Cloud Native, AI, and the Rise of Platform Engineering | Live from KubeCon Atlanta | TSG Ep. 965

Hey everyone. So evidently, when it comes to cloud native size matters. You're watching Textron Gang.

Hey everyone. It's Alan Shimel and we are at CubeCon right here on the floor at our Techron booth. Mm-hmm.

Giving you a special edition of Textron Gang with a couple of Textron gang members, some old, some new, some Blue, and me. Um, we've got a lot to talk about today. Let me introduce you to our gang members on the far right.

Well, you know him, he's everyone's favorite Yankee fan. Mm-hmm. Mike Vizard.

Hello. Secondly, making his debut on the Text Drug Gang, though he's a, he's my co-host on our Platform Engineering Show podcast, which you should check out. Mm-hmm.

And he wore that little special sticker right for me here. DevOps is dead. Trick or treat.

He, he keeps that sticker on. He may not make another gang. org.

Yeah. Luka Tyler. Thank you.

And then last but not least, she's been on hiatus a little bit. Yes. But we've got her in here live at the Q Con.

Mm-hmm. Our good friend, hope Lynch. She hope.

How are you? Doing well, Al thank you for having back in, back in the saddle. Yeah.

Mm-hmm. So, guys, we're at CubeCon. I guess we gotta talk cloud native, CNCF Mike, what are we kicking it off with?

Well, you gotta pay attention here. 'cause these numbers are kind of like dealing with the Department of Labor Statistics. So what, So do we gotta fire the Statistic?

It may come to that, I don't know. 6 million developers, but they changed how they count that a little bit, because what they're saying is these are anybody who's using any one pro product or project from the community as a whole. So when you read the numbers, they also say there's nine coin, 3 million backend developers.

But underneath that, because the base is somewhat different, they're also noting that the percentage of folks using containers is down. And the percentage of folks using Kubernetes is down from like a high of 36% of the community now down to 30%. So in that, there's definitely been some softness, but you know, you have to also kind of put that up against the total of the numbers.

So that, 'cause it's not a pure apples to apples comparison, but I don't know, I'm gonna start with hope here. Mm-hmm. Um, what are you seeing out there in terms terms of container adoption Kubernetes?

What the, what, what are the folks talking about? So, what the people I have spoken to are talking about is that Kubernetes is still very, very hard. Right.

And I think there is a certain amount of just exhaustion on the part of some, right. And when you consider the economic pressure, a lot of companies are under, and the returns they want to see, if they have technology that their teams are complaining about that's not getting configured, that's not paying off, they probably are less likely to make continued investments. Right.

Um, there are still organizations that are all in on Kubernetes, but I know for certain products I have discussed, people have said, if I need to use Kubernetes, I'm not gonna use the product. Right. So it is actually a blocker for some adoptions.

Yeah. There are some good news in the numbers for folks like you because they also said, well, I think it was like 54% of the folks using cloud native technologies are also using DevOps. But they also pointed to IDP usages up.

So from like 23% to 27%. And in my mind, IDPs is code word for platform engineering. Mm-hmm.

Yeah. So, you know, it looks like that's, you know, catching hole. Yeah.

I mean, a quarter of the population. Does that ring true to you? Absolutely.

Yeah. I think, you know, it's growing a lot. Um, the whole platform engineering space, we've seen, um, you know, I I was just talking about there's this morning, like, um, I think back in 23, there was a prediction by, uh, garner that by 2026, so in two months from now, we would have like 80% of organizations having mm-hmm.

Uh, some sort of platform engineer initiative in place. Mm-hmm. Well, and Dora, the Dora report actually came out, um, like, I dunno, like a month ago mm-hmm.

Saying that actually 90% of organizations they survey already today have a platform in place. So mm-hmm. Even, you know, a, uh, I think like a forecast that seemed very optimistic back in 23, even to the enthusiasts like me in the, in the market, in the industry, um, it was actually beaten.

Um, so, so I think that speaks volumes, you know, to the fact that yeah, platform engineering, you know, keeps being on the rise. Um, but, you know, going back to the Kubernetes thing, I think there is also an interesting, um, you know, like, especially here at CubeCon, a lot of people that there's still, I think like some confusion around like what exactly platform engineering is and, and, and what's, you know, what's the level of context that we're talking about here? Because I had conversations today with a lot of people that are like, Hey, well, I run Kubernetes.

