Become a Creator today!Start creating today - Share your story with the world!
Start for free
00:00:00
00:00:01
/platform engineering: rethinking complexity in tech
41 Plays14 days ago
Everyone’s talking about AI, but are you overlooking the real driver of developer productivity? On this episode of The Forward Slash, Dillon Courts, Callibrity’s Director of Platform Engineering, shares what he’s learned about building high-performing platforms, the real limits of the “everyone does everything” DevOps model, and why self-service tools are changing the game for software teams. He also explains why leaders who focus on eliminating inefficiency are pulling ahead of those who just follow tech trends.
Recommended
Transcript
Cultural vs. Software Change
00:00:00
Speaker
Can we learn from their research and if we were doing things the products way in its opinionated way, we're actually more, we'll be better off. It's a really fair question.
00:00:12
Speaker
It's way harder to change culture and change people's patterns and ways of working than it is to just acquiesce and and change software. Like that's just easier to do for some reason. People really do not like to change the way that they are doing things. So to answer your question, like, are we just doing it the wrong way?
00:00:29
Speaker
Probably in a lot of cases. Sometimes it's easier to spend millions of dollars on custom development than to change the way you're working.
Introduction and Guest Background
00:00:45
Speaker
Welcome to the forward slash podcast where we lean into the future of IT by inviting fellow thought leaders, innovators, and problem solvers to slash through its complexity. Today we're talking to Dylan Quartz. Welcome, Dylan.
00:00:57
Speaker
Thanks for having me. Dylan is the director of platform engineering here at Caliberty, where he spent the last seven years has it been the long seven years that's awesome helping clients navigate the complex space between software development and operations. He's been in tech for 15 years, though his career started in the unlikely world of mobile casino games, mostly slot machines, which surprisingly taught him plenty about user behavior and bad product decisions.
00:01:23
Speaker
Today, he focuses on platform engineering, developer experience, and cloud native technologies outside of work. He's a dad, ah Midwesterner. i didn't know that was a thing that you have to label yourself now. I guess I have to put that on my LinkedIn.
00:01:35
Speaker
And spends his free time playing basketball, mostly to undo the damage of sitting through too many meetings. I want to hear about casino games. How did that, like, what what was that like?
Tech Insights from Casino Games
00:01:50
Speaker
Yeah, I mean, like most software engineers, I think, i wanted to start my career in game development, and the only job I could land was making casino games. So that's where I ended up.
00:02:03
Speaker
um Also, I think there was a component of it. I wanted to stay in the Midwest, and that's where the opportunity was rather than going to one of the the coasts to work some for some of the bigger AAA studios. But the...
00:02:16
Speaker
It was interesting experience trying to, there's a lot of psychology in casino games that people don't really realize. the i like You're trying to get people to continually spend money.
00:02:28
Speaker
And to do that, you have to get inside their heads and make them think they're having fun even while they're losing, which which can be difficult to do. So yeah i did a lot of research in that space trying to figure that out and At one point, I was putting together different algorithms and like running through, we wrote this in Python, and running them through like simulations where it would spin 100,000 times and figuring out, okay, what was the max payout somebody got on a given spin? What was their total payout percentage?
00:02:58
Speaker
total payout percentage How many ups and downs did they have? Did they go five spins in a row without ah without a win or was it more like 20? And different players also liked different types of games. So like we would push out different types of games. Some players like to win a little bit very often. Some players like to win whole lot, but less frequently. So putting out different types of games and seeing which types of players are attracted attracted to different types was definitely an interesting experience.
00:03:26
Speaker
So in your experience, you know, kind of since that since then and being in that world, you're you're working for, you know, enterprises and bigger companies and whatnot and doing kind of internal products or or even their their customer-facing ah products.
Microservices in Gaming
00:03:42
Speaker
Do you see that level of like observability, of behavior observability in those apps typically? Or is it is it a different level for those kind of games? I think it's...
00:03:54
Speaker
probably more of an afterthought for a lot of software that's built today. i mean, we're getting better at it. There's a lot of conversations around product and user behavior and tracking that. And there's a whole discipline built up around it.
00:04:06
Speaker
That was my first introduction to it was seeing that and just how valuable that is. So we were doing simulations beforehand and then also tracking that behavior in real time too, to make sure that our simulations matched up to the real time behavior.
00:04:19
Speaker
So I think it, You just see differing levels of maturity depending on the organization that you're in. I think games games have a particular high stake in it because it has to be fun. If your game isn't fun, you don't have a game. So they care a lot more probably about that that piece of so but piece of things.
00:04:37
Speaker
Now I don't want to talk about platform engineering. I want to talk about the psychology of our users and how do we measure. it I thought do think that it is interesting that you you all were sophisticated enough to You were doing the the um simulations and then you were verifying your assumptions in your simulations and how those simulations ran by comparing it to actual real world behavior. That's a pretty sophisticated approach. I like that.
00:05:02
Speaker
Yeah, it was a it was a really cool system. It was a great way to start my career too just because I got to touch not just that front end component where you're making the video game and the experience for the user, but also like you said, the back end piece and modeling user behavior and then tracking that user behavior. We built like an entire microservices system in Python, back when microservices were cool.
00:05:24
Speaker
And then we we sent everything through to, we used the Elastic Stack at the time. So it was the Elk Stack to stand everything up. And we ran everything through there. Every single spin was recorded.
00:05:37
Speaker
ah its inputs, its outputs, we had all that data, and then we were able to run analysis on it.
00:05:43
Speaker
It's quite a difference from my first you know professional programming job as an intern in college. I ah spent time working for a cellular telephone billing company and my opportunity to code was changing one character in a COBOL file and I had to bring a stack of papers to the meeting that was about a foot like deep.
00:06:06
Speaker
right And then we would sit there and look through the diffs in those paper stacks. That's a little bit of a different experience, I think. Well, my first professional experience was mostly editing, like read me docs that were HTML and some markdown here and there.
00:06:21
Speaker
So not, not too far off. This was my second professional experience, which was a lot more interesting. It definitely taught me I did not want to be in the world of COBOL. I didn't, that was like, what are you kidding me? So it was basically changing. Like when you call for directory, I said, need that's not a thing for anymore. You probably have no idea what i'm talking about, but when you used to call 411 and,
00:06:40
Speaker
you'd say, I need the number for Bill Jones. It would cost 75 cents. I had to change that to 85 cents. And it took a stack of papers literally a foot tall to to go through and verify every little change that that caused in the bills. And it was, oof, it was gross.
00:06:56
Speaker
Anyway, so let's talk about platform vision engineering. That was cool. i'm I'm glad we touched on the casino games thing because that is pretty pretty fascinating. i want to hear more about that. But we're here to talk about platform engineering.
00:07:09
Speaker
so What even is it, right? That's what the kids would say. What even is platform engineering?
