Hey Reader,

Big shifts are happening in the tech industry. AI is having a huge impact on Software Engineering jobs.

Many of my students have asked me, how is this going to affect DevOps jobs?

Is it still worth it to become a DevOps engineer in 2026?

In this newsletter I'll share some data that came in this week that will help you relax tonight.

And yes, it is DEFINITELY worth it to pursue a career in Kubernetes and DevOps in 2026.

In fact, for DevOps, AI is creating MORE jobs that pay even better than before. I'll show you why.

If you want to land a six-figure job in the tech industry, the time has never been better.

​→ Click here if you need help with landing a DevOps job​

​

AI is creating more jobs for Kubernetes & DevOps engineers

A few weeks ago I created a video where I explained why.

You can CLICK HERE to watch that video to get the full breakdown.

But I'll summarize it very quickly for you

• AI runs on Kubernetes
• AI is growing
• Therefore, Kubernetes is growing

68% of companies report being understaffed in AI & ML operations.

If you know Linux and Kubernetes, you're good for the next decade at least.

The amount of jobs is growing. But how long is that going to last?

Isn't DevOps & Kubernetes engineering at risk for being replaced by AI too?

​

Why DevOps jobs are safe from AI replacement

Quesma just released OTelBench. This open-source benchmark tested 14 LLMs on OpenTelemetry tasks that would be easy for a Site Reliability Engineer.

The results are eye-opening.

AI failed 71% of the time.

The best AI model can only trace your failed login 29% of the time.

Observability is how engineers see what's happening inside complex systems when something breaks.

Instead of guessing why your app crashed, you can trace exactly where requests failed.

OpenTelemetry is the industry standard for collecting this data. It tracks requests as they flow through dozens of services so you can pinpoint problems fast.

It's essential infrastructure work, and someone has to set it up correctly.

The benchmark results are harsh.

Claude Opus 4.5 led at 29%. GPT 5.2 followed at 26%.

These weren't complex systems. They were clean, 300-line microservices.

Tasks that any human engineer would find simple.

Here's the most epic fail. Models couldn't tell apart two user journeys.

They tracked HTTP calls but missed the business context. They mixed successful logins with failed token requests in one flawed trace.

They viewed the timeline as a flat list instead of two distinct trees.

This shows the gap between "code that works" and "observability that helps fix issues."

​

Here’s what this benchmark means for your career:

AI can create OpenTelemetry code that compiles. But, it often generates broken traces in production.

You still need to fix issues when alerts sound, making your debugging skills vital.

Models failed completely with 0% success rate in Java, Ruby, and Swift.

If your distributed system has multiple languages, AI finds it hard to manage everything. The complexity makes it challenging.

Your polyglot skills provide you with an edge that automation won't replace.

Context propagation is still impossible.

Models can wrap HTTP calls in spans, but they lack the ability to reason about business logic.

Your understanding of systems—not just code—will keep you in demand.

The training data gap is structural.

The best code bases are in private repos at Netflix, Airbnb, and Apple. AI hasn’t seen enough examples of good work, so your hands-on experience is invaluable.

​

The hard truth for anyone selling "AI-powered observability."

If a model can’t instrument 300 lines of clean Go code, it’s not ready for your production environment with 50 services during an outage.

This doesn’t mean AI is useless for SRE tasks. Some tasks hit 55% pass rates. But the gap between marketing and real capability is huge.

The industry needs more benchmarks like OTelBench, not vendor demos.

We need reproducible, open-source evaluations of actual work.

Until then, you’ll still be writing your tracing code yourself. And you’ll need to review every line an AI suggests against your trace topology. Not your compiler.

​

What this means for your future.

Your DevOps and SRE career is safer from AI than the headlines suggest.

Yes, AI can generate most of the code. But it can’t handle the context needed for large distributed systems. It can't think about business logic. It can't consider user journeys.

It also can't see the links that make observability helpful during incidents.

Engineers who have a deep understanding of these systems will not be replaced soon.

They are the ones who can build, debug, and maintain real infrastructure.

The path to job security is clear: build real skills on real systems.

Not tutorials. Not certifications. Hands-on experience with Kubernetes, distributed tracing, and observability.

This is work that AI can’t do.

Inside KubeCraft, you don’t watch theory videos. You create production-grade Kubernetes environments from the ground up.

Then, you deploy real applications. This way, you build a portfolio that highlights your skills beyond AI.

Your homelab becomes interview-proof. Your GitHub becomes employer bait.

The skills you build lead to six-figure roles that will remain in demand.

→Apply to see if you qualify for KubeCraft​

​

P.S. The engineers who will thrive in the AI era understand what AI can't do. This includes distributed systems, context propagation, and real-world observability. If you want help building those skills quickly—with mentorship, accountability, and a clear path to interviews—book your free call here.

​