An infrastructure engineer’s essay on what the AI adoption push actually does to the job — and the honest part is that he’s watched this movie before. The setup: companies now want AGENTS.md or INSTRUCTIONS.md in every repo so any project is agent-contributable, which is slightly funny because he’s never once gotten a human teammate to read the README, yet now everyone writes better docs than ever — aimed at a robot. The obvious question: does this make engineering redundant?

We’ve been here before

  • Kubernetes killed Ansible — kind of. He hasn’t written a playbook in years, not because config management stopped mattering, but because the unit of work moved up from “the machine” to “the workload.” He still decides what image runs, what it talks to, how it scales, what happens when it fails.
  • Nobody says Kubernetes/Fargate/Cloudflare Containers replaced infra engineers. Each replaced a specific layer of manual work (hand-building images, hand-patching boxes, knowing which node a workload landed on), and the engineers moved up one layer every time. AI is doing the same thing, one layer higher.

What changed day to day

  • Uses Claude daily to generate Helm charts and Terraform modules. What AI removed isn’t the thinking — it’s the lookup work: no more reading AWS provider changelogs to figure out what changed between v5 and v6.
  • Iteration gets output into shippable shape, and then it becomes the example for next time, especially with an AGENTS.md in the repo pointing at it.
  • The same thing happened one level down before: raw Kubernetes YAML → Helm charts → now directing Claude to write the chart.

What hasn’t changed

  • Still needs to know what a good Terraform module or well-structured Helm chart looks like.
  • Still needs to SSH into a node when something genuinely goes wrong and killing the pod isn’t an option — “the layer above doesn’t remove the layer below, it just moves how often you have to touch it.”
  • Still decides the actual shape of things: maintainability a year from now, deployment and versioning. “AI does the time-consuming part. I still give the direction.”

The skill you trade away

  • The honest tradeoff: faster at building and debugging than two years ago, and visibly rustier at the fundamentals underneath that speed. His HCL syntax recall is gone — he hand-wrote a four-level nested for loop (flattening a map of maps of subnets across accounts) in an hour years ago; Claude writes the equivalent in seconds now, and he’d have to sit and think to reproduce it from scratch.
  • Not a hypothetical cost — “one I can feel happening in real time.” Same as engineers who came up after Kubernetes never learned to hand-roll a server image, and were fine because they never needed to.

Where this goes next

  • The uncertain part: how long “I still give the direction” holds. Right now he has the context and the agent doesn’t — not beyond what’s written in a repo’s AGENTS.md. The company-wide AI pushes are exactly the attempt to close that gap: give the agent the whole context, every decision across every repo for years.
  • If that works, an agent with a genuine long-term view of the entire infrastructure might plan better than he does — the same way he can’t out-debug a tool that has read every changelog for every provider.

The conclusion: Kubernetes didn’t replace infrastructure engineers, it replaced a layer of their work and moved them up one. AI isn’t replacing engineering either — it’s busy eating the layer just below “give direction,” and he’s not fully convinced that’s the last layer it eats.