I'm in charge of Kubernetes, so I'm, I'm running the whole platform. And it's like, yeah. You know, Kubernetes obviously is a platform in itself, but it's also not always the whole story, right?

Yes. Yeah. Take a crack at that for us.

What is the difference between platform engineering and DevOps in your mind? 'cause some folks would say, platform engineering is a DevOps mulligan. We didn't get it right the first time.

We're trying it again. Yeah. Well, you know, I think like, you know, DevOps ultimately was, was really about, you know, shifting left and, and, and, you know, this core candidate of like, Hey, you build it, you run it, right?

Mm-hmm. Um, which I think is something, uh, in a, in a, in an, in a, in an industry where we saw a huge increase in complexity and also in the size of the engineering, um, organizations, right? In the last like 10, 15 years since develop became a thing, you know, some organizations when 10 or a hundred x in some cases in terms of really the size.

Um, and, and then, you know, not to mention the complexity, you just need to look at the users need to look at the CNCF landscape to see just like the explosion of complexity, right? So in a world that's like so much bigger and so much more complex, this idea of like, Hey, you're responsible for doing everything, you know, and shifting everything left in the developer doesn't really work. And, you know, we talked about it in the, in our platform engineering podcast before, right?

Mm-hmm. Ultimately, you have to, you know, you have platform engineering. Um, really, it, it, it's sort of like, I, I think the industrialization moment in a way mm-hmm.

Of, of our market, right? Of our community where we're going beyond, um, you know, this, this, you know, this this traditional role of like, okay, now we shift everything onto a persona and actually back to, hey, actually it makes sense probably to specialize, right? Mm-hmm.

Um, and have assembly lines mm-hmm. Which are the golden paths essentially, right? Mm-hmm.

Um, and so I think, I think it's, you know, yeah. You know, we, we ran this like campaign a few, a few years ago, and this, and this idea of like, Hey, DevOps is that, and, and that was definitely like inflammatory, uh, maybe too, too much. So, um, but, you know, uh, certainly I think you have seen, um, I think you can, you can think about from engineering as an evolution of DevOps, um, and, you know, and I think a good one.

Um, But In some organizations though, they still have separate DevOps teams and a platform engineering team that is enabling all of them. So it's really not an either or proposition. No, no.

Yeah. So it's interesting because, you know, I hear that it, uh, anecdotally it goes like this. Mm-hmm.

We had a full stack team of developers that beat up the IT ops team that deploy Kubernetes, and then they stuck them in the corner somewhere and let 'em run Kubernetes. Mm-hmm. Then the IT ops team looked at Kubernetes and said, wait, this thing scales.

This is great, and I can centrally manage it. And then they went to the rest of the developers and said, we want you to write on Kubernetes. And that's when everybody said, say what?

Yeah. Yeah. So I can see that happening.

You better give your shimmy steak on this. All right. The famous Shimmy steak.

Yeah. I gotta throw some stones all around here. Love it.

First of all, Mike, you never lead talking about size and softness. They just don't go Okay. Just a little, just a little less, You know, there's a lot of medicines out there, but Yeah, I get it.

I get it. But size and softness, it's, this is, this is a hard problem too. But that being said, yes, there's no doubt that there's more platform engineers now mm-hmm.

Than there were four years out as, and there's more people who are involved in platform engineering, including SREs, including security people. Mm-hmm. Yes.

Including DevOps engineers and including good old fashioned CIS admin ops. Mm-hmm. Mm-hmm.

But the point I think we're trying to make is, is platform engineering getting a bigger piece of a shrinking pie? Mm-hmm. Or is the pie getting bigger still a platform engineering's getting its share, as is DevOps, as is Argo and GI Ops, all, all of these places.

I think the thing we gotta confront, and this is the God's honest truth, no matter how long the line was to get into this sport mm-hmm. Is that this show has been flat in America now for three years. Yeah.

Mm-hmm. Mm-hmm. It isn't rowing.

Mm-hmm. 60% of the people who come here mm-hmm. It's their first time here.

Yeah. It's their first time here. Mm-hmm.

How come they're not coming back? Yeah. Mm-hmm.

The show's not growing. Mm-hmm. 60%.

This is math. It's simple arithmetic right here. Yeah.

60% of first timers, the show doesn't grow next year. It's the same size show. It's 60% of first timers.

