News World That's Worth

Anthropic "Claude Code" Leak Unveils Hidden Features: AI "Dreaming" and Stealth Protocols Exposed Last night, a significant leak occurred when Anthropic Claude Code project was accidentally uploaded to the NPM registry. The incident triggered a massive wave of technical deconstructions from the global developer community. Alex Kim , a prominent AI developer, is among the experts who have shared a detailed breakdown of the source code, revealing several unconventional features hidden from public view. Inside the Leaked Source Code The analysis uncovers sophisticated mechanisms that highlight Anthropic’s unique approach to AI tool development: Anti-Distillation Defense: Claude Code includes advanced logic to prevent users from scraping its outputs to train rival models. This includes the ability to decode obfuscated text to bypass direct interaction logging and the insertion of "dummy" tool calls to confuse data-harvesting bots. Stealth Identification: ...

Axios Hit by Critical Supply Chain Attack: Malware Injected into 100M-Download Library Axios , one of the world's most popular HTTP clients with over 100 million weekly downloads , has fallen victim to a sophisticated supply chain attack. Threat actors successfully injected a Remote Access Trojan (RAT) into versions 1.14.1 and 0.30.4 . The malicious code remained active for approximately two and a half hours before being detected and removed. The Mechanism of Attack: Dependency Poisoning The attackers utilized a "Dependency Poisoning" technique. They first published a malicious package named plain-crypto-js and subsequently modified the Axios source code to include this package as a dependency. While Axios typically uses GitHub Actions for automated builds and publishing, these compromised versions were uploaded directly from a manual account associated with the email ifstap@proton.me . How the Malware Operates The primary Axios code did not contain the malware direct...

WHOOP Raises $575M in Series G, Valuation Hits $10.1 Billion as Global Expansion Accelerates WHOOP , the human performance company, has officially announced a massive $575 million Series G funding round, propelling its valuation to $10.1 billion . The round was led by Collaborative Fund , with significant participation from a diverse group of institutional giants including Qatar Investment Authority, Mubadala Investment Company, Abbott, Mayo Clinic, and Macquarie Capital . The Power of Star-Studded Backing Beyond institutional support, the round saw personal investments from global icons such as Cristiano Ronaldo, LeBron James, and Rory McIlroy . WHOOP states that this convergence of healthcare institutions and elite athletes reflects the brand’s growing influence as it transitions from a niche fitness tool to a globally recognized health platform. Explosive Growth and Financial Stability Operating on a subscription-based model, WHOOP now boasts over 2.5 million members worldwide. In...

Oracle Begins Major Workforce Reduction Amid AI Infrastructure Spending Surge Oracle has officially notified its employees of a new round of layoffs, though the exact number of affected positions has not been disclosed. Impacted staff in the United States and India began receiving termination emails on Tuesday morning local time, with the cuts taking effect immediately. As of May last year, Oracle’s global workforce stood at approximately 162,000 employees . The Financial Pressure of AI The move follows a report from financial firm TD Cowen , which previously estimated that Oracle might need to lay off up to 30,000 employees and divest certain assets. These drastic measures are seen as a strategic pivot to offset the massive capital expenditures required to build and maintain the company's expanding AI infrastructure . Historically, Oracle's primary revenue came from selling database management software. However, the company is aggressively pursuing a leading position in the...

OpenAI Shatters Records: Valuation Soars to $852 Billion After Massive $122 Billion Funding Round OpenAI has officially updated its latest funding figures, revealing a staggering total of $122 billion  a significant jump from the $110 billion previously announced in February 2026. This capital injection brings the company’s post-money valuation to an unprecedented $852 billion , marking the largest investment round ever recorded for a private startup. A Consortium of Tech Giants and Institutional Investors The round was led by SoftBank , with major participation from venture capital powerhouses Andreessen Horowitz and D. E. Shaw Ventures . Crucially, OpenAI’s core strategic partners   Microsoft, NVIDIA, and Amazon  all maintained their involvement, further solidifying the company's position at the center of the AI ecosystem. A First for Retail and Institutional Inclusion In a strategic first, OpenAI confirmed that $12,000 million of the additional funding was secured t...

