DevOps in the Age of AI: Front-Row to the Transformation
Arguing With a Probabilistic Parrot
It’s 2 a.m., the CI pipeline is green, my coffee is cold, and I’m arguing with an AI that swears a library called flibber-jibber exists.
Ten years ago I’d still be hand-scaffolding a repo; tonight the boilerplate is done—by a machine that confidently makes things up.
That’s what the “age of AI transformation” looks like from my seat in DevOps: not science-fiction robots, just a quiet revolution in the way we build software...
The AI Wave Hits DevOps
The last couple of years have turned large-language models from parlor tricks into serious tools.
As Thomas Ptacek puts it in his Fly.io essay You’re All Nuts, we’ve moved past pasting prompts into a chat box.
Agents now lint, compile, navigate codebases, even run tests—essentially doing the grunt work that once ate hours of an engineer’s day.
Goodbye Mundane, Hello Multipliers
Tools like Cursor and Claude Code have quietly stolen the most mind-numbing parts of my job.
Documentation that no one reads? Gone.
Repo scaffolding and endless copy-paste? A few prompts and it’s ready.
With AI laying the groundwork, I can spend my time on the pieces that actually matter.
My personal productivity has jumped somewhere between 2× and 5×—the difference between hoping to ship an idea this sprint and actually shipping it before the coffee goes cold.
The Secret Sauce: Knowing the Craft
But this isn’t “push button, receive software.”
For people who don't understand code, who never absorbed the pillars, principles and patterns that were baked into me long before AI existed ... it will be a very different experience
Years spent learning design patterns and clean-architecture principles now pay dividends: I can tell an LLM exactly how to layer a DDD application or enforce a service-locator pattern, and it will write that code ten times faster than I ever could.
Without that grounding, newcomers can still coax code from AI, but it’s often brittle or incoherent.
A friend recently asked me about wanting to learn to code and where to start, he mentioned AI. I told him to first learn the basics and then empower them with AI. i even pointed him at one of my favorite books: The Pragmatic Programmer—because solid fundamentals will outlast whatever flavour of AI is hot this year.
The model is a turbo-charged apprentice; you still need the master carpenter’s eye.
Imperfect Partner, Midnight Arguments
AI is fast, but not flawless.
I’ve had full-scale domestics with AI Models at ungodly hours—debating why it hallucinated a library that doesn’t exist or reinvented a pattern we already use.
And sometimes it tries to be too clever.
There are nights when it would have been quicker to open the file and make the change myself.
Instead the AI dreams up an overly engineered solution, as if it’s trying to impress a future code-review committee.
Sometimes the best fix is still a human typing three characters and hitting save.
Then there’s the lost-child moment: ask an AI tool to change directory paths and it panics like it’s been dropped in the woods without a map.
Every turn it stops and asks, “Where am I?”
What follows is duplicated files, mysterious folders, and eventually me, shouting at the screen like a parent whose toddler has wandered off in a supermarket.
I once found an entire copy of my own app nested neatly inside a sub-folder of… the same app.
Inception, but with more rage and fewer Oscars.
And this, circling back to Ptacek’s article, is the point worth remembering:
we are still responsible for the code we commit.
AI is like another engineer making a pull request—you still review it.
The machine may type at lightning speed, but the merge button is still ours to press.
The Role Re-Defined
Not everyone is convinced.
Some colleagues and industry voices dismiss the hype, worried about job displacement or the quality of machine-written code.
Ptacek argues that ignoring these gains is like ignoring the Internet in the ’90s:
the floor has been raised for everyone, but the ceiling still belongs to those who understand the craft.
Opportunities and Risks
The opportunities are obvious: faster delivery, richer automation, time to focus on the work that actually moves the needle.
But there are risks—over-reliance, security blind spots, and the temptation to let fundamentals atrophy.
AI accelerates; it doesn’t absolve.
Human at the Helm
Back to that 2 a.m. scene: the CI pipeline is still green and the AI has finally conceded my four-space argument.
I’m the one who presses deploy.
AI is a powerful tool—my tool—not a replacement.
In the age of AI transformation, the craft of engineering matters more, not less.
If anything, the future belongs to those who can both code and command the machine.