CD Platforms Adapt to AI-Driven DevOps

Hey everybody, we're back at the Open Source Summit here in Minneapolis, and we're having a little chat about, well, CD platforms, because that's a subject that's close to our heart. And my friend Max here is going to give us an update on a lot of the projects. How are you doing?

Hey, nice to meet you. Nice to meet you. I'm pretty much excited to discuss about what are the new future updates, and I'm really excited to share what I learned so far in this particular conference.

Right. So what are the latest and greatest updates to the CD projects? And I know you're kind of deeply involved in that.

So, what's got you excited? So far, in this whole year, right at the far end of 2025 and to the beginning of 2026, we have seen tremendous amount of updates, to be honest. Everything is more getting integrated with AI.

So AI is almost the new thing which we are focusing and hearing every other day. It's more about the adaptation of AI and how we can actually integrate and make it better with the open source world. And as a part of the CDF contribution team, I always see that as an opportunity to emerge, grow, and excel with the AI technologies.

Some of the latest things which we are working upon is the software, so supply chain security, SBOMs, and the projects which are the main core projects are Jenkins, CDF events, or Tellius, and those are the three main projects which we have been focusing and working upon. I think people are trying to figure out if the rise of AI means that they're going to need a new CI/CD platform, or can they extend their existing ones to deal with all this code that's going to be generated by the AI agents. I don't know if you have an opinion about which way that's going to fall, but what's your thought here?

I would say it's just still 50/50, to be honest. So far, there have been so many new nuances of technologies. Every other day, we see new updates, new features getting rolled out by the companies.

But what I think is going to remain static is the open source world. The contributions, where all the players and the ecosystem players, not just only developers and engineers, but overall professionals who will be contributing to those projects. Those are the people who are going to stay in the back end and continuously evaluate those features and actually build the product, not just only for a specific organization, but to overall as a community as a whole.

Mm-hmm. With AI, I would say there will be many pieces of code, snippets, et cetera, and there will be many new software deliveries which will be happening. But overall, I would say the backbone of this thing will be still remaining the same.

Yeah. So what should people be thinking about as they kind of approach this whole subject? Because I see a lot of people are, there's a lot of paralysis of analysis, shall we say.

So, what should people be focused on? What fundamentals and what should they be doing now to get started, or at least, or maybe accelerate their efforts if they're already starting? Yeah.

So, I would say, just be a lifelong learner, to be honest. That would be my main key takeaway. I honestly started working in the open source team a decade ago, and still today I'm actively contributing on various such projects.

Prior to this, contributions, so far I've seen, there have been a tremendous amount of dynamics which have got shifted in the IT industry. But anyone who wants to contribute and learn, the big fundamentals remains the same. Just deep dive into those specific technology or a skill set, and then just continuously work on those things.

One thing which I have learned so far in my career is start low, start slow, and just continuously upgrade yourself. So that has been my mantra so far. Yeah.

Are you at all worried that AI somehow or other is going to replace software engineers? This is one of the key thing, or I would say one of the most controversial take in the industry right now. I would say there could be repercussions of AI in the future, where people would actually would have to hire software engineers or other DevOps professionals, or even data engineers, et cetera, to again see and fundamental thing, what actually the AI pieces of code has done, and what AI has generated.

And this has been a emerging trend in the company so far. I have seen in the FAANG/MAANG companies and in various other tech companies, that whatever the code which has been deployed and rolled by the AI in the production, companies are not able to debug it. And now that's where the whole concern is.

We have seen a phase where companies have started to do a reverse engineering in terms of hiring back the engineers again. Mm-hmm. But it does seem like the job of coding is changing, where I'm not writing as much code as much as I'm reading it and proofreading it and kind of trying to understand it, but how do you maintain the context with code you didn't write in the first place?

Seems like it would be difficult to debug something that I didn't intimately know. One good thing, for myself, which I can advocate about, is having a strong fundamental knowledge about the domain. So if I'm actually well-versed with any open source project, let's say Jenkins, then I would deep dive into that project so much that I'll know in and out about any specific feature or a specific technology which I'm learning.

And that's what I would recommend anyone. Try to go deep dive, increase your knowledge, expand your fundamentals as much as possible, so that in future also, if anyone is going to replace you, might not be able to actually replace you. Yeah.

It feels like, to meMaybe too many people are using AI like a crutch, and they're relying on it so heavily that they don't understand the code that they're passing through, and they're not really checking the quality or how efficient it is. So, do we need to kind of revisit our fundamentals in software development best practices? Yes, I do agree with you.

And one of the key things which I have noticed is companies are continuously trying to focus and rethink and evaluate on the guardrails and governance of any specific measurements with AI. So companies are now actually integrating some of the test cases solely on the AI-driven things. So if there's any pieces of code which has been automated using AI or wipe coding, then they have been specifically putting some guardrails and exclusive test cases so that the test cases, when they are getting passed or when they are going for quality checks, they are not just simply getting pushed in the production.

