How Dutch insurer Univé built an AI-ready workforce
A vendor case study from OpenAI shows how Univé used leadership, governance, and employee-led innovation to scale ChatGPT Enterprise across the organization.
AI & ML
Frontier models, research results, agents, and developer tooling.
Two API Settings Tripled OpenAI's ARC-AGI-3 Scores
OpenAI reports that enabling two API settings—retaining reasoning and enabling compaction—tripled GPT-5.6's scores on the ARC-AGI-3 benchmark.
Hugging Face Details Timeline of Frontier Lab Agent Intrusion
The post offers a rare chronological record of how a frontier lab's AI agent was breached in July 2026, with implications for agent security.
NVIDIA Cosmos-H-Dreams Targets Real-Time Simulation for Surgical Robotics
A new NVIDIA project aims to bring real-time generative simulation to surgical robotics, according to a Hugging Face blog post.
OrcaRouter Launches OrcaCyber Zero 1.5 With 1M-Token Context
The gated cybersecurity model reports perfect Cybench scores and strong CVE-Bench results while offering a 1M-token context window.
Interconnects launches Artifacts Hub and Adoption Dashboard
A new Hub and Dashboard from Interconnects aims to make the open AI ecosystem easier to curate and measure.
OpenAI Signals Data Maps ChatGPT Adoption by Country
New country-level usage figures from OpenAI's Signals initiative show how ChatGPT adoption and behavior differ around the world.
OpenAI Adds Education Plugins to ChatGPT Work and Codex
OpenAI is targeting classrooms with new plugins designed to support teaching, learning, research, and building.
OpenAI rebuts Apple lawsuit, says claims are baseless
OpenAI publicly pushes back against Apple's legal action, correcting employee-related claims and releasing internal messages as evidence.
Baseten Joins Hugging Face Inference Providers
Hugging Face adds Baseten as an inference provider, giving developers another route to run open models on dedicated GPU infrastructure.
How Avatarin Built a 24/7 Multilingual Retail Agent with GPT-Realtime
A two-week deployment at Yamada Denki shows GPT-Realtime handling round-the-clock multilingual customer service in retail.
Open Models Laguna S2.1, Inkling, Kimi K3 Show Proliferating Training Capacity
Three new open releases illustrate how widespread the ability to build strong models has become.
GPT-5.6 cuts costs and boosts efficiency for enterprise AI
OpenAI's GPT-5.6 lowers pricing for Luna and Terra while improving model and inference efficiency, delivering more value per dollar.
DeepMind launches Lyria 3.5 in Flow Music with new creative controls
Google DeepMind's Lyria 3.5 arrives in Flow Music, promising upgrades in musicality, lyrics, vocals, and creative control.
DeepMind unveils two Gemini Robotics models for reasoning and control
Two new Google DeepMind releases advance robot abilities in complementary ways: one improves video understanding and collaboration, the other adds whole-body dexterity.
Google's Science One Ties Autonomous Research to Chain-of-Evidence
Google Research's Science One framework aims to make AI-driven discovery auditable by requiring a verifiable chain of evidence.
Robotics, Week-Long Coding, and OpenAI's Accidental Hacker
Import AI 466 links robotics' bitter lesson, longer-horizon AI programming, and an OpenAI incident to warn that AI capabilities are outpacing societal readiness.
Idle GPUs Are the New Grounded Aircraft: Why Wasted Compute Costs You
Hugging Face argues that underutilized GPUs represent a massive hidden cost for AI teams, urging better scheduling and monitoring.