TLDR DevOps 2026-03-11

Most teams are letting AI decide how they build software. Then they wonder why AI agents don't follow their rules.

Packmind captures your playbook from Slack, PR comments, and scattered docs to build a shared engineering playbook. Then, it automatically enforces it across repos and coding agents.

Don't replace your team's expertise. Turn it into guidelines that they — and your AI assistants — can follow.

Try Packmind (Managed or Open Source)

Kubernetes launched an AI Gateway Working Group to develop standards for networking infrastructure that supports AI workloads, focusing on proposals for payload processing inspection and secure egress routing to external inference services. The group, operating under SIG Network, is actively developing specifications to handle the unique requirements of AI traffic, including prompt filtering, response validation, and token management across Kubernetes clusters.

OpenAI recently raised $110 billion at a $730 billion pre-money valuation, including a $50 billion Amazon investment granting AWS exclusive distribution of Frontier's stateful AI runtime while Azure retains exclusive stateless API access. The deal expands OpenAI's AWS spending by $100 billion and formalizes a multi-cloud AI deployment split between Azure APIs and AWS stateful agent infrastructure.

Red Hat engineers developed a credential provider plugin for CRI-O that enables Kubernetes pods to authenticate to private registry mirrors using namespace-scoped secrets instead of node-level credentials, solving a security isolation problem in multi-tenant clusters. The solution, which requires Kubernetes 1.33's KubeletServiceAccountTokenForCredentialProviders feature gate, uses pod service account tokens to retrieve secrets from each pod's namespace and generates short-lived authentication files for CRI-O, maintaining tenant isolation while preserving the performance benefits of registry mirrors in air-gapped and cost-sensitive environments.

linkerd-destination is the core of Linkerd's control plane. It handles service discovery, policy distribution, and Layer 7 rules. The tool reacts to Kubernetes events, enriches endpoints, streams updates to proxies, and ensures high-scale mesh reliability and observability.

Spacelift introduced Templates to deliver governed self-service infrastructure, enabling platform teams to define immutable, version-pinned configurations while developers deploy standardized stacks without Terraform expertise, preventing drift and maintaining consistent lifecycle control across environments.

Every new monitoring tool, service, or "quick integration" adds another dashboard to the pile—and before you know it, you're spending more time switching between tools than actually building. With Microsoft Azure, your entire stack lives in one workspace: infrastructure, data, AI, and more. See how Azure brings it all together →

Promptfoo is an open-source CLI tool and library that helps developers evaluate, red-team, and scan LLM applications for security vulnerabilities before deployment. The tool supports multiple installation methods, including npm, brew, and pip, and enables developers to compare performance across major models like GPT, Claude, Gemini, and Llama using declarative configs that integrate with CI/CD pipelines.

Hermes is a self-improving AI agent that creates and refines skills from experience, maintains memory across sessions, and can run on infrastructure ranging from a $5 VPS to GPU clusters while supporting 200+ models through providers like OpenRouter and OpenAI. The open-source agent works on Linux, macOS, and WSL2, featuring a one-command installer and the ability to operate independently in the cloud while users interact with it via platforms like Telegram.

AI coding agents can improve code quality if teams use them to aggressively refactor, experiment with architectures, and eliminate technical debt that would otherwise be too time-consuming to fix. Agents drastically reduce the cost of refactoring and prototyping, so teams can adopt a “compound engineering” loop where continuous small improvements compound to produce higher-quality codebases over time.

incident.io keeps its stack simple to scale efficiently: GCP for cloud, GKE Autopilot for Kubernetes, Cloud SQL for Postgres, GCP Pub/Sub for messaging, Argo CD and Buildkite for GitOps and CI/CD, Terraform/Spacelift for infrastructure, and Grafana for monitoring and alerts.

New Relic's Java Agent uses the JVM's instrumentation engine to monitor applications with minimal overhead by operating as a "pass-through observer" that modifies bytecode at runtime rather than storing data locally. The agent includes built-in safeguards like a circuit breaker to prevent memory spikes and automatically propagates transaction context across multi-threaded operations using techniques like Thread Local Storage and a token-based API for complex async workflows.

Most evaluations are set up to fail before they start — wrong environment, wrong metrics, wrong grading criteria. Get the free Buyer's Guide →

Amazon is requiring senior engineers to review and approve any AI-assisted code changes after several outages were linked to the use of AI coding tools.

AWS Transform custom is an AI agent that automates large-scale code transformations such as AWS Lambda runtime upgrades, handling planning, code changes, dependency updates, and validation to reduce manual effort and technical debt across hundreds of repositories.