DevOps Platform Debate, Copilot Alternative and Did your AI lose its mind – DevOps Chats Podcast EP7

Hi everyone, this is Alan Shimel and Mitch Ashley, and you're listening to DevOps Chats. All right. Hey, Mitchell, first of all, apologies to all of our readers or listeners or watchers or whatever they are out there.

You know, we've been on the road, we were out in Paris at CubeCon, and, and then I was away a week and you were away the week before CubeCon, um, you know, with our family. So you, we, we, we had a little gap there, but week or two, but we're back. Uh, thanks for hanging and waiting for us to get back.

We appreciate it. Yep. Well, it's not like there are another 16 other strong podcasts to listen to.

There's always one, but, uh, yeah, there's a bunch and we've added some more. So if you do get a chance, if you're listening to this, whether it's on YouTube or, or Apple Podcasts or Spotify, or wherever you listen to your podcast, you know, check out all the Textron podcasts. There's some really good ones out there.

Um, so Mitch, I, I, maybe we should start. Cube Con was amazing. It was, uh, biggest Cube con ever, I believe.

Uh, just under 13,000, something like that. Mm-Hmm mm-Hmm. Uh, it seems the European Cube coupons are bigger than the US ones at this point.

Stronger. They're in bigger cities too. I think that may just be the Covid post Covid scheduling, but Yeah.

Could be. I mean, very well attended. It'll be interesting.

'cause the next us one, of course, is in, I think, October, November, and that is, uh, salt Lake City. Yeah. So, we'll, we'll see how that shows.

But Paris was a smashing success, and it was, it was a good reason to spend a couple extra days before or after. Right. It was, I've tried to get there a couple times and finally made it and got some time to go do the sites and all of that.

You know, one of the things I was, many things that, you know, I know we all kind of took away from that. One of them was, I remember hearing from a lot of people in the US from their, from vendors saying, yeah, we're not gonna go to Coup Con in Paris. We're not planning on going this year.

But I think I saw everybody there. I, I can't listen names. No, I mean, there saw, I mean, for instance, there were a lot of folks, Dell Dell was not there.

We've, Dell's been a sponsor of ours in the past. They'll be back at Salt Lake City. There was, but I, I think, you know, CNCF has something like 180 projects now.

Yeah. So the amount of projects, people using them and companies affiliated is so great that maybe you don't miss Adele or a particular company. Um, there was no lack of big companies.

There was no lack of small companies for sure. Absolutely. Um, my observation was, to me, this was the first cube con where I think DevOps felt comfortable in its skin.

Mm-Hmm. We, we didn't see the, you know, I think people recognize now how DevOps and Cloud native kind of go together. I, I think we even recognize how platform engineering plays into the whole DevOps cloud native thing.

Agreed. In spite of what some people may still think. Um, and so it was, it was, it was a tour de force for modern software factories.

Right. It was how, how we do software today. It was e excellent stuff.

Um, and a lot of energy people that are on the projects were energized. They were Yes. Hopping.

It wasn't just the buzz of the crowd people that are contributing. No, I, I agree. Um, and of course, look, all of our interviews from, uh, cube Connor available on Tech Drunk tv, and there's a whole bunch of them over the course of the two days, three days.

So go check it out. I at Text Drunk tv, and you'll, you'll be right up to speed with them. Mitch, I wanted to jump into what we got going on in DevOps though for this week.

Right. And, um, well, why don't you kick it off? What did, what did you find that kind of tickled you?

Well, one of the things is I've, I've talked to, uh, tab nine a few times, and they're kind of transitioned from their own, let's say it's source they use for generating code to now moving to LLMs. Uh, and, you know, they, they, they came out with their con generally available product. Uh, and I kind of knew that was headed that way for writing code.

And there's also been some other announcements, other companies about, you know, not just creating a copilot, but really a kind of a software engineer as AI to do complete development tasks. Things are moving pretty quickly. It's not just co-pilot at Microsoft and, and Google that are, you know, doing these kind of things.

