Dell’s offering, the XPS Googlebook, mirrors the design of the XPS 13 with a light gold shimmer finish and Google’s software integration. Priced at $999 through a Best Buy special offer, it includes a fingerprint reader and two USB-C ports. It is powered by a Qualcomm Snapdragon X Elite processor, available with 16GB or 32GB of RAM and 512GB of storage, and features a 13.4-inch 2.5K touchscreen display measuring 12.7mm thick and weighing 2.2 pounds.

The HP Googlebook 14 starts at $1,299 and includes both facial recognition and a fingerprint reader. It offers up to 1TB of storage, the highest capacity among the initial models, and features a dark blue chassis similar to the HP Spectre. Connectivity includes two USB-C ports and a headphone jack, with a 14-inch 2.8K OLED touchscreen, 14mm thickness, and a weight of 2.74 pounds. It is powered by a Qualcomm Snapdragon X Elite processor with 16GB or 32GB of RAM.

Lenovo’s Googlebook 15, starting at $1,149, features the largest display at 15.3 inches and includes an HDMI port, unlike the other models. It is the only device in the initial lineup with a white chassis featuring a satin finish. It includes an Intel Core Ultra 5 Series 3 processor, 16GB of RAM, 256GB or 512GB of storage, a 15.3-inch 2.8K OLED touchscreen, and measures 15.95mm thick with a weight of 2.84 pounds. It provides two USB-C ports, one USB-A port, one HDMI port, and a headphone jack, along with face authentication and a fingerprint sensor. An updated starting price was added on September 28th.

The introduction of Googlebooks represents a strategic expansion into hardware by Google, aiming to establish a standardized platform for AI-integrated laptops. The initiative includes partnerships with major manufacturers and emphasizes long-term software support, which could influence consumer expectations for device longevity and ecosystem integration.

A small stationery retailer in Japan, Write Love Pens, removed AI-generated promotional images after criticism from an online forum dedicated to fountain pens. The shop had used algorithmically produced visuals — including a depiction of a pen with two clips attached to its cap — to market curated surprise boxes. Following a wave of negative commentary on a popular Reddit thread focused on fountain pens, the business issued an apology and substituted the digital artwork with hand-drawn illustrations created by the owner. The transition was confirmed by the shop’s operator, who cited community feedback as the catalyst for the change. No further public statements have been released.

The reserved domain example.com, designated by IANA for documentation purposes, received a technical update on 28 September 2026 that introduced multi-language support through JavaScript-driven content rotation. The change replaces the previous static English page with dynamic display of content in English, Arabic, Chinese, French, Russian and Spanish, cycling every five seconds. Each language transition employs a character-by-character opacity animation where individual letters fade in sequence using staggered CSS transition delays. The update also incorporates an SVG book icon within the redesigned interface.

The shift follows a series of infrastructure modifications over the domain's 27-year history. Earlier in 2026, the site transitioned from Cloudflare to a new hosting arrangement, though specific provider details remain unspecified in public records. This update aligns with broader patterns of periodic revisions, including earlier changes in January 2025 when the Server header disappeared after a move from EdgeCast CDN to Edgio, and a subsequent migration to Cloudflare later that year.

IANA's official statement explains that such modifications aim to reduce bandwidth consumption while maintaining the domain's primary function as a placeholder for documentation examples. The organization emphasizes that example.com serves solely as a reference resource for illustrative purposes, not as a general-purpose endpoint for testing or monitoring systems. Traffic to the domain is largely automated, with most users not retrieving the JavaScript file that enables the new language features.

The domain's evolution reflects its technical purpose rather than commercial development. Its earliest recorded version appeared in the Wayback Machine on 20 January 2002, displaying a list of reserved domain names. Subsequent updates included the addition of RFC 2606 documentation links in February 2003, expansion to include .edu domains in July 2010, and a 2013 redesign that shifted from redirecting to IANA's site to hosting an independent page via EdgeCast CDN. A 2019 update modernized the font stack and refined copy, while a 2025 infrastructure switch removed server identification headers.

The recent multi-language implementation represents the most visible functional change in recent years, though it maintains the domain's core principle of providing a non-operational example for technical documentation. The transition animation technique, while visually distinctive, serves primarily as a demonstration of CSS capabilities rather than a user-facing feature. The domain continues to carry warnings against reliance for operational testing, as server availability can be inconsistent.

The changes underscore the technical stewardship required to maintain even seemingly minor internet infrastructure elements. While the visual presentation has evolved through multiple iterations, the underlying purpose remains consistent: to serve as a standardized reference for documentation without requiring registration or active service provision.