One of the things I'm trying to wrap my mind around is, so let's say there's 10 members of a DevOps team. Each of them will have their own agents. Each of them will probably have five, six agents.

Mm-hmm. Then there'll probably be agents that are working on behalf of the company where it's been assigned a task like you're the test agent. How will all this get orchestrated in some way that's meaningful?

Because, what it seems to me, my agents will need to negotiate with your agents, which in turn will negotiate with a corporate agent. It seems like the orchestration for all of that is a work in progress. Yeah.

So last week, there were significant changes, which happened in our company and we were simply thinking and considering this particular aspect. If Nikunj Doshi is an employee of any XYZ company, then how is Nikunj Doshi going to virtually or digitally going to replicate in terms of the skills? So, we got this news that companies are no longer actually focusing on building agents, but rather than building skills.

So now, the entire game has been focused on establishing the skill sets and building those skill sets with the AI, rather than just having simply the AI agents. Mm-hmm. So with that, if Nikunj Doshi is a software engineer or a DevOps engineer, then Nikunj Doshi's agent will be specifically skilled and designed for Python, AWS, Red Hat OpenShift, Kubernetes, Azure, GCP.

All those major cloud and DevOps techs which Nikunj Doshi is actually skilled with. So it's not just simply about an agent, it's more about a virtual skill builder. And the skills are where the agent essentially takes on the personality of a task or a function, and that's how it learns what to do.

Is that also going to maybe help bring down some of the costs of using this stuff? Because I talked to a lot of people and they're struggling with token maxing, and they're starting to figure out that the cost of some of this stuff is a lot higher than they thought it was going to be. So, do we need to rethink what it is we're putting in the context window for the AI agent to keep those costs under control?

Yeah. So, last week only, I just read this news. There's a company which is majorly renowned for streaming analytics and content delivery.

They just ran into this trouble. They were simply focusing on having token maximizations. They had adopted AI who would do the job of three or five software engineers.

Now, as soon as their token got maximized, they saw their bills. When their CFO saw the bills, it was like the bills were going more than almost 500K. And a salary of a software engineer for a month would be just around, I would say, pretty much $5,000 to $10,000 a month.

So you think how much the companies are investing in AI and what kind of rethink strategies they would have to think. Mm-hmm. So ultimately, yeah, you might wind up spending more on an AI agent in coding than it costs to hire the human developer in the first place.

But I'm not sure that's the math metric we're really looking for, because I'm assuming the AI agent, well, doesn't get sick, doesn't go home and it's always working, dare say. Yeah. But, as we kind of look at all this, though, is it really a fundamentally different era or we've just kind of accelerated the pace of the previous era?

Yeah. To be honest, in this last five years, especially post-COVID, we have seen fundamental changes, especially at the beginning of 2024, all the way to this mid of 2026. We have seen not an era actually, but more of these dynamic shifts getting continuously.

We evolved from AI, now to gen AI, now agentic AI, and moreover, in the future, we might see artificial super intelligence and any other AI coined terms. Yeah. So it's like, AI was already there, AI automation was already there.

AI automation is also going to remain in the future. It's just that the terms are getting changed, and then new terms are getting coined continuously. Mm-hmm.

So, what do you see people doing that just makes you shake your head a little bit and go, "Folks, we need to be a little bit smarter than that"? Obviously, yeah. Moreover, more than being smart, people have to be adaptive and continuously learn new skill sets.

If I, as an engineer, I always have to learn so many different skill sets every now and then. Keep myself getting to learn all those technologies every now and then. " And so how do you make the time to go learn all that stuff?

Yeah, that is there. Yeah. Sometimes, instead of having to put nine hours a day, I have been putting 12 hours a day, where I have been preparing myself for cloud and DevOps certifications and somewhat AI certifications as well nowadays.

So I do get that point. There is a churn rate and a saturation level of a human being, but sometimes we have to think it in a broader perspective that how we can actually invest in our futures rather than just thinking for a short-term goal. All right.

Everybody needs a nap at some point. So let's say that you have been granted supreme powers and you could fix anything in the DevOps world, anything in the CD world, CI. Is there something that you're seeing out there that we should really all focus a little more attention on?

Sure thing. If I have a magic wand, the three things which I want to fix is, the very first thing is the end-to-end deployment. Many times end-to-end deployment has so many micro stages, right from rollovers, cutovers, then pre-deployment, then passing those test cases.

If I were to get a future as a superpower that I can see and visualize how is my actually workflow going to look like in the production area, rather than just looking at the Git codes, et cetera. This is the number one thing which I would like to see in the near future. All right.

Hey, Max, thanks for coming by. Oh, yeah, Nikki.