The people that are in market are innovating too, and, and really jumping on AI code development code generation. Yeah, I, I agree with you. Um, what I think we're seeing, so first of all, the, the, I was talking to a, another friend of mine who was out at Google next this week, and Google announced some more AI stuff today, or, you know, Mm-Hmm.

I got the name of the new project, but it combines pictures and video and Transcripting, the Gemini. Yeah. No, it's not Gemini.

It's something else. And, okay. I just don't remember off the top of my head that this second, but that being said, the, the pace of evolution here in AI is just faster than anything I've ever seen.

Mm-Hmm. Right. Um, it is just morphing and continuing to evolve, which, which they all do.

Right. But the pace of it today is just, I mean, ridiculous. Ridiculous.

And, um, I, I think one of the kind of evolutionary steps we're seeing now is the modularization of LLMs. Mm-Hmm. I just made up that term.

Okay. I'm not even quite sure it's real. Well, quick picture research will create a category and sub subscriptions to it, right.

Mod the modularization of, of LLMs where, you know, you, you're gonna have your, your front facing like Salesforce Einstein or something like that. Right. And then behind it gets plugged in different LLMs, s SLMs, what have you, as well as whole different a AI engines.

Mm-Hmm. Right. And so you as the end user, get to pick what's the right LLM for me for this task.

And, and maybe it's good for this task, but not to task after it and what's the right, you know, one for me here. Yep. What's the right, what's the right LLM to have respond to this part of the prompt, right?

Mm-Hmm. Kind of subdividing what parts of the prompt would be best be satisfied and, and what's the right AI as like we see in clouds, right? Azure does certain things maybe better than A-W-S-A-W-S does certain things better than Google Cloud gcp.

Um, it could be, you know, you pick the right platform for the tool, you pick the right ai mm-Hmm, excuse me. You pick the right platform for the task. You pick the right AI for the task.

It's all about picking the right tool. Think about it, you know, applying it to software engineering is, is every augmented code or even just self coding, ai, turn of AI bot engine, whatever it is, the right cha right thing for every tool. I mean, I could see LLMs or s SLMs that are, um, very good at refactoring legacy applications right?

For the cloud or are un really good at unraveling complex code that might be in a monolith application, or really good at, let me deploy the whole environment for you in a Kubernetes, you know, AWS instance or whatever you, you could easily see, you know, today we have co-pilots for everything, but I think it's more likely co-pilots or pilots for specific areas and types of tasks that you're really good at as well as having generalized LLMs out there. Agreed. Agreed.

So, I, I think that's where we're heading. Um, you know, a plug and play architecture for AI and LLMs and I Mm-Hmm. I think that's, you know, I think tab nine's on, on the right path with that.

Yeah. I think they're doing some good things there. It's nice to see somebody else besides, you know, with Copilot and Microsoft and GitHub and others who are doing good stuff too.

But, you know, absolutely. On that note, I will tell you I downloaded from my iPad, the Microsoft copilot. It's free, I assume it's based on open ai, if it's Microsoft.

Mm-Hmm mm-Hmm. And I've been playing with that a little bit. Wow.

Wow. It, it's, it's a real help. Uh, you know, it's not gonna replace me, but it helps when I have tasks.

I, I, I've actually started going to that first now Mm-Hmm. And it, it, it, it has a, it's pretty damn good. And it's free.

And it's free. So I don't know for how long it'll be free, but, um, it's pretty darn good. Pretty darn good for 9 95 a month onto your office 365 and you too can have whatever.

Yeah, yeah. What would they, they, what'd they spend, they spun something off, off from Office 365 now, didn't they? Um, I'm not sure.

I, I don't know if I know about that. No, it's not like Word or PowerPoint obviously, or one of those, but it's another, oh, no, not copilot. Uh, whatever it is there.