Mm-hmm. Mm-hmm. We're, we're losing, what Are we, because I would argue maybe the fact that they're first timers is a good thing.

'cause it shows that maybe there's new blood in the Platform, just people are coming in. But where are the old people going? Are they going to platform con?

Yeah. Yeah. Of course he says, but You know, something else that I thought of when I came to the show, I think the mix of vendors, sponsors has changed it over the years.

Yes. In the beginning it was very, very focused on a direct Kubernetes connection. Right Now it is like more like a general technology conference.

Yeah, That's true. Because Kubernetes has been embedded into so many different things. It's in Argo, whatever else you've used, and it's a, it, there's so many abstractions above it, then you may not see it.

So all these people are here Yeah. Selling something that sits on top of Kubernetes Well, Or Other, or other stacks too. Right?

Like, I think, you know, you look at the, at the fastest growing, I think, you know, or like the biggest project, uh, right after Kubernetes is OpenTelemetry, right? Mm-hmm. There's like so many observability folks, there's so many security folks, and I think it speaks to your point, hope of like really the, the, the maturity mm-hmm.

That is also increasing. So I Also moving devs to ops, Right? Yeah.

Mm-hmm. Exactly. Mm-hmm.

Which I think, you know, more maturity, and I think in, in a sense you could say, you know, quality over quantity also of the conversation. To your point about that, so people are using OpenTelemetry for things other than Kubernetes. So that's beyond cloud.

Neither My mind observability is, is a world unto itself in independent of coup. It is. No doubt about it.

Yeah, it is. But here's the thing about coup. First time I saw coup in action was in, uh, Austin at the Docker Con Uhhuh 2016 or so.

Mm-hmm. It was frigging hard. And you know what, here we are at Coup Con 2025, it's still frigging hard.

Yeah. It, it's hard. It is.

It's not gotten easier. Yeah. And I think to a certain extent, it reminds me when HTML first came out mm-hmm.

I was no coder, but I learned HTML and I fancied myself. Yeah. A software engineer, nearly famous, right.

Nearly famous. That was before Shimmy set. Right.

But, but here's the thing. All of a sudden they came out with CSS mm-hmm. And the Java script.

Mm-hmm. And I was back to just sitting here by the camera. I, my coding days were over Short Stits.

Right? Yeah. It's a short, it is a short Fiver, but bla of glory.

But with, with Kubernetes, right. If you had that skillset, you still are in high demand because it's still so damn hard. Yeah.

And so, have they made it hard on purpose? There's an argument about that. You know, one of the things that no matter how much I squinted that landscape that you were talking about, I can't find the one that says AI agent for Kubernetes.

And No. And why not? Why not?

It's gotta be coming. Yeah. But I, I, I'll, I'll put half Of the boots here are promising already, you know?

Ah, Yeah, yeah, yeah. I gotta comment about that too. But let me just make one other thing.

Mm-hmm. The p in Kubernetes platform is not the same p in platform engineered Allen. Right.

They're different keys and, and people who think that because they're running Kubernetes, that somehow they've got themselves a, a platform for developers to work on, or, or the rest of the, you know, youth workforce mm-hmm. I think a delusional at some level. Mm-hmm.

Right? It's not, it's, it's not the same. Mm-hmm.

Um, you mentioned ai, though, I gotta tell you the truth. The last two years of, of Q Con, whether here or in Europe mm-hmm. It was all ai all the time.

Just like every other conference we go to. Yeah. I am seeing less AI here today.

I, I don't know about you guys, but seeing less ai, there's, There's two flavors of the conversation. There's AI for Kubernetes in the stack, and then there's AI workloads running on Kubernetes. Mm-hmm.

I see a lot of conversation about AI workloads on Kubernetes. I'm not seeing a lot of AI for the stack kind of conversations. Yeah.

Yeah. So what about AI for IDPs? Yeah.

I Would say, I was just gonna say, we see the same exact thing, right? Like platforms for AI or AI for platforms, and, you know, AI for platforms ultimately is about embedding LLMs and having conversational interfaces for developers to the platform. That's cool.

Mm-hmm. But that's not gonna, you know, like change your life, it anybody, whereas I think, you know, platforms for AI is essential. You know, we were talking on the show, uh, the last couple of weeks, right?

