Hey Ala,

The way you prompt AI will decide whether you're employable in two years.

Claude and Codex have solved the ability to write code in proper syntax and style.

This has led engineers to lazily delegate all of the decision making. If you use AI in your development workflow and sometimes feel bored waiting for AI to complete, you're probably one of those engineers.

There are two ways you can prompt AI to write code.

1. Goal oriented - "Build me a web app that lets users upload files, deploy it to Kubernetes, and make it production ready."
2. Instruction oriented - "Write a Go API with one POST /upload endpoint that saves files to an S3 bucket. Package it with a multi-stage Dockerfile that runs as a non-root user. Then write a Kubernetes Deployment with 3 replicas, memory and CPU limits, and a readiness probe on /healthz. Put a ClusterIP Service in front of it."

See the difference?

The first prompt hands over every decision - language, container build, replica count, pod failures, etc.

The second prompt makes every one of those decisions itself and delegates the coding.

Both give you working code, but one will result in you in an uncomfortable position begging the AI to fix errors as they arise because you don't understand how the code works.

To write the second prompt, you have to have the experience to know why a multi-stage build matters. Why you don't run containers as root. What a readiness probe is for.

That knowledge is what companies pay six figures for, not typing YAML.

Automation is the process of converting a repeatable process into an independent controllable system.

So stop trying to automate building something for the first time. Do the fundamentals well so that when you're in a senior devops engineering role, you know exactly how to build what you need.

Then you can give clear, explicit instructions to your AI agent to speed up the workflow that you could already do by hand.

Mischa

P.S. This is exactly why we're so obsessed with homelabs inside KubeCraft. A homelab is where you break things on purpose and gain the experience that others are relying on AI for. If you want to learn how to use your homelab to land your next DevOps role, apply here.