Loading...
Loading...
Browse tech news organized by topic. Topics are automatically detected and ranked by activity.
Russian President Vladimir Putin suggested that the war could end soon and proposed involving a former German chancellor as a mediator. The European Union has rejected the idea of including Gerhard Schröder in talks to resolve the conflict in Ukraine, reflecting concerns about his close ties to Russia and the political acceptability of intermediaries. The exchange underscores tensions over who can legitimately facilitate negotiations and illustrates broader EU caution toward Russian-led initiatives, even when framed as pathways to peace. Limited reporting leaves key details—timing, specific roles, and responses from Ukraine or Germany—unconfirmed.
Take-Two says a paused Switch 2 port of Borderlands 4 remains technically feasible but was delayed so Gearbox could fix PC performance problems that hurt launch reception. The publisher cancelled preorders and reiterated its priority remains PlayStation and Xbox, though it has supported Switch 2 with other titles. Meanwhile, Capcom’s new IP Pragmata—which included Switch 2 from the start—sold over 2 million copies in 16 days, highlighting the console’s appeal when platforms are prioritized in marketing and launch plans. However, investor worries persist: rising memory-chip costs have dragged Nintendo’s stock down, raising concerns about Switch 2 pricing and profit margins despite strong software prospects.
Developers are seeking ways to keep local edits hidden or uncommitted in shared repositories—whether to conceal AI-assisted changes or to avoid committing experimental tweaks. One common technique is using git update-index --assume-unchanged or skip-worktree to ignore local file modifications, allowing developers to maintain personal edits without affecting commits. At the same time, new Git clients like Rebased aim to streamline workflows, offering interfaces that simplify rebasing, staging, and managing local changes. Together, these trends highlight a push toward better local-change management tools and workflows that balance individual experimentation with team collaboration and repository hygiene.
Nvidia’s announced strategic partnership and potential $2.1 billion investment in data-center developer IREN signals a major push to scale AI infrastructure. The deal includes a $3.4 billion commitment to acquire and deploy Nvidia’s next-gen Blackwell chips, a five-year option to buy IREN shares, and plans to jointly build up to 5 GW of compute capacity. Markets reacted with short-term volatility in IREN stock, but analysts see the pact as validation of IREN’s access to power, rapid GPU deployment, and stronger long-term positioning in the AI infrastructure race. The move ties Nvidia’s chip and cloud strategies directly to large-scale data-center rollout.
Developers running local LLMs face linked issues around token-level streaming, server parsing, and new community checkpoints. Users ask whether OpenWebUI honors the preserve_thinking flag or if front-ends rely on back-end servers to expose unpruned “thinking” tokens for live streaming and debugging. A llama-server bug with chat-template-kwargs JSON whitespace can silently disable preserve_thinking for Qwen3.6, underscoring the need to validate config strings. Meanwhile, z-lab’s new gemma-4-26B release has prompted compatibility and performance queries as hobbyists test quantization and inference trade-offs. Together these stories show the interplay of UI support, server robustness, and model releases for local deployment workflows.
Chinese robotics firm Unitree unveiled the GD01, a production-ready, piloted mecha that combines large-format hardware, custom sensors and control systems to deliver a transformable civil transport at a reported starting price of ¥3.9 million. Investor reaction framed the launch as a showcase and marketing catalyst for the startup’s engineering stack. The announcement fits into a broader shift in mobility and robotics: legacy vehicle makers and startups alike are reallocating resources toward robot-focused hardware, while companies such as Zoox emphasize beginning with bespoke sensor architecture when designing autonomous platforms. Together these stories highlight a trend toward integrated sensor-driven vehicle design and commercializing novel robotic form factors.
OpenAI’s recent model changes—retiring GPT-4o and GPT-5.1 “Thinking” while pushing ChatGPT 5.5 and Gemma—are disrupting creators who say replacements feel ‘watered down’ and undermine longform storytelling workflows. At the same time, an odd glitch where ChatGPT began fixating on “goblins” forced OpenAI to investigate and apply mitigations, underscoring how unintended behaviors can surface from training signals or system prompts. Together these stories highlight tensions between product evolution and user expectations: developers want versioning, migration tools, and stable behavior, while platform operators must balance rapid iteration with monitoring, prompt engineering, and fast corrective action to preserve trust.
