Alongside Dapr Agents v1.0 and the CNCF AI Conformance Program updates, KubeCon Europe 2026 delivered a third piece of production AI agent infrastructure: agentevals, a new open-source project from Solo.io that brings continuous behavioral scoring to agent deployments. The problem agentevals addresses is deceptively simple to state and surprisingly hard to solve: how do you know if your production AI agent is still doing what it’s supposed to do? What agentevals Does Most AI agent evaluation today happens at development time — you run evals before deploying, decide the agent is good enough, and ship it. What happens after deployment is typically monitored through logs and user feedback, not through continuous automated assessment. ...

On March 26 in San Francisco’s Financial District — two days from now — something notable is happening in the AI agent security space: Gen (NASDAQ: GEN, the parent company of Norton, Avast, and LifeLock) is co-hosting an exclusive post-RSA event with the OpenClaw core team. This is the first confirmed public partnership between the OpenClaw team and a major enterprise cybersecurity vendor. And it matters beyond the event itself. ...

饲养龙虾. Sìyǎng lóngxiā. “Raising lobsters.” That’s the phrase that took root in Chinese tech communities to describe the act of setting up and nurturing a personal OpenClaw AI agent. And for a few months, it was a national phenomenon — enthusiastic, grassroots, and spreading fast. Now, according to a sweeping NBC News feature published March 24, the craze is running into its first serious friction: government security concerns, corporate pullbacks, and a mainstream media that still can’t quite tell OpenClaw from OpenAI. ...

JetBrains has been quietly building something bigger than an IDE upgrade. On March 24, the company officially confirmed JetBrains Central — described as “the control and execution plane for agent-driven software production” — with Early Access opening in Q2 2026. If you’ve been following the JetBrains Air IDE (which this site covered earlier), Central is the layer above it. Air is where individual AI coding agents work. Central is where you manage, coordinate, and scale many of them simultaneously. ...

In a single week, OpenAI pulled off one of its most dramatic pivots yet: killing off Sora — the AI video generation app it launched just six months ago — while quietly completing pretraining on its next-generation flagship model internally codenamed “Spud.” The double announcement is more than product housekeeping. It signals OpenAI’s strategic posture heading into its IPO: ruthless focus on frontier model capability at the expense of creative consumer bets. ...

On March 26, Gen Digital — the NASDAQ-listed parent company of Norton, Avast, and LifeLock — will co-host an exclusive post-RSA event in San Francisco’s Financial District with members of the OpenClaw core team. The event, “The Future of Safe AI Agents,” marks what appears to be the first confirmed public partnership between the OpenClaw team and a major cybersecurity vendor. What’s Being Demoed The centerpiece of the event is Gen’s Agent Trust Hub (ATH) — a free security platform launched in February 2026 designed to help individuals and organizations govern AI agent behavior before and during deployment. ...

In China, the community idiom for setting up your own AI agent has a flavor entirely its own: 饲养龙虾 — “raising lobsters.” It’s grassroots, organic, and a little absurd in the best way. And it tells you something important about how a technology with deep American roots became a Chinese phenomenon within months. What Is “Raising Lobsters”? OpenClaw, the open-source AI agent platform created by Austrian programmer Peter Steinberger, has swept China with remarkable speed since its November release. More than 600 million people in China — over a third of the population — now use generative AI, according to a Chinese government-affiliated research group. OpenClaw usage in China is reportedly almost double that in the US, per American cybersecurity firm SecurityScorecard. ...

JetBrains confirmed today that JetBrains Central will open for Early Access in Q2 2026 — and it’s a meaningfully different product from the JetBrains AI-assisted IDE tools we’ve covered before. Where JetBrains Air focuses on AI assistance inside the IDE for individual developers, Central operates at a higher level: it’s described as a “control and execution plane for agent-driven software production.” Think of it as the management layer that coordinates multiple parallel AI coding agents working simultaneously across real codebases. ...

OpenAI made two announcements on March 24 that, taken together, reveal a company in sharp strategic pivot mode: it killed Sora, the AI video app it launched just six months ago, and confirmed through internal sources that it has completed pretraining on a new flagship model codenamed “Spud” — one Sam Altman has privately told staff can “really accelerate the economy.” Sora Is Dead The shutdown of Sora isn’t a quiet deprecation. OpenAI is pulling the standalone app entirely. Sora launched in late 2024 to significant fanfare as a consumer-facing AI video generation tool, but it struggled to find its place in a crowded market that also includes Runway, Pika, and Kling. Multiple outlets — including The New York Times, NBC News, TechCrunch, Axios, CNN, and Variety — confirmed the shutdown on March 24. ...

LangSmith Fleet formalizes two agent authorization models: Assistants (on-behalf-of user credentials) and Claws (fixed service-account credentials). Picking the wrong one creates either security gaps or broken functionality. This guide helps you choose and implement correctly. For background on why this distinction matters, see: LangChain Formalizes Two-Tier Agent Authorization in LangSmith Fleet Decision Framework: Which Model Do You Need? Answer these questions before you write a line of config: 1. Does the agent access data that belongs to the individual user interacting with it? ...