Viktor 00:00:00.000 I suspect that people will be moving towards AI Foundation Conference or whatever it's called, and similar. 'Cause we're privileged people, kind of like, we go to multiple conferences. Mm-hmm. But a normal person feels fortunate if their company approves one conference a year. Right? Mm-hmm. that's a normal thing. And then, then it's not kind of, "Okay, uh," if the question would be, "Would it be cool for a person to go to both KubeCon and AI something something?" Then the answer is yes. Mm-hmm. But if you can choose only one, I suspect that, um, AI will be growing and KubeCon will be, uh, going down. This is DevOps Paradox, episode number 377, KubeCon North America 2026 review
Darin 00:01:41.526 Well, here we are. Well, Viktor and Whitney are there at KubeCon NA in Salt Lake City. I am still sitting in beautiful, wonderful Texas, where it's a nice balmy 85 degrees, I think. What's it like in…
Viktor 00:01:54.946 Degrees
Darin 00:01:56.607 it's the last gasps of summer in Texas. What's it like in Salt Lake?
Viktor 00:02:02.315 It's actually nice. It's really nice, yeah. Yeah, the
Darin 00:02:04.339 so it's fall. It's a proper fall
Viktor 00:02:06.918 Yeah, I mean, you can, you can easily go out without a jacket. Yeah
Darin 00:02:10.485 Nice. we could talk about weather for 30 minutes. We won't.
Viktor 00:02:14.290 I probably know, have more information about whether it's Salt Lake City than KubeCon is Salt Lake City
Darin 00:02:20.961 That's why Whitney. is
Viktor 00:02:22.029 We have Whitney. Exactly. Exactly
Darin 00:02:24.182 that's why we have Whitney here. in case you haven't heard Whitney, Whitney Lee, she's been on a lot of times, been with us on all the KubeCon reviews. so I have a list of things that I've just sort of… I tried to keep track, and some of it was good information, some of it was, like, eh information. the only thing that really jumped out at me, like big to me, is that Nats graduated That's the only thing that seemed big to me
Viktor 00:02:51.592 I think at least from those that I've seen, potentially big that's agent substrate. Yep. Or OpenShell maybe too, but running agents on Kubernetes generally speaking.
Darin 00:03:02.757 So that actually brings up-- I have that further down the list. So OpenShell was donated from NVIDIA. It's in sandbox. Agent Substrate is in sandbox. And then I didn't put the date on it, but Nirmata has already written a post, OpenShell versus Agent Substrate versus Kubernetes Agent Sandbox So they're ready. But I looked at him and I went, "Yeah, so what? What, what is the so what on these things?"
Viktor 00:03:37.638 I mean, I have mixed feelings about it, generally speaking. The good thing is that, is basically that it makes snapshot of what your agents have. Mm-hmm. So it, they can be shut down until they're needed again that's kind of cool, So they're not meant to… Normally, you would run an agent, and an agent would do something, and then it would keep existing until you choose to destroy it, right? Uh, this is kind of, "Hey, yeah, you, you destroy it when you want, but in between, if it's not used, we can shut it down, put it to sleep," right? Uh, and when it wakes up, it will have whatever it needs. that's the part I like. The part I don't like is that, at least in my case, agents have very, very variable resource needs. Mm-hmm. Next five minutes, I will need 30 cores, and then I'm going to drop to 0.1 core, and then I'm going to go to two cores, right? I believe that was, like, one of the selling points of OpenShell, is that it, the first thing it does is observe the agent and what it wants, and then it ge- it hands you, like, the policy that you can, it recommends. Yeah, but what do you recommend in my story, in my example of, okay, is it 32 cores? Yeah. Is it 0.1 core? Right? Kind of what is it? Right? It's- And, and it's, it l- Kubernetes still makes a lot of sense for those things. It's just, it's not necessarily designed for, I need you f- to run forever and ever, and I have no idea what you need. It's humans' a- attempt to control the uncontrollable, basically. Yeah. Yeah. Yeah. So I'm still slightly inclined more towards this is my VM that is for more than one agent, and then, yeah, you need 32, and then somebody else needs one, right? And so kind of like they're on average, um, when there are many of them, they-- I know wha- what they need on average for all of them as a total, right? What kind of limits and guardrails do you think are necessary in gener- generally