What's left for infrastructure-as-code after AI moves in?

**Ryan Donovan** (0:08)
Hello, and welcome to The Stack Overflow Podcast, a place to talk all things software and technology. I'm your host, Ryan Donovan, and today we have a fun episode from the fine folks at IBM talking about what's left for the DevOps team after AI moves in. And my guest for that is Rosemary Wang, who is a developer advocate at IBM. So welcome to the show, Rosemary.

**Rosemary Wang** (0:30)
Thank you, Ryan, I'm excited to be here.

**Ryan Donovan** (0:32)
Yeah, so before we get into AI engineering apocalypse, we like to get to know our guest a little bit. How did you get into software technology?

**Rosemary Wang** (0:40)
I got into software technology because of the cloud apocalypse. You know, my first job was in cloud engineering and network engineering. And what was pretty funny about the situation was that cloud was sort of new at the time. And it's not new now, we take it for granted. But at the time, getting servers on demand was a very strange concept.
And so I started out doing automation for private cloud deployments. And at the time, I was very interested in solving intractable automation problems. Basically, how do I get these values from this spreadsheet to paste into the UI as quickly as possible? Because I don't want to waste my time doing it.
And so that's sort of how I fell into software engineering, mostly because I wanted to automate as much as I could and spend more time doing the things I liked and building. And I eventually got moved to network engineering, because I guess folks were like, hey, Rosemary is great at automating things that can't be automated. Let's see if maybe she can automate networking. And the really hilarious thing is that the conversation that I heard 10 years ago from the networking space, which was we're basically automating ourselves out of jobs. It's the cloud engineering apocalypse. Like now it's the AI apocalypse that's upon us. And we like to make apocalypses happen every decade.

**Ryan Donovan** (2:00)
That's right. Every one of these apocalypses gets a whole new set of jobs afterwards, right?

**Rosemary Wang** (2:04)
Exactly. But perhaps that's the point of the apocalypse. It's like it blows up the paradigm of how you work and maybe something new and something better comes out of it. So I'll try to be optimistic.

**Ryan Donovan** (2:14)
You would see the optimism around the DevOps field after AI automates themselves out of a job.

**Rosemary Wang** (2:22)
Yeah, that was the funniest thing that I remember about learning DevOps early on when philosophically everybody was trying to understand that and apply that to the enterprise. A lot of folks who were thinking about DevOps was like, we don't want a single point of failure in our organization handling operations. It's everybody's responsibility to both do development and operations. And I think there were many different fields that came out of this mentality. You have SRE, Site Reliability Engineering.
You now have platform engineering, which is a whole other concern. You know, folks are still saying, we've gone the other direction, where we're shifting specialist knowledge back to a certain subset of people. But I think the interesting thing about AI that we didn't really think about was that it makes us closer to DevOps and that now anybody can create something and deploy it. They may not know exactly what they've deployed or know exactly what they vibe coded or not vibe coded, but they are able to deploy it very quickly. And that's at the core of the interesting AI movement in which maybe we've gotten a little bit closer to DevOps because the information that we need in order to deploy this specialist knowledge has become a little bit more accessible, and we're able to do it more quickly.

**Ryan Donovan** (3:39)
I've had a lot of conversations with folks that talk about software engineers should be very close to DevOps if DevOps should even exist as a separate field. Do you think that the way that AI is making engineers closer to DevOps is beneficial or is it a harder situation to deal with?

**Rosemary Wang** (3:56)
I think that we have in this hype cycle for AI, we've made it a little harder to deal with ourselves because we never learned this lesson of what guardrails should be in place, what enforcement should be in place. It's like how someone learns infrastructures code and infrastructures code is not necessarily DevOps, but the idea of the practices of DevOps being applied to infrastructure, it's very easy to just deploy something. There's this joke of like terraform, you only live once, in which you ride a whole bunch of terraform and you deploy it, and you see how it goes.