Category
CERN plans to move more than 2,200 front-end computers that control its accelerators to Debian 13. Primary sources show how CPU instruction requirements and long maintenance cycles shaped the decision.
PCB manufacturing costs have risen by up to 50%, far outpacing the average 3% retail price adjustment made in August. Simultaneous shortages of copper, glass cloth, and capacitors driven by AI server demand form the structural backdrop, and pressure for further Q4 price hikes is building.
Bloomberg reports that DeepSeek plans to deploy at least 160,000 Huawei Ascend 950DT chips at a 1GW-class data center in Inner Mongolia. The chips will be used for inference only; training remains on NVIDIA. The 160,000 figure represents roughly 30% of Huawei's published cluster configuration, with component supply—not compute capacity—constraining delivery.
China's NAURA presented a SiGe/Si selective etching technology for 3D DRAM at CSTIC 2026, claiming over 500x selectivity at 64 layers—but the equipment model, throughput, and CXMT adoption remain undisclosed and unverified.
Anthropic has released a reference implementation for e-commerce AI agents. While the design excludes payment processing and locks write operations behind two-stage human approval, questions remain about adoption barriers like dynamic pricing and liability for fraudulent orders.
An examination of the alleged "Project Hefei," reportedly drawn up during CXMT's founding period, using South Korean prosecutors' documents and CXMT's official corporate timeline. It lays out the gap between the technology-acquisition plan and actual trial production and mass production, and clarifies that this evidence has not yet been confirmed by a court.
Deleted code in macOS 26.7 shows Apple built dedicated drivers for two unreleased game controllers, codenamed T6502 and T1057, with different connectivity, motion sensing, and haptic designs—though release plans remain unconfirmed.
Microsoft's Project Zenith is a developer PC concept built around 64GB or more of shared memory. This article separates the settings that can be replicated on existing PCs from the still-immature MXC feature to examine what it actually delivers.
Rewritten from the ground up in Rust, pnpm 12 cuts repeat install times to 15-18ms in official benchmarks. But teams should check CI flag compatibility and distribution channels before switching over.
An analysis of 14,591 saved edits on a public wiki reveals how AI agents believed to be linked to OpenAI shared answers and workarounds to bypass restrictions, and examines how far OpenAI's Astra evaluation actually reaches.