speaking for agents? for agents, I, I don't know. I'm not in a group of paranoid people. Uh-huh. So, uh, it needs, uh… Let's put it this way. I think that agent, just as people need to be able to do what they need to do. Yes. Uh, so that, that's the limit. Don't give them more than what they need- Mm-hmm … but don't give them less either. And I know that majority of companies was giving people less historically all the time, kind of like, "Oh, yeah, you cannot do anything, but you're safe," right? Um- When you're really, I guess, with humans and with the agents, you're wasting both time when they can't do what they need to do, and they're all the time waste been trying. Yeah. And then in some ways, you're opening up to security problems because it's gonna try to figure out, people and AI are gonna try to figure out how to go around. Correct. And it's also, I, I think that this is the story almost the same as with people. I cannot know in advance everything that a person or an agent needs. Mm-hmm. Like how, how do I know? Oh, you need a Go compiler. I, okay, cool, cool. But, you will decide to do something with Python afterwards because it's easier to calculate some average of something, something, right? I, I still think there's a world that you just, uh, for high variability, you observe it for a long time, and then from that information, put on guardrails that are, you know, loose but sufficient or, you know, uh, I'll say, um, maybe loose isn't the word, but there's a lot. And then you at least have something, and then maybe you also have some sort of way for an agent to go outside of those guardrails per a human-approved reque- request. Yeah. It's just that I feel that that's a wrong approach. What's the right approach? Uh, so that, that's basically essentially the same approach as what we've been doing for decades in terms like kind of we're going to decide what the, we're going to technically enable and disable certain things to be done, right? For decades. What, what's an example of that? Uh, you cannot SSH, uh, but you can SCP. Okay. Right? Got it. Uh-huh. Uh, blacklist, whitelist type of approach. Right. Well, I think that the right approach, and with experienced humans as with agents, is kind of let's figure out, let's talk and explain to each other what you should or shouldn't be able to do. Mm-hmm. Right? I feel, I feel that that works better, kind of you're not allowed to push to main. Mm-hmm. Which is very different from the ability, permission to push to main does not allow you to do it. Mm-hmm. Because there will be a time that it should push to main
Darin 00:08:21.725 We should push to main all the time, period.
Viktor 00:08:24.311 Oh, the-- we can have a long instruction-- discussion if you want now about why we shouldn't be using PRs
Darin 00:08:30.556 Yeah, we don't have that much time. one of the other things that I saw was… Actually, a couple things. Let me, let me get some of the soft points out and then we can talk more tech. Currently, CNCF has 231 projects They are counting a 21 million developer population, which feels a little high to me, but you know, what do I know?
Viktor 00:08:53.109 accounting agents now? Oh, then, then it's- You said 21 million … it's, it's too low. The number is too low. That's a single company only
Darin 00:09:01.853 Two other projects that I saw were the Open Cluster Management, OCM is in incubating, and then Kubray was donated. I didn't look to see where it's at. Are you excited about OCM at all? That seems reasonable for somebody that's running that kind of scale It's a open source way to manage multiple clusters, whether it's self-hosted or managed, meaning one of the big three type deal. So you have one hyper layer, I'm making things up now, one meta layer, can't use any of the names anymore, to manage all your clusters at once. That's the way I read
Viktor 00:09:44.347 So their flavor? Yeah. So it feels like cluster API type of something
Darin 00:09:50.283 Somewhere in that family
Viktor 00:09:51.937 I don't know. I don't care. I mean, kinda… h-here's where I'm going with this is do you care whether tool A or B is used if you got what you asked for today?
Darin 00:10:08.008 No
Viktor 00:10:09.397 Right, kind of like if, uh, if you say, "Hey, can you, can you calculate, uh, the averages in from, or numbers from this Excel sheet?" And gonna-- And then do you enter kind of should it do it with Python? Should it do it with Node.js? Should it write a shell script, or do you not care?