Platform Engineering and DevOps
00:07:14
Speaker
here is that Is that just, you know, it it seems like for years and years we were talking about DevOps and now it's like that doesn't happen anymore. we're talking about platform engineering. do we just change the name of something? Is that what happened?
00:07:27
Speaker
Yeah, I think there's a short answer and maybe a long answer to this. The short answer to platform, what is platform engineering is it's building a platform for your developers to write software on.
00:07:40
Speaker
that's the That's it in a nutshell. Now, getting into your comments around DevOps, I think it is a reaction to DevOps in a way. the The pendulum swings a lot in our industry, and we saw the pendulum swing with DevOps toward developers doing almost everything or having to know almost everything, maybe a better way to phrase that, because we still had operations teams, right?
00:08:05
Speaker
But there was a lot of complexity involved in that developers needing to know Kubernetes and understand YAML. um There was a lot of YAML sprawl, having to understand how things work in the cloud and and get well outside of their IDEs, the number of tools that developers have had to be responsible for and know their way around.
00:08:25
Speaker
In the modern age of software development has become astronomical and there's a lot of context switching. And I think we've started to see that developers are pushing back on that and not not as interested in a lot of that work too. I think think platform type work isn't for everybody.
00:08:42
Speaker
And there's a productivity issue with it as well, where developers just aren't able to push as much code that's providing business value because they're spending time doing those day to day operational tasks quite a bit.
00:08:55
Speaker
Now, i i I do think too, though, we're not abandoning DevOps. like DevOps isn't going by the wayside. If you think back to what DevOps originally was, it was more of a cultural movement than new...
00:09:09
Speaker
this new team that you would stand up and a lot of organizations have gone towards DevOps teams, but that was never really the original intent. The original intent was for it to be a cultural movement. You had the the three ways in DevOps, like the first being flow and systems thinking, like really thinking about how you get your code all the way from ideation phase all the way into production and thinking about the steps in between. And not all of those steps are can be solved technically too, right? There's a lot of people problems in there.
00:09:39
Speaker
And the second way being decreasing feedback loops as much as possible. So having to have that that quick iteration so that you can make those changes and and see if that's meeting your user needs as quickly as possible.
00:09:52
Speaker
By decreasing, you mean like kind of tightening them, making the the time span of that feedback loop tighter. Is that what you mean by that? Yep, absolutely. Yep, I'm shortening that up as much as possible. So if I make a change in my code,
00:10:03
Speaker
And I want to be able to see that running in a production environment, trying to keep that down to 10 minutes or less is a good benchmark as opposed to hours or days or even weeks or months in a lot of organizations.
00:10:16
Speaker
Yeah. And then the last piece being, um third way being the cultural improvement piece or continuous improvement piece where you're just constantly looking for ways to get better at your craft, better at delivering software.
00:10:30
Speaker
So you you mentioned like the the the developers, I'm going to take, to go back a little bit. You mentioned the developers kind of pushed back on this. But one of the, I remember at the time, and if you read books like Accelerate, I think it talks about this a bit, was like, you know, kind of that that notion of you build it, you run it. There were some benefits to developers, to you know, getting in their hands dirty with some of those those aspects of of caring and feeding for software solutions, you know, back in the day.
00:10:59
Speaker
people would, you know, developers were focused on developing and then they would just develop software and they would kind of throw it over the wall for someone else to run. So the the theory, and I guess there was some proof to it, that if you had to run your software, that pain that happens when it doesn't work very well and you're the one having to wake up in the middle of the night, that that that feedback loop was there to kind of force you to write better software, more robust software and that sort of thing. So are are we losing out on that aspect if we're just building these platforms and we're going kind of going back to you're not really running your software anymore, someone else's?
00:11:34
Speaker
I don't think so, because I think you you are still running your software you're still responsible and accountable for your software, i think is is the key piece here. So
Self-Service in Engineering
00:11:45
Speaker
just because we've offloaded some of those operational tasks doesn't mean that development teams aren't still responsible for development or sorry for running those applications.
00:11:55
Speaker
So there's one of the key components of platform engineering is the self-service aspect. So the idea is you're you're making it so that your developers can self-service and run their applications in production develop them, get them to a place where they're running and then operate those as well and observe them and make sure things are running appropriately.
00:12:15
Speaker
But doing the heavy lifting for creating like guardrail or a path for developers to do that is really what platform engineering is is all about. So that's a main pillar of it is that developer experience and getting it to a place where developers are able to do that all end to end without having to go to the operations team and have their hand held. The operations team or the the platform engineering team in this case is more of like ah an escalation point if they need help.
00:12:45
Speaker
But they're they're really trying to build developer experiences, but still the responsibility of running that application and production falls to the development team. So I don't think you lose out on that. I think it's still there. it's just an evolution of that and taking some of the heavy lifting and removing it, removing that burden from the developers.
00:13:02
Speaker
So the maybe the the emphasis, like you know back in the day we would have sysadmins, right? and you know they were And I think people have ah phrased it more like, you know um treat them as cattle, not pets, right? But they were doing a lot of, ah and you've heard the term, server hugging back in the day. Like, this is my server and I want to run i want it to run really well and everything has to be smooth. So they were more optimizing for the hardware's experience as if it were a living being, as opposed to the developer's experience. Do you think that's kind of the shift in the mindset? Is that where it's almost an empathy for the developer, not the piece of iron sitting in the corner buzzing along all day long?
00:13:44
Speaker
Yeah, absolutely. I mean, that's a ah big part of it is treating platform engineering or your platform, your internal platform, like a product. That's a huge part of it. it's It was funny, I was thinking about this more and there's that meme of um there's that meme of the astronaut who's looking at the world and there's another one behind him with a gun to his head. And was thinking like a good meme would be, oh platform engineering is just product development.
00:14:11
Speaker
and yeah And the guy behind him is like, it always has been. Because that's really what it is at the end of the day is we're developing a product for the developers internally because that's what it is. It's product development. So having that mindset is incredibly important as you're developing these these platforms.
00:14:25
Speaker
I do like that meme. It's one of my favorite ones. so
00:14:31
Speaker
So we kind of got here you know but to platform engineering. You're you're saying it it's almost a reaction to that. um that that pendulum swing over to like you build it, you run it. And and we kind of had a bit of an adverse reaction. We found we're doing things that, you know, maybe i didn't get into software engineering to to be a sysadmin all the time, right? So there's a little bit of ah it's like a correction in the market, right?
00:14:55
Speaker
Okay. Well, how did you get here? How did this become the thing that you you want to wake up every day and say, I want to go do platform engineering things today? Before I answer that, I do want to ah back up a little bit on the platform engineering like definition point and just make sure I mention, like we talked a lot about the developer experience piece of platform engineering. There's also the productivity piece. like That's one of the main goals to try to reduce the cost to to doing software development and also the speed at which you can deliver that. and
00:15:29
Speaker
There's also a big compliance and security piece to it, like trying to do things in a standardized way that are not going to cost the company too much money or too much risk going forward, especially when we are getting into this AI world, which I think we'll we'll get into a little bit further in this conversation.
