Viktor
00:00:00.000
What I'm saying is if you can combine both, then you have something very, very, very valuable. If you can combine experience with curiosity, then we're talking. Then nothing can stop you. This is DevOps Paradox, episode number 302, using AI to help with your programming tasks.
Darin
00:01:21.651
On a recent Friday episode of DevOps Paradox Live on YouTube, usually Friday afternoons, Central Eastern time, wherever you want to call it. We took a look at the Windsurf AI code editor. It was interesting. Not totally surprising. I guess we could have said that all along for GitHub copilot, which I've played with like once way back when I haven't tried it since.
Darin
00:01:53.938
Well, I know I am missing a lot, so I've got to start baking off and cursor is another one and there's numbers of other different options. Here we are 2025. If we're not using AI right now, as we're coding, are we just being dumb?
Viktor
00:02:09.652
depends how you define it, right? I think that AI is very useful, but I don't fall for false expectations, I don't think that it's making anybody twice as productive, I do not think it is replacing. Experienced engineers. I think it's yet another tool that we can use to do our jobs better.
Darin
00:02:38.391
I've seen some posts on Reddit that people started using it and then everybody starts smashing them. It's like, Oh, you should be able to do that by yourself. You shouldn't have needed help. Look, I've been writing SQL for a long time, a long time, 92, 93. I finally got grouped by a few years ago, completely burned in my head, but It's not something I do every day. And even if I do something every day, if something can get me there five minutes faster, 10 minutes faster, a day faster, shouldn't I be using that help?
Viktor
00:03:15.307
you definitely should. That's the point of it. And I think that that help is now AI,
Viktor
00:03:25.508
But it was Like, for example, IntelliSense alone, nobody talks about it, it's a great thing. it made us much more productive than before. That thing alone. You know, when you type dot, I can give you the list of functions in that class or whatever the language is, right? That alone is a big help, It doesn't look like that because we are all used to it, but it is. If you don't think it is, then yeah, open nano and do the same thing.
Viktor
00:03:58.635
Or vi. I know vi, I intentionally say nano instead of vi, because I've seen people doing some funky stuff with vi that looks like magic. So okay, I'm putting it in a completely different category, right? But yes, right, or normal text editor. So yeah, what we're, so back to your question, what were we doing before? We were using IntelliSense, we were using Stack Overflow, we were using Google, We were using wikis and so on and so forth. What is AI helping us? Well, it's doing all that a bit faster, I still don't see the big difference between that and Stack Overflow, if we exclude the time it is required to get to the solution. Is Stack Overflow always a correct one? No. Can you rely on it 100%? No. Will you introduce mistakes in your code just by copying and pasting from it? Yes. Now, if I change the word Stack Overflow with AI, wouldn't that sound very similar to what people are saying right now for AI?
Viktor
00:05:07.757
Exactly, because it is very similar. I AI is trained, when we use, I mean, it's not trained only on Stack Overflow, right? But it is trained on code, basically, it is essentially doing the same thing to what Google is doing for us. You know, when you Google something, I don't know, go this and that, Google finds the solution in Stack Overflow, and most likely, it will show Stack Overflow as the first result. it is finding the closest match among the existing solutions it has indexed. Now, again, if I replace that with AI, am I not getting exactly the same story?
Viktor
00:05:50.982
So if you say, and I'm ignoring now the quality of AI for a second, right? If you say I am against using AI to help me with coding, you're effectively saying you're against Googling Google. And Stack overflow as well.
Darin
00:06:09.105
It's interesting that you stack it up that way. no pun intended, because if you believe that AI assistance isn't useful, then I double dog dare you. I I'm going straight for the double dog dare. A double dog dare you to when you're trying to code something, you can't find it, don't go to Google or go to Stack Overflow,
Viktor
00:06:30.131
exactly. It's the, the end result is very, very similar. The main difference is in speed of obtaining those results.
Darin
00:06:40.457
but you got to be careful. Right? Because what if your company hasn't signed off on using AI
Viktor
00:06:52.583
Okay. Then there you go. Some block the, uh, AI as well. Now there is a fundamental difference in that respect, right? There is no guarantee that when you ask AI something based on your existing code. Your code will not go somewhere, Because there is a round trip. I'm asking you something, it goes somewhere to some model, and it gets back data, right? There is no guarantee that in that request going there, your code did not go with it. I'm not saying that that's what's happening, but there is no guarantee, right? While in Stack Overflow, you almost certainly did not reveal any company secrets in Google Search box, right?
Darin
00:07:39.649
Well, maybe, maybe not. It would take a little bit longer because you don't have enough context. And that's the key part to the AI stuff is if you've got full context, you could reveal a lot more things than you probably meant to.
