It is time for Operations teams to Code or die – Cloud Field Day 22 Delegate Roundtable

Welcome to the final element of Cloud Field Day 22. We are heading straight into the very last of the delegate round tables. Our last opportunity from to hear from this awesome delegate panel that has been assembled before me, uh, for the last, it seems like a week, but I'm sure it's a lot less than that time.

And I think everybody's looking a little bit tired as if it has been a marathon here for this week. Uh, and and another of my own grand traditions, I'm going to throw a good friend under the bus as our topic for this one. Uh, our topic is actually one that's not new.

This is absolutely something we've talked about in the years past, and I'm sure Eric has previously said, code or die code or die, not code or die. That's, that's the operations team alone. Is there a place for an IT professional, particularly an infrastructure IT professional who is unable to write code to make processes happen again?

It, and I think Eric, if, if I'm not too off face, is this the center of your premise? Yeah. So previous discussion, yeah.

Yeah. If I could say, uh, to, to start my defense of the affirmative on this code coding things to achieve efficiency in my day to day, to get rid of the mundane and to automate what I knew I could trust to accelerate my delivery was game changing. Oh, I hate, I feel like a marketer really badly.

Now, let's, let's just delve, let's delve into the topic here, but it was crucial and pivotal to my, sorry, I had to be an LLM for a second. It being able to script not be a true developer. I'm not a developer.

I'm not built like a developer. I have tried to do a learn Python, learn r learn every coding language. 18 minutes later I rip up the book and then I just copy and paste code until I figure it out.

I'm a very autodidactic learn learner, but that code taught me to think differently about how I did things. And that, that was actually really more what it was, is that the process of documenting and then co-defined something, changed the way I thought about the activity. And it made me an architect and it also saved me a ton of time.

So I think for people now, I wouldn't suggest that you have to go do a dev bootcamp, but if you're not using some level of scripting and automation as part of your day to day, you're leaving yourself behind. What does it mean to code? Well, this is it, right?

And even script versus developer, right? Like I'm, I'm a scripter. Uh, nothing.

I, I don't treat myself as developer. I write actual web development web applications too. But I still think of myself as a scripter.

I'll jump on that one. Ned. Yeah, Go for it, Larry.

I mean, I do it this every single day, right? Working with clients, helping them develop a practice around center of excellence on automation, of course. And adopting that SL SDLC process and building, um, everything around automation in that format.

Meaning we followed proper source control policy, you know, branching strategies. We can go on and on for days on this stuff. But getting these, lemme just say that the hardest challenge is what Eric was hitting on, is taking a traditional infrastructure person and trying to make them follow that mentality to do development, because it is not an easy transition.

And Eric, I think Ned, we've talked about this maybe a couple years ago. Um, I personally started out as a developer and then I took a different path. I don't do any development anymore.

I just get to talk about it. Um, but that is the biggest challenge that we have is getting people to think about that. Because it is, like Eric said, you can write a script, you know, the number of times we come into a engagement and they're like, oh, you're just writing bunch of scripts.

No, we're not, we're developing either using some terraform. I mean, it could be anything, but if it comes down to we also have to develop Python code to do a thing, we're still following that same process. Yeah.

Like what do you do for people that don't know their fu from their bar? Like literally, that's the one thing that used to always bother me. People are like, fu and bar, if you are not a, I, I still to this day don't even know what it means.

It's the most, it's, it's just, it puts people off if they don't know what it is. But it's funny, it's because it's natural and common language, and we do know what it means. But like, that's immediately when you talk to somebody about, oh, do you know about strongly typed this?

I'm like, this is strongly typed to me. Like I, I don't know what strongly typed means, but I know how to copy paste code and I know what it it's expected to do and I can figure out how to make it do that thing. But There's a gulf between that and what Larry is describing of Right.

Actually automating the SDLC. I mean even what, what is SDLC software development lifecycle, right? The time from I decide I need my software to do something right through the process to, I'm absolutely certain it will do that thing in production.

That's, that's the software development lifecycle, right? And there's a whole lot of discipline in there that I as a, as a infrastructure developer, never do. Like I might have Tests, documentation Tests and, and documentation.

Absolutely. I'm writing modular reusable code, right? These are all things WW we don't do.

I, I take my script that I used before I grab it, I cut out the bits. I don't use this time. I write some new code in it, and then I go, I go back and I optimize the original version, but I don't optimize the change version.

You know, this, this is me being a terrible developer. I'm a terrible developer. Hey, at least you admitted it so I can work.

