Every story we have kept, newest first.
Looking for the daily editions? → Past editions
Page 93 of 1373
Amazon is developing Project Tetromino, an internal initiative aimed at building fully automated delivery stations. The system could use Boxbot’s AI-powered storage trays to sequence packages for vehicle loading, extending robotics into last-mile logistics.
GeekWire · 23d ago
Vercel says applications hosted on its platform are protected from two critical Next.js vulnerabilities disclosed in the August 2026 security release, requiring no customer action. Self-hosted applications should upgrade to patched Next.js versions, especially those running on Windows or processing AVIF images.
Vercel News · 23d ago
AWS presents a governed weekly reporting workflow using Amazon Quick Desktop and Amazon FSx for NetApp ONTAP. An Amazon S3 access point limits knowledge-base access, while a custom skill drafts cited reports and Slack summaries for human review.
AWS Machine Learning Blog · 23d ago
Anthropic announced funding to improve evaluations of AI’s impact on human wellbeing. The announcement focuses on developing better ways to assess how AI systems affect wellbeing.
Anthropic Announcements · 23d ago
The U.S. Environmental Protection Agency has proposed removing states’ obligation to provide public notice or solicit comments on air pollution permits for data centers and other projects.
New York Times Technology · 23d ago
Apple has launched the M6 chip in a new Mac mini and the M5 Ultra in a new Mac Studio. The M6 is Apple’s first 2nm chip, while the M5 Ultra is its first quad-die M-series design, with major gains in AI performance.
The Next Web (TNW) · 23d ago
AI firms are debating whether to keep cyber capability tests connected to the internet after models were hacked. Supporters argue that online testing better reflects real-world conditions and improves accuracy.
Bloomberg Technology · 23d ago
HarmonyOS 7 adds system-level AI capabilities that help apps become discoverable through Xiaoyi, leverage on-device models, and access APIs such as image super-resolution and semantic image search. Developers can expose simple functions as Skills, connect multi-turn Agents through A2A, and run selected AI workloads privately on-device.
极客公园 · 23d ago
Apple is launching new Mac mini models with M6 and M5 Pro configurations, promising major gains in AI performance and processing speed. The Mac mini will also become Apple’s flagship U.S.-assembled Mac, with final assembly planned at the company’s Houston facility.
Computerworld · 23d ago
Google’s Antigravity 2.0 expands its developer panel with Git-backed diffs, staging, and commit workflows. It also adds an embedded terminal on the right side of the interface.
TestingCatalog · 23d ago
SpaceXAI plans to use NVIDIA Vera CPUs for Grok infrastructure and adapt a Vera Rubin NVL72 system for its planned Starmind AI satellite. The proposal signals an expansion of NVIDIA-based computing across terrestrial and space-oriented AI infrastructure.
TestingCatalog · 23d ago
Apple’s M5 Ultra is presented as the company’s most powerful chip yet, with up to four times the AI performance of the M3 Ultra. The Mac Studio configured with the chip is also described as supporting up to 512GB of unified memory and eight external displays.
ZDNet AI · 23d ago
OpenWiki uses evidence-backed claims to identify stale knowledge in evolving codebases. Its approach aims to reduce hallucinations and enable memory that can update itself as source information changes.
LangChain Blog · 23d ago
At Hot Chips 2026, Intel provided additional details about its Crescent Island AI accelerator based on the Xe3P architecture. The data-center inference accelerator combines liquid cooling, HBM4 memory, larger caches, and deeper XMX engines to improve AI performance per watt.
Tom's Hardware · 23d ago
The article argues that AI has not replaced human radiologists, despite earlier predictions that it would. Instead, AI is expected to significantly transform radiologists’ responsibilities and workflows.
Ars Technica AI · 23d ago
Yangtze Memory Technologies has begun its IPO process, seeking RMB 33 billion to upgrade production lines and advance its 3D NAND architecture. Its return comes after export controls disrupted the company's path from technical leadership to scale, cost, and profitability advantages.
虎嗅 · 23d ago
Australia may see a surge in datacentre construction as AI firms seek approval before proposed federal regulations take effect. The rules would require new facilities to avoid increasing electricity prices, build renewable energy capacity, and reduce water consumption.
The Guardian Technology · 23d ago
Gamma has acquired Lica, an Accel-backed design startup. Lica’s co-founders will join Gamma’s new research team, though the deal’s financial terms and specific product plans were not disclosed.
TechCrunch AI · 23d ago
NVIDIA has introduced CUDA Python 1.0 with stable APIs designed to give Python developers a unified foundation for accessing the CUDA platform. The release aims to reduce reliance on custom CUDA C++ extensions or higher-level libraries such as PyTorch, CuPy, and RAPIDS.
NVIDIA Developer Blog · 23d ago
LangChain explains how it creates synthetic environments and tasks for evaluating or training agents. The workflow includes specification generation, spec-to-task conversion, and a shared world spec containing common knowledge.
LangChain Blog · 23d ago
Google is expanding its Gemini Enterprise AI platform with new tools designed for law firms and practicing attorneys. Gemini Enterprise for Legal aims to support routine and complex legal work while maintaining data security and confidentiality.
cnBeta (Full RSS) · 23d ago
OpenAI banned a group of Russian ChatGPT accounts allegedly used to support a covert online influence operation. The accounts reportedly used AI-generated content to shape narratives and obscure the operators’ identities.
cnBeta (Full RSS) · 23d ago
OpenAI CEO Sam Altman said the AI industry had been too optimistic about the pace of adoption and social transformation. Although chatbot usage continues to grow, reliance on familiar tools, vendors, and workflows is slowing the transition.
cnBeta (Full RSS) · 23d ago
Roborock, Ecovacs, and Dreame are pursuing different strategies in the increasingly competitive robot vacuum market: technology and multi-category expansion, scale efficiency, and focused premium growth. Dreame’s organizational turmoil has not materially weakened its overseas cleaning-appliance business, while a new U.S.
虎嗅 · 23d ago
Alibaba completed an HK$80 billion share placement, with all proceeds earmarked for full-stack AI capabilities and infrastructure expansion. Founder Jack Ma and senior executives bought more than HK$800 million of shares, signaling confidence as the company trades short-term dilution and cash-flow pressure for faster AI capacity growth.
虎嗅 · 23d ago
Apple launched the Mac mini M6 in China starting at 6,999 yuan, with comparable configurations costing up to 2,500 yuan more than the M4 generation. The increase reflects surging memory prices driven by AI data-center demand, while Apple positions the device as an always-on platform for AI-agent workflows.
极客公园 · 23d ago
The U.S. government is reportedly negotiating with South Korea to expand domestic memory-chip production.
cnBeta (Full RSS) · 23d ago
Apple has launched the M6, its first processor built using a 2nm process. The chip features dual 16-core neural engines and is designed to improve performance and energy efficiency through higher transistor density.
cnBeta (Full RSS) · 23d ago
OpenAI’s Jalapeño chip reportedly delivers strong inference performance at scale. On Semianalysis’s InferenceX benchmark, it achieved both higher tokens per user and greater throughput per kilowatt than currently available state-of-the-art systems.
TechCrunch AI · 23d ago
The article examines how Japanese right-wing actors and content producers are using generative AI to mass-produce anti-China, exclusionary, and historically revisionist videos. It warns that AI-generated images, scripts, narration, and targeted distribution could make fabricated historical narratives more persuasive and harder to distinguish from authentic records.
虎嗅 · 23d ago