Viktor
00:07:53.089
Yeah. But, you know, if that's a concern You have quite a few options, right? You can say, okay, I don't trust JGPT, but I run my stuff already on Azure, so I trust Microsoft. I can use Azure OpenAI. I don't trust Azure even though I'm using Azure, maybe. Cool. You can host your own model, right? So, that privacy thing. Privacy is not a discussion whether you should use AI or no. Privacy is a discussion whether we should use this AI or that AI. Should it be a managed solution or a self managed solution, right? that's a discussion related to privacy issues. AI itself, just like anything else, can be public, can be private, whatever else in between.
Darin
00:08:51.122
Speaking of in between, we were talking about not being able to use something. Let's go the other direction. Salesforce recently, Mark Benioff, the CEO, and there's a link to this article down in the episode description, I'll make sure it's there. It led with, Salesforce will not be hiring any more software engineers in 2025 amidst significant productivity boosts from AI.
Viktor
00:09:15.299
I have two possible comments on that. First, I don't believe that's the reason, I would like to see the proof that actually he did plan to hire people and then, went to HR and say, cancel that request. think it's a marketing stunt, kind of like. second, even if that's completely true, I think it's just a temporary bump. Because we've seen many advancements in the software industry that all prophesized the same outcome, right? Oh, now that you can drag and drop, we don't need software engineers to write code anymore, Now that use some higher level languages, you don't need that many engineers anymore. Oh, now that we are using AWS instead of running in our own data centers. We don't need that many engineers anymore, right? I could probably continue with examples forever, but you do agree that we heard those things many times before. This is not the first time. And the result of all that, excluding maybe some temporary, bumps, is that we ended up hiring more people. Because what is happening with advancements like that, unless you're the first, is that Yes, you can easily come to the conclusion. If I do this. People will be more productive, hence I need less people. That's a very, very valid assumption that misses only one important thing, and that's that the goalpost is moving away now, So our expectations of what we expect from software are going to be bigger than they were before. Not because we do it, but because if we don't do it, competition will do it. So the amount of work will increase. At the same time, and that means that we will still need, let's say, the same amount of people, not less, Because all those ideas ignore the fact that we will be expected to do more, Expectations will change. you remember, let's say 30 years ago, We both witnessed that part, right? What was required to do in a normal company, right? Something that probably three engineers could do what hundred engineers were doing 30 years ago, right? your whole system is three servers. Oh, your whole application is, is one application and so on and so forth. So if nothing changed from 30 years ago, yeah, whatever hundred people did 30 years ago, three people could do now. Yeah. Yeah. Yeah. But the expectations are not the same, so we still need the same number of people, or more. That was long rambling and I see in your eyes that you disagree with something, I'm just not sure what.
Darin
00:12:14.765
Well, it's not so much that I disagree. I want to go back to this article and I'm going to read something and we'll throw it up against what you just said. Are you ready? So when asked if Salesforce would have, and we're picking on Salesforce because I just remember this happening in December, this could be fill in the blank for many other companies. When asked if Salesforce would have more or fewer employees in five years time, you were talking about that. He, Mark Benioff said, he thinks the company will probably be larger. But he went on to say, we're not adding any more software engineers next year because we have increased the productivity this year with AgentForce, that's their AI thing, and with other AI technology that we're using for engineering teams by more than 30 percent to the point where engineering velocity is incredible. I cannot believe what we're achieving in engineering, finally. And then we will have less support engineers next year because we have an agentic layer. Hold that thought for a minute. We will have more salespeople next year because we really need to explain to people exactly the value that we can achieve with AI. So we will probably add another 1000 to 2000 salespeople in the short term. So we're not hiring engineers, but we're hiring salespeople to sell the things. Now, again, that's Salesforce. That's part of what they do. More revenue. I mean, at some point there have to be humans there, but. This whole agentic layer, again, we had Patrick on a couple of weeks ago and he started going down that path a bit. I think it will get there. I think there are like crew AI and a few of the other agentic platforms you can do things with. I still think it's a little early.
Darin
00:14:03.570
It's okay. It's very early, but in five years, that could be a different thing. I think.
Viktor
00:14:12.160
Yes, it will be a different thing, absolutely. We will be more productive, whether that's 10 percent or 90 percent or 500, 000%, I don't know. I don't think that anybody can predict right now accurately, right? We will be more productive. I have zero doubt about that. We are already more productive than we were five years ago. And not only because of AI. For many different reasons, including AI. I still don't see, judging from the past, that that results in less demand for skilled engineers.
Darin
00:14:52.525
So if I'm coming out of college now and I'm trying to get into a place, I'm not skilled yet, but I could be right. Because now, used to 20 years ago, you would have had to have basically done internships at the big companies in order to get a foot in easily, more or more easily. But now, between open source, between AI, if you can do those things, you understand how to do those things, that might get you in a little bit faster. I don't know.