Yeah. It's, it's, it's also like having, sorry, go ahead, Nala. Oh, sorry.

I feel like, um, what's being described in this moment is my life two years ago. 'cause I got into cloud when I decided I was done, um, being a network engineer. 'cause I got bored.

And so I became a cloud network engineer and then an architect, and then somehow ended up becoming a DevOps engineer. So I had to learn how to code and, you know, automate and all these kind of things. And yeah, it, it was tough for me in the beginning.

Chatt PT did help me, but like, it was tough. I had to go learn a bunch of stuff. Now it's easy.

But yeah, it, um, it was a lot. So that's something that I was gonna kind of bring up, not to bring up AI once again. Mm-hmm.

But I think going to your point of, you know, you kind of just need to recognize that there's an opportunity to, you know, gain efficiency and then kind of have the idea. Um, but if you have those two things, you know, with ai, you can kind of ask it to write the Python code or write the script. Um, and it will hopefully get you there.

And I think as you know, AI gets better, it will increasingly get you there. So I think the definition may be of, um, knowing how to code or being a coder is kind of expanding to someone who can just recognize opportunity and then be able to, you know, use the tools, whether it's GPT or, or whatever to get you there. Yeah.

Maybe. And learning how to feed the dodo bird, uh, right now and getting those skills, and then all of a sudden it's not there anymore. Is, is there script, does scripting, do scripting languages exist in a world of generative ai or does that just become a layer of abstraction that you don't need to do?

Just like in the past many things you had to reach down into and, and reach into the guts to actually make something happen. And it's been abstracted and so we don't need it anymore. I, I would actually, I would like say, let's restate your original premise to say, rather than code or die, it's use code or die, right?

Because your point, it's like, I may not be able to write the code, but I can ask chat GPT or would my preferred LL underwrite the code for me? And then I can use SDLC me mechanisms, CICD, all those things to manage the code. And I don't necessarily know, need to know how to code myself, but we should all be using code to manage the infrastructure.

Somebody said this recently, and I forget who it was, it was the original attribution was you aren't gonna be replaced by ai, you're gonna be replaced by, replaced by somebody using ai. Yeah. Yes.

Mm-hmm. And I thought that was that that's actually a great way to encapsulate the problem. Well, which is kinda strange because you follow LinkedIn these days and you're hearing all sorts of like software shops getting rid of the junior developer nowadays and basically replacing it with AI generated code.

You only really need, uh, according to these companies, the high level, top level developers nowadays, because you don't need to hire an intern anymore. You don't need to hire a junior developer anymore. They're, it's cheaper for a lot of folks just to go running to chat GPT to generate this base 80% code that they need to just get the project uplifted to begin with.

Maybe what you really need is to be able to write a great query. Hmm. Yeah.

That's, well they prompt engineers, right? That people are getting paid a lot to do that. And I think as we're, as we play with it and work that we're learning ways to phrase things and to back to the, you'll be replaced by somebody who knows how to use ai.

I think it's just getting in there, playing with it, figuring out things that can do and how you can, I mean, it's just been amazing while we've been walking, watching people present, like with Catchpoint, I had this question to ask about, you know, who, who might acquire you, et cetera. And I thought, well, I'm just gonna ask chat GPT and it'll structure my question. And then when we asked them, uh, who might acquire you, they then asked, what did chat GPT say?

And it was, he said, that's, yeah, those are pretty much the people that, uh, we've been talking to or been talking to us. I Think what's interesting about writing a great query is you have to have the knowledge and context to structure that query in a way that's gonna get you the result you mm-hmm. You want.

And someone who is less familiar with the industry could ask that question and get results that are obviously false and not know how to interpret that result. Exactly. If I'm building out an enterprise architecture, and I know based on my own experience what goes into building a resilient architecture, I can phrase my question in such a way to say, help me build an architecture that supports this number of nines of availability, is able to sustain, uh, a regional outage for this many days and has a DR plan with an RTO and an RPO of these.

And I know all this terminology and all of these things to ask for, but if I ask somebody else, okay, I need you to design an enterprise architecture for me, go ahead and use chat GBT, they're not gonna know to phrase the question in such a way that chat GPT is gonna include all that detail in the answer. It really gets back to a problem of, as most of us are fairly senior in, in our roles or what, you know, what level of knowledge we have in technology. How does someone who's just getting started get started if no one's willing to pay them to be junior?

