All Updates
Page 486 of 1664
May 29, 2026
Nvidia Exec Madison Hwang on Non-Linear AI Careers
Madison Hwang, Senior Director of Product Marketing at NVIDIA, discusses the importance of non-linear career paths in the AI era. She highlights her unconventional background, which includes professional culinary training before joining the tech industry.
BYD Launches 4nm Xuanji A3 Autonomous Driving Chip
BYD has officially released the Xuanji A3, China's first 4nm autonomous driving chip. The chip supports L3 and L4 autonomous driving and comes with a one-year accident liability coverage policy for urban navigation.
Chinese-made EVs arrive in Canada, including Shanghai-built Teslas
Under a new trade agreement, Chinese-manufactured electric vehicles, including those from Tesla's Shanghai Gigafactory, have begun arriving in Canada. This marks a significant shift as tariffs are reduced from over 100% to approximately 6%.
Liquid AI releases LFM2.5-8B-A1B edge model
Liquid AI has launched LFM2.5-8B-A1B, an upgraded edge model featuring a 128K context window and 38T tokens of pre-training. It is designed to handle complex tasks and tool calling while remaining efficient enough to run on entry-level laptops.
Stepfun Open-Sources Step 3.7 Flash LLM for Agents
Stepfun has released Step 3.7 Flash, a 196B-parameter sparse MoE model specifically optimized for agentic workflows. It features high-speed inference of 400 tokens/s and native tool-calling capabilities.
URIEL: Robotic Airborne Systems for Sustainable Tropical Logging
URIEL proposes a novel logging method combining heli-logging, robotics, and AI-driven post-harvest silvicultural treatments. The study demonstrates that this approach can virtually eliminate collateral forest damage while remaining economically viable.
Rethinking LLM Evaluation for Public Comment Analysis
This research proposes an Interpretive Audit Pipeline to address the limitations of standard accuracy-based LLM evaluation. By treating inter-model disagreement as a diagnostic tool, the framework identifies ambiguous public inputs that require human intervention.
Longitudinal Study Reveals User Habits in LLMs are Sticky
A longitudinal analysis of 12,000 Microsoft Bing Copilot users shows that individual behavior patterns remain remarkably consistent over time. The study also warns that popular research datasets like WildChat-4.8M may be skewed toward power users, misrepresenting typical user-AI interactions.
Behavior-Aware Corrections Improve Off-Policy TD Learning Stability
This research introduces BA-TDC and BA-TDRC, two new methods that replace traditional auxiliary covariance matrices with behavior-aware Bellman matrices to stabilize off-policy temporal-difference learning. The paper provides a theoretical framework for these corrections and demonstrates their effectiveness in improving robustness across various reinforcement learning benchmarks.
OpenAI Investigating Widespread ChatGPT Login and Usage Outage
OpenAI is currently investigating technical issues preventing users from logging into ChatGPT or engaging in conversations. Downdetector reports indicate approximately 5,000 users have been affected by the service disruption.
OpenAI launches Rosalind Biodefense for public health security
OpenAI has introduced Rosalind Biodefense, a specialized platform providing vetted developers and U.S. government partners with access to GPT-Rosalind. This initiative aims to leverage frontier AI to accelerate biodefense, public health research, and pandemic preparedness.
Microsoft launches Foundry Local for on-device AI execution
Microsoft has released 'Foundry Local', a platform allowing developers to integrate AI features directly onto user devices. This approach eliminates cloud dependency, reduces network latency, and removes token-based usage costs.
Anthropic to Release Mythos-class Opus 4.8 Model
Anthropic announced the upcoming release of its new 'Mythos-class' model, Opus 4.8. The model will be available to all customers within the next few weeks after implementing enhanced safety measures.
Anthropic releases flagship Claude Opus 4.8 model
Anthropic has officially launched its new flagship model, Claude Opus 4.8. This release marks a significant update to their high-end AI model lineup.
Huawei's Ultra-Luxury ZunJie S800 Opens for Pre-orders
Huawei's high-end brand ZunJie has opened pre-orders for the S800, a luxury vehicle targeting the 2 million RMB market. The model features the new Tuling Longxing platform and will be officially released in late June.
Run Docker Containers Directly Inside Vercel Sandbox
Vercel Sandbox now supports native Docker execution, allowing developers to build containers, install system packages, and run services like Redis or Postgres. This update also adds support for FUSE filesystem drivers and VPN clients to enhance development environments.
Obayashi Corporation validates RICOS AI-CAE for wind load prediction
Obayashi Corporation has begun validating a new AI-CAE solution developed by RICOS to predict wind loads on buildings. The system utilizes AI to analyze wind direction and building geometry, aiming to significantly streamline the architectural design process.
May 28, 2026
Claude 4.8 Released: Outperforms Mythos with Multi-Agent Support
Anthropic has launched Claude 4.8, which demonstrates performance exceeding the Mythos model in specific benchmarks. The update introduces robust support for running hundreds of sub-agents in parallel for complex, long-running tasks.
White Rhino CTO on end-to-end autonomous delivery scaling
White Rhino CTO Xia Tian discusses the transition to mapless navigation and end-to-end models for L4 unmanned delivery vehicles, aiming for full automation by 2026.
Fujitsu partners with OpenAI and Anthropic for AI expansion
Fujitsu has announced simultaneous partnerships with OpenAI and Anthropic. The company aims to integrate these leading AI technologies with its own proprietary AI solutions to enhance its enterprise offerings.