Google’s Web AI initiative is focused on maximizing the potential of running client-side AI, which marks a strategic shift from supporting model training in JavaScript to focusing purely on inference within the browser using its new LiteRT.js runtime. TNS Senior Editor Richard MacManus reports on Google’s vision for an agentic internet and the core beliefs underlying this vision.

In other technologies we’re exploring today:

• TNS interviewed full-stack developer and Angular consultant Sonu Kapoor about why frontend developers should treat machine learning models as a local assets. He explains that tools like ONNX Runtime Web allow models to run directly in the browser. Find out how that benefits the frontend.
• This week at KubeCon, Platform Engineering Labs is demoing Formae, an open source Infrastructure as Code (IaC) platform. It’s designed to solve the complexity and brittleness of traditional IaC tools like Terraform and YAML. TNS Senior Editor Joab Jackson reports that Formae uses Apple’s Pkl configuration language — which enforces schemas and type annotations — and an agentic backend to guard against system drift and error-prone configurations. Read Joab’s full report from KubeCon.

This is the first of two articles. What happens when your machine learning system quietly breaks, and no one notices? Unlike traditional systems that crash loudly and obviously, AI systems tend to fail silently and subtly. They don’t always throw errors, log stack traces or trigger alerts. Instead, they degrade — quietly, over time — until the damage is done. This silent failure...

Running Kubernetes at scale is complex. This IDC Spotlight shows how a modern private cloud platform helps to automate, secure, and simplify private-cloud operations, turning Kubernetes from chaos into control.

Join us on November 20 to explore how leading companies are using Kafka to deliver real business impact. As a bonus, Aiven will be offering a $20 eGift card to anyone who attends and participates in the survey.