00:15:48
Speaker
But that's a big part of this platform engineering definition as well, is that compliance and security piece and the developer productivity and like cost reduction and in terms of not letting development teams just spin up all these resources in the cloud that are going to cost your organization a ton of money. So putting some guardrails in place.
00:16:04
Speaker
Now, so risk you're saying it's risk mitigation, cost control. um what I mean, ah guess having developers have to do everything, that would mean that then you have to educate them how to do and know all of those extraneous things as well as how to develop good software. So you'd have to teach them all of the sucks compliance and everything. Now granted you should know some aspects of it but you're saying that platform engineering allows us to almost codify and kind of bake that sort of those sort of things those guardrails into the the platform form and that's how we kind of enforce or or assure ourselves that these things are happening in the right way.
00:16:43
Speaker
that the idea? Yeah absolutely. I think it's talking about that pendulum swing again, we're getting back towards maybe a little bit more specialization. We're realizing that there's just, there's too much in modern software development that's required for a single individual or a single development team to to understand and be able to do well.
00:17:03
Speaker
I feel like that sort has happened on the on a smaller level when, you know, there was a trend being in one of those pendulums. Everybody has to be a full stack engineer, right? Like we've got to and now it feels like there's people are kind of settling back into like, I'm um' a UI developer, right? Or I'm a ah services developer. I'm an integrations developer. There seems to be more of that. Again, that trend seems to be fading. Would you would you agree with that?
00:17:28
Speaker
I do think that's true. Yes. i think you're always going to see that in the startup environment for sure, just because they're trying to do more with less, but especially in larger enterprises and larger organizations, I do think we're starting to see a lot more specialization.
00:17:44
Speaker
I think it makes sense for, again, keep coming back to it, the complexity that the, the software world has gotten to.
00:17:52
Speaker
yeah I mean, it's, it's hard to know. everything about backend engineering and everything about front end engineering is hard to keep up with just one aspect with the market changing every you know week like it does.
00:18:03
Speaker
So, okay.
00:18:07
Speaker
So answering your question about like, why am i drawn to this field? So we were doing some prep for this and you asked me that question and and like, I almost had an existential crisis. i was like, what am i It's a great question. Why am I doing this? Like I've been doing this for seven years now. I don't even know why anymore. So it,
00:18:23
Speaker
Maybe do some soul searching. So I appreciated that question. And it made me think a little
Dylan's Journey and Motivation
00:18:27
Speaker
bit. So I think one is just a natural evolution of my career to this point. I have a background in app development.
00:18:35
Speaker
I think because I worked for a startup early on, I also got more exposure to the infrastructure side. more than probably most developers do. So I think I gained a skill set that maybe a lot of other application developers don't have. And I just found myself doing a lot of the tasks that were related to operations and DevOps type work early on in my career.
00:18:57
Speaker
And then I also started to feel like I was having more of an impact at like a macro level by doing the doing the the platform engineering and DevOps type of work.
00:19:12
Speaker
think I think I've shied away from application development over the years because it feels so narrow in its scope. You are generally shipping a lot of features, but you you don't you only have impact on your one product or application that you're working on.
00:19:26
Speaker
Whereas ah in the platform engineering space, you're you're working to enable the entire organization to be better at shipping products and applications. And that that impact feels much broader to me. And I think I really enjoy that.
00:19:39
Speaker
Well, I owe you an apology then because my assumption was that you looked at like salary surveys and you're like, hmm, these platform engineering folks make a lot of money. I'm going to say that's what I am so that they have to pay me more. I thought that, I really thought that's what it was.
00:19:54
Speaker
It feels a little, and and now that you explain it it, there's a little deeper meaning behind what that decision for you. So, i you know, I misjudged you. I apologize, Dylan.
00:20:05
Speaker
Well, I'm not going to pretend I didn't have something to do with it. Right?
00:20:10
Speaker
I think there's there's more to it than just that, though, too. I think also I'm just hardwired to hate inefficiencies. I ah can't stand it. I can't stand like just like I get a visceral reaction when I i see inefficiencies going on. So platform engineering is a good fit for me there.
00:20:27
Speaker
like i hate menial tasks like doing laundry or documentation. I hate doing it, cutting the grass, doing the dishes. like Anything that I feel like I can automate, or things that are are just repetitive over and over, i feel like I have to do something about that.
00:20:43
Speaker
um Do you have a robot lawnmower? I don't have one yet, but i would like to. Well, the other problem with me is I hate spending money. So yeah sometimes I'll just do things even anyway, just do the menial task because I don't want to spend the money to automate it.
00:20:58
Speaker
But I think- The neighbor has one of those things. It's nuts. and Does it work it works though. It mows his grass every other day or something like that. So he has a beautiful lawn. And it looks good?
00:21:08
Speaker
Yeah. Doesn't miss spots? It does a good job, yeah. I'm about to look into that. um Another thing about like me hating the inefficiencies I always want to automate things, like I mentioned earlier. So one of the fastest ways, i think people at Coliberty know this, to get something automated is probably to make me do that task.
00:21:26
Speaker
One example is i I do a lot of the scheduling for this podcast, and I got very fed up with having to do that on a regular basis. So I wrote a little Python app to automate that away. I've said in the past, if you...
00:21:38
Speaker
if you If you want to figure out how to code something up or whatever, you you know, you want something done, just tell developers on the internet that it's impossible and then you just sit back and they'll they'll do it for you for free. It's very true.
00:21:53
Speaker
It happens. I think it is very true. I think, not every developer, but I think a lot of them are wired that way too. Developers are lazy, like you've said that a lot. Yeah, we should be. It's a good it's actually a good thing for developers to be lazy. So I appreciate the The automation, especially the the scheduling automation, is pretty cool. often I actually haven't seen how you did it, I'll take a look at it whenever we get a chance to. think that leads into another point about why I'm drawn to this field.
00:22:20
Speaker
I'm incredibly utilitarian, and I'm a huge proponent of like good enough. So app development, while it's fun, a large portion of that to me, ends up feeling like you get into a lot of form over function conversations about making things pretty for your end customers.
00:22:39
Speaker
And it's just not that interesting to me. Once something works and is good enough, I feel like I've i've solved a high level problem. So I think like the stakes are a little lower when you're doing platform engineering and your customers are developers.
00:22:52
Speaker
Because a lot of them think similarly, I think. So they are more willing to accept you know something that maybe is not beautiful as as they are using that platform, but it works really well.
00:23:05
Speaker
And this this podcast scheduler is a great example of that. I got it like 95% of the way there. it doesn't It doesn't add Zoom links, for example, to the the calendar invite, but I can do that with like the click of a button and I only have to do it one time, but it automates away like 98% of what I have to do and probably saves me 10 to 15 minutes for each schedule that I have to do.