Like, really how, you know, now you have all this like MIT studies and like horror stories of like, oh, all this like AI pilot that's failing in the enterprise area. Well, I think a lot of it is, is really because you don't have an underlying solid foundation and platform that lets you go beyond this initial, like, impressive demo and into day two, day 100 operations. Mm-hmm.

If You, if you go visit Red Hat down there, they're talking about now you can use Lightspeed to talk to more than one Kubernetes cluster. Mm-hmm. Someday soon.

Someday. Soon someday. You, you, your pause will spread.

Know what this soon, on that note, on that note, let's pull the plug on this, on this one. Mm-hmm. We're gonna come back.

We have our next segment. What's it on, Mike? We're gonna talk About AI workloads on Kubernetes AI work.

Imagine that you're watching techron Gang Discover Techron Group, the epicenter of tech innovation. We are your go-to for reaching IT leaders and practitioners worldwide. Our secret impactful content that sparks awareness, engagement, and top quality leads with us.

You'll access editorial websites, streaming videos, virtual events, custom content analyst research, and more. Join our satisfied clients. Let's revolutionize your tech journey.

Contact us today and tell your story to the world in the most powerful way with Textron Group. Hey folks, we're back, and we're gonna have a little chat now about, well, AI workloads on Kubernetes as promised, uh, the folks at the CNCF are trying to build a standard framework for deploying these workloads on Kubernetes. So that ensures that they're portable.

This is turns out to be a big issue because everybody who's built these things are little tiger teams that built something that has a lot of proprietary extensions in it, and mm-hmm. These things may not scale and they may not be easily distributed. And there's all kinds of reasons that are, um, driving a conversation around this, not the least of which is this Cuda framework from nvidia, which is locking everybody in.

And that's kind of the backstory of this whole movement in the CNCF. 'cause this is the first step to trying to figure out how do we create a framework that is actually truly portable. And that's where the backroom conversations are happening these days.

But we've been talking about this in the NVIDIA and Kuda before, it's becoming an issue with folks. And the problem is, like the CIOs and the IT people understand the issue, but the people who wrote the software in the first place were like, were, we did what? Locked into who, Eh, But isn't Cuda Open source?

Open source? It's Open source, but it's led by Nvidia and basically they own It, not part of a foundation. Right?

Yeah. Right. So, Luca, let me ask you, have you seen kda in platform engineering yet?

No, not that much. Not that much. I think, you know, we're, we're seeing a lot of, um, you know, the fastest growing segment in our community right now is data engineers.

Mm. Um, but they're more approaching it for, I think to your point, right? Uh, from an infrastructure first angle.

So they're not necessarily the people that are the users. Mm-hmm. Uh, and so they are like Cuda heavy, and you know, they're not the MLOps people.

They're not the, or the ML engineers Right. Or the data scientists. Right.

They are the kind of like the infra, uh, for, for PLA for for AI people. Mm-hmm. Um, and so we're not seeing too much Cuda yet.

No. Pope got, So where I recently have seen conversations about Cuda is, uh, Linux, uh, OS and CUDA integration, which was, you know, a step forward. Yeah.

But it still isn't solved. The other problems really, uh, that we're dispensing. Right.

Um, because that is at the OS layer above that, it's not really gonna have an impact. But part of this conversation is on the assumption that there's gonna be a separation between the people who train AI and the people who deploy the inference engines, which are smaller. And that's gonna lend itself more to IT ops platform engineering teams.

And they're the ones that were concerned about the portability and the mm-hmm. And how distributed these things are. So this is an evolving conversation in my mind, but I think going into 2026, this is gonna become a bigger hot button issue.

Mm-hmm. Yeah. For sure.

I think it is. But just again, a couple of thoughts on this shimmy thoughts. Number one, $5 trillion is enough.

I don't wanna see them taking over any other markets, all right? Mm-hmm. And if, if, if, if they wanna donate Cuda to the CNCF or the lf, I'm first in line getting behind it.

Mm-hmm. But hello, freeze over first, You know, who's not here giving us keynote speech. It's the only conference in the world that he hasn't given a keynote speech in.

Jensen won. That's exactly right. I didn't see one motorcycle jacket.

Right. So I, to me, I, I think this is, you know, time to throw up the stay away from me vampire, I, I would prefer not to see Cuda mm-hmm. Become the, a defacto standard here to, to deploy this.

I'd prefer not to see Cuda become Yeah. Ingrained into backstage or any IDPs or, or platform engineering in general. Mm-hmm.