Viktor
00:15:27.990
Yeah, but also, you know, young people have other qualities that we were always missing, They might be more, uh, better understanding what the heck is going on and how they can leverage it to their advantage than old timers like we are. I mean, I could just as well argue that actually AI poses bigger threat to us, older people, than young people. Because we don't know what the heck is going on. We go on Reddit, those of you who, those of us who actually know that it exists, and say how AI is just a passing thing. It's horrible. Young people are not, do not have that reaction. So if you're worried about anybody, I would say you should be worried about older people rather than younger people.
Darin
00:16:18.523
So what you're saying is if we're older, I'll go and put myself in that bucket. I can either go ahead and get on the AI, let's put it this way. I was going to say bandwagon. Let's not say that. I should go ahead and get on my AI bicycle or in my AI car. Not literally. And start riding along with it. Not the hype train, but the actual things.
Viktor
00:16:43.902
Let me rephrase it in a more generic terms. If you're more experienced, You have value for being more experienced, but you should also be keeping up with the trends, Yesterday that was AWS, today that's AI, And tomorrow it will be something else. More experienced people have sometimes problem adapting to new trends. And younger people cannot, do not have that experience that we do have. But statistically, they're more likely not to be defensive against
Viktor
00:17:28.336
because it's, it's part of human nature. when you are a few years old, right, children, like 3, 6 years of age, they learn new language infinitely faster than you and me would learn a new language now. That's kind of proven. There is no discussion even about that, right? And the reason why they learn new language infinitely faster than adults is because is because they are empty. They're like sponges, They do not need to unlearn things. They don't have a conflict with what they already know. They're not trying to put it in a box that is not the correct one, and so on and so forth, right? There are certain things that benefit from experience, but also And sometimes it is easier to learn something if you're experienced, right? Because things build on top of each other. But sometimes it's easier to say, I know nothing about anything. I'm just going to see what is this thing, what is happening right now, AI. That's what I'm going to focus on. And you might get there faster than we do.
Darin
00:18:39.784
I think the key part in that is using. We have the data scientist and other people in place to help us build the models, but the actual training of models is infinitesimally small versus the actual usage or the inference of the models. I mean, there's AI stuff happening underneath your scales, your toothbrushes, all these other things that are happening. It's like, I don't know that one, that data there, but you know, it's that, that's where it's going to matter. So how can we better understand how to use the tools that are put in front of us and use. Now, if you get the chance to help build great, but I'm going to throw An analogy here, that's sort of like the small DevOps team. Okay. Yeah. We're DevOps podcast. I have to go there. The small DevOps team that's actually managing the app, two, three people that's 70 or 80 people actually wrote. Right. We're going to go back to mainframes on that. And there's one operator. And then you had hundreds of developers, but now it's like, okay, how can we use the new tools? I got a new toolbox for Christmas. I got some new tools in it. Do I know how to use the tools yet? No. Do I need to learn how to use them? Yes. Otherwise I wasted my money. Ooh. Now here, here's the twist. Let's, let's rewind back to Stack Overflow for just a second because this just triggered in my mind. With Stack Overflow and with Google, it didn't cost me anything. But now AI does cost me something in hard cash. Is that a true or false statement? Is it, I mean, is it really true?
Viktor
00:20:20.987
to be honest, first of all, I don't know how much it costs. I'm using OpenAI, which just happens to be bundled in, you know, I get a bill for whole Azure, so I haven't even checked how much cost me over there. I'm just guessing, and you probably have more information than I do, that if it does cost something, it is acceptable. I don't know how much is co pilot, GitHub co pilot.
Viktor
00:20:55.798
Yeah. And if what the sub guy says, no, Salesforce guy says 30% increase in productivity, which I, I don't believe, but let, let's roll with it for a second. 20 bucks for 30% of increased productivity. I'd take it anytime of the day. Heck, I would pay 20 bucks for 5%.
Viktor
00:21:24.279
you are potentially asked to pay. And by the way, there are three options, just to be 100 percent clear, right? You can host it yourself on your Mac. It works okay. Maybe not that good, but good enough, right? So first of all, there are three options. Second, is increasing of your productivity worth less than Netflix? Netflix is approximately that price, right? Yeah.
Viktor
00:21:55.519
Yeah. So why is Copilot worth less than Netflix, in case you're judging the price?
Darin
00:22:04.500
I think it gets into, and this takes us back to the other section of, okay, well, what if I can't use it because of fill in the blank legal reasons or whatever? Another great reason to work in open source. Well, you could self host it. But you'd have to have the hardware to self host it.
Viktor
00:22:19.747
excluding companies where you shouldn't be working. Excluding those. If a company says, hey, we cannot use it because of this and that reason, right? I feel that the company should self host it then. It's not each of us working in that company should self host it. That would be very inefficient.