Monitoring of the domain's performance history reveals patterns of intermittent updates separated by extended periods of stability. These modifications, while technically significant, occur within the narrow scope of maintaining a reserved resource for technical writing purposes rather than representing broader shifts in internet governance or commercial web practices.

Scientists have identified two materials that may enable next-generation spintronic devices by combining semiconductor behavior with magnetic properties at room temperature. The discovery, reported by an AI-assisted research team, centers on two compounds with near-zero net magnetism but distinct spin-sorting capabilities.

The search focuses on materials that fall between ferromagnets and antiferromagnets. Ferromagnets, like those on refrigerator doors, generate external magnetic fields that interfere with nearby components. Antiferromagnets lack such fields but cannot distinguish electron spin states for data storage. Researchers aim to develop Luttinger compensated (LC) materials, where opposing magnetic atoms cancel each other's net magnetism while allowing spin separation.

One candidate, YBaMnFeO5, was computationally designed as an LC semiconductor. Simulations suggest it could have a 2.35 eV band gap with spin-separated energy windows of 1.0 eV for holes and 1.4 eV for electrons. The material reportedly maintains magnetic order up to approximately 420 K in raw simulations, or 490 K after calibration. However, its crystal structure requires precise atomic arrangements that may degrade during synthesis above 950 K, potentially compromising spin-sorting properties.

The second material, KV[Cr(CN)6], was first synthesized in 1999. It belongs to the Prussian blue family and features chromium and vanadium ions arranged to cancel magnetic moments. Computational analysis indicates it has a 2.1 eV band gap with spin-separated windows of 2.6 eV for holes and 1.6 eV for electrons. Crucially, its structure inherently maintains atomic site segregation, unlike the designed candidate. A 1999 sample remained magnetically ordered up to 376 K (103 °C), exceeding room temperature.

Simulations using hybrid functional methods (HSE06) predict both materials could operate without significant stray magnetic fields, reducing interference with adjacent components. However, practical implementation faces challenges: the 1999 sample contains water in its pores, resulting in a minor residual magnetic moment (0.125 Bohr magnetons per formula unit). Simulations disagree on the impact of hydration on spin-separation windows, with more accurate methods suggesting reduced but still viable performance.

The findings emerge amid broader efforts to advance spin-based electronics. A 2025 study previously identified room-temperature LC semiconductors as an open challenge, noting that many predicted candidates lose magnetic order below ambient conditions. KV[Cr(CN)6] represents a material that already meets this criterion, having been synthesized decades ago. Researchers emphasize that direct experimental validation of its spin-sorting properties remains necessary.

Both materials require further study to assess manufacturability and real-world viability. The computational work, including simulation files and analysis tools, has been made publicly available for independent verification. This development contributes to ongoing research into magnetic semiconductors that could enable faster, more energy-efficient data storage and processing technologies without compromising device compatibility through stray fields.

Developers can now install the third macOS Golden Gate 27.2 beta, which Apple began testing two weeks after issuing the second beta. The update is accessible through System Settings, General, Software Update, and the option to enroll in macOS Golden Gate betas. This build adds support for Siri in additional languages and provides new brightness and volume sliders. Apple anticipates a broader release of macOS Golden Gate 27.2 toward the end of October.

The latest beta follows Apple's recent rollout of iOS 27, which introduces expanded Siri capabilities and other refinements. That update began distribution on September 13, 2026, though availability may vary among users. Concurrently, Apple announced plans to modify the Full Disk Access setting in macOS due to emerging privacy concerns linked to AI agents such as Meta's Muse and OpenAI's Dots.

Code analysis by developer "pdfu" revealed private frameworks in iOS 27 and macOS Golden Gate that enable third-party AI models, including Claude, to integrate with Siri through a feature termed Model Delegation. This suggests a strategic shift toward modular AI components within Apple's ecosystem.

The updates reflect ongoing efforts to enhance system functionality while addressing technical and privacy considerations associated with artificial intelligence integration.

A new add-in-card from JGINYUE brings AMD's B650 chipset to a PCIe expansion board, offering users additional connectivity without replacing their motherboard. The card, priced at 388 yuan or approximately $58, includes four M.2 slots for NVMe SSDs, a 10 Gbps USB Type-C port, two USB 3.0 Type-A ports, and a 2.5 Gbps LAN interface. It also features a 6-pin PCIe power connector intended to supply auxiliary power for the USB functionality, even when the card operates solely with M.2 drives. The design incorporates a 4-pin fan header and two mounting standoffs to support an optional 40mm fan for cooling, though the fan itself is not included. JGINYUE has indicated plans to release a version supporting 20 Gbps connectivity in the future. The product reflects a growing trend of expanding chipset capabilities through modular expansion solutions.