Yeah. That was the really strong thing I got from, um, that conversation about we we're not hiring junior developers. Well, you don't get senior developers without hiring junior developers.

You're gonna start somewhere. The senior developers are now going back to being junior developers. 'cause like John indicated is that, that you can ask it to write this code.

And I went through that. I experimented, I told it, and Ned, I'm gonna touch on this as well, is you have to ask it the right question. So now we're prompt engineering and I said, Hey, give me a cloud formation template with two regions and I just want your standard three tier architecture.

Okay? And it wrote it. Mm-hmm.

And I went to go plug it in and it didn't work. It didn't have my routes, didn't have my tables, didn't have my thing. I was like, I want this deployed.

So now I had to write a long thing and by the time I got done of it writing the code and trying to write it, I think I could have wrote it myself as a junior or something, you know, I could have wrote the code and get it done. But it's just my understanding of it was, and the validation of it working is a junior engineer's perspective. Because now the senior guys have to validate the code.

It's gonna work and make sure it's gonna, When we used to have a, remember back in the days of L-M-G-T-F-Y, let me Google that for you. And you would just like go there, put in your Google search and then send that to people. Yeah.

That's what the, that's what we need to let me GBT that for you because you know, there's the bit of that cursive. Lemme check that to domain real quick. That the seniors have cursive knowledge and we've already gained the knowledge.

We, the problem I see is that the gap where we're pushing away junior network engineers, junior developers, junior, everybody, we won't feel that pain for 15 years. And then it's gonna become very obvious and it's gonna take 15 years to fix at that point. But I think we're thinking of junior developers as they're defined now, once again, see it, it's not the, uh, j AI that's gonna replace 'em, it's the people who, who know how to use it.

And so I think the industry will, will shift and the roles will shift. And so the people who are junior will be using generative AI in a different way to do different tasks that help these more senior people to accomplish what they need to do. And so I think it will evolve.

And so right now, we don't know. I'm sure some people do how it might play out, but I think it will change. And so we need to be skating where the puck is going rather than talking about where it is now A Lot.

I think a part of that too is accounting for, you know, the AI and the tools to also get better and to get more intelligent, more helpful, um, to the point where maybe you don't have to know all the keywords or all the exact points. Um, so I think like, you know, going back to selector, um, as an example, like, uh, you know, a problem comes up and, you know, they obviously know exactly what's going on, so they kind of know the things to look for. Um, but I think with more intelligent lms, it could potentially lower the, the barrier, uh, to entry for someone to be like, okay, well explain this to me.

What should I look for next? Right? Um, and then kind of go through it that way.

Why, why do we write software? You're asking me specifically, I'm just saying like, hard word. Think about it.

Make things Go to make things do, Right? We write software because we're trying to accomplish usually some business goal. The value's not in writing the software, the value's in accomplishing the business goal.

And if you gave me a pile of bricks and you're like, make a building, I mean, I guess I could do it eventually, or I could like buy a building at the end. I have a building either way. One is much easier to do than the other.

I'm just gonna buy the freaking building. AI is just gonna get better at building the building for us. I look at, so I started using copilot a year and a half ago, and I would ask it to, you know, write some Terraform code for me, and it would be like 50% wrong.

It would not realize it has to create variables for the input variables it put in the resource block. And so the code would absolutely not run. Today I built an entire website and I created the Terraform code to deploy the infrastructure for it just by talking to chat GPT.

And it did it almost flawlessly. And when it got something wrong, I told it what it got wrong, and then it went back and fixed it. That's a year and a half of progress.

Extend that out. 10 years. I can say I need a website that sells widgets and it takes this kind of input.

It has these kind of customers and I need it available in this region. And then the AI of 10 years from now is just going to build that whole website for me. And it'll be a conversation where we tweak it together until it works.

And I won't write a single line of code and I can live without coding, Eric. And you don't, you don't need to know the specific, um, outcomes. So you talked about you, you named maybe five outcomes, right?

That you would do now to make something happen? Well, you're not gonna need to know them. You're gonna talk in terms of what you just said, which is, I wanted a website that does this and that.

You're not gonna have say, you know, configure this server, bring this up, make sure that when it gets to this thing, it fall, uh, failover happens, et cetera. Mm-hmm. I think let's, uh, the most important delegate that we haven't invited to the table is right here.

No Cha, I need you to tell me when we discuss the premise of code or Die and that I believe that every ops person should learn how to code to some degree. You agree, disagree. And what's your opinion on that?

