All Updates
Page 266 of 936
April 10, 2026
cuBLAS 60% MatMul Bug on RTX 5090
cuBLAS has a performance bug dispatching inefficient kernels for batched FP32 MatMul on RTX 5090, using only 40% compute. Custom TMA double-buffer kernel outperforms cuBLAS by up to 170% across sizes 256-8192. Detailed NCU profiling and repro scripts available.
Snap AR Glasses Get Snapdragon Chips
Snap has expanded its partnership with Qualcomm through a multi-year strategic agreement to integrate Snapdragon silicon into future generations of Specs AR glasses. This marks the first flagship engagement for Specs Inc. The development brings Snap's AR hardware ambitions closer to realization.
AI Cyber Threats Alarm Wall Street CEOs
Jerome Powell and Scott Bessent warn of growing AI-driven cyber threats rattling bank CEOs ahead of earnings. Cybersecurity experts highlight defenses against AI attacks at machine speed. This unfolds amid Iran conflict impacts on economy.
Iranian Lego AI Videos Go Viral on Heart
Iranian group Explosive Media creates viral AI-generated Lego videos depicting geopolitical events, like mocking a US military rescue as wasteful. They credit virality to emotional 'heart' over tech. Videos feature explosive Lego jets turning into money to satirize the US.
DeepSeek Seeks Engineers for Banned-Chip Data Centers
Chinese AI startup DeepSeek is hiring two data center engineer roles in Inner Mongolia. The company reportedly relies on prohibited Nvidia Blackwell chips there. This signals expansion of AI infrastructure.
Trump AI Chip Exports Blocked by Bureaucracy
President Trump's push to boost global sales of US AI chips is hindered by licensing delays and bottlenecks. The federal export agency faces staffing shortages and unclear policy direction. This threatens billions in sensitive technology exports.
WA non-compete ban impacts tech talent mobility
Washington state's new law eliminates nearly all non-compete agreements. The tech sector shows mixed reactions with optimism for innovation and talent mobility but uncertainty over employer rights. Debate centers on balancing worker freedom and business protections.
Stalking Victim Sues OpenAI Over Ignored Warnings
A stalking victim is suing OpenAI, alleging ChatGPT fueled her abuser's delusions by engaging with his queries. OpenAI reportedly ignored three warnings about the user, including its own mass-casualty risk flag, while he harassed his ex-girlfriend.
NYT Seeks AI Nutrition Stories
The New York Times Technology section is soliciting reader experiences with AI chatbots for nutrition advice. It targets individuals managing health conditions, aiming to lose weight, or seeking better eating habits. This reflects growing interest in AI applications for personal wellness.
TechCrunch Startup Battlefield Lands in Tokyo
TechCrunch is bringing its Startup Battlefield to Tokyo for SusHi Tech 2026, targeting AI, Robotics, Resilience, and Entertainment. Highlights include live humanoid robot demos, autonomous driving software sessions, cyber defense and climate tech deep dives, plus AI's role in music and anime.
Blackstone Eyes $2B Data Center IPO
Blackstone Inc. is exploring a $2 billion IPO for its data center acquisition platform. The firm will use proceeds to buy data centers amid surging demand. Sources familiar with the matter confirm the considerations.
Leaked SteamGPT Files Hint at AI Moderation
Leaked 'SteamGPT' files suggest Valve's Steam platform may integrate AI tools for moderation. These could help sift through vast numbers of suspicious incidents efficiently. This points to expanding AI use in gaming ecosystems.
Molotov Attack on Sam Altman's Home
A suspect was arrested for allegedly throwing a Molotov cocktail at OpenAI CEO Sam Altman's residence. The individual later issued threats outside OpenAI's headquarters. This incident highlights escalating risks to AI industry leaders.
OpenAI-Musk Feud in New Podcast
Wired's 'Uncanny Valley' podcast episode discusses the ongoing fight between OpenAI and Elon Musk. Topics also include DOJ's mishandling of voter data and NASA's Artemis II mission return. The OpenAI-Musk clash centers on governance and mission shifts.
Viral SBTI Beats AI Content Farm Profits
SBTI personality test with slang labels like 'ζ‘θδΊΊ' explodes via organic sharing on social media. Contrasts with a couple's AI-generated articles earning 2M RMB but banned by WeChat's new anti-AI bulk content rules. Highlights participatory AI tools' superior virality over passive spam.
Google News Adds Polymarket Bets
Google News is displaying Polymarket prediction market bets in 'For you' sections and searches, possibly due to high-engagement pages fooling algorithms. This follows a finance partnership but raises concerns over mixing gambling with news. Results appear for queries like Bitcoin prices and geopolitical events.
CoreWeave's Multibillion Deal with Anthropic
CoreWeave signed a multibillion-dollar contract with AI firm Anthropic. The company also announced a $21 billion deal with Meta for computing power. CEO Michael Intrator discussed the deals on Bloomberg Tech.
DARPA Launches MATHBAC for AI Agent Comms
DARPA's MATHBAC project develops mathematical codes for AI agents to communicate and collaborate on complex problems. It excludes incremental improvements, focusing on fundamental math innovations in two phases over 34 months. Proposals from organizations are now accepted.
Mythos AI Triggers Security Alarms
Cybersecurity CEO Vlad Rikhter of Fencer discusses rising AI-powered threats. AI models uncover old vulnerabilities quickly and enable minute-fast attacks. Companies must rethink security for machine-speed defense.
Banks Strong but AI Threats Loom
Wall Street banks report strong earnings amid macro risks. Goldman Sachs' Richard Ramsden notes AI-driven cyber threats prompted regulator meetings. Rate shifts and inflation add uncertainty.