News archive.
Earlier AI news and analysis, kept for reference. Current coverage lives in News & Analysis.
CEOs who move fast gain edge in fragmented geopolitics
McKinsey argues that competitive advantage now flows to leaders who place bets quickly as global markets splinter. Here's what that means for your strategy.
Central Europe's AI gap widens as businesses treat it like a side project
McKinsey warns Central European companies are falling behind because AI remains a peripheral initiative instead of a core business redesign. Early movers are already restructuring around it.
Your AI pilots are stuck. Data readiness is the bottleneck
McKinsey finds companies scaling AI hit a wall: disconnected data systems. Leaders now prioritize governed, reusable data foundations to move pilots into production.
NVIDIA DAQIRI Streams 100+ Gbps Detector Data Straight to GPU
NVIDIA released DAQIRI, a networking library that bypasses the kernel to send high-speed sensor data directly to GPU memory for real-time inference. CERN is testing it to search discarded particle collision data.
NVIDIA's CCCL Runtime Brings Modern C++ to CUDA Development
NVIDIA released CCCL runtime, a new C++ API layer for CUDA that replaces implicit device state with explicit dependencies and adds owning/non-owning type pairs. Safer kernel launches and memory management without rewriting existing code.
Telcos still operate AI at level 2–3 autonomy; reaching level 4–5 requires shared platforms
Most telecom operators use AI for predefined network tasks but lack the shared reasoning models, digital twins, and governance layers needed for autonomous agents to discover novel solutions. NVIDIA outlines the platform architecture.
Meta cuts video call bitrates 20% with AV1 codec on billions of devices
Meta deployed AV1 video compression across most Android and iOS phones in Messenger and WhatsApp, achieving at least 20% bandwidth savings in offline tests. Here's how they solved power, binary size, and device eligibility at scale.
Run GitHub triage locally with Gemma and Qwen—no API bills
Hugging Face built a real-time PR classifier on local models, cutting OpenAI API costs to zero. See how they benchmarked Gemma-26B and Qwen-35B against GPT-5.5 on 330 labeled issues.
Hugging Face ships weekly releases using open AI and deterministic safeguards
Hugging Face automated its Python library releases from every 4-6 weeks to every week using open-weights models and GitHub Actions. Here's how they keep AI-drafted release notes accurate without vendor lock-in.
OpenAI launches free security tool for open-source code patching
OpenAI's new 'Patch the Planet' program offers free vulnerability detection and remediation for open-source software. Learn what's included and how to use it.
OpenAI's GPT-5.5-Cyber tops Anthropic's Mythos on security tasks
OpenAI claims its new GPT-5.5-Cyber model outperforms Anthropic's Mythos on a cybersecurity benchmark. The company has not released independent test results.
OpenAI Launches Daybreak Framework for Cybersecurity Defense
OpenAI introduced Daybreak, a new AI framework designed to strengthen cybersecurity defenses. Learn what it does and how security teams can integrate it into their workflow.