Darin 00:10:26.745 I don't care anymore That last word is what
Viktor 00:10:31.097 general. Yeah, yeah, yeah. Kind of, now what the direction I feel that we should be going is, okay, I have-- I'm more experienced, for example, and I will have specific requirements like this is the latency, it could be fault-tolerant, it should be this, or it should be constantly managed, blah, blah, blah, whatever the requirements are. So that will make potentially Claude do better than somebody es-else asking for something similar. But I'm still not choosing the tools. that's the point that I think is important. Not necessarily today, but I'm sure that that's where we are going. Did you get Cl- Uh, Darin, did you get your cluster yet? Okay. That's it? My cluster is running.
Darin 00:11:15.793 or not? Is it up and running? Do I have it and is it running correctly?
Viktor 00:11:20.078 Yeah. And I, I repeat the same story over and over again. You're a manager now, and if you would be managing people, you would probably have the same reaction, kind of like, "I trust you guys," kind of. Did they get what they wa- asked? Yeah? Yeah. Cool.
Darin 00:11:34.700 Two other sort of soft items that I saw. There's a new certification, the Certified Kubernetes Network Engineer Just feels like yet another reason to spend money.
Viktor 00:11:46.067 They have a new certification every coupon now. I have a business idea, Darin. Um, yeah, uh, how about we start a business where actually we would take open weight models and then make them certified and, uh- … serve those? Yeah. Kind of like, uh, I have, uh, Kimi that is certified in XYZ. Huh?
Darin 00:12:11.920 That's the next level of the… I've been watching some reels or something like on LinkedIn about AI interviewers, where the interviewer is AI, and then what people are doing as the interviewee. I don't know. I d-
Viktor 00:12:26.908 What are people doing as the interviewee
Darin 00:12:29.471 Oh, like doing s- stupid stuff like sitting there butt naked or, which they didn't show everything, or just wearing like balloons on their head or something that's like… And the AI responds of something like, "Hey, look, this is a professional business thing. You should take it seriously." Look, if you want it to be serious, there'd be sh- a human talking to me, not a machine. Well, leave that
Viktor 00:12:55.790 But, uh, th- think of it this way, sorry for going back to that. Especially in bigger companies, you will have, uh, different screenings, right? You wanna reduce the number of people from 500 candidates to 50, as you're progressing towards that one person that you're really, those interviews become more concrete, right? So in my head, it does makes to- make sense to kind of use some sort of AI to discard first 90%.
Darin 00:13:25.245 Okay
Viktor 00:13:26.190 Because th- those same, uh, y- the, the HR person would not be any better, the, the, the point I'm trying to make, right? Tech lead coming later and speaking to you, that will be better. But the HR kind of like, "I don't know. I, I was told that, uh, this company does something about Kubernetes. Do you know Kubernetes?" There's-- I hear you. There's something that feels bad about- … uh, w- using it with people like that,
Darin 00:13:50.500 Let me flip it around, right? Yeah, I, I agree. So if instead of it being a human avatar the AI, make it like a Muppet or
Viktor 00:14:00.885 yeah, 100%. Yeah
Darin 00:14:02.965 don't, don't try to make it look human, just make it look what it is. It's a machine. So make it, you know, "Hey, who would you like to h- have interview you today? Would you like Stitch? Would you like Animal? Would you like Kermit?" You know? "Who, who do you want to interview you today?" I think that would be fun
Viktor 00:14:20.395 Yeah, that I agree. Kind of don't make it something that it's not
Darin 00:14:24.980 the last thing I'll call out and then we'll go through all of Whitney's list. Um, they seem to make a big deal about user groups. Is, I was like, I thought there'd been user groups all along. Was I wrong in that?