Microsoft is tackling Windows 11 File Explorer sluggishness through both short- and long-term fixes. Engineers are combining selective preloading of critical components with deeper architectural work—streamlining startup sequences, reducing disk reads, modernizing legacy dialogs, and cutting unnecessary animations—to address issues like slow folder navigation, thumbnail rendering, and context menus. Concurrently, an optional April update (KB5083631) patches UI glitches such as the dark-mode white flash, fixes explorer.exe termination and memory leaks, preserves folder view settings, and improves startup-app readiness and taskbar stability. These incremental updates are rolling out to Insiders and opt-in users now, with broader distribution planned on Patch Tuesday.
In Festus, Missouri, approval of a $6 billion data center ignited a swift political backlash: voters ousted four incumbent council members days after the council greenlit the project. The rout reflected community concerns about transparency, developer influence, land use, power demands, and environmental impacts despite promised jobs and tax revenue. The episode underscores rising local scrutiny of large tech infrastructure and the political risks for municipalities that fast-track deals without robust public engagement. It signals that data center developers and local officials must prioritize transparency, community outreach, and impact mitigation to secure social license for future projects.
Xiaomi continues to broaden its consumer electronics footprint with innovations across home appliances, smartphones and vehicles. The new Mi Home Air Purifier 6 debuts a dual-filter, dual-blade design and multi-stage purification matrix promising rapid removal of PM1, formaldehyde and allergens while remaining quiet. Elsewhere, leaks suggest Xiaomi (or a Redmi-offshoot) will push battery capacity boundaries in phones with a rumored 10,000mAh class device later in 2026, reflecting a broader industry pivot to ultra‑endurance handsets. Automotive plans also advance: spy shots of the Kunlun N3 extend‑range SUV hint at Xiaomi’s entry into mainstream electric vehicles under a new sub-brand, reinforcing a strategic push into high‑power, long‑range hardware across categories.
A dispute between Bambu Lab and community developer Pawel Jarczak over an OrcaSlicer integration has reignited Right to Repair debates. Bambu Lab issued a cease-and-desist, saying third-party integrations produced tens of millions of unauthorized cloud requests daily; Jarczak removed his fork that restored direct control of printers. Repair advocate Louis Rossmann pledged $10,000 to help Jarczak fight potential legal action and urged the community to pressure for the fork’s restoration as resistance to vendor lock-in. The clash highlights broader tensions between manufacturers protecting cloud-based ecosystems and users asserting control, repairability, and open modification rights.
The emergence of generative AI since ChatGPT’s 2022 launch sparked a frenetic industry scramble as tech giants, startups and investors raced to build or back competing models. Internal documents from the Musk v. Altman trial reveal that Microsoft’s 2018 reluctance to fund OpenAI was tempered by fear of ceding advantage to rivals like Amazon, illustrating how strategic rivalry shaped early alliances. That competitive dynamic—driven by promises of massive efficiency gains and profits, plus concerns about risk and public impact—continues to influence product priorities, partnerships and policy debates as companies jockey to turn AI hype into everyday value for users.
Reports indicate Google and SpaceX are in talks to develop orbital data centers, exploring the feasibility of launching and operating server infrastructure in low Earth orbit. The discussions focus on leveraging SpaceX’s launch capabilities and satellite expertise to host cloud computing hardware off-Earth, aiming to reduce latency for certain applications, enhance connectivity, and experiment with novel cooling and power solutions. The collaboration signals growing interest in space-based infrastructure as tech firms seek new ways to scale compute and diversify data center locations, though technical, regulatory and cost challenges remain significant hurdles.
A watchdog report accuses Meta of financially enabling Israeli settler attacks on Palestinians by allowing fundraising and payment tools on its platforms to be used to support extremist settler groups. The report cites examples of pages, groups and payment links on Facebook and Instagram that promoted or funded violence, and criticizes Meta for inadequate moderation and enforcement despite prior policies and warnings. The findings raise questions about platform liability, content moderation effe
Nvidia and the open ML community are converging on AI-driven tools to tackle quantum computing’s error and calibration challenges. Nvidia released large Ising Calibration and smaller Ising Decoding models—weights published on Hugging Face—claiming faster, GPU-accelerated error correction and calibration by mapping hardware tuning to Ising optimization. This underscores a trend of multimodal and domain-specific models (echoed by Baidu’s ERNIE‑Image also appearing on Hugging Face) becoming widely shareable, while raising reproducibility and benchmarking concerns: Hugging Face staff caution that provider-side optimizations can mask true performance, so researchers should evaluate models on controlled infrastructure to validate quantum-AI claims.
