The AI & DevOps field guide
The AI and DevOps articles on this site, arranged into five reading paths, so a question that spans ten posts reads as one sequence instead of ten separate search results.
By Matthews WongReviewed on
Key takeaway
Five ordered reading paths: MCP and AI agent security, LLM inference serving, Kubernetes for AI, AI SRE and agentic ops, and software supply chain security and regulation. Each path is split into stages, and every article on it carries one line on why it sits where it does.
Five paths, one question each
Pick the path that matches the problem in front of you. Each one lists its stages and how many articles it walks through.
- 01MCP & Agent SecurityPrompt injection, the 2026 MCP incidents, access controls, sandboxes and vetting agent skills.Stages Start with the threat model · The 2026 incidents and advisories · Controls around MCP access and agent credentials · Sandboxes and the supply chain of agent tools25 articles
- 02LLM Inference ServingvLLM and SGLang, disaggregated serving, llm-d and KServe, autoscaling and cost per token.Stages Pick and run an engine · Disaggregated serving and the KV cache · Serving on Kubernetes, and the gateway in front · Autoscaling and cost per token25 articles
- 03Kubernetes for AIThe 1.36 and 1.37 upgrades, DRA and gang scheduling, Gateway API routing and agents in the cluster.Stages Upgrade first: what 1.36 and 1.37 change · GPUs: DRA, gang scheduling and AI conformance · Traffic: Gateway API and inference routing · Agents on the cluster24 articles
- 04AI SRE & Agentic OpsIncident-investigation agents, MCP servers for ops tools, agents in CI, and the IaC beneath.Stages AI SRE agents on call · MCP servers for the ops stack · Agents in CI and code review · The infrastructure code agents act on25 articles
- 05Supply Chain & Regulationnpm and PyPI worms, model signing and AIBOM, secrets in agent commits, the CRA and EU AI Act.Stages Worms and poisoned packages · Models and datasets are artifacts too · Secrets and code that agents add · Regulation and standards with dates attached23 articles
Where each path begins
The first article of each path, for when you want one place to begin.
- Security9 min readPrompt Injection Attacks on AI Coding AgentsMCP & Agent Security: The base case: commands hidden in code, issues and web pages, why filtering fails and how containment reduces the risk.
- AI9 min readRun Local LLMs with Ollama: Models, API, and ModelfileLLM Inference Serving: The simplest local baseline: pull an open-weight model and call it through an OpenAI-compatible API.
- DevOps12 min readKubernetes 1.37 Release: What Breaks When Upgrading From 1.36Kubernetes for AI: The 1.37 overview for platform teams: defaults that flip on, deprecation dates and a preflight script.
- DevOps12 min readHolmesGPT Kubernetes Alerts: Runbooks Beat the Model ChoiceAI SRE & Agentic Ops: Start here: a two-person SRE team routes Prometheus alerts through HolmesGPT into Slack, and finds runbooks matter more than the model.
- Security10 min readSoftware Supply Chain Security: SBOMs and SigstoreSupply Chain & Regulation: The foundation: generate SBOMs, sign artifacts with Sigstore and cosign, and verify SLSA provenance in CI.
A reading list, not a summary
Each hub is a reading list. The articles do the explaining; the hub puts them in an order where each one builds on the one before, and says in one line why it is there.
An article can sit on more than one path when it belongs to both. Under every article a path lists, a link leads back up to the path and on to the next step, so you can walk it from either end.
The review date is the day the lists were last checked against the blog. Each article carries its own publication and update dates; those say how current its facts are.