Viktor 00:14:39.158 Right. End users has, is becoming more and more of a thing, and like the end user tab and stuff. Yeah. they technically existed before, but now they're getting like a more formalized role and there's like the, um, technical advisory board. The tab, I believe, is made up of end users. Mm-hmm. So the governing used to be completely community, and now there's a lot more end users and there's, um… Uh, they showed end users reference architecture that they're sharing with each other. And it's interesting too, because the end users are often not tech companies. Like, they're like Mercedes or CVS- Yeah … or something. And so it's weird to be like at this big tech conference and then watch a CVS presentation about the, their architecture. Because every software company that, that ever considered sponsoring KubeCon is sponsoring KubeCon. How do you grow from there? Right? Kind of, and it's a- Yeah … okay, but there, there are all other companies in the, in the world, kind of like, let's, uh, le- let's get Mercedes. Yeah.
Darin 00:15:41.099 Have you heard an attendance number?
Viktor 00:15:42.793 mean, can… I wanna go back to that. Actually, there, that, that's a spicy topic- Oh, that's, uh- … I'm gonna get to that in a second. Yeah, yeah, yeah, yeah, yeah. But, like, so, uh, if you're an end user, you're basically saying, like, instead of choosing a vendored product, I'm choosing to use this, the open source projects, but then I'm gonna give money directly to CNCF. Is that what's happening? Instead of giving money to, like, Red Hat or whatever? Yeah, or probably both, but yeah. Uh-huh. Yeah. And what's the advantage of that? Um, because I can. Be- Because- I don't get what they get from it. What do you mean you don't? Spotlight on a keynote. Is that it? Yeah. Just ego? Yeah. Kind of like, those are rich people. Kind of their ego costs more than yours. That's the thing, right? Kind of like- Do you know what they get out of it, Darin?
Darin 00:16:29.172 I think Viktor's, his snark, he's correct.
Viktor 00:16:32.619 Yeah
Darin 00:16:33.582 they see it as a marketing ploy because let's be real, KubeCon not a tech conference anymore.
Viktor 00:16:41.760 I do think CBS said something about, like, trying to hire world-class engineers, so that's, like, something. But that still seems expensive for what they're paying to be on the stage.
Darin 00:16:52.160 it's, think back to, Capital One, Netflix, what was the picture sharing site that was huge? I don't know if they still are or not. Pinterest, right? They, they sh- they were always at all the shows talking. It's like, "Hey, and we're hiring." So the figure of thing, the further they can get up the pipe, meaning the f- bigger stages that they can get eyeballs in, maybe they can find the people they need. But in this land of AI, I don't know that we need to do that anymore
Viktor 00:17:23.939 Take that money and make their salaries higher and the people you're trying to hire, that'll do it Oh, I guess if I said it, start making cars that people wanna buy I'm just trying to I'm not even trying to criticize. I'm trying to understand, honestly
Darin 00:17:41.659 we're trying to apply logic to the situation. I don't think that's gonna help.
Viktor 00:17:46.631 Yeah, this is g- this is going too far. the subject. Find me something without logic. Oh, ask, ask us about the number of people at KubeCon again.
Darin 00:17:54.602 Yeah, go, go ahead and, uh… So how many people are there?
Viktor 00:17:59.500 On this stage, Chris Aniszczyk, he, he said, "People always want to know how many people are at KubeCon." And then he threw up this number like, "This is how many people attended CNCF events worldwide." And then he never said how many people were at KubeCon. Yeah. He totally dodged the question.
Darin 00:18:16.189 Yeah, I, I didn't write that number down because
Viktor 00:18:18.377 No, I didn't
Darin 00:18:19.174 make sense. I don't care about that number." Because you think about it, if, if they're doing… If they count all of the KCDs, a- anything that's even tangentially related to CNCF, it's like that number's so fake to begin with. to anybody at CNCF listening. just give us real numbers,
Viktor 00:18:40.519 Yeah, but you know, kind of it, it does matter. It says that, you know, it's similar like when you have shareholder meetings of public companies, and, uh, this quarter we decide not to disclose the results, then you know where to… That's the best way to know the results, right? Um, yeah, I mean, the number of masks have been lower than- I'm trying to Google it now- last year … and I can see that they had expected 9,000, but I'm not seeing how many actual. I… The n-
Darin 00:19:11.573 you?
