A Co-op workplace surveillance controversy shows why AI monitoring changes the employment relationship even when its accuracy claims sound modest.
Google Cloud’s On platform frames agent-led migration as an enterprise control, cost, and accountability problem—not just a faster way to move workloads.
The Long-WAM paper argues that longer context for world-action models must preserve useful state, not merely store more observations.
One prompt from idea to interface
Start with ready-made website templates, then customize the interface with AI.
Chronological feed
The newest reporting and analysis, ordered by publication time and labeled by editorial desk.
Across the newsroom
More reporting, launches, research, and industry developments from across the AI landscape.
An analysis of the OpenAI–Chatham capital-markets AI proposition, the evidence still missing, and the controls that would make embedded expertise useful.
OpenAI’s distillation-disruption framing raises a practical security question: how can model providers detect unauthorized capability transfer without treating legitimate customers as attackers?
Ai2 open-sources AstaBrief, an 8B scientific report model that trades frontier scale for faster cited synthesis, local control, and inspectable training.
Google DeepMind outlines encrypted cloud memory with device-held keys, authenticated channels, public software records, and independent audit goals for personal AI.
H Company’s Holo4 models combine GUI actions, code, MCP, and APIs; the release makes interface switching and reproducible trajectories central to agent evaluation.
A new source-aware verification approach for MCP agents asks not only whether an answer is true, but whether the agent used the right source for the decision.
NVIDIA’s OpenShell and Sentry reference design treats autonomous agents like untrusted software, combining sandboxing, DPU monitoring, policy enforcement, and audit trails.
The Allen Institute for AI’s OLMo-core 3 release focuses on the code beneath an open mixture-of-experts model, where reproducibility depends on engineering details.
Editorial desk
Frontier models, benchmarks, scientific advances, and the research shaping what comes next.
Hugging Face’s report on fine-tuning Nemotron for IOI and IMO examines how training changes reasoning performance and why gold-level results need context.
Editorial desk
Autonomous systems, copilots, orchestration, and the new operating layer for AI work.
Hugging Face’s ThinkingBox case study shows how database agents can report success while leaving the underlying state unchanged or wrong.
Editorial desk
Adoption, funding, strategy, startups, and the economics of putting AI into production.
Jump Trading is scaling quant research with ChatGPT, a case study in how AI changes financial work without removing the need for independent verification.
Editorial desk
Governance, privacy, cybersecurity, regulation, safety, and public-sector decisions.
A Co-op workplace surveillance controversy shows why AI monitoring changes the employment relationship even when its accuracy claims sound modest.
Editorial desk
Chips, data centers, energy, cloud systems, robotics, and physical AI.
Google Cloud’s On platform frames agent-led migration as an enterprise control, cost, and accountability problem—not just a faster way to move workloads.
Editorial desk
Coding agents, APIs, open source, platforms, and practical tools for builders.
Google Playground lets people create and play custom games with AI, shifting generative media from prompts toward interactive rules and authorship.