00:23:25
Speaker
But it's just a little Python app and I just run it locally. like Will I ever get to publishing it out on some infrastructure so that it can be used more broadly or will I ever get to a point where I add the Zoom link in?
00:23:37
Speaker
Maybe. it's just going to be driven by need though. For me right now, like it's good enough. So from a utilitarian perspective, I'm not overly motivated to go make that better. So would you say that we are the world's first AI-driven podcast? is that i mean Is that a claim we can put on our um website? ah not Not yet. right right now i just Right now I just wrote a and a REST API. wait So I would have to hook it up to a chatbot or something. are there Well, hold on now.
00:24:05
Speaker
I don't know. Back in the day, if you had like an if statement in your code, that was enough that was the bar for AI. like you could You could say that. Yeah. I think I do have one or two if statements. So if logic trees count, that's I feel like that's artificially intelligent enough to me. So we'll have to update our branding.
00:24:24
Speaker
and Fantastic. So part of your job as the director of platform engineering here Caliberty, one of the things that that you must do is kind of keep yourself apprised of the goings on in the market.
Trends and AI Impact
00:24:37
Speaker
And you recently attended a conference of some sort for platform engineering. Is that right? I PlatformCon in New York. So it's a week-long conference, but they have been doing a live day. So they did a live day in London and a live day in New York. And I got to go to the one in New York.
00:24:53
Speaker
the kind of The people who put on the conference, they put on the kind of put on the live day in London, and then they flew overnight on a red eye and then put on like the live day in New York. kind of felt sorry for them, but I only just did the New York portion.
00:25:05
Speaker
Wow. that's That's quite a schedule for them. So anything like, you know, the the we talked about the pendulum. What's next for platform engineering? what what are they kind of what are What are all the buzz right now within the platform engineering world?
00:25:21
Speaker
There's a couple of things. ah One is platform engineering is definitely on the rise. It's it's been overshadowed somewhat somewhat by AI's kind of sucked all the air out of the room and sucked a lot of the investment, I think, out of organizations as well. But nonetheless, it's still on the rise. You can go look out there at what kind of adoption it's getting across the industry and you'll see most for the most part you're seeing like 80% plus of organizations are looking at this or starting to execute on platform engineering in some capacity and you know that's that's risen from probably only 50% or so a couple of years ago that being said there's still under investment just because of the the AI boom I think right now and
00:26:08
Speaker
If you're a CEO or like a large organization, like where are you going to spend your money? Are you going to spend it on the efficiency side? Are you going to spend it on a potential like AI products or integrating an AI? Because that's what all of your competitors are doing and trying to see returns there.
00:26:30
Speaker
That's interesting. Yeah, I mean, AI is kind of, like you said, taking the air out of the room for everything. that you' Nothing can really get a word in edgewise, so to speak. What about
00:26:41
Speaker
barring AI? a bit Let's take AI out of the picture. What about the the, you know, the sell of platform engineering for for businesses? what what is the What is the pitch?
00:26:52
Speaker
and And is that effective? Or have we been telling a good story when it comes to platform engineering to say that this is worth the money to invest this in in this platform engineering thing?
00:27:05
Speaker
I think you have to make the case from a dollars and cents, just like anything else. Like that's going to be one of the top ways is if you can make the argument, we're going to save 25% of developer time.
00:27:18
Speaker
And then you can add up the salaries of all your developers and you can take 25% of that. And you can say, Hey, like for an investment of a seven or eight person platform engineering team, look at all this money we're going to save.
00:27:30
Speaker
If we're able to implement a good platform, you know, in six months to a year. I think that's a big part of it. Another big part of it is developers like to work at organizations that have really good developer experiences. So if you want to be able to attract top talent to your organization and keep that talent, then you're going to have to invest in this because there's other organizations that are doing it. And so your top developers that are going to be shipping applications and products that are driving revenue for your business you need to be investing on the platform engineering side in order to, think, to win those folks over.
00:28:10
Speaker
So I think that brings up an interesting kind of line of questioning here. So with, with AI and its promise of like, Oh, we're not going to need developers anymore. And so,
00:28:24
Speaker
Maybe as could it be that it's not just the fact that AI is just taking the air out of the room, it's it's lessening the the need. like I don't really need to keep the developers happy. I don't need to worry about attracting top talent per se anymore, at least not as much of it.
00:28:41
Speaker
now that AI is entering the picture and writing code for me, I don't care. There is no such thing as a smile on an AI code agent's face. like I don't need them to be happy to do what they do. They just do it. right So is is that part of it?
00:28:53
Speaker
Do you feel like that's impacting or has been impacting it?
00:28:57
Speaker
It's an interesting question. So this is one of the trends as well as just AI in general and platform engineering. I've got two other two other ones we'll get to, i think at some point, ah one being internal developer portals and one being treating platform engineering like a product. But on the AI front, I think I'll answer your question in like a roundabout way. So AI is being used in platform engineering in a couple of different capacities.
00:29:27
Speaker
One is people platform engineers are just like using LLMs to just like developers are right to to build platforms, to ask questions about Terraform, to ask questions about what cloud resource to use in this particular instance. Maybe it will it'll write some basic scripts for them, just like developers are doing to to write a lot of their code right now.
00:29:50
Speaker
I think there's also ai being used to like augment platforms. So doing things like anomaly detection, you're seeing that a lot with like APM tools, for example, like Datadog will do that.
00:30:03
Speaker
It'll detect, Hey, usually have 10,000 requests per minute coming in and now we're only seeing 3000. This is probably something you should look at. And then agentic workflows is the interesting point that I think ties in to your question. So platform teams are for the most part being asked to support this. And I saw something really interesting the other day ah from ah from a former colleague and he posited this question. He's like, we are Currently trying to figure out how to make these agents productive in terms of writing code, because we want them to potentially replace our developers.
00:30:43
Speaker
And what we found is that the best way to make them productive, there's a couple of things we got to do. One, we have to have really clear guidelines about how they need to contribute to this code. So that might be like a markdown file in the repository, for example, or readme, so that the so that the agent can can look at that and understand how it needs to create pull requests and what the structure needs to look like.
00:31:04
Speaker
We need super clean, like isolated environments in order to keep that feedback loop low. So an agent can go spin up an environment, make some changes to the the code base, and then test those changes in isolation, and then iterate on the solution and potentially create a pull request.
00:31:23
Speaker
And then we also need like good documentation and and code comments to help them understand. And then also like very small contained tasks. So we don't want to send an agent out to do this this very large implement this entire feature. we just want to send it to implement and maybe like one endpoint or some small contained task.
00:31:41
Speaker
Now, what's really interesting when you think through all of those things, if you replace agents in all of those recommendations with humans, human developers are exactly the same thing, are exactly the same way.
