Is Vim Dead? A DevOps Engineer's Take on AI vs

Here's a wild admission from Mischa van den Burg, a DevOps instructor who's mentored over 1,000 people into tech careers: he almost never writes code by hand anymore. Not with Vim, not with VS Code, not with anything that involves typing out syntax character by character.

Instead, he built his entire AI application—CubeCraft AI, which queries two years of coaching calls running on a local Kubernetes cluster—by dictating to AI models. Zero Vim involved in the actual code creation.

So when one of his students asked whether AI speech-to-text models will render Vim obsolete, you'd think the answer would be obvious. Spoiler: it's not.

The Question Nobody Wants to Ask

The student, William, framed it bluntly: "For the last 30 years being a skilled Vim user brought serious productivity gains. But now as speech-to-text models are getting so good, why not just talk for almost everything you do on a computer?"

It's the kind of question that makes old-school developers uncomfortable because it challenges a skill that took years to master. But it's also completely reasonable. If AI can generate code from plain English descriptions, why spend months learning keyboard shortcuts that turn your text editor into a modal editing maze?

Van den Burg's response reveals something interesting about how AI is actually changing development work—and it's not the clean story that either Vim evangelists or AI maximalists want to hear.

Writing Code vs. Navigating Hell

Here's where things get nuanced. Van den Burg admits he rarely writes code manually anymore. "Writing code hardly will be done by humans any longer," he says. "You still need to be able to read the code, but the writing of it almost always happens by just dictating to the models."

But that's for greenfield development—building new things where you can describe what you want and let AI generate it. DevOps work involves something messier: debugging production systems at 3am when everything's on fire.

"There are environments that I worked in that are still just running Linux virtual machines, Linux servers everywhere," van den Burg explains. "You need to be able to go in there and configure something if something goes wrong. Like ideally you'll be doing this with Ansible, but... there is still a situation where you could be on a command line with nothing but your Vim skills and the man pages to fix something."

This is the part AI demos conveniently skip. Sure, you can dictate a Python script to generate user reports. But when you're SSH'd into a production server with no GUI, limited permissions, and a config file that needs fixing right now—that's not an AI-friendly environment. That's a Vim environment.

The Certification Lock-In

Van den Burg lands on a pragmatic argument that has nothing to do with which tool is objectively better: professional certifications still require Vim proficiency.

For DevOps engineers, the CKA (Certified Kubernetes Administrator), CKS (Certified Kubernetes Security Specialist), and CKAD (Certified Kubernetes Application Developer) exams are career gatekeepers. And they're hands-on, time-limited tests where you have two hours to solve real problems in real environments.

"These exams are so freaking hard and you have to do it with high time pressure and any advantage you can get by doing the work faster and solving more questions during these exams is going to be an advantage," he says.

So here's the weird catch-22: even if AI makes Vim less useful for daily work, it's still career-critical because the certification process hasn't caught up to how people actually work in 2025. Learn the tool you won't use much because the test requires it.

That's... not great? But it's honest.

Navigation Still Matters

There's another dimension van den Burg highlights that's easy to miss: navigation speed. Even when he's not writing code, he still uses Vim constantly.

"Just the fact that I can just do VR and open up my recent files like this to me is already very useful," he demonstrates. "The fact that I can do this without using the mouse and being able to go through the information, edit it very easily, nicely. That is something that I've built up over the years and I do not wouldn't like to lose this skill."

This gets at something AI hasn't solved yet: rapid context switching through existing codebases. AI is great at generating new code from descriptions. It's less great at "jump to the definition of this function, check these three files, make a quick edit, and get out" workflows that define debugging and code review.

You could argue that AI will eventually handle those tasks too. Maybe. But right now, there's still value in being able to move through code at keyboard speed without waiting for an AI to interpret your intent.

The Six-Month Window

Van den Burg ends with the most honest thing anyone can say about AI tooling predictions: "This is my current opinion and we'll see where it is in six months from now."

That timeframe isn't random. AI capabilities are moving fast enough that confident statements about what will or won't be obsolete age like milk. The Vim question isn't really about Vim—it's about which technical skills maintain value when AI can automate more of the work.

Right now, the answer seems to be: skills that involve high-pressure problem-solving in constrained environments (production debugging), navigating existing complexity (code review and maintenance), and jumping through certification hoops (professional credentials).

Six months from now? Could be completely different. The fact that someone who built an entire AI application by voice still teaches Vim in his courses tells you something about how muddy this transition actually is.

For anyone learning DevOps today, the practical advice is almost frustratingly moderate: learn enough Vim to be dangerous, but don't overthink it. You might not write much code with it, but you'll probably need it when things break and when you need to prove you know your stuff on an exam.

The future might be voice-to-code. But the present is still modal editing on a command line at 3am trying to fix a config file before your boss wakes up.

—Tyler Nakamura