Agentic AI Matches Human Economists on Accuracy — Beats Them on Consistency
A new paper from the Federal Reserve Board ran the same economics research task given to 146 human teams through three agentic AI systems. When AI models then rank all...
Dispatches
Timely news and updates from the world of agentic AI.
Latest
OpenAI Tightens Grip on Enterprise Agent Development with SDK Overhaul
A significant update to OpenAI's Agents SDK adds native sandboxing, configurable memory, and a new computer-use harness — making agent deployment easier for enterprises building inside OpenAI's ecosystem, and less...
Visa and Mastercard Are Building the Rails for Agentic Commerce
The world's two largest card networks have both moved into agentic AI within weeks of each other — and the implications for how AI systems spend money on your behalf...
NVIDIA Wants to Be the Operating System for Enterprise AI Agents
At GTC 2026, Jensen Huang declared the agent inflection point has arrived. NVIDIA's new Agent Toolkit — and its open source OpenShell runtime — is a serious bid to own...
OpenAI's Operator Agents Are Now Handling Real Business Workflows — Here's What That Means
The shift from AI assistant to AI employee is no longer theoretical. Operator-class agents are being deployed inside enterprises today, and the implications for how businesses are structured are only...
EU AI Act's August Deadline Approaches — And Most Agentic AI Deployments Are Not Ready
Full enforcement of the EU AI Act begins in August 2026. For organisations running AI agents on sensitive data or in high-risk workflows, the clock is ticking — and the...
MIT Researchers Find a Smarter Way to Build Leaner AI Models
A new technique called CompreSSM compresses AI models during training rather than after — cutting compute costs significantly while actually outperforming models trained small from the start.