Viktor 00:19:13.299 That's definitely less than EU Oh, yeah. A lot. Mm-hmm … and I think it's less than US as well. The expected number, I think it was 10,000. Was it 10,000? Something like that. Yeah, usually they hit. So the expected number is lower than the, the previous year number, and then the number is less than expected number. The actual is probably lower than expected. Mm-hmm. And I can tell you, being here, walking by the rooms, looking in them, uh, most of them are about half full, at that. Yeah, yeah. Yeah. Yeah. I mean, what was it, Amsterdam, kind of like people were queuing- Mm-hmm. Lines out the door … in front of rooms. Mm-hmm. And I doubt that they had less rooms than, than here. Mm-hmm. So, uh, it was… You can notice swarm of people. Uh, not this one. The expo floor felt all right. I heard vendors were okay. Yeah, yeah, yeah. They're happy. Yeah, yeah. Yeah. That, that's what I heard as well, yeah
Darin 00:20:02.771 Was there anything that was on the floor that was not AI?
Viktor 00:20:07.829 Coffee
Darin 00:20:09.638 I guess that's all we'll say about that.
Viktor 00:20:11.614 Puppy petting area. Yeah. I mean, that is obviously-- It's so obvious that it's only… The only thing, the, the only thing that matters right now, at least from the vendor perspective, is AI, right? And then we discuss for what, right? since this is KubeCon, you know, that's the typical stuff that you normally
Hear At Kubecon 00:20:30.905 observability, well, AI for observability, uh, infrastructure, well, AI for infrastructure, and so on and so forth, right? one thing- I'd say observability for AI is more important than AI for observability. Yeah, both of them. Kind of like, yeah. Uh, uh, observability for AI with AI. There we go. Yeah. Um, where was I? I forgot. But yeah, it's, it's all AI
Darin 00:20:59.156 Whitney, what's on your list that I didn't bring up?
Viktor 00:21:01.599 Inference serving
Darin 00:21:04.029 Oh, okay.
Viktor 00:21:05.595 I think everyone's talking about it and nobody's doing it, as best I can tell. Yeah. Yeah. And, and-
Darin 00:21:15.962 Models are too stinking big
Viktor 00:21:18.164 yeah, kind of like there are a couple of obstacles. Finding a piece of land where you're gonna do it. Building a huge, massive building where, where it's going to happen. Uh, buy GPUs that you cannot buy. Um, and then, uh, then making a projection of return of investment, let's, for let's say three years, and that's where, uh, being religious very helps a lot. Because then you start praying that nothing changes in those three years and does not mess you up completely, Uh, ignore the fact that we have no idea what, what's going to happen next. Uh, it's complicated. Now, if you do it right, I know some stories. If you do it right and you're at sufficiently big scale, uh, self-running inf- self-managing is, the inference is a good thing, it saves a lot of money. I heard a statistic that s- like from CNCF, that 7% of organization are deploying their own models. Whether they're like serving inference at scale, I don't know. Yeah, but th- there is, there is a massive difference between, uh, the number of companies serving models- Mm-hmm … and the number of companies serving models people actually want to use. Yeah. That, that's, that, that's a massive difference, right? Mm-hmm. Uh, yeah, you can, you can-- Why, why wouldn't you serve Qwen, but that's not the same as- I feel- … uh, me replacing Opus. I feel like if I was in charge of a company, I would be ready to make the switch to serving my own infr- inference. Like there's a lot to know, there's experiments to run. Yeah. I would be doing at least that much because once Anthropic IPOs, who knows how the cost is gonna go, or if, if open models get good enough that I can switch out and save money, I'm gonna be ready. Yeah. But I'm not gonna actually do it until it, it's worthwhile, and it's not yet. Or you do it, it j- just-- The, the problem is like, think of it like early days of Kubernetes, right? Mm-hmm. Was Kubernetes beneficial and good from day one? I would argue yes, right? It got significantly better over time, but it was useful from day one. Mm-hmm. Now, was Kubernetes managed with kubeadm, uh, where you bang your head against the wall- … for everybody from day one? I forgot about those days. Well, well, not necessarily, right? If you manage to do it, it is likely going to be a successful story, but many failed in the process.