A tech blogger has adapted Blu-ray M-Discs for use as a physical, offline backup system distributed to trusted individuals.

The approach treats the discs like a modern analog of key-sharing: copies of irreplaceable digital assets are burned onto M-Discs, sealed in durable cases, and entrusted to friends or associates for safekeeping. The creator described the method as a "last-resort, decentralized backup" to mitigate risks of single-point failure, such as fire or hardware degradation affecting both primary storage and conventional backups.

The process involves curating essential data — including digitized documents, music, and curated photo archives — onto three M-Discs, which are then placed in B&W Outdoor Case Type 500 containers. Each case holds three to four discs, padded with inert PE foam to prevent movement during transit. Seals bearing unique serial numbers and barcodes are applied to indicate tampering, and visual inspections are scheduled during periodic replacement cycles.

Labels for the cases were custom-designed and printed externally. A desiccant packet was included for moisture monitoring. The author emphasized that data within the discs is password-protected and may include credentials intended as a digital legacy tool, though access to such information remains at the custodian’s discretion.

The initiative reflects a broader trend of exploring analog alternatives for digital preservation amid concerns over cloud service reliability, hardware obsolescence, and centralized storage vulnerabilities. While the method relies on physical trust and long-term custodianship, it does not address scalability or institutional verification of data integrity.

The project’s documentation includes specifications for materials and sourcing, though no timeline or recipient list was disclosed. The creator invited feedback on potential refinements but did not outline plans for public distribution or formalized handover protocols.

Microsoft's latest Windows 11 26H2 update has introduced a defect in the AC-3 audio decoder that can cause applications relying on Windows' built-in audio processing to crash or fail to launch. The issue, documented by the company after the September 22 update, affects all three current development branches of Windows 11 — 26H2, 25H2, and 24H2 — which share a common code base. Microsoft attributes the problem to the AC-3 decoder component, which is primarily used by legacy software that still depends on the operating system's native audio encoding. Modern applications that implement their own decoding logic are unaffected. The bug has been observed in games, media players, and certain productivity tools, though the company has not specified which programs are most vulnerable. No patch or workaround has been released, and users are advised to uninstall problematic applications until a fix is available. Microsoft has confirmed it is developing a resolution for a future update but has not provided a timeline. The issue follows a series of audio-related problems introduced in recent patches, including a September Patch Tuesday update that required an out-of-band fix for RDP and Hyper-V, though those did not fully resolve prior audio instability. The company has not indicated whether the AC-3 decoder flaw is isolated or part of a broader pattern in its audio stack maintenance.

A Florida woman is facing felony charges after an AI chatbot entry she made was flagged as a potential threat and reported to authorities. Carli Michelle Heller of Bonita Springs allegedly wrote on September 26 that she would attack the local sheriff's office, describing the plan as a "shoot up" of the facility. The entry was created using Anthropic's Claude chatbot, which she referred to as a personal diary.

Claude's safety mechanisms detected the content and escalated it to a human reviewer. After determining the statements constituted a credible threat, the reviewer forwarded the information to law enforcement. Deputies located Heller at her residence and took her into custody without incident. An intelligence detective from the Lee County Sheriff's Office then assumed responsibility for the investigation.

Heller now faces a charge of making a written threat of violence under Florida Statute 836.10, which classifies such communications as a second-degree felony when they involve threats of mass violence or terrorism and are transmitted in a way that others can view them. The statute applies to both written and electronic records, regardless of the medium used.

Anthropic has stated it may disclose user information in limited emergency situations when it believes there is an imminent risk of death or serious physical injury. This policy was cited in the company's handling of the case.

The incident adds to growing scrutiny of how AI platforms handle potentially dangerous user input. It follows a separate lawsuit filed by the state of Florida against OpenAI and Sam Altman, which alleges the company contributed to real-world harm through its chatbot's outputs. That case references the 2025 Florida State University shooting as one example of alleged harm.

Earlier, British Columbia authorities sued OpenAI after an 18-year-old former student, Jesse Van Rooselaar, was identified as having discussed gun violence in conversations with the company's systems. While OpenAI's safety team flagged those discussions, the company did not escalate them to law enforcement, citing insufficient thresholds for intervention.

In a related development, investigative documents revealed that human reviewers working on Microsoft Copilot's image editing feature have access to user prompts, uploaded images, and AI-generated modifications. While some of this material has included sexual, disturbing, or potentially illegal content, reviewers are instructed only to assess output accuracy, not to monitor or flag the content itself.

The case underscores broader concerns about the intersection of artificial intelligence, public safety, and digital privacy. It highlights the point that user-generated content shared with AI systems may not remain private, particularly when it involves potential threats to public order or individual safety.