00:31:53
Speaker
But now all these organizations are racing to and invest and finally doing this correctly when over the years there's been a huge underinvestment in developer experience and developer productivity when they when when humans need the exact same thing.
00:32:06
Speaker
It's kind of fascinating. I mean, some of it I guess might might ah get down to like just the the speed at which these these agents can do things faster than humans. I was just thinking, when you when you mentioned like the isolated environments for testing, you imagine...
00:32:23
Speaker
an AI agent doing some coding and their verification step is deploying into a shared environment, which can be very brittle when you run tests. And there's like, maybe there's some transient failure. So they modify some code, deploy to a shared environment, run their tests, something goes wrong. Oh, well back to the drawing board and they go change their code and then they deploy and that error goes away or a different one shows up.
00:32:45
Speaker
they can spin on this thing like just crazy and keep churning and churning. The same thing happens as human beings, but we might stop and be like, wait a minute, is is Bill running his performance test again? You know, that kind of thing. We might think outside the box there.
00:32:58
Speaker
But yeah, they you can get them into ah quite a cycle there if ah if you're doing this to the ah AI-based agents. That's an interesting concept. I and hadn't thought about that. It's fascinating.
00:33:10
Speaker
Yeah, I found that to be somewhat mind-blowing when I thought about it more. And I was like, this is great because if nothing else, at least organizations will start investing in developer experience for their agents.
00:33:21
Speaker
And then that will trickle down to developers as well because developers aren't going away. like the of your question was, is AI going to replace all these developers? And I don't think so. I think it's just, we're going to get more productive and there's going to be more software written and just like people who had the same have had the same concerns with every big change in technology.
00:33:42
Speaker
right Cloud came along, we were wondering, are operations going to completely die? And they haven't, they've just morphed. I think our roles will just morph a little bit. I think we'll get more productive as a society. We'll be able to output more and hopefully that rising tide will lift all ships.
00:33:59
Speaker
Yeah. i want to like the One of our recent episodes, we talked about how kind of this whole this whole you know hyper um specificity that we have to do with with our agents, right? with With AI, with LLMs, you know, you have to get into like very meticulously explaining ah exactly what I want you to do.
00:34:20
Speaker
Well, that's requirements, right? So maybe that will be the byproduct is we'll get, we'll get better at writing, you know, good requirements and, and being able to specify software a little more clearly in in the future because of this exercise. Cause you just have to with these AI agents right now, we got kind of lazy and like, just create me a login screen and in human beings, you know, we'll, we'll figure it out and we'll, we'll do something that's 90% there and then present it to you. And you'll say,
00:34:48
Speaker
I wanted the button to be purple. Sorry. and Okay, yeah, we'll fix that. But an agent won't do that. It's so funny. like why Why is that? why Why are humans hardwired that way? like If I tell you to go make a login screen and it looks terrible, like I'm going to assume that's your that's your fault. like like That's what humans do. They're like, oh, that's I definitely explained it correctly.
00:35:10
Speaker
It's definitely the developer's fault that it didn't get done the right way. But then if you go tell an agent and the agent does it wrong, like you will assume, least I do, like culpability on that one. Like, okay, i did my prompt wasn't good enough. I need to go back and fix this. like There's some psychology aspect to that that's very interesting to me. yeah Why is the computer infallible?
00:35:28
Speaker
yeah right like It must be my fault. I apologize, Mr. Computer. Let me me straighten my act up real quick for you. It is an interesting phenomenon that's going on. It's like the people don't care about having to write ah hyper ah you know specific requirements now. we call prompts.
00:35:47
Speaker
But they don't worry about that now. They're not you know fighting back about it um moment maybe I don't know, again, maybe does it come down to the productivity thing? Is it because if I sit here and talk to you, Dylan, and explain, okay, I want my login screen to look like this and make sure it appeals to this and it matches ADA requirements, blah, blah, blah, blah.
00:36:05
Speaker
And I spend all that time and then you go off and you know you take two months to do it. Okay, that's fine. But as if I take that time and explain it to ah an AI agent it and it takes 20 minutes,
00:36:19
Speaker
That's a fair point. right Yeah, maybe that maybe that's it because I'm getting the payoff and it's worth going through the trouble because I get get my answer back in 20 minutes. and That's actually a great thought. It is worth it because the speed is so much quicker.
00:36:32
Speaker
Yeah. That's good thought. I do want to touch briefly on a couple of other trends that I saw at PlatformCon.
Product Development Approach
00:36:42
Speaker
One is treat it like a product. This one's been going on for at least 18 months now since I've really been invested in the platform engineering space and been looking at it closely and people have been shouting this from the rooftops, but it just it keeps coming back over and over again.
00:36:57
Speaker
Like you need to treat your platform team like it's an internal product and like it's long lives and it needs to have a cross-functional discipline to it so that means you need product representation on that team you need to go out and interview your users and understand what they really need those are the developers you need to do usability testing like that is absolutely critical to the success of your platform you're not going to have a good platform unless you've done that part of the work. So again, like back to the meme, like it really is at the end of the day, just another form of product development. It's just an internal product.
00:37:34
Speaker
Yeah. I mean, mean, I guess when we would do, and and I've done, it wasn't called platform engineering before, but I was a principal engineer and we would set up some of these tools that's reusable tools for the development team. And then we would want them to use them.
00:37:48
Speaker
And you know you can get into ivory tower situations where you're like, I think this is the awesome way to do whatever. And then you kind of force that on um the developer. So yeah, it would have been nice to maybe ask, what do you think, person who has to use this thing every day, all day, what do you think is the best way to present this to you? That's that's like a good point.
00:38:10
Speaker
Absolutely. And like I mentioned, the cross-functional, there's a marketing element to it as well. It's not just if you build it, they will come. You need to make sure that you're marketing to your developers and going out and having those conversations and and partnering with them. And hey, did you know that you can do this with our platform? It's really cool.
00:38:30
Speaker
So that all needs to happen, just like you would do with any kind of product you're marketing out to your external customers. What about the D word documentation? dude Do we have to worry about that?
00:38:42
Speaker
Well, just have the agents write that. Yeah, I write that. But yes, you do need good documentation.
Internal Tools and Developer Experience
00:38:48
Speaker
yeah Hopefully, i mean, like a lot of organizations right now are moving towards like internal rag chatbots.
00:38:54
Speaker
So... Make sure that your platform it has documentation that's hooked up to that so that we can ask things in natural language. Because it's weight um you put stuff out on Confluence or whatever your internal tool is nobody and nobody reads that stuff. Nobody's able to find anything ever.
00:39:13
Speaker
it's It's exhausting. you've got to have the same conversation over and over again. So I think that I love the the internal chatbot revolution that's going on right now. Maybe we'll actually finally get to use the fruits of all of our labor having to write all this documentation for years and years that, you as you said, nobody ever read and they would just say, hey, Bill, how do I do this?
