Good Friends, Systems Thinking, and a Hello World for AI

It is a good day when two friends have a thoughtful conversation about where our industry is going. It is especially nice when something you built together comes back into that conversation months later.

My friends Andy Lapteff and Ryan Booth sat down on The Art of Network Engineering for AI, Systems Thinking, and the Future of Network Engineers. Ryan is one of the original voices in network automation, and I appreciate his measured approach. Andy brings the questions many engineers are asking while trying these tools for themselves.

I’m mentioned in this episode; Ryan is the guest. Their conversation is worth sharing in its own right, and the callbacks to our earlier work made it particularly meaningful to me.

The two callbacks

• 1:06:15 — Hello World for AI, NetBox, and MCP. Andy recalls our session connecting an LLM to a network source of truth. He describes asking about circuits in Atlanta through natural language and getting useful answers from NetBox. That is the kind of first success I want people to have: a question they understand, answered using their own network data.
• 1:11:02 — Our AutoCon subnetting game. Andy mentions the specification-driven development session we did together at AutoCon, where we built a subnetting game. He points listeners to AONE’s resources, where the game is available to play.

Those examples connect two things I care about: helping someone get started and leaving them with something they can actually use.

Systems thinking is already part of the job

Around 21:01, Ryan uses changing a default gateway to explain the difference between a small configuration task and a change across a large environment. The command is only one part of the work. Applications, dependencies, undocumented systems, and other teams all enter the picture.

That example will feel familiar to anyone who has prepared a network change. My takeaway is that our operational experience matters here. We already spend time asking what depends on a service, what could break, and how we will know a change worked. Those are useful questions to bring to AI-assisted work too.

Building a script is the beginning

At 32:08, Andy talks candidly about building scripts and scheduled jobs to help run AONE. Getting individual pieces to work leaves another job: understanding how they depend on each other, spotting failures, and keeping the whole operation running.

I like that he includes the untidy parts of learning. A working demo is exciting. Making it understandable and maintainable for the next person takes more thought. For me, this is where specifications, documentation, and explicit expectations earn their place in the process.

Trust grows through evidence

The discussion around 48:06 asks how we decide to trust an engineer or an agent with responsibility. Ryan talks about testing ability and gradually increasing the work entrusted to it. Andy connects that with familiar automation practices: begin with read-only work, test small changes, and learn from the results.

That is a useful lens for a first MCP project. Start with a bounded question. Look at the evidence behind the answer. Understand the tools before expanding what they can do.

Give yourself a reason to start

At 1:02:10, Ryan encourages learning through a hobby or personal project. A subject you enjoy can give you the motivation to keep experimenting without carrying every complication of your production environment into the first exercise.

The subnetting game and our Hello World session are two approachable starting points. You can play, ask questions, make mistakes, and learn something concrete along the way.

Thank you, Andy and Ryan, for the conversation—and for the kind mentions. I’m glad these earlier sessions are still helping people find their way into the work.

Watch, read, and try it

• Watch Andy and Ryan’s full conversation, or read AONE’s episode notes.
• Revisit You Don’t Need Python Anymore: Hello World For AI, published April 22, 2026, and watch our original build.
• Play AONE’s subnetting game and browse the community resources.
• Explore Ryan’s work and art at abstractryan.ai.
• Continue with my MCP collection or more AONE appearances and mentions.

The podcast episode is dated October 7, 2026; the YouTube upload is dated October 9. All timestamps above refer to the YouTube video, whose timing differs from the podcast audio. The two mentions were checked against YouTube’s auto-generated transcript.