This month, I launched my Big New Thing: A membership program to accelerate DevOps/SRE/Platform Engineer adoption of AI agent techniques. I'm so excited about what we're creating that I'm having a hard time sleeping.
I'm calling it the Agentic DevOps Guild, or just "the Guild" (no, not That Guild), and we've already had two weeks of meetups and course video updates, with more videos dropping this week.
Guild members get immediate access to these benefits:
- Join our weekly meetups (Summaries and links available if you missed)
- Access my new GitHub Actions and AI for DevOps courses as I make them
- Access the Guild community chat, where we discuss topics between meetups
- Access to upcoming live workshops I'm building on GitHub Actions Security, sandboxing, remote agents, and more
- A 1-on-1 onboarding call with me for accelerating your AI adoption*
*only for Guild signups in March/April
I listed out some goals I have for Guild members. These are the outcomes I want to help you achieve this year:
- Accelerate Your AI Dev Learning
- Become Your Team's AI Lead
- Apply AI-Assisted Engineering to DevOps
- Adopt A Competitive Edge During AI Layoffs
- Create AI Skills and Custom Agents to Reduce Toil
- Adopt Sandboxing and Apply DevSecOps Principles to Agents
- Improve the Safety and Accuracy of Your Agentic Systems
- Learn the Latest Patterns for AI in Your DevOps, PE, and SRE
The program will evolve as I learn what works and what doesn't, and how best we can improve everyone's outcomes for adapting to the rapidly changing environment.
If you want to get more serious about adopting Agent tools and learning what is (and isn't) working for others in the DevOps / SRE / Platform Engineering space, then this is the program for you.
This is all incredibly new, but the potential is undeniable
I've been working on this idea, in some form, for a year. Ever since we launched the Agentic DevOps podcast at KubeCon 2025 in London, I've realized that the pace of LLM improvements and the tools we have to use them has been increasing, yet very little of the developer discussions has been around using these tools in infrastructure and CI/CD. Heck, a year ago, people were saying, "I won't let AI near my infrastructure."
What a difference a year has made.
Now we're seeing a bunch of tools, startups, and news around accelerating CI/CD automation and SRE troubleshooting with LLMs. We've got CNCF projects like kagent to interact with our Kubernetes clusters.
KubeCon itself is changing. Just six months ago, at KubeCon Atlanta 2025, only 2-3 talks out of hundreds even mentioned using AI to do the work of DevOps/PE/SRE. All people were talking about was running LLM inference and custom line-of-business agents. Nirmal and I discussed this gap several times on our two podcasts in 2025, so I knew we were a bit early and that I was betting 2026 would be the year of Ops engineers coming on board the LLM bandwagon.
Then, suddenly, at KubeCon EU last week, controlling infra via MCP and Agents, and even "talking to your infrastructure," was everywhere. I plan to send you a future newsletter catching you up on everything Agentic at KubeCon.
The road ahead won't be easy, but I expect it to be transformative
Our problem is twofold.
- In my expirence, it takes considerable effort to learn how to safely and correctly use a AI stack like Claude Code, AGENTS.md, SKILL.md, sandboxes, MCP, MCP Gateways, and least privilege auth for Agents. Expect to be BAD at it for a while. There's a lot to learn and practice, but what I'm seeing bleeding-edge engineers do has convinced me that our jobs will look very different in a few years. These are the exact skills we'll be practicing in the Guild.
- There is a significant disconnect between what business leadership thinks AI can do today and what we Ops Engineers can accurately and safely accomplish with today's tools and protocols. I've already had multiple Guild members tell me that management expects DevOps/SRE/PE engineers to move at the same pace as a solo developer can with current tools, but that's not the reality of what it takes. AI DevOps tools are in their infancy, while general development tasks are much further along with current tooling and Agent harnesses. A friend says he sent this xkcd to his boss when they expected magic out of the Ops team. It's a perfect representation of the difference between a dev using a local Agent harness and an ops trying to quickly create or recover infrastructure safely and securely with those same tools:
One of my jobs in the Guild is to identify what's working with members and document it for others with video and examples. No one person has all this figured out right now. No one's an expert, and the tools are changing so fast you wouldn't be an expert for long anyway.
So I hope this didn't sound too salesy, 'cus this is my main focus for 2026. I hope for some of the stuff we're learning to eventually land on my YouTube, and of course, we'll be talking about it a lot on the podcast, because I'm tired of the hype and I want to do the real work of building systems to enable Agentic DevOps, and I lie awake at night thinking about what I'm going to experiment with tomorrow.
If you have questions, just hit reply!
