AI agent evaluation shifts from clean traces to cohort baselines
LangChain, Conviva and CoreWeave executives said enterprises are rethinking how they test agents, with cheaper judge models and mo...
By Renata Fuchs 1 month agoSection
AI
Google is reportedly designing a Gemini-specific AI inference chip
The Information says Google’s Frozen v2 chip could improve inference efficiency by 6 to 10 times, with deployment planned for 2028...
By Renata Fuchs 1 month ago
Neill Blomkamp releases AI-generated short film Nightborne
The District 9 director used Seedance 2.0 for a 13-minute sci-fi horror short and says he plans a feature in the same format.
By Wei-Lin Zhao 1 month agoZillow says AI ROI depends on measuring before deployment
At VB Transform 2026, Zillow and Glean executives said enterprise AI gains depend on persistent context, cost controls and pre-exi...
By Renata Fuchs 1 month ago
Microsoft expands Azure AI chip supply with AMD Helios
Microsoft plans to add AMD's Helios platform to Azure in 2026, while Anthropic is reportedly testing AMD hardware for Claude workl...
By Wei-Lin Zhao 1 month agoJumpCloud survey finds lower AI confidence as agents reach production
JumpCloud says self-rated AI maturity fell from 40% to 23% in six months as IT teams encounter governance gaps around production a...
By Renata Fuchs 1 month ago
White House weighs pressure campaign against Chinese AI models
Axios says Commerce, the NSA and the White House have considered sanctions, warnings and liability rules that could push U.S. comp...
By Renata Fuchs 1 month ago
Hugging Face says autonomous AI agent breached internal systems
The AI platform says a malicious dataset led to credential theft, while its own AI tools compressed incident analysis from days to...
By Renata Fuchs 1 month ago
Moonshot halts new Kimi K3 subscriptions as capacity tightens
Moonshot said demand for its new Kimi K3 model reached the limits of available compute within 48 hours, prompting a pause on new p...
By Wei-Lin Zhao 1 month agoData engineer warns RAG will not fix weak AI data pipelines
Naveen Ayalla argues enterprise AI failures are often rooted in ingestion, validation and access-control gaps rather than LLM limi...
By Colin Brandt 1 month ago
Alibaba previews Qwen 3.8 as open-weight AI race intensifies in China
Alibaba says its 2.4 trillion-parameter Qwen 3.8 model will ship with open weights soon, targeting coding and productivity workloa...
By Colin Brandt 1 month ago
Google DeepMind tests video generation as a shortcut for computer vision
GenCeption adapts Alibaba’s Wan2.1 video model for depth, segmentation and pose tasks, with strong benchmark results from limited...
By Wei-Lin Zhao 1 month ago