00:39:35
Speaker
Maybe Bill can be a chatbot now and he'll go read that for you. I like it.
00:39:42
Speaker
Absolutely. And then the other trend that seems hot is the internal developer portal trend. And this is an interesting trend. I think It's worth noting, if you're having the space, you already know this, but developer portals are not the same as developer platforms. So you can think of and developer portals as like the front door, the entry point to your overarching platform. So a developer portal might have capabilities like the place you go as a developer to initiate some kind of self-service action. Maybe that's to create a new repository. Maybe you go there to actually deploy things. Maybe you can go there to look at observability metrics on your application.
00:40:29
Speaker
You can go there and look at maybe all the services or all the applications that are out there running and who owns them. That's that service catalog concept. There can be documentation there. And that's all great. However, it's it's just that single pane of glass front door piece.
00:40:46
Speaker
It doesn't contain like the actual engine behind it. So I think some organizations have made the mistake of going and buying a internal developer portal and expecting that to be their platform.
00:40:57
Speaker
But that's like um some analogies might be, i don't know, buying a a car with no engine in it. So you can't actually go anywhere, but it looks pretty and like everything is there to hook up to make the car go, but you have to actually have that engine underneath and that's the actual platform itself.
00:41:15
Speaker
Another analogy might be ah you you buy a front door and you call it your house, for example, but there needs to be a lot more behind that. So when we talk about like the platform behind it,
00:41:28
Speaker
let's take a self-service action as an example. So I'm a developer, I want to stand up a new application and I wanna create a repository. And that repository, I want to be written in a specific language, let's say it's Java, and I want it to have a deployment that goes out to a production environment or a non-prod environment. And I want observability already baked in with logs and all that.
00:41:55
Speaker
I would go to my portal and ask for those things and maybe go through a wizard. What language would you like to do? ah Where do you want to deploy this? do you want to observ What observability do you want baked in? You can check some boxes.
00:42:09
Speaker
But then when I hit create, everything that happens behind the scenes to actually do that has to be built and is generally built somewhat custom to your organization. So that's actually building the the platform part.
00:42:21
Speaker
portals, I think they can save a lot of time, especially if you've got a platform engineering team that doesn't have a lot of front end skill sets, which is the case in a lot of organizations. But it's not ah a solution that you can just pull off the shelf and expect to to radically change your organization overnight.
00:42:40
Speaker
So i'm trying to like think of a a piece of what would be a piece of the platform. And you mentioned observability, maybe maybe having all of the um visualization and and the um the tools that where you can observe all your running software, that's part of the platform. But, you know, and and that would need to be wired together as part of like bootstrapping a new application. But the the developer portal itself isn't going to be like, here's your visualization and metrics and all of that platform. That's not necessarily included in the in the developer portal.
00:43:12
Speaker
Correct. It's going to pull those visualizations and whatnot, probably from another source. And it's going to ingest data from another source. it's just an opportunity. We talked a lot about like sprawl earlier, technology sprawl and platform engineering being a solution and an answer to that. So internal developer portals are a part of that, just creating that one location that developer might be able to go and view everything that they need to do.
00:43:36
Speaker
Some developer portals have the capacity even to have, for example, the the tasks that a developer might be working on that day and personalized to you. So you can see like, here's the Jira tickets that are assigned to you, for example.
00:43:48
Speaker
Okay. So it's almost like their personal productivity tool as well. It can be. So this is ah this is an interesting thing. and i've And I've always, everywhere I've gone, we've always tried to optimize this whole idea of like creating new projects.
Challenges in AI Compliance
00:44:03
Speaker
And there's always been that thought in the back of my head lurking is like, okay, but how often do we really do this? Has there been any analysis? Like is the is the industry thinking about this? Because, you know, i mean, when when we were all talking about microservices where it was like basically...
00:44:21
Speaker
database table as a service. and Okay. Yeah, you're probably spending up a hundred of those a week, but that's kind of settled down. We're going back to more like monolithic, you know, domain driven, you know, apps. It doesn't have to be micro as in less than 10 lines of code, right? Like we're, we're not going crazy anymore.
00:44:41
Speaker
So what's, what's the payoff there with with us optimizing this like bootstrapping effort so much? I think this comes back to the, because that's always the example you hear, but I think this comes back to the ah treat it like a product.
00:44:56
Speaker
If that's not something your organization is doing regularly, you hit the nail on the head. It's not ah high value task that's worth automating. And one of the things that you need to do first is a self-service action.
00:45:07
Speaker
That might not be the most important thing. Maybe there's a huge friction point around doing deployments in your organization. And that's the thing you need to focus on. That's why it's so important to go talk to your your end users, your developers, your development teams, and actually understand what it is they're doing regularly and where the friction lies.
00:45:25
Speaker
So i think for some reason that is the use case. Oh, we're going to go create a new application, a new repository, and we're going to do it all automatically and it's going to have everything hooked up. But like you said, most organizations aren't doing it that often.
00:45:38
Speaker
Development teams are probably doing that once or twice ah a year, maybe bootstrapping a new application, depending on what's going on. So it's probably not the place that you want to invest a lot of your time.
00:45:49
Speaker
and think, you know, some of it is like, if we can get them to do everything right from the get go, then we, you know, we've basically wired all this stuff together, all the things we want them doing the right the quote right way, right?
00:46:01
Speaker
um Then we'll be good. So that may be part of it, I guess, that, you know, if we're laying down, you know, we're generating all the code or, or you know, laying everything down, wiring everything together for them,
00:46:13
Speaker
okay, we know it's done correctly. I guess the the missing piece that may be maybe a better payoff over time would be ah how do we kind of migrate or augment or...
00:46:28
Speaker
know introduce new practices into legacy systems that were, maybe they were generated by the platform before, I'm sorry, the portal before in the past, but we have a new practice. Every app now has to have a chat bot and integrated into it. How do we now retrofit all those things that we did before? i think that That may be part of those or something that's compelling to organizations more now as we roll out new things, how do we inject those into everything we've built before and and bring them along with it?
00:46:57
Speaker
That could be interesting. That is interesting. I think it's a harder problem to solve. That's probably another reason the self-service task of creating a new repository is one of the first things a lot of teams try to tackle because it's greenfield, right? You don't have to actually support anything that's currently going on. This is brand new. No one's using it. So I know we can take as long as we want to develop this.
00:47:18
Speaker
can do whatever we want to do so think that's a big part of it as well yeah it's fun too it's not fun to solve those hard problems like oh you mean you've got thousands of users in production and i can't interrupt their lives by just mucking around with your code oh that's not fun i'm go build a greenfield then yep absolutely When it comes to the you know more of the what's next, I know we talked about kind of what are the current trends, but do you you know putting getting out your crystal ball, so to speak. I say so to speak a lot.