I, I think Microsoft is figuring out as many different ways to milk that cow as they can. Mm-Hmm. Well, you can bundle it all together, which is great, but then you want upsell.

That's kind of hard. So pull a bundle apart in a few places. Yep.

Absolutely. com. AI might be lying to you, which it could be applicable for I any range of number of conversations.

He was describing Mal, but not on purpose, but not on purpose. It wasn't malicious. Well, not with Malid intent.

Let's put it with ma with an absence of malice as the movie said, we'll have to have ratings on each on the LLMs. Now, absence of malice. Um, but he was talking about his own kind of hobbyist playing with, I dunno, whatever copilot or some code generations, I assume, assume it's copilot.

And, uh, how, and I've experienced this too, when you're asking, uh, jet GBT to do a certain set of tasks and you say, no, I do this, add this to it, or do it this way out, like I've said, create an outline like this and put five sentences under each bullet point or something like that. And it doesn't always do that. It kind of does its own thing.

It says, well, I'm gonna write a paragraph for each of those. Like, I just wanted a bullet point. And that was the same thing was happening with code.

He had described a scenario where I want the math done this way. Right. A certain, for whatever reason, he wanted his math, a certain approach or style or whatever, and it was refusing to do it, just wouldn't, wouldn't follow instructions.

And you kind of wonder, okay, so is that it knows better, it just doesn't know another way, it doesn't know what you mean. You know, why, why does it, why does, you know, generative AI kind of hit those roadblocks where it just stops listening. It doesn't do what you ask it to do.

I, I don't think it's, it's the terrible twos and it's throwing a tantrum if that's what you're getting at. But I didn't hear any ticking his screen. Right.

But, um, you know, the thing about computers, Mitch, I learned a long, long time ago, don't blame the computer. Right? It, no, it doesn't make mistakes.

Adding up numbers. If there's a mistake, it is because you put the wrong number in or something, or you gave it the wrong instruction was wrong or whatever. Right?

Yeah. I I think it's the same thing here. If it's not doing what you tell it to do, you know, it's not the laws of robotics coming into play here.

No. Right. Where you shall do no harm or don't have to harm the least, or, or, you know, second in law and zero law rot provided doesn't harm a human.

All that Right. Asimov laws of robotics. But, so I, I would attribute it more to that.

I, I don't think it's vengeful. I don't think it lies on purpose. I don't, you know, look, look at the person behind the computer, and I think you'll find your answer there.

So very, I, I agree with you. I was also thinking me, not that it's emulating a human brain, but we as human kind of, we get stuck down one path when we're thinking about something or a way of solving a problem, and we kind of stay there. We hang out there on that approach.

Like, I'm gonna find the answer when in fact, the answer might actually be a diff sort of backup a second. And let's go down there. And I wonder if some of our, as we enter prompts into, into degenerative AI systems that conditions it, that gives a context, right?

And that conditions it what it's gonna answer. So it could be, you said something three or four prompts ago that kind led it down that path, and that's why you're stuck there. So we need some, maybe we need some skills to how to unstuck.

Like, well, here's this, here's an example of code that I'm talking about that does what I'm talking about. Could you write it this way and see if those kind of things would unh it move it down the path you want to go down? Maybe?

Um, I don't know. I mean, so I don't pretend to be an AI expert or engineer. You know, John Willis is off doing all his research and, and finding out all those kind of tough to know questions.

So I, I don't know what's going on under the covers there. Mm-Hmm. Um, I mean, I'm sure there's a perfectly logical explanation, uh, but I guess we'll have to wait and see, or maybe next time we ask, we see John, let's ask him, let's ask him.

Hey, I have a bit of trivia though. You mentioned ASBOs laws. So I started working in AI in the late eighties doing lisp and prologue programming.

And I used to teach a prologue class to people, uh, college students, uh, with advanced degrees coming in that were going into write tools for software development. And, uh, I had a lot of fun. The, one of the bigger programming assignments was to implement as MA's laws of robots.