Darin
00:22:38.205
I think enough companies that do self host the other things, I believe this is going to be on their radar in the next few years. But now they've got to get hardware. They've got to figure out how to operate it, right? It's just
Viktor
00:22:51.130
it can be early. This is just starting, right? This is not life changing today. There is time.
Viktor
00:23:00.318
It might be life changing. So, there is a difference in, for you as an engineer, as an individual, this is not life changing. It's not like, hey, I'm not on it today, I'm out of the job market. That's not the case. What might be the case is that if you as a company, from the business perspective, is not. At least experimenting with it, then you might be in deep trouble. Not because it will change your business today, but because when it does change your business, it might be too late.
Darin
00:23:34.994
There seems like with this AI play, used to, we would have, just think of the standard distribution curve. Right. You'd have the early people. You'd have the extremely late people. Then you have the big bell curve in the middle. I think AI is following that model, but I feel that that bell curve in the middle is going to be really compressed on each side, and the tails are going to get really long on both sides because it's the people that are trying, trying, trying, Oh, we've got to go do it. And then you've got the laggards that will never get there. And that's going to be the hard part.
Viktor
00:24:14.302
Yeah, but still more, not necessarily for you as a user, right? I don't think that we are even close to it now. It's not life changing yet. For you as a company investing in it, it might be too late. Yes. Yes. Right, because it takes much less time. So let me give you an analogy, right? When cloud appeared, specifically AWS, yeah, five years later, you can still say, yeah, five years later, which is a long time ago, right? Or even 10 years later, you know what? Now is the right time for me as an individual to jump in. I could argue you should have been using it earlier, but it's okay. you're not going to die because you haven't. Now for companies, it's a very different story, because if you didn't jump within the first five years, now you're in a situation that you cannot catch up. You cannot become AWS now, But you can start using AWS now. a very different perspective.
Darin
00:25:25.369
Let's rewind to the beginning of AWS, or at least my perception of the beginning of AWS. We had EC2, we had S3. That was,
Darin
00:25:36.455
In 2012, again, using the five year mark here, in 2012, there were a lot of services that were built on top of those two items.
Darin
00:25:46.387
So I could have started early, We could replace this with Kubernetes, with Mesos, with everything else. I could have started early
Darin
00:25:59.862
either a competing service or just using the thing, right? But let's play this out. Let's say, okay, I'm going in 2007, I'm going to use AWS to host my database servers. Great. I've got EC2 instances. I know how to do volumes. All that's easy. But then within that five years, I don't remember if the timeline is correct or not, but within those five years, now AWS has come out with a managed database service. Did I waste my time doing the work five years previous, or was it actually smart?
Viktor
00:26:32.349
answer is yes to both or no to both. So somebody coming 10 years later or whenever, it would take less time to get to the current situation than you investing in all those middle steps, But on the other hand, if you're already there, if you're already used to EC2 and S3s, and probably there were already, there were subnets and VPCs and stuff like that, right? You're already ahead. You might have invested more than somebody new becoming and having to get to that point, but you're ahead of the curve. You adopting RDS? It's easier than somebody coming from on prem to adopt RDS. So you're at the advantage, but you might have spent more, time and effort to get that advantage, but it's still advantage.
Darin
00:27:28.986
So I guess the parallel to that, going back to as a software developer, should I be trying out, These AI editors or AI assisted editors. My vote is yes. Be careful, right? You got to be careful, but go ahead.
Viktor
00:27:47.321
Let me rephrase that question. Should you as a software developer, and you give me the answer, trying new things?
Viktor
00:28:01.267
There we go. That's the answer. And you can now apply it to AI, you can apply it to anything else. It's in your self interest. Forget about your company. Forget about what's there. It is in your personal self interest to be on top. To know more than your competition, to be more prepared than others, and so on and so forth.
Darin
00:28:27.275
So once you get older, you don't have to worry about not understanding what's going on.
Viktor
00:28:35.916
There is always this balance, just to be clear. And I mentioned that at the very beginning, you cannot easily replace experience either. So I'm not saying that, hey, you're experienced in this or that. In DB2, that's worth nothing. And somebody knowing nothing about anything is just starting to learn AI and that person is worth more. No, you're worth, both of you are worth, just from different perspectives. What I'm saying is if you can combine both, then you have something very, very, very valuable. If you can combine experience with curiosity, then we're talking. Then nothing can stop you. Because, most people have one or the other, not both. The moment you get experience, many people stop being curious, and the other way around. While they're, they're curious when they're not experienced.
Viktor
00:29:53.533
Yeah, you're, you're, you're still okay, right? We need experts and we need also people capable of learning new things. So I'm not saying that you're in a bad shape if you're one or the other. I'm just saying that if you want to be in an amazing shape, then you do both.
Darin
00:30:11.597
hosed in both scenarios to an extent, I guess. So what do you think? Head over to the Slack workspace. Look for episode number 302 and leave your comments there 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.