00:47:52
Speaker
Get out your crystal ball, if you will. And ah you let's let's try to predict the future here. what's what's your What's your call? What do you think are going to be the next kind of big challenges in this in this world of platform engineering?
00:48:07
Speaker
I think it's, it's AI just like everything else is, is AI right now. It seems like that's where things are. AI really? You think that's going to have an impact on the future? Come on It's weird. i'm already We're already seeing it a little bit. I mentioned internal developer portals earlier. I'm seeing several like ports one that we follow closely, and they recently put out MCP server for port. So you can use natural language to execute some of those self-service actions you might create.
00:48:35
Speaker
like That's the idea behind that. So I think platform engineers... just like software engineers and developers definitely need to start to wrap their head around this AI revolution that's going on.
00:48:48
Speaker
You need to familiarize yourself with LLMs. You need to familiarize yourself with agentic workflows, what that is, how you might support them. We talked earlier about agents that are doing development, like platform engineers are going to be the ones that are tasked with getting those environments set up for those agents to operate in and making sure that that's done securely and and in a way that's compliant for an organization as well.
00:49:14
Speaker
There was some interesting talk at PlatformCon about trying to make LMs are inherently non-deterministic. And compliance is all about given these sets of inputs, you're always going to get these outputs.
00:49:28
Speaker
Like that's the idea and so that they can pass an audit. So trying to make something that's non-deterministic by nature deterministic is going to be a big challenge of platform engineering, I think, going forward over the next couple of years. All right.
00:49:44
Speaker
I like that. I think that's, I think that's very true. That's going to be a problem. It's fascinating to me that like, that's all you're hearing a bunch of, and I'm sure it will evolve, but like the buzz is like, you know, natural language is the new UX, right? You hear a lot of people are just creating these products. That's just a box and you start typing it and you say, well what do you want to do?
00:50:05
Speaker
Right. Um, hopefully that will evolve to a little better and a little more rich experience. i don't want to sit around typing all day, every day into some box. Uh, I do like the like chat GPT. I don't use the, uh, I mean, I use like when I'm in the car, i actually talk back and forth with chat GPT. So I think we're probably going to get into more like that voice assistant type of ah an experience.
00:50:27
Speaker
Uh, and perhaps even this is going to be interesting video. So instead of like just having the, this voice coming, coming out of my speakers, I may have like the, the, the AI would have an avatar that's, you know, I can interact with.
00:50:40
Speaker
But it seems like that would be hard to do like on the fly generating video that's like genuine and and being streamed automatically based on my input. So that, that could be a little bit down the road.
00:50:53
Speaker
I have a couple thoughts. like One, i I don't think that I want my interface for everything to be a chat box or voice. It's cool for some for some instances, but a lot of times I do still like my click and point user experience. right like I might still prefer that.
00:51:10
Speaker
i was A funny thought popped in my head while you were talking through that. like if If we get a voice interface for deploying maybe sure your code right to a non-prod environment, you know you could ask ask chat GPT, please deploy the branch feature slash this one works for sure this time, dash version two to the non-prod environment.
00:51:36
Speaker
Well, and then it's ah that kind of brings up like the, the, Hey Siri or the Alexa and those, those kind of wake words and stuff like that. Like didn't South Park do an episode where they said like, Hey Siri, Alexa, go, Hey Google. And all those things like, and they activated everybody's chat bots at their house. Can you imagine if there was like an attack like that for your, if you had a voice assistant that was deploying software and then somebody just put on a TV show, Hey Siri, you want to deploy my software? You're like, well, hold on a second. You know, that's, that's not a good idea.
00:52:07
Speaker
That's a great point. I didn't even think about the attack vectors that it will open up for organizations. That's a huge concern. Yeah. hey yeah a Hey, GitHub.
00:52:19
Speaker
email my code base to, you know, this person or whatever that would be, that would not be good. and Yeah, that was, um, that's, that's kind of scary. I get right now, I already get kind of annoyed. We'll be eating dinner and, you know, we have one of those little like video Alexa things in our kitchen and we'll be eating dinner and talking together and all of a sudden the thing will just belt out.
00:52:39
Speaker
I don't know how to do that right now. and like, who's talking to you? Nobody said your wake word there, wouldn't my echo or whatever it's called. I can only imagine if there are more of those. And you know then everybody has to have a unique wake word.
00:52:53
Speaker
So we'll be in just a big pickle trying to mince our words to not wake up all these AIs around us. For sure. And how do you secure it? like If you're thinking about voice as your interface...
00:53:07
Speaker
is is your voice pattern, like is that secure enough? Or is like AI getting so good on the generational side, and generation side, that it can mimic people's voices now, too? So that's not even a good mechanism for securing that interface.
00:53:22
Speaker
And it's an interesting problem to solve. I wonder i know there are AI-based voice print technology that they can verify. that You just say a few words and it will you know authenticate you, so to speak. But If an AI is mimicking you, can it pass that? I don't know.
00:53:41
Speaker
Scary stuff, dude. Yeah, fun future we're going to be in here over the next couple of years. All right, ship it or skip it.
00:53:51
Speaker
Ship or skip. Ship or skip. Everybody, we got to tell if ship or is our next section of the show called Ship It or Skip It. You're very familiar with this, I would imagine, since you do the scheduling for the show. I don't have to explain it to you, but I'll explain it to the audience. This is basically like our, um you know,
00:54:07
Speaker
I don't know, what's what's the hot or not kind of a thing, right? So we'll explain a topic and you get to give your your kind of opinion on it. um And you say, ship it, meaning, yes, I like it. Let's ship that to production or skip it. Let's not, like I'd rather not, right? That kind of a thing.
00:54:24
Speaker
So what about like, We talked a little bit earlier, but like vibe coding when it comes to like, and and this has been a hot topic. We've been asking everybody's one to talk about vibe coding because, you know, it's what we like to talk about here as developers.
00:54:37
Speaker
Vibe coding, but with infrastructure as code. How have you found that to be? I think it it's ah it's exactly the same as vibe coding development from my perspective in terms of the outputs that you get and how productive that it makes you. The same lessons or the same thoughts apply to vibe coding and development.
00:54:59
Speaker
You need to keep it small in terms of the context of what you're asking your LLM to do and You need to make sure they're using it more as an assistant and not just copy pasting things or just letting it do things and not looking at it because you you absolutely will atrophy in terms of your skill sets and understanding what exactly it's doing if you just let it do a bunch of stuff and don't have any idea how it works or what's going on under the hood. so I think it's great as an assistant.
00:55:27
Speaker
It depends on how you define vibe coding, as I think a lot of your previous guests have said too. But like the way that I define it is more, you're using it more as an assistant and asking questions and then taking that output and applying it to the program that you're trying to write. So on the infrastructure side, I think it works the same. The only hiccups that I've really had with it, and I think this applies on the development side too, is anything that's new.
