LLMs and Infra As Code future thoughts

I think a bit about the feeling I had when I started using Juju… and its hook based reactive system in bash… I know most have moved onto python “Ops” library… but I do wonder with the advent of LLms are people ready to start writing things that look more like prompts instead of Python or bash? How far am I from writing instead of a bash hook… simply a LLM prompt … backed by LLM infused error handling and triaging?

Feels like 7-8 years away, yet I see very little in the sys-admin space about this thought or idea… instead taking LLms to create lists or articles… am I alone here in being a bit eager to quit writing so much code and instead lean on simpler statements like

“instantiate nginx, certbot, and a npm project that is fullstack based at this git url within a k8s cluster (or vms)”

etc etc?