Remember when every website had its own username and password? Each website you visited had different log in rules, and, as a user, it was all so tedious and hard to keep straight!
Thank goodness those experiences are uncommon now. And Keycloak is part of the reason why.
In this episode of 🌩️Thunder, Alexander Schwartz, Keycloak maintainer at IBM, walks through an example Keycloak flow. He shows how a login screen gets customized, how tokens keep you signed in without re-entering your password every five minutes, and how admins can run the whole thing from a simple UI or automate it with a REST API. Heck yeah.
The bottom line is that Keycloak enables you to run your own identity and access management system, free of vendor lock-in, in a highly customizable way.
Kagent allows users to easily build and run AI agents in Kubernetes. It is an open-source programming framework designed for developers and platform engineers.
In this episode of ⚡️Enlightning, Lin Sun, CNCF TOC member and kagent co-creator, explains how kagent’s Agent custom resource lets you declaratively configure everything about your agent, its instructions, which large language model it uses, and which MCP server tools it can call, all with the same YAML and kubectl apply workflow Kubernetes users already know.
Kagent is a Cloud Native Computing Foundation sandbox project.
InfoQ covered my KubeCon EU talk with Viktor Farcic (Upbound), “Choose Your Own Adventure: AI Meets Internal Developer Platform.” The article walks through how AI agents are becoming a new layer of the internal developer platform, and how OpenTelemetry traces let you see what an agent did instead of guessing.
Kubernetes clusters can carry hidden security issues in infrastructure config, workload settings, RBAC permissions, and containers.
Kubescape maintainer Ben Hirschberg joined me on 🌩️Thunder to explain how Kubescape scans a cluster across five dimensions and surfaces what’s wrong and how to fix it.
It checks infrastructure, workload, and RBAC configuration against OPA-based policies (things like unencrypted secrets or containers running as root), catches vulnerabilities with Grype, and watches runtime behavior with eBPF, then provides recommendations about what to fix first for maximum impact.
Serverless can save you money, but the tradeoff is that it is weird. Your CPU gets frozen between requests. Your code runs in a tiny, isolated sandbox with extremely limited resources. And every millisecond of initialization time costs money.
In this episode of Datadog Illuminated, AJ Stuyvenberg, Staff Engineer on Datadog’s Serverless team, explains how his team rewrote Datadog’s serverless agent from scratch, in a language none of them had ever used in production.
The rewrite cut the binary from 55 megabytes to 7 and dropped cold start time from around 400 milliseconds to 50-60.
In this episode of 🌩️Thunder, Ed Warnicke, Distinguished Engineer at Cisco and Network Service Mesh co-founder, explains how NSM connects individual workloads at Layer 3, regardless of whether they’re in different clusters, different clouds, or running on-prem.
Application service meshes connect entire systems to each other. Network Service Mesh connects only the workloads that need to talk, and those workloads don’t know or care where each other is running.