Google Disrupts the AI Video Market: New "Veo 3.1 Lite" Offers High-Quality Generations at 1/8th the Cost Google has officially expanded its generative video lineup with the launch of Veo 3.1 Lite . Positioned as the most cost-effective model in the series, Veo 3.1 Lite aims to democratize high-fidelity video creation by significantly lowering the barrier to entry for developers and creators. Unbeatable Pricing Structure The standout feature of Veo 3.1 Lite is its aggressive pricing. Google has managed to slash costs significantly compared to its more powerful siblings: 3x Cheaper than Veo 3.1 Fast . 8x Cheaper than the flagship Veo 3.1 (Large) model. For a 720p video , the price is set at just $0.05 per second . In comparison, Veo 3.1 Fast currently costs $0.15 per second (slated to drop to $0.10 on April 7), while the flagship Veo 3.1 remains at $0.40 per second. Performance Without Compromise Despite the "Lite" branding, Google confirms that the generation sp...

Google Drive Level Up: New AI-Powered Ransomware Protection is 14x More Effective As the world increasingly shifts toward cloud storage solutions like OneDrive, iCloud, and Google Drive , the threat of Ransomware has evolved to target these digital vaults. Recognizing this growing risk, Google has officially unveiled a powerful new defense system within Google Workspace to safeguard user data from malicious encryption. The AI Advantage According to the latest Google Workspace Updates , the tech giant has integrated an advanced detection and recovery tool powered by state-of-the-art AI. Google claims this new iteration is 14 times more effective at identifying ransomware threats compared to its previous security protocols. Damage Control & Instant Recovery The core objective of this feature is to minimize data loss. When the system detects suspicious activity such as bulk file encryptions it triggers several automated responses: Automatic Sync Suspension: The system temporarily ...

GitHub Silences Copilot "Spammy" Pull Request Footers After Developer Backlash GitHub has officially disabled a controversial feature that appended promotional suggestions to the end of GitHub Copilot pull request (PR) messages. The decision comes after a wave of complaints from the developer community, who reported seeing tens of thousands of automated "spam" messages cluttering their workflows. The Escalation of Automated Messaging Martin Woodward , Vice President of Developer Relations at GitHub, confirmed that the feature has been deactivated. While Copilot has included footer suggestions for some time, the issue escalated recently. Previously, these messages only appeared in PRs fully authored by Copilot. However, a recent update caused Copilot to inject these footers into every pull request where the AI was mentioned, leading to an exponential increase in automated noise. Indirect Advertising Concerns The backlash was further intensified by the content of t...

  Apple Intelligence Spreads to China? Recent iOS 26.4 Update Sparks Rumors Amid Regulatory Hurdles Reports regarding the rollout of Apple Intelligence for Apple users in China have been circulating since last year. However, the official launch remains stalled due to the stringent approval processes required by Chinese regulatory authorities. The iOS 26.4 Incident The tech community recently buzzed with excitement when several Apple users in China reported receiving notifications to install Apple Intelligence immediately after updating to iOS 26.4 . Despite the lack of an official announcement, screenshots of the setup process quickly went viral. However, the excitement was short-lived. Industry expert Mark Gurman from Bloomberg noted that the feature was swiftly retracted, likely appearing as a technical error. According to Gurman, while Apple has finalized the system and is "ready to go," the company is still awaiting the final green light from Chinese regulators. Local...

Meta Unveils "Instagram Plus": A New Premium Subscription with Stealth Viewing and Advanced Story Tools Meta has begun testing a new subscription tier for Instagram, aptly named "Instagram Plus" Currently rolling out to select users in specific markets, this premium service offers a suite of advanced features designed for power users who want more control over their Story interactions and privacy. Exclusive Features of Instagram Plus: Custom Story Lists: Users can now create multiple categorized lists for Story viewers, moving beyond the single "Close Friends" (green circle) limitation. Re-watch Insights: For the first time, subscribers can see exactly who has viewed their Stories multiple times. Viewer Search: A dedicated search bar within the viewer list allows users to quickly find specific names. Extended Expiry: Stories can now remain active for longer than the standard 24-hour window. The "Ghost Mode" Highlight: Perhaps the most anti...