I, I think, you know, we, we've gotta be really careful of a lock-in when you got one company that's $5 trillion market cap mm-hmm. And, you know, I, I grew up so Chinese open source competitor, then Maybe, Well, no, but, you know, why can't the lf Okay, why can't the LF and and a consortium Yeah. Sum up with a guess predator.

So I think it would be safe to assume that they are thinking about that very issue. Then the question is becomes, are they gonna like, pull together a MD and everybody else into something to build a unified thing that forces Nvidia in? Or will Nvidia come around and say, look, we're gonna give Coda to you guys, but we're gonna wanna have our Hey guys, Vince will fly first.

Mm-hmm. But my mind continues to go back though with to what you were talking about in the beginning, portability. Right.

My mind says, Hmm, if that problem can be solved, great problem to solve. But it seems like a really hard problem. I, I mean, something in me just does not believe it.

You know, you might assume that there are a lot of primitives in kuda that are optimized specifically for GPUs versus other processors. Yeah. Which I also think is part of this conversation.

'cause I think people are saying, I don't need to run GPUs for every inference workload. And I, and I may not even wanna use it for training. 'cause it's too hard to find them.

They're costly, they're expensive, and there's other ways of skin in this calf. Mm-hmm. Mm-hmm.

Mm-hmm. Yeah. I, I, I agree with that.

Um, now to be fair though, if you were to talk to Jensen Wong and the Nvidia team mm-hmm. You know, they do have a vision that could, is not just for training. Mm-hmm.

No, No. Yeah. It's for a burns as well.

It's for mm-hmm. And I, I would speculate that it's not just even for inference, but it's for AI enabled applications wherever they are. Yeah.

Mm-hmm. Including the CPUs they're building and the DPU that they're building. So they very much have a processor.

I I, you know, there's a, there's a path to go to 10 trillion right Off For sure. Yes, for sure. I mean, the reality right now is like, you know, they, they were, they're the leaders by far.

Mm-hmm. They have the entire stack, they're controlled entire ecosystem. So it's gonna take a while.

But I think what we've learned from like Google Open Source Kubernetes is that, you know, the people behind, and I think this is, honestly, this is what China's doing really well. Mm-hmm. Like, they're behind.

And so they're just like open sourcing, open sourcing, open source, and catching up really fast. Did did we learn anything from those recent price hikes from Broadcom on VMware? Did we?

Come on, think about it. Yeah. Yeah.

It's called Lock It. Lock it. Yeah.

Mm-hmm. But the interesting thing about that is when you look at people who are walking away from VMware, they don't cite price necessarily as the biggest issue. Mm-hmm.

It's about freedom, about moving to the cloud mm-hmm. And stuff like that. So I, I think they cite not the actual cost, but the notion that someone could do that to them is what they're trying to put a in for it's freedom.

So it's a Breach of trust really, was what you're saying. Yeah. Yeah.

Mm-hmm. But I, I just look, I think one of the great things about this conference, about the CNCF and I'm, they're not perfect, God knows. Mm-hmm.

CNCF is not perfect. They gotta figure out how to make the lines work better. But, um, but one of the great things about it is, where else can Broadcom sit down with AWS Yeah.

With, with Microsoft and Google mm-hmm. And agree to cooperate on an orchestrator or a platform, or on GI ops or on observability mm-hmm. Or IDPs Yeah.

Or any of these things. Mm-hmm. That's what makes this special.

Yeah. I, I think we're just waiting for, you know, the United States government to show up as a shareholder in certain processor companies and kind of sort this out for us. Yeah, yeah.

You know, it's interesting. I, I interviewed the CMO of SUSE today. Mm-hmm.

And we spoke a little bit about, uh, digital sovereignty mm-hmm. Which they're very big on. Right.

They're a European company. They're building European data centers with European support staff independent from, whether it's the us, China, or anyone else. Mm-hmm.

This is an important issue to the eu. Mm-hmm. It's an important issue here too.

Yeah. Right. You wanna be free from the government sticking their head Yeah.

Into your pockets. Mm-hmm. You gotta think about that.

So, if the GPU and the software framework came from a US company, you're pretty much outta luck when it comes to sovereign clouds. If you gotta come up with a different answer. Yes, they do.

Yes, they do. And they do make no mistake. They do.