It's definitely a hot topic. I think coding skills can be a game changer for ops folks. It empowers them to automate tasks, troubleshoot more effectively, and understand the infrastructure on a deeper level.

It's not an absolute must, but it can definitely be a huge advantage. What do you think? But that's back to the prompt engineering or knowing what to query, because you can lead it in infrastructure On a deeper level In one way or another.

You Gave it a good shot. But, so there, I, I want to carry on with when Ned started going, but didn't quite get to because Ned is an IT guy. He's not a business guy.

So when he is talking about a business outcome and says, I need a website, that's not a business outcome, that's a No, you're right. Outcome, right? I was gonna Go there too.

Yeah. Yeah. It's like, you know, the, the business outcome is I need to sell more widgets.

Mm-hmm. Right? Chat GPT 76, I need to up my sales by 36% in Japan in the next three months.

That's it. That's, that's my prompt to go, go do the whole thing. Then we don't need to code and then we're probably all gonna be taken over by the machines.

Well, well, so lemme ask everybody a question. Who, who here started in infrastructure rack and sack servers? I know Eric, I know Ned gone, right?

Yeah. Yep. So after that, then maybe virtualization came, right?

You guys did that as well? Yep. Yep.

Cloud came, you guys did that as well, right? Yeah, I get it. We're all though, right?

Registration, right. So all that's, so right now you're not racking in second service though, right? But what we could probably all agree on is the core concepts of all that operating systems, networking more or less.

Same concepts, right? Same thing with software engineering. I mean, there was a time where, you know, people were writing their own compilers, people were doing their own garbage collection.

Things just get more and more abstracted. But it doesn't mean that those skills necessarily go away. It's everybody's point about you're gonna go, you're gonna ask generative ai.

Hey, go write this thing for me. Cool. You need to know and understand what that thing is.

And per definition of any cognitive scientist, you need to know the thing to do the thing and to know the thing. You have to be doing the thing, right? We, we lose our ability to know how to do something.

We're no longer doing it, therefore we're gonna have to keep doing it in some capacity. I'm done. So I'm, I'm gonna push back on that a little bit and say, um, yeah, I learned how to put a server in a rack and I learned how the, the handshake works with TCP, but what is the actual valuable skill that I honed and learned through all that experience?

It was the ability to reason through a problem and create a logical series of steps, whether it's a script or just, you know, a runbook that will help me solve a problem. Those are like the skills I was actually developing. And I can develop those skills whether I'm working in it.

If I'm working as an accountant, I could learn it if I was a carpenter. Like we as humans, just, we have the ability to reason and work through problems and adopt to new technologies that will help us accomplish our goals. And I think we get a little too bogged down and focused on, well, I learned this thing when I was your age, so you should learn it too.

And it's like, well, I don't, I don't wanna know how to change my own oil and honestly, my car doesn't even have oil in it anymore. So That's where the coders come in. 'cause it's all code now.

That's right. You, you're just gonna code the oil until it's not right. It's just been up upgrade.

Well, that's the, the risk that we also have too. That we are a bit of a, an echo chamber audience. 'cause we're so deep in everything inside why it works, how it works, when it works.

We, we, it's tough. It's, it's even like I always complain about, people say the work from home thing, everybody's like, it's inhuman to make me go back to an office as they're going to the grocery store where that person doesn't get to work from home. Right?

So there are 80% of the population in the world does not have the advantages that a lot of us have in this room. And that's the, the benefit where we can use it to understand the mapping of the business, but in the end, hopefully it is always gonna trickle to a point where other people gain advantages from what we are dabbling in, messing around with. And I, I want to just get the, just as touch meta on this, right?

Because one of the things that we at, at Tech Field Day like is diversity in our panels. And as I look around the room, this panel reflects absolutely those of us who've been through that, that series of progressions. And there's some younger people there, but this is a call out to younger people who have been able to put up with the railings and the complaints and the noise from all of us old folk.

If you are involved in it infrastructure, it in general. And if you have this sort of ability to discuss topics, and particularly if you don't look exactly like most of the people here, I would love to hear from you. com.

I would love to have people who don't look, sound and think like me here because it is that unique human nature that is so vital as we are learning new skills, as we're building new things, continually updating, I want to continually update our awesome delegate community because we need, uh, more new thinking that's completely outside of what we're talking about with coding or die. You don't need to be able to code, although it doesn't hurt if you can code, I'd love to learn from you how to be a better coder.