  Mistral AI Secures $830M Loan to Build Massive "Grace Blackwell" Data Center in Paris French AI powerhouse Mistral AI has officially announced a significant $830 million financing package from a consortium of leading financial institutions. The list of lenders includes Bpifrance, BNP Paribas, Crédit Agricole CIB, HSBC, La Banque Postale, and MUFG , among others. This capital injection is earmarked for a massive expansion of its high-performance data center infrastructure in Paris. The Power of NVIDIA’s GB300 The centerpiece of this expansion is the deployment of NVIDIA’s Grace Blackwell architecture . The Paris facility will boast: Total Compute Power: 44 Megawatts (MW). Hardware: 13,800 units of the state-of-the-art NVIDIA GB300 GPUs. This deployment places Mistral at the forefront of AI hardware, utilizing the most advanced chips available for training and inferencing next-generation Large Language Models (LLMs). Strategic Sovereignty: The 2027 Roadmap Mistral has re...

Indonesia Becomes First in SE Asia to Ban Social Media for Under-16s, Following Global Trend Indonesia has officially commenced the enforcement of a landmark law prohibiting minors under the age of 16 from using social media platforms. This move makes Indonesia the first nation in Southeast Asia to implement such a strict digital age limit, following through on a policy framework announced earlier this month. Platforms Under Scrutiny The new regulation targets major digital platforms, requiring them to implement robust age-verification systems for users within Indonesia. The affected services include YouTube, TikTok, Facebook, Instagram, Threads, X (formerly Twitter), Bigo Live, and Roblox . The Indonesian government has granted these platforms a transition period to align their technical infrastructure with the new legal requirements. Austria Joins the Global Movement Following Indonesia’s lead, Austria has become the latest nation to announce plans for a total social media ban for ...

Microsoft Announces "Performance-First" Overhaul for Windows 11: Fixing File Explorer Lag in 2026 After years of persistent user feedback regarding system sluggishness, Microsoft is officially pivoting its strategy for Windows 11 . Moving away from its previous focus on adding new features, the tech giant has unveiled a major performance roadmap spanning from Version 25H2 through 26H2 . The goal is simple: to make Windows 11 feel as fast and fluid as promised. Addressing the "Experience Gap" Despite Microsoft’s previous benchmarks claiming Windows 11 outperforms Windows 10, many users have reported a different reality. The File Explorer has been at the center of criticism, with users citing slow load times, sluggish search results, and a noticeable "stutter" in the right-click context menus. Structural Efficiency and WinUI 3 The 2026 update focuses on deep-level structural improvements. Microsoft plans to reduce the OS's "baseline" resource...

OpenAI Ad Experiment Pays Off: ChatGPT Ad Revenue Hits $100 Million Annualized Rate OpenAI has revealed impressive early results from its advertising pilot program, which launched exclusively for U.S. users last month. In just six weeks, the company has reached an annualized revenue run rate of over $100 million from in-app advertisements, signaling strong market demand for AI-driven ad placements. Untapped Potential for Growth According to an OpenAI spokesperson, while approximately 85% of U.S. users meet the criteria to be shown ads, the current ad load is conservative appearing in only about 20% of eligible sessions . This indicates significant room for revenue expansion as the company fine-tunes its delivery frequency and expands its user base. A Strong Start with SMEs The platform has already attracted roughly 600 advertisers , with nearly 80% consisting of small and medium-sized enterprises (SMEs) . These businesses have shown a keen interest in ChatGPT’s unique ability to co...