So just more, more fans are from the, from the, of course, More Shimmy fans. Shimmy. So KU is a very complex piece of software.

This is not gonna go away tomorrow. No. But then, but then what's the solu, you know, so you mentioned this like, you know, the portability and the Kubernetes aspect, but then isn't, isn't, isn't there a risk here in like, you know, again, going back to this, you know, Kubernetes is, you mentioned like flatlining a little bit, and there are people that are seeing, you know, some limitations.

Mm-hmm. So like, are, are we trying to touch a much bigger trend to something that's like smaller and has some limitations just because it's the core of this community over here? It's true.

Um, there are a lot of AI workloads running on Kubernetes, but it's still early days. Yeah. Very.

It may not be the platform for AI workloads. We'll, Well, I, so, so I think that's the mission of the CNCF Exactly. To make sure that it is the platform for AI workloads.

Mm-hmm. And, and, and that may well determine whether this thing ever grows more or not. Right.

But that's the bet. They should bet on that. Right.

That's, if it works, great. If it doesn't work, Then agreed. It's not Working.

Or Right. But the only people in the world that are more afraid of Kubernetes than developers are data scientists. So go figure.

There is not. All right. Yeah.

Let's take a break. I think we, we've hit this one 'cause it was too close to the first one. We're gonna come back, we've got our third segment here on the gang today.

Mm-hmm. Mike, what is our third segment? We're Gonna talk about AI maybe, or maybe not replacing SREs.

We'll see AI SREs. I, I've done two interviews today on that. You're watching Textron Gang.

Hey folks, we're back. And we're gonna have a little chat about AI and SREs, because, well, at the show, I think there's at least three vendors talking about AI and SREs here. At least that I met, uh, dev Tron, Causley and Commodore.

And I think there's a couple of more floating around. I know you had some conversations. I, I spoke to Causley and another one.

Mm-hmm. Uh, about A-I-S-R-E. Uh, they changed their name.

Me, SMO, SMO sma. Yeah, of course. So, I mean, on the face of it, you can have a little conversational chat with this AI thing that'll help you figure out whether or not, um, I don't know how to fix an incident faster or why my code sucks and this isn't running.

And, you know, that seems harmless in itself. But as always, I also feel like maybe these folks are over promising and under-delivering somewhat. 'cause anytime I see the word ai, everybody says it's gonna do these amazing things.

And last time I checked all these LLMs, they're, they're mimic engines at the end of the day. Yeah. Hope.

What are you seeing? What's real, uh, For SREs? I honestly feel they're such a critical part of any organization that immediately trusting it to AI feels to me like a huge risk.

Right? If there are, like, with developers, small repeatable tasks that are low risk that it could take on where failure would not be catastrophic, fine. Let them have at it.

Or sending alerts, um, you know, working with observability, integrating alerts, that would be fantastic. But I just, um, I just can't imagine today with the current state of ai right. That it could replace any good SRE in an organization.

And especially if you go to organizations that are highly regulated, right? Uh, the exposure they have for, uh, for breaches, for lack of alignment to regulations, the penalties are enormous. They are absolutely enormous.

And if I were weighing that risk, I wouldn't take the risk. I might say, you know, let's look at it again in 10 years, right? Yeah.

10 years. A long time. It Is a long time.

But finance and regulatory bodies often move pretty slowly. Yeah. Lu or s e's one of the biggest growing segments in the platform community, isn't it?

Yeah. Are they there because they're afraid AI's gonna replace 'em? No.

No. I don't think so. I think it, it goes back to what we were saying earlier, right?

Like AI for platforms or, or platforms for ai. I think here again, it's actually, well, the question is not necessarily, well, can AI replace sre, but what can SRE do to make sure that AI is actually, you know, more available, scalable, um, and actually enterprise grade. Mm-hmm.

Um, and, and again, move it beyond this initial like, demo thing that's like cute, but doesn't actually scale. Mm-hmm. I think, I think Hope has its spot on.

I think most SREs are like, yeah, if you can get rid of some of this Scott work, I hate doing, using ai, great. Like generate a report for what I just did. Mm-hmm.

Super. But the minute it becomes, you know, a question of like trusting it to execute a task that's, you know, mission critical. I think most SREs are gonna be like, no, I'm gonna do that myself because last time I checked, they won't come and fire the AI agent.

They're gonna come and fire me. Right? Mm-hmm.