And, uh, here was the scenario, your robot is stuck on this planet, and it's moving in a circle, if you remember from the book, that whole thing. And here's what you need to do. And you have to, you have to build the logic and prologue, uh, to get the robot unstuck.

And there were certain variables of things that you were gonna count, almost, kinda a little bit, like a little bit of a d and d adventure on top OFMs. Uh, but they wrote, uh, and some of them were really creative. I really enjoyed, they're like, that's the best programming assignment we've ever had.

Or I also have the who's ov Right. You have that reaction to it. Well, you do get some of those.

I mean, Hey, just a quick plug, right. So in RSA in, uh, DevSecOps and AI event this year, which is Monday May 6th, and we have free Expo pass that'll get you into our seminar on Monday, May 6th. If you're interested, we have David Bryn as the keynote speaker.

Fantastic. David is a, a multiple Hugo Nebula Award-winning sci-fi author. He's probably one of the greatest living sci-fi guys at this point.

So, so many of the giants have passed away. Um, and, you know, in addition to the Uplift series and the Postman and kiln people, and, uh, you know, so many great books, he was also one of three authors s well, that's part of uplift. Mm.

There's a whole uplift. I think there's six or nine books in the uplift series. Yes.

Uplift War. Yeah. Starting Rising.

Um, star Tide Rising Rainbow Reef, and a bunch of others. And it's funny, I read all of them years and years in the nineties and when they first came out, and now here we are with David Keynoting our DevSecOps event, but, um, he was chosen by the Asimov family as one of three authors. It was Brynn Benford, and not Bradberry.

I forgot the third guy. Mm-Hmm. Greg Benford has passed, passed away as well, um, to continue the, the foundation series.

Could you imagine? So he wrote the, the sequels, he one of the sequels. Yeah.

Pretty cool stuff. That's, that's amazing. That's great stuff.

Yeah. Actually, he'll be doing a book giveaway. I don't even remember which book it is that we're giving away on Monday, May 6th.

I had to think. But the first 100 people, I think, for getting a book. Yep.

I think we have a hundred bucks to so cool stuff. Excellent. Anyway, what else do we have on DevOps, Mitch?

Hey, um, not to stir up old arguments, but I know you had a, uh, you recorded a recent, um, what was it? Uh, pod not podcast show CD pipeline show a video show with the CDF and the CD f continuous delivery goes, you know, did DevOps get it wrong? And Platform engineering gets it, right?

Is that how that, so that's the thing. That wasn't the, the topic was platform engineering and cd, and we had a, a diverse set of opinions on the panel. And, you know, I'm the moderator.

I'm supposed to let people talk. Yeah, yeah. But I, you know, I I I got a little crazy, um, did a few buttons get pressed?

Well, you know, the first thing was the, the platform engineering guy started off by saying, well, the reason platform engineering had to replace DevOps was because the DevOps tools just didn't work. And engineers weren't happy. And, and Eng and developers weren't happy because DevOps tools aren't good.

So DevOps is no good. Uhhuh. And I tried to be nice.

I said, look, this isn't me speaking, this is, but if you know anything about DevOps and you've ever listened to Patrick Debar or John Willis, or Damon Edwards, or Jean Kim or any of the other people, Andrew Schafer, Andrew Clay, Schafer, any of the people about DevOps, they will tell you that if you think DevOps is about tools you don't know DevOps, you missed it, you missed it. Right? The tools are interchangeable.

Puppet today, chef tomorrow, this, you know, the next day. But it really is about culture and about breaking down silos. And he kind of poo-pooed the culture thing.

Right? Interesting. 'cause it's all about tools.

Tools and workflow and, and how unhappy people were. I said, well, are they happier now than they were 10 years ago, 20 years ago? He says, well, who cares about 10 years or 20 years ago, I'm talking about today's developers who only want to use Cloud native.