Developers are tackling practical trade-offs in advanced AI assistants like Codex and Claude by controlling when built-in skills run, offloading low-risk work to cheaper models, and masking AI fingerprints in output. Tools like CodexSaver (an MCP utility) route routine coding tasks—tests, docs, explanations, lint fixes—to a lightweight DeepSeek model while keeping high-risk decisions in Codex, cutting cost ~48% with modest latency. At the same time, Claude skills such as “humanizer” aim to remove telltale AI traits from text, raising questions about provenance and safety. The trend highlights optimization of cost, latency, and control, alongside new concerns about detectability and governance.
Mayor Zohran Mamdani has pledged to open five municipal grocery stores—one in each borough—with the first slated for 2027 on a vacant parcel at La Marqueta in East Harlem. Framed as a response to rising grocery prices and food access gaps, the initiative targets neighborhoods with high SNAP and public assistance use. Details on funding, governance, supply chains, staffing, eligibility rules and performance metrics remain scarce. Supporters call it an experiment to ensure affordability; critics warn of thin grocery margins, potential taxpayer losses and unfair competition with private retailers. The proposal spotlights public-sector solutions to urban food insecurity amid debate over feasibility.
Google is tightening Android sideloading by tying most off-Play app installs on certified devices to “verified” developers, citing far higher malware rates from internet-sideloaded APKs and rising scam-driven installs. The plan adds identity verification (and a small fee) plus new friction in the sideloading UI, including a mandatory reboot and 24‑hour cooling-off period aimed at preventing coercion and impulsive installs. After backlash, Google introduced an “advanced flow” that power users can enable in Developer options, with allowances that can persist across devices, while ADB installs remain exempt and limited distribution options are promised for hobbyists.
Elon Musk buys a fifth of his own Cybertrucks WIRED tracked where members of Elon Musk’s short-lived Department of Government Efficiency (DOGE)—a cadre of young technologists who were placed into federal agencies—have gone after the group fragmented. DOGE was blamed for mass firings, damaged agencies like USAID, and operational disruptions (for example at SSA). Some operatives remained in government, rising to influential posts within the Trump administration, while others returned to the privat
Blockchain analytics firm Elliptic raised $120 million at a $670 million valuation in a round led by One Peak Partners with backing from Deutsche Bank and Nasdaq, underscoring growing institutional interest in crypto compliance tools. The London-based company says it screens over a billion transactions weekly, highlighting demand for transaction-monitoring as markets mature. At the same time, large crypto treasuries remain active: Michael Saylor’s Strategy Inc. reported a $2.54 billion Bitcoin purchase in a single week, bringing its holdings to roughly $61 billion. The twin developments reflect robust capital influx and rising institutionalization across the crypto ecosystem.
A growing body of research suggesting vaccines can deliver broader health benefits—such as lowered dementia risk from shingles vaccination—has collided with escalating turmoil in U.S. vaccine governance. Under HHS Secretary Robert F. Kennedy Jr., the CDC’s advisory machinery has been repeatedly reshaped, with attempts to narrow childhood vaccine recommendations and reconstitute ACIP prompting court injunctions that paused policy changes and questioned the legality of appointments. Meanwhile, reports that CDC leadership delayed or blocked publication of positive COVID-19 vaccine effectiveness findings have intensified fears of political interference. With measles exposures rising in under-vaccinated states and leadership uncertainty at CDC, trust and guidance remain unstable.
诺基亚在英国上诉成功，阻止宏碁和华硕就视频流媒体专利提起诉讼 Asus reported Q1 revenue of NT$208.37 billion, beating the estimate of NT$200.13 billion, with operating profit of NT$11.90 billion, according to Caitong News. The stronger-than-expected top line suggests continued resilience in the PC and components vendor’s business amid a competitive consumer electronics market. For the tech industry, Asus’s results signal demand stability for notebooks, motherboards, and gaming hardware, and may influence supplier orders, inven
Android 17 introduces Pause Point, a new anti-doomscrolling tool that forces a mandatory 10-second pause before opening apps users mark as distracting, giving people a moment to reconsider impulsive usage. The release also highlights Google’s collaboration with Meta and Apple to enhance creator-focused features—bringing advanced video-editing tools to Android and better cross-platform file sharing similar to AirDrop—as well as upcoming apps like Adobe Premiere for Android. Together, these updates reflect Google’s push to balance user wellbeing with powerful creative capabilities and broader ecosystem interoperability in its latest mobile OS.