Darin 00:23:45.211 What else? Was it just inference? else on your list?
Viktor 00:23:49.578 I mean, it's all inference or agents. Yeah. You pick. You pick. Uh, Vetni has a list. She can go through it. And I think we, uh, we went through the big things- Yeah … I wanted to talk about. Yeah. There were like, um, showboaty demos, like, um, some little robot that kept falling over, and there was, um, Lynn Sun did her drone demo that she did. She did it last ti- time, six months ago, and it failed, and she did it again this time and it worked, and that was cool. Yeah. But- See how much I improved? I mean, I would argue that first, the NVIDIA, like, little duck demo was not a success, so they have a chance to do that again in Barcelona next time and get it right.
Darin 00:24:30.043 So let me ask this question, because you mentioned Barcelona is actually next March 15 through 18, something like that,
Viktor 00:24:38.305 Mm-hmm.
Darin 00:24:38.727 in 2027. does anybody really need to go?
Viktor 00:24:42.558 first of all Barcelona is awesome.
Darin 00:24:45.890 Okay, I understand that. So
Viktor 00:24:47.525 If your company will you pay your vacation to Barcelona, who are you to say no? I think there may be, there may be a time and a place where survey inf- inference becomes a good thing to do. Yeah. In which case, yes. And then I'm a little confused, actually, a question for y'all. We have the Agentic AI Foundation, which is Linux Foundation, and then we have s- a, you know, CNCF Kubernetes stuff. Like, w- what's the line between what belongs in the AAIF and what belongs… Like, why is- I don't know … Agent Substrate over here and not over there? why is Agent Substrate not in Apache Foundation? Yeah. I th- you, as a project, you have the right to donate to any foundation you want. Oh, okay. Right? Kind of, there are many more than those two. I think because it's all- It's just a question of- … under the Linux umbrella. Yeah … what do you think is the right place for your project, right? Mm-hmm. And for many of, for significant number of projects, that for CNCF. And now the question is, will that be CNCF or the AI Foundation project, right? I suspect it's the latter. Yeah. If you're going to an open source conference, is KubeCon the one? And if not KubeCon, then what conference? I suspect that people will be moving towards AI Foundation Conference or whatever it's called, and similar. 'Cause we're privileged people, kind of like, we go to multiple conferences. Mm-hmm. But a normal person feels fortunate if their company approves one conference a year. Right? Mm-hmm. that's a normal thing. And then, then it's not kind of, "Okay, uh," if the question would be, "Would it be cool for a person to go to both KubeCon and AI something something?" Then the answer is yes. Mm-hmm. But if you can choose only one, I suspect that, um, AI will be growing and KubeCon will be, uh, going down. Agentic AI has, Foundation has, like, MCP protocol, and, like, AgentsMD and Goose. I'm, I'm thinking, like, if as infrastructure people, should we still be going to the KubeCons and not the AAIFs? 'Cause that seems a little more developer-focused, like AI developer application-focused. I mean, I would rather go to the, to the AI conference. Yes. That doesn't exactly answer the question, but okay. I mean, so think of it this way, right? We've been yelling at, at heavens for years, kind of like, "Okay, this is, this should be real. This should be stable. This should be used in production," right? Kind of- Mm-hmm … w- we need to get to the solid point where actually this is a real thing, right? That's, that's the, that's the goal that everybody has, and we reached that goal, right? And, uh- We did … yeah. Good job, everyone. And that means kind of like you move on somewhere else, right? Because stable and inno- in- innovative is, is rarely the same thing.
Darin 00:27:46.058 So you mentioned Barcelona is EU next year in March. Uh, North America next year is currently and should be New Orleans
Viktor 00:27:55.759 Yep. Yeah
Darin 00:27:56.958 the, uh, second week of November. In fact, this week, n- next November.
Viktor 00:28:00.561 Are you gonna go to that, Darin? Are we g- the three of us can do this all in person for that one?