Right. Because if you've played with AI at all, it will give you an answer and you look at the answer closely and say, is this the answer? Oh, you're absolutely right.

That's wrong. You know? And if no one is making those checks, you are now adding additional overhead to the job of the SRE.

You are decreasing their satisfaction with their work. Um, it, it's a huge risk to me. Yeah.

I, I gotta tell you, let me give you a shimmy's take on this one. You know, I hear all this stuff with the ai mm-hmm. And I'm reminded of when I was a little child, and we all probably heard this from, we were small children, right?

When you did something wrong or something, your parents or mom would say, you better be good, otherwise the boogey man's gonna come sink you away. Right. This one's coming to take you away.

That one's coming to take you away. Mm-hmm. AI is today's boogieman.

If you don't do your job good, or you don't mind your Ps and Qs, or you don't work cheap enough, or you don't, you are gonna be replaced by ai. Mm-hmm. I'm, I'm here to tell you the good news AI's not replacing you today, but like we said before, the person you who is replacing you is the person who uses AI better.

So my whole take on this AI SRE thing is no SREs aren't being replaced by ais. Mm-hmm. SREs are being supplemented by ai.

And if you're an SRE who learns how to supplement your worth with ai mm-hmm. You're gonna be more valuable. If you are an SRE who says, come back in 10 years, you've better go get on unemployment right now.

I don't think The SRE would be making that call, but I I'm with you. Alright. I kind of look at it this way though.

And sometimes, and I wonder, it's like by the time I check the AI's homework, I could have done it myself. And I'm not looking at it to see, you know, is this thing overly verbose? I i, is this thing flawed?

You've Got trucks issues, You got trust issues, but There's a training, there's a long training path there, and you have to stick with it and train it and tune it and train It. This really comes down. Do you trust the ai?

Mm-hmm. Well, And combine, I think, you know, because essentially what ai, right? Like LLMs right?

When we're talking is LLMs. It's just this like probabilistic, you know, you know, but you can, and you know, and until now we've, we've done everything with this like, deterministic approach and, and really think it's like, it's about combine those two things, you know? Mm-hmm.

Because like, the problem is like when people, I see like a lot of companies building all this like, pretty complex, um, you know, AI driven, uh, pipelines and it's like one model, checking another model, blah, blah. And that makes a lot of sense. But it's like some simple, you know, there's a lot of things where you already have all the rules, just like have a deterministic step of that pipeline and just compliment it, you know?

Yeah. Um, and so I think like the, the, the two things together at the end of the day is the, is the more powerful thing. I, I will tell you what every SRE does know, it's, it's one thing to be wrong.

It's another thing to be wrong at scale. Yes. Yes.

At ai Scale, scale, an AI scale is eventually yes. But I, you know, at the end of the day, I think again, this whole AI SRE thing mm-hmm. Is just another, uh, boo boogie head.

Mm-hmm. It's just another, you know, you better behave yourself otherwise, you know, you're not getting nothing for Christmas. Yeah.

Santa Claus isn't Coming. So First it was developers, now it's SREs. Who's next?

Security people, journalists. Lawyers. Who?

Lawyers. Lawyers. Lawyers, doctors.

Mm-hmm. They're all, you know, they're all in the crosshairs. Dick Chimney has a lawyer pass.

Oh. Oh. Hey.

I wouldn't have mind. You know what Shakespeare said, kill them first. Anyway.

Hey, I think that's gonna, I Think we should end on that note. That's gonna wrap this year for Tech Extra Gang. Hey, we are gonna be live all week on text, on tv from here in CubeCon.

We'll also be taping, uh, gang shows, or Whenever we getting the jackets, the leather jackets. That's how you get D and Young to come, you know, it's like you have the gang. We're gonna get you a gang, but we gave your jacket to the ai.

Mm-hmm. Okay. Sorry.

No. Do we have gang color as You're an imaginary fan? Yeah.

Well, we're gonna have to, we'll, we'll have to work there. We're gonna get your Jagged with a DevOps Yes. Thing right over here though, right?

Yeah. Yeah. Hey, enjoy Text Strong gang.

Enjoy Textron tv. If you're watching this, you're probably not a cute con. I'm sorry.

It's a great show, but we're gonna do our best to bring it to you. Until then, though, this is Alan Shimel for Text Strong. We're out.

Mm-hmm.