TLDR DevOps 2026-03-23

Learn how to solve common issues with multi-cloud networking - either active/active or failover - with ngrok's k8s Ingress/Gateway controller.

ngrok's k8s Ingress/Gateway controller routes, shifts and balances traffic between different clouds, giving you fine-grained bi-directional private communication between any k8s clusters regardless of environment (even locally).

No more learning cloud-specific tools and implementations, and no more worrying about if your securities policies are consistent.

👉 Get the guide and wrangle those clouds🤠☁️

Kubernetes 1.36, expected April 22, introduces security and platform improvements, including stronger Linux user namespace support, Gateway API adoption as Ingress-Nginx retires, WatchCache performance updates, enhanced Dynamic Resource Allocation, OCI artifact volumes, and manifest-based admission control configuration.

Datadog released version 4.0.0 of its Terraform provider, introducing predictable monitor access controls, a unified AWS integration resource that replaces four legacy resources, and enhanced security standards, including one-time read application keys. The upgrade also moves the provider to Terraform protocol v6 to support future schema improvements, though teams can continue using v3 configurations until they're ready to migrate.

Kubernetes SIG Network released Ingress2Gateway 1.0, a migration tool that translates Ingress resources to Gateway API configurations and now supports over 30 common Ingress-NGINX annotations (up from just three before 1.0), ahead of the March 2026 Ingress-NGINX retirement deadline. The tool includes controller-level integration tests that verify behavioral equivalence in live clusters and provides warnings about untranslatable configurations to help teams safely modernize their networking stack.

Kubernetes is getting a new Agent Sandbox project designed to handle the shift from short-lived AI tasks to long-running autonomous agents that need persistent identity, isolated environments, and the ability to suspend and rapidly resume. The project introduces a Sandbox CRD and SandboxWarmPool feature that eliminates cold-start delays by maintaining pre-provisioned pods, solving the problem of managing stateful, singleton AI workloads at scale.

Curried functions are elegant and common in functional programming, but their main advantage—partial application—is not unique and can be replicated with other styles using simple syntactic techniques. Tuple-style functions are often more intuitive, composable, and better aligned with how functions conceptually map inputs to outputs, despite currying's aesthetic and niche advantages.

Agents write code from static files and docs, without seeing actual service behavior. With mirrord, you give agents access to your live staging environment so they can test against real APIs, databases, and queues. The result: integration-tested PRs, not broken deploys. Try mirrord free.

AI agents write code faster than ever but they're working blind. With mirrord (5k GitHub stars), they get real-world context (actual API responses, database state, queue payloads) so they produce working code, not guesswork. Every change gets tested against real staging instantly. monday.com cut dev cycle time by 70%. See how it works

Browser-Use is an open-source Python library for AI-powered browser automation. It was launched under MIT licensing. The library offers both a free local version and a cloud API for production use.

Crossplane's function-kro is a new composition function that brings kro's YAML+CEL authoring model into Crossplane's pipeline architecture, which allows developers to define Kubernetes resources using simple declarative syntax with CEL expressions while maintaining compatibility with Crossplane's broader ecosystem of composition functions and operational controls. The function has been donated to the Crossplane community at crossplane-contrib/function-kro and offers full feature parity with the latest kro release, joining Crossplane's existing support for languages like KCL, Go, Python, HCL, and YAML templating.

Cloudways built CW Copilot, an AI-powered SRE agent that monitors over 90,000 servers and analyzes incidents to generate automated diagnostics, remediation steps, and fixes, reducing support workload and speeding resolution using LLM reasoning, orchestration via Ansible, and DigitalOcean serverless inference.

AWS Lambda Managed Instances now supports Rust, allowing developers to run high-performance Rust functions on Lambda-managed EC2 instances with built-in scaling, routing, and load balancing while benefiting from improved utilization and EC2 pricing models.

The Aurora Superintelligence Platform uses deterministic agents and expert human oversight to deliver trustworthy, real‑world security outcomes. Discover how Arctic Wolf sets a higher standard.

Prevent infrastructure mistakes by combining GitHub Copilot custom instructions, GitHub Actions pipelines, and Managed Identities to enforce conventions, automate deployments, and eliminate credentials.

Azure SRE Agent is now generally available with guided onboarding, Deep Context learning, and integrations across logs, code, incidents, and Azure resources to enable automated investigations, faster root cause analysis, and workflow automation across operational systems.

Grafana Cloud now offers built-in observability for Model Context Protocol (MCP) servers through integration with OpenLIT's auto-instrumentation tool, allowing developers to monitor AI agent interactions with external data sources using pre-built dashboards that track tool performance, latency, and errors.