Claude Users Report Rapid Quota Depletion: Anthropic Engineer Confirms "Peak Hour" Resource Management Over the past week, several Claude users have noticed a significant change in how their usage limits are calculated. Many reported that the standard 5-hour rolling quota is exhausting much faster than usual, hindering long-term projects and continuous workflows that previously ran smoothly. Confirmation from Anthropic While Anthropic has not issued a formal press release, Thariq Shihipar , an engineer at the company, clarified on social media that these observations are not a glitch. The shift is part of a strategic resource management plan designed to maintain system stability during periods of extreme demand. Key details of the temporary adjustment include: Dynamic Quotas: Claude has implemented a new calculation for the 5-hour usage cap specifically during "Peak Hours." This affects all user tiers, including Free, Pro, and Max subscribers. The Peak Window: T...

Android 17 Hits Platform Stability: Beta 3 Unveils System-Wide Windowing and Improved Quick Settings Google has officially released Android 17 Beta 3 , marking a major milestone as the OS reaches "Platform Stability" This means all developer APIs and app-facing behaviors are now finalized, with the remaining development phase focused exclusively on bug fixes and performance polish before the final public rollout. The Headline Feature: Universal "Bubble" Windows The most significant addition in Beta 3 is the expansion of "Bubble" mode (or small window mode). Users can now open any application in a floating, non-fullscreen window that stays on top of other apps. This multitasking enhancement is available across all form factors, including standard smartphones, foldables, and tablets. On large-screen devices equipped with a taskbar, users can trigger this mode by simply dragging an app icon to the far-right corner of the taskbar , instantly transforming i...

NASA Sets New Launch Date for Artemis II: Crewed Moon Mission Targeted for April 1, 2026 After facing several technical setbacks, NASA has officially announced a new target launch window for the historic Artemis II mission. The Space Launch System (SLS) rocket is scheduled to lift off on April 1, 2026, at approximately 6:24 PM EDT . Overcoming Technical Hurdles The Artemis II mission has already seen two significant delays. Originally slated for February 2026, the launch was scrubbed due to a persistent helium leak within the propulsion system. This issue forced NASA engineers to roll the massive SLS rocket back to the Vehicle Assembly Building (VAB) for extensive inspections and repairs to ensure the safety of the crew. The Launch Window Opportunity NASA officials emphasized that the April 1st date remains subject to various factors, including weather conditions and final systems checks. If the mission is unable to launch on the first attempt, a backup launch window has been ident...

  From $500,000 to $400: How Reco Used AI to Port JSONata to Go in Just 7 Hours In a remarkable display of AI-assisted engineering, the security firm Reco has shared its journey of porting the JSONata library from JavaScript to Go . By eliminating massive architectural overhead, the company successfully slashed its annual server costs by $500,000 , all for a one-time AI token cost of just $400. The Architectural Bottleneck Reco’s core engine, written in Go, relies heavily on JSONata to evaluate security policies for internal events. Since JSONata is natively a JavaScript library, Reco previously ran it in separate containers, communicating via RPC (Remote Procedure Call) . While functional, this setup was incredibly inefficient. For large-scale clients, Reco had to maintain up to 200 replicas , costing the company $300,000 per year in server fees alone. Previous optimization attempts such as embedding the V8 engine into the Go process or using simpler libraries like GJSON failed...

Google Research Unveils "TurboQuant": A Breakthrough Algorithm for High-Efficiency AI Data Compression Google Research has published a pioneering paper titled "TurboQuant" a new compression algorithm designed to drastically reduce the size of data transmitted during AI processing. This innovation specifically targets the bottleneck of high-performance AI workloads by optimizing how models handle massive amounts of information. The Vector Challenge At the heart of modern AI, Vectors serve as the fundamental mathematical representations used to link and process data. As AI models tackle increasingly complex tasks such as high-resolution image generation or multimodal reasoning these vectors grow exponentially in size. This results in heavy memory consumption, particularly within the KV (Key-Value) Cache used during model inference. TurboQuant addresses this by compressing essential vector data during processing without sacrificing overall accuracy or performance....