What you’re saying is that starting a service in kubernetes as a dev is ok, what other people say is that operating a k8s cluster is hard.

Unless I’m mistaken the managed kubernetes instances were introduced by cloud vendors because regular people couldn’t run kubernetes clusters reliably, and when they went wrong they couldn’t fix them.

Where I am, since cloud is not an option ( large mega corp with regulatory constraints ) they’ve decided to run their own k8s cluster. It doesn’t work well, it’s hard to debug, and they don’t know why it doesn’t work.

Now if you have the right people or can have your cluster managed for you, I guess it’s a different story.

Most megacorps use AWS. It's regrettable that your company can't, but that's pretty atypical. Using AWS Kubernetes is easy and simple.

Not sure why you think this is just "as a dev" rather than operating in production — K8s is much more battle-hardened than someone's random shell scripts.

Personally, I've run K8s clusters for a Very Large tech megacorp (not using managed clusters; we ran the clusters ourselves). It was honestly pretty easy, but we were very experienced infra engineers, and I wouldn't recommend doing it for startups or new projects. However, most startups and new projects will be running in the cloud, and you might as well use managed K8s: it's simple.