Okay. So that was, that was, and the profess was Strike two professor said that was Strike two. That was strike two.

You know, I, I I gave him the benefit. Mitchell, you took a deep breath. Took a deep breath.

Well, no, I was kind of getting worked up and then, okay. You were by now. All right.

Yeah. So, you know, that was that we don't care about the history of it. And I, I guess the third thing was DevOps engineers were doing it wrong.

And that's why you need platform engineers. Hmm. Okay.

So at that point, I kind of went, you know, DevOps is about culture number two. Um, the, the DevOps engineer is a made up name because people were looking to hire people with DevOps. But DevOps is about cross-functional teams.

Right. And if you think DevOps is all about DevOps engineers, again, you don't know DevOps flow. And number three, if you don't understand the history of something, you're bound to repeat its mistakes.

Mm-Hmm. And so when you take all of that, you know, I I, I started off with a very enlightened thing. I didn't say that platform engineering is just a name, the s**t that's been being used for 20 years already.

Mm-Hmm. Right. We've always had ops folks and engineers setting up the workflows, right.

The processes and the guardrails. Right now we call it platform engineering. Yeah.

But, and, and I, and platform engineering's a real thing. I think it's, it's a great thing. I think it's, again, part of this continuum of how we develop and deploy software today.

Um, you know, and, and yes, CD tools like spinnaker, which had two people on the panel from the Spinnaker project Deploy C, DF. Yeah. Tools are, are just tools.

You use the right tool for the job's, there'll be more tools, right. You use the right tool for the job. That's what a big part of life for me was understanding, right.

Use the right tool for the job. You work a lot easier and a lot faster. And ultimately, here's the thing.

That's the goal of platform engineering. It's the goal of DevOps. It's the goal of Agile.

It's the goal of ITSM. It's the goal of all of these kind of software or IT frameworks, which is go faster, be better. Mm-Hmm.

And, and so I'll leave it at that, but if you're interested on this conversation, it was pretty fascinating. It was the CD pipeline, tech Strong TV show. It's also its own podcast.

Probably be out, I think it comes out by the time this will come out, it'll be out as well. Yeah. So you can check that out.

Second week of April. Third week of April, somewhere in that range, even before then. Yeah.

Something like that. One thing I would say is the reason why I like to suggest people read the Phoenix project isn't just 'cause it's a good read. It is.

But name one tool they talk about in Phoenix Project, there aren't no tools. They might talk about CI/CD or they don't use the term necessarily, but it's all about the how, not the what tool. And it, it, in in part not fully, but in part pulls together a lot of things that contributed to what DevOps became.

Uh, it's cross functional teams, smaller portions of work so you can, you know, get things done faster and find mistakes sooner, and do quality improvement, iteration, all of those things. There's so much more to it. But it's, it's how you do software development.

Not, or even operations, you know, and operations. It's not about this is the right tool. If you aren't using Jenkins, you're not doing DevOps.

You know, it's not anything like that, though. There are a lot of Jenkins users out there. So if you need help, I brought that up as well.

Oh, you did. You know what, okay. What?

Yeah. And, and the thing, you know, by the CD's own survey, 40% of people are still using Jenkins. 40% plus.

Yeah. I'm still using Jenkins for CI/CD. Mm-Hmm.

It's all about tools, but, you know, I, I have a term for these things now they're DevOps deniers. Mm-Hmm. Ooh.

Right. DevOps de like climate change deniers and flat earthers and all that good stuff. Um, it is what it is.

I'm gonna leave it at that. I think that's a good place to end this for this episode of DevOps Chats. Mitch.

Yeah, I think you've, uh, you've exercised the, uh, new term generator pretty heavily on this episode. Yeah, well, it's my own custom. LLM.

Um, all right. So don't be a DevOps denier. com, listen in on the next DevOps chats.

But that's it for now. This is Alan Shimel and Mitchell Ashton. And you've just listened to DevOps Chats.