Iran is proposing new rules to assert control over seven undersea fiber-optic cables traversing the Strait of Hormuz, demanding permits, tolls and that management, repairs and maintenance be handled by Iranian firms under domestic law. Framed as bolstering sovereignty and security—echoing recent IRGC threats to submarine infrastructure—the plan targets major data routes connecting Europe, the Gulf and Asia. Experts warn the move could raise legal, commercial and geopolitical risks, disrupt redundancy and latency patterns, and complicate operations for cloud providers, carriers and international network operators reliant on these strategic digital chokepoints.
Stellantis has expanded its strategic partnership with Chinese EV maker Leapmotor (零跑汽车), deepening cooperation after acquiring ~21% of Leapmotor and forming the Leapmotor International JV (Stellantis 51%). The two will boost capacity at Stellantis’ Figueruelas, Spain plant with new lines to build an Opel C‑segment BEV (potentially starting volume production by 2028) and may produce Leapmotor’s B10 SUV there as early as 2026. They’ll also pursue joint procurement to leverage China’s EV supply ch
A federal jury concluded Live Nation and its Ticketmaster unit illegally monopolized the live-event ticketing market, finding the company overcharged consumers by about $1.72 per ticket. The verdict, stemming from a 34-state prosecution and lengthy trial, spotlights alleged anti-competitive tactics like exclusive venue contracts and clauses that block rivals. It opens the door to damages, potential divestitures or a breakup, and sharper regulatory scrutiny of tech-enabled platform dominance. The decision could reshape concert ticketing, affect venues and artists, and signal broader enforcement against gatekeepers in entertainment and online marketplaces.
OpenAI has set up the OpenAI Deployment Company, a majority-owned subsidiary backed by over $4 billion and partnerships with top investors and consultancies to embed Forward Deployed Engineers (FDEs) into enterprise clients. The move includes acquiring AI consulting firm Tomoro and onboarding about 150 experienced FDEs and Deployment Specialists to help design, build, and operationalize AI systems. By aligning engineering, workflows, and model roadmaps, the unit aims to accelerate safe, scalable AI adoption in large organizations. The initiative underscores a shift: competitive advantage now hinges on operationalizing advanced models through hands-on deployment expertise, not just model development.
OpenAI is advancing voice intelligence by adding new voice-focused models and features to its API, enabling developers to build richer audio experiences. The updates include improved speech recognition, natural-sounding text-to-speech, and tools for real-time voice interactions, aimed at conversational agents, transcription services, and voice-enabled apps. By integrating these capabilities directly into the API, OpenAI simplifies development workflows and supports a wider range of use cases from accessibility to customer support. The move underscores broader industry momentum toward more capable, low-latency voice AI that blends speech understanding and generation in a single platform.
YouTube is introducing a way to limit or effectively disable Shorts by adding a Shorts feed timer that can now be set to zero. Rolling out across mobile and web (initially to parents and gradually to all users), the setting hides the Shorts tab and shelf and stops Shorts autoplay after the limit is reached. The change builds on earlier 15-minute minimums and parental controls, responding to user frustration, creator concerns, and regulatory scrutiny over addictive design. If widely adopted, the toggle could shift discovery and engagement dynamics for creators and advertisers while giving viewers more control over their YouTube experience.
Researchers and practitioners are exploring practical ways to add JIT compilation to existing C-based interpreters without full rewrites. Laurence Tratt’s yk prototype demonstrates that modest code changes—hundreds of lines added, few altered—can retrofit JITs into interpreters like PUC Lua, delivering up to 4× speedups on some workloads and roughly 2× on average. Coverage varies by interpreter idioms, and retrofitted JITs don’t yet match hand-tuned engines like LuaJIT. Discussion centers on techniques (tracing, on-stack replacement, dynamic codegen), trade-offs in portability, debugging, and maintenance, and the promise of lowering the barrier to JIT-enable many legacy and embedded runtimes.