John Capobianco is back, and this time the two of us are sitting in Munich building a subnetting game from scratch, live, with spec-driven development.

I've been vibe coding automations for the show for about six months. Some of it works. A lot of it falls over, and I've spent too many late nights chasing bugs in software I couldn't rely on.

John wanted to show me a more structured way to build with an LLM: a constitution first, then a spec, then a clarify step where the model comes back with questions, then a plan, then the code.

Along the way we get into what a spec really is (closer to a network change plan than you'd think), why the tests get written and run without you asking, the token bill that caught John by surprise, and the CCNA study pain that nearly made both of us quit over subnetting. Eric Chou stops in partway through with his own take on why he doesn't get stuck anymore.

It's a real build, so it's messy in the good way.

John Capobianco: https://artofnetworkengineering.com/guests/john-capobianco
Eric Chou: https://artofnetworkengineering.com/guests/eric-chou
Spec Kit: https://github.com/github/spec-kit
NAF / AutoCon: https://networkautomation.forum/

Send us Fan Mail

Support the show

For everything AONE: https://artofnetworkengineering.com/

How much time did you invest in your career last year?

At AutoCon 5 in Munich, Andy sits down with Mike Bushong for a candid conversation about career growth, automation, AI, and staying valuable as network engineering continues to change.

Mike pushes back on one of the industry's most common narratives: learn automation and AI or get left behind.

Technology is changing, and learning new skills creates opportunity. But Mike argues that career development shouldn't be driven by fear. It should be about creating optionality; giving yourself more possible paths as your career and the industry evolve.

Then comes the foosball story.

Mike once worked with an engineer who desperately wanted to move into leadership but wasn't making progress. Over the course of a year, that engineer spent roughly 500 hours getting very good at foosball.

Mike asked him a much harder question:

How many hours did you spend improving your career?

That leads to the central idea of this episode: your expectations and your effort have to match.

Andy and Mike explore how network engineers can keep growing without chasing every trend, why deep technical knowledge may become more valuable as AI commoditizes access to information, and how gaining a little exposure across different areas creates "surface area" for future knowledge to accumulate.

You don't need to improve along every axis. You don't need to become an expert in everything. But you do need to keep growing.

Start small. Build exposure. Go deep where it matters. Let knowledge compound.

And every once in a while, ask yourself:

500 hours of foosball. How many hours on your career?

Full show notes, transcript and links: https://artofnetworkengineering.com/episodes/500-hours-of-foosball-20-hours-on-your-career-mike-bushong-on-staying-valuable

Send us Fan Mail

Support the show

For everything AONE: https://artofnetworkengineering.com/

AI is rapidly changing software development, but why are network engineers still hesitant to trust it?

In this sponsored episode, Andy welcomes back CBT Nuggets trainer and Data Knox creator Knox Hutchinson to discuss the intersection of AI, networking, and modern operations. Before building Transit AI, Knox spent years as a software developer delivering enterprise applications for financial institutions before transitioning into networking, a perspective that gives him a unique view of why AI adoption has looked so different across the two disciplines.

Together they explore:

• Why software engineers embraced AI years before network engineers
• The real reasons network professionals distrust AI, and whether those concerns are justified
• How enterprise software differs from "vibe coding"
• Why AI should investigate network problems, not make changes
• How junior engineers can troubleshoot complex outages with confidence
• The role of guardrails, policy enforcement, and local LLMs in enterprise networking
• A live demonstration of Transit AI troubleshooting a DMVPN outage in under a minute

Rather than replacing engineers, Knox argues that AI can eliminate cognitive overload, accelerate troubleshooting, and help teams collaborate more effectively, while keeping humans firmly in control of production networks.

Links:
- Download Transit AI's free-forever SSH, serial, and API client:
https://transitai.app/aone

- Test drive Transit AI's investigation-only troubleshooting capabilities with their 14-day free trial.

For all things AONE, check out https://artofnetworkengineering.com/

*Transit AI is an AI-assisted SSH client designed to help network engineers investigate and troubleshoot infrastructure while preventing AI from making configuration changes.

Full show notes, transcript and links: https://artofnetworkengineering.com/episodes/cursor-for-network-engineers-meet-transit-ai

Send us Fan Mail

Support the show

For everything AONE: https://artofnetworkengineering.com/