Darin 00:28:05.892 here's the thing. This is too much information, but, you know, it's in the past, so I can say it. I had a doctor appointment today,
Viktor 00:28:12.875 Угу
Darin 00:28:13.254 to schedule my one-year doctor appointment, to go to my next doctor appointment, and guess when I scheduled it for
Viktor 00:28:19.352 Mm-mm. For New Orleans?
Darin 00:28:21.292 For New Orleans. Well, actually, I don't remember if it's my blood work for New Orleans or if it's the actual doctor visit after the blood work, so I have to look. So
Viktor 00:28:29.793 time for you to change it, so you're pretty locked
Darin 00:28:32.464 I… Well, I could take a look at it. I, you know, I got home and realized… 'Cause I was like, "Oh yeah, it's a year out. Just book it whatever day you want," and didn't bother to look at that far ahead in my calendar, and of course it would be that week. I'm going to try. I am going to try to be there in person, ' cause New Orleans, it's an easy drive. It's only, like, nine hours from the house. And of course Viktor rolls his eyes that, "Nine hours? What are
Viktor 00:28:59.550 I'm not even going to comment anymore on, on, on those statements kind of. I don't wanna be offensive
Darin 00:29:05.888 Nine hours just gets me to the south side of Dallas.
Viktor 00:29:09.082 Okay, let's put it this way, kind of like it, uh, I, I, I get, I will get there faster than you
Darin 00:29:15.526 Yes, you will. I understand, and that's fine. Doesn't hurt my feelings any at all At least there's no hurricanes that could pote- well, there are potential hurricanes that could hit New Orleans in November, but that's very doubtful.
Viktor 00:29:29.605 Hey,
Darin 00:29:30.272 late
Viktor 00:29:31.533 we, we run away from it faster with a plane than a car
Darin 00:29:35.418 Yeah, especially down there, the interstates are not that great. anything else we need to say about this year's conference? How… So if you think about what we had in, where were we in the spring? Amsterdam. We're in Salt Lake now. It was all AI. Did the AI get more obsessive in Salt Lake than Amsterdam? Or does it feel about the same?
Viktor 00:30:01.925 It's like nothing doesn't have AI. You almost don't even need to say AI anymore. Yeah.
Darin 00:30:07.453 Oh, that's interesting
Viktor 00:30:09.051 It's just like the AI version of whatever we had before Yeah. The-- And that's, that's the real problem the way I see it, right? Within this, this environment is that I, I don't have a feeling that companies here are kind of ready to rethink how we should do something with AI. Mm-hmm. But, if we change the color on our, on, on this thing to orange that is very similar to Entropics, I think, I think that's AI. change all our job titles to forward deploy engineer, then we did it. There we go. There we go. Yeah. But I mean, that's- Our DevOps engineers are now FDAs … there is definitely story. AI is, AI is huge, right? Kind of we need to figure out security, we need to figure out sandbox, and we need-- There, there, there are a lot of things to figure out. It's just that you get a strange feeling that, the right solution is the same as the solution we had five years ago. I mean, that's- Mm-hmm that's what you hear, right? And I feel that that's a wrong approach. Kind of like, why, why are we even talking about it if you already have a solution before it even, uh, was born? AI amplifies. It amplifies any problem that you already had in your organization. So in some ways you do. If you didn't have security solved well, and now you have AI and you have 10 times the traffic, now you have 10 times the security problems. And so- That's all true. Uh, but where I'm going is, yeah, let's say that that's true, and let's say that that's for 90% of the cases and cool. Mm-hmm. But again, we already have it. Yes. So kind of like- Do, do it … can we focus on the things that we don't have? Yes. Right? Kind of, uh, why, why do we need another solution if this is exactly the same as the previous solution, just slightly different, right? this is opportunity to rethink things. I mean, Kubernetes did that. Yeah. Kubernetes did not copy everything that we have before. Just okay, past exper- exper- experiences and now let's figure out new things. Our service mesh, oh my God, controllers. Oh my God, right? All the new things or semi-new things, not all, but many of them were born out of this foundation. And I have a feeling that now we are in that second phase. Okay, so whatever we have must be good for the next wave. Mm-hmm. It wasn't good for this wave, just to be clear. We c- we, we are not going to reuse existing solutions for Kubernetes, but now that actually next wave came, that I think that existing solutions are the right solutions. I'm philosophical here
Darin 00:32:39.280 anything else we need to talk about? Does this pretty much wrap it up?
