Kubernetes 1.37 Preview: What’s New for Cluster Operators
Platform teams preparing for the Kubernetes 1.37 release (codenamed Garhwal) face a shift away from legacy cluster assumptions. With cgroup v1 support fading, IPVS mode in kube-proxy entering deprecation, and static pod config restrictions tightening, 1.37 reinforces cluster security and operational observability—headlined by the long-awaited graduation of metrics.k8s.io to stable.
Sandboxes Are Becoming the New Security Perimeter for AI Agents
As autonomous AI agents move from chat interfaces to executing arbitrary code, browsing external endpoints, and invoking cloud APIs, traditional security boundaries are collapsing. To prevent prompt injections, poisoned code execution, and indirect sandbox escapes, platform engineering teams are shifting security to the runtime level—leveraging lightweight microVMs, gVisor, Kata Containers, and the Kubernetes Agent Sandbox project to turn execution runtimes into the new security perimeter.
Kubernetes Storage Is Cloud-Native’s Hardest Scaling Problem
While Kubernetes abstracts compute and memory with elastic ease, persistent storage introduces physical boundaries around IOPS, latency, network topology, and node locality. As stateful databases and AI inference workloads scale across cloud-native clusters, managing data gravity through Container Storage Interface (CSI) volume expansion, shared vs. local storage trade-offs, and throughput capacity planning has become cloud computing’s toughest operational hurdle.
How VictoriaMetrics Cuts Observability Costs From $65,000 to $1,700 a Month With an Open-Source-First Stack
Roman Khavronenko, co-founder and Engineering Manager, and Aliaksandr Valialkin, co-founder and CTO at VictoriaMetrics, explain why telemetry volumes have exploded over the past decade and
Kubernetes Just Got A Chatty New Assistant, And It Speaks Fluent YAML
Red Hat has introduced an open-source Model Context Protocol (MCP) server for Kubernetes and OpenShift, allowing AI coding assistants to interact directly with cluster resources using natural language. Unlike wrappers that execute command-line utilities, this Go-based tool connects directly to the Kubernetes API server, delivering high-performance, RBAC-enforced cluster troubleshooting and resource management.
How Cerbos Secures Probabilistic AI Agents With Deterministic Authorization
Emre Baran, CEO and co-founder at Cerbos, explains why authorization is the one security layer most teams still hand-roll in code. Externalizing it into human-readable
How Sysdig Is Taking Cloud Security Headless for AI Agents
As attack windows shrink to machine speed, security operations are shifting from reactive dashboard triage to programmable, agent-driven workflows. Sysdig’s headless cloud security leverages the Model Context Protocol (MCP) to embed real-time CNAPP runtime telemetry directly into AI coding agents like Claude Code, Cursor, and Codex—enabling automated vulnerability prioritization, threat investigation, and contextual remediation directly within existing developer environments.
Civo’s Konstruct: A GitOps-Powered IDP in Minutes, Not Months
Integrating Kubernetes, GitOps controllers, secrets synchronization, and DNS automation into a functioning Internal Developer Platform (IDP) often demands months of platform engineering friction. Civo’s Konstruct accelerates this architecture by delivering an out-of-the-box, GitOps-powered IDP that unifies Argo CD, Crossplane, External Secrets, and multi-tenant management clusters into a pre-integrated, production-ready framework.
AWS Patches SDK Flaw That Let A Region Field Steal Credentials
Unvalidated input in regional endpoint templates across seven AWS SDKs allowed malformed region strings to redirect signed API requests to external servers. By manipulating URL parsing with @ and # separators during AssumeRoleWithWebIdentity calls, security researchers at Pi Inc. demonstrated how a single unvalidated input field could expose plaintext bearer tokens and lead to total production cloud compromise.
Small Language Models Beat Giants On Enterprise Cost Per Task
Enterprise AI budgets are facing a reality check as massive frontier models are routinely over-provisioned for repetitive tasks. Discover how small language models (SLMs)—leveraging model distillation, cascade routers, and lower inference latency—are outperforming mega-LLMs on cost per task across classification, structured extraction, and request routing workloads.
Talos 1.13 Adds VRF Routing and Live CPU Hot-Add for Kubernetes
Talos Linux treats Kubernetes nodes as API-managed appliances, eliminating SSH, shells, and package managers to ensure immutable node predictability. Discover how Talos 1.13 advances host-level management—introducing declarative Virtual Routing and Forwarding via VRFConfig and live udev CPU hot-add support to streamline multi-tenant bare-metal routing and rebootless VM scaling.
Intezer: 100% Alert Coverage. How AI + forensics catch what Human SOCs miss
Mitchem Boles, Field CISO at Intezer, argues the SOC model is broken: human teams only reach 30-55% of alerts, and real attacks hide among the