Viktor 00:32:43.796 I think that wraps it up. You know, uh, Viktor and I are gonna go two-stepping tonight. Wait, line dancing. I don't know. Line dancing. No, no, no, no, no, no, no, no. We're gonna go to Thirsty Thursdays. That was, that, that was inaccurate description. We are going to a place where somebody might be line dancing- … and somebody might be having beer. Not mutually exclusive things. Yeah. But not necessarily, uh, mandatory to be done in pairs
Darin 00:33:13.830 all the pictures in my head just went, "What is happening? I
Viktor 00:33:17.666 Picture Viktor with a cowboy hat, a big gold belt buckle, some
Darin 00:33:22.164 Okay, cowboy hat, yes. Again, Texan, I c- I get used to cowboy hats, but the belt buckle would trip me out. And to see him actually… So Viktor, in one of your videos soon, one of your Monday videos, you're going to have to have yourself two-stepping in the video
Viktor 00:33:35.512 It's, uh, line dancing Wait, wait, wait, wait. Can, can it be in a podcast?
Darin 00:33:41.743 Could we do it in a podcast? We could do it on a live stream, I guess.
Viktor 00:33:44.302 but does that mean that you are now all of a sudden okay with me being drunk in a podcast swearing and all the things that are required?
Darin 00:33:51.703 I didn't say anything about drinking or swearing. I said
Viktor 00:33:54.616 But how, yeah, but how would you expect that to happen? Don't you know me, Darin?
Darin 00:34:00.639 you're a grown boy, you can figure it out Oh, dear. Okay, so the next one, Barcelona, week of March 15th-ish. I don't have the real calendar in front of me. The review epis- yeah, it… I tell you what, every time we turn around, it feels like we're doing this, and it's about every six months, right? And this, I guess it'll be five months. the review episode, if everything goes according to plan, will come out on March 24th of 2027. Yes, I took a d- big deep breath, because that just doesn't seem possible. I'm sorry, but I just… Of course,
Viktor 00:34:37.347 Darin
Darin 00:34:38.749 uh, you wanna talk about living in the future in this a little bit of too much inside baseball, I'm already scheduling episodes for 2028 Yeah
Viktor 00:34:49.640 And then he complains about his doctor scheduling a year in advance. Yeah. There we go. Uh-huh. Ha ha. How do I deal with that? "But I'm in control," he says. It is- Yeah it's an important distinction.
Darin 00:35:05.535 purely an illusion that I think I'm con- in control. I know that for sure. So anyway, Whitney, thanks for being with us again.
Viktor 00:35:11.958 Thank you for having me
Darin 00:35:14.076 Viktor, you're flying back, I would assume tomorrow, Saturday-ish.
Viktor 00:35:20.180 Yeah
Darin 00:35:20.699 and then, then you just sort of sit at home and wait until March is what I'm hearing
Viktor 00:35:25.414 Uh, with my guys
Darin 00:35:27.832 Yeah
Viktor 00:35:28.617 and girls. Okay, I don't wanna be
Darin 00:35:30.724 Are you, are you naming them now?
Viktor 00:35:32.409 Huh? No, no, No, no, no? no. No names. The-- I had, I had names when there were a few of them, but now there are, uh,
Darin 00:35:42.095 Legion?
Viktor 00:35:42.649 to name
Darin 00:35:44.761 There are too many names. Okay, so anyway, that's enough of this. Uh, thanks for listening to this review for KubeCon in a 2026 We hope this episode was helpful to you. If you want to discuss it or ask a question, please reach out to us. Our contact information and a link to the Slack workspace are at devopsparadox.com/contact. If you subscribe through Apple Podcast, be sure to leave us a review there. That helps other people discover this podcast. Go sign up right now at devopsparadox.com to receive an email whenever we drop the latest episode. Thank you for listening to DevOps Paradox