00:55:50
Speaker
So I tried to do a cross-plane implementation ah couple of months ago. And cross-plane has been going through some pretty major changes lately. and The LLM I was using, I think had been trained up through, i don't know, maybe the ah early 2024. So a lot of the the documentation had changed because they did a new version release.
00:56:10
Speaker
And even though I kept pointing the LLM at the ah URL for the new documentation, it still kept getting confused between the previous version and the new version. And I burned hours and hours just arguing with the thing about how it's not using the right version and trying to tell it over and over again to use the right version when I could have just gone and read the docs and done it myself. So you got to be careful you don't get in those loops, I think. But that's true of of doing it on the software side as well.
00:56:36
Speaker
Yeah, absolutely. And any sort of current event. um I have found that i'll that I'll kick it into like a deep research mode when ah when I get into those situations and that seems to help because it'll bring in the the the relevant context um from the interwebs and inject it into the conversation. So it grounds itself a little bit, but even then I get i get mixed results with that as well.
00:57:00
Speaker
I do like it and I'm not You know, I didn't, I'm not a big infrastructure as code guy. I like infrastructure as code and I can, I'm dangerous. I'm just good enough to be dangerous with it. But it, I think it was helpful for me. I had to do some terraform stuff not long ago.
00:57:14
Speaker
like give me an example of this. And then like it would give you the example and then, hey, can you explain what's going on there? but i I think it's really good about that and like teaching. and It's a good teaching tool for me. So I do like that aspect of it, of generating recent code and then tell me what the heck this thing does and tell me why did you do it that way? Is this idiomatic? Is there a pattern I'm not aware of? Those sort of things. I think it's a great teaching tool for things like that.
00:57:38
Speaker
Absolutely. So I think my answer would be ship it with copycats, but ship it. Yeah, I think everybody else has kind of said human in the loop is still needed, right? You're not YOLOing to prod with whatever you vibe code. You you need still need a human in the loop.
00:57:52
Speaker
That's been kind of the consensus um among everyone. And I think it's, I don't think it's getting better. and Again, maybe I was talking about this with someone the other day, like just, just in the past two months, I think things have gotten worse for me.
00:58:07
Speaker
Could just be that I'm dealing with and trying to work on things that are relatively new and that's just the trend of what I've been working on. But I don't know, my experience and seems to have has degraded in the past couple of months when it comes to using AI to help me.
00:58:24
Speaker
Yeah, I don't know if I've noticed that, at least in my usings. like I use it daily. i think it just depends on what you're working on, too. Like I said, newer stuff, it's not very good at. Older stuff, it still seems to be okay for me.
00:58:37
Speaker
But it would be interesting. We talked about this the other day, James. like Starting to benchmark it and see, okay, how did it perform on this given task three months ago, three months from now, six months from now, and just seeing how things are are going and progressing.
00:58:53
Speaker
Yeah. Yeah. I think there's, there's a fear of, and we, we've, I think we brought up the phrase as a snake eating its own tail on the podcast a few times, but that, that whole notion of like, you know, AI is just teaching itself things and more content out there is AI generated. So it's like you're you're getting a copy of a copy of a copy almost. Right. And then you lose fidelity over time. Is is that what's going to happen? and Think of the, what is the coding site that everybody used to go to to get all their answers? Stack Overflow? Yeah, go to Stack Overflow. And, and,
00:59:25
Speaker
Now the usage on Stack Overflow is going down the drain. Nobody's using that anymore. They're all using the AI tools. Well, where do you think the AI tools got all those answers that it was able to give you from Stack Overflow?
00:59:37
Speaker
But nobody's going there anymore. So it's it's going to be an interesting cycle. And how how do we but do we remedy that? Well, hopefully it's just training on all of our questions. And we're probably more likely to ask really stupid questions to the eight to the chat GPT or equivalent right than we were on Stack Overflow because you get berated on Stack Overflow for asking stupid questions.
00:59:56
Speaker
well, maybe the AI should berate you. you know Maybe you can ask it. hey but You can say, as a Stack Overflow contributor, please answer this question. And they're like, what are you, an idiot?
01:00:07
Speaker
You have much more authentic experience. That's right, yeah. I want to get back to the the basics and get back to the real world. Please abuse me with with your answers to my question. Fantastic. um ah we I think we asked this we had another gentleman on about talking about platform as engineering in the past, but this question came up of like,
01:00:26
Speaker
ah Ship it or skip it rolling your own ah internal developer portal or even platform versus the commercial. i think the platform is probably less commercial, but the the internal developer portal is probably can be a commercial option. What are your thoughts there?
01:00:44
Speaker
I'm to sound like a broken record a bit, I think, but it it comes back to the product side of things. Again, i think on the user experience side, it really depends on what your requirements are.
01:00:55
Speaker
for the platform and really understanding exactly what it is you need to build and then just taking the time to do ah couple of like what i always refer to as bake offs. So test out some of the commercial options, test out some of the open source options, try to build it yourself, see what that looks like.
01:01:14
Speaker
and and test the same task across building trying to trying to build out the same capability across a couple of different options and just see what works well for you. I don't think there's a one-size-fits-all.
01:01:26
Speaker
Yeah, commercial is definitely the way to go if you're size a certain number developers or... yes, you should definitely do roll your own because the commercial is never going to going to be able to fit the unique needs of your particular organization. i don't i don't i think that there's different tools that are good for different parts of the job. i think that's what I'm learning more and more like the the further I get into my career too is the tools aren't really overly important and like having those good requirements and really understanding what you're trying to do going back to first principles is the important part. And then they're just tools to to get the job done and build the experience that you want to build.
01:02:03
Speaker
All I'll buy that answer. i might take a little bit of a contrarian viewpoint just for just for giggles. One thing that i've I've found commercial offerings, so this is coming from a custom software developer. This is what I do for a living, right?
01:02:16
Speaker
But one thing I do find about commercial offerings, presumably, right, someone has, if they're doing their their homework, someone has gone out and learned very deeply about this thing, this domain.
01:02:29
Speaker
And they've developed a product that kind of fits the best practices and the kind of the the smart way of doing things, right? And that can be argued that's not the case in a lot of places, but lot sometimes it is.
01:02:40
Speaker
So what I found when, you know, especially if it's something that's, it's not my core business as a company, right? Okay, I probably need to buy a commercial product for that a lot of times. What I found is,
01:02:53
Speaker
Whenever there's a bit of a paradigm mismatch or whenever there's a friction with that commercial product, you know whenever you have to do a hyper customization, people tend to just hop straight into the a customization loop and like, okay, let's make it do exactly what we want.
01:03:08
Speaker
Nobody really stops to think, okay, hold on a second. We're doing it this way, but the people who studied this domain very intimately and they know this world inside and out better than we do structured their products such that it works this way, which is counter to the way we do it.
01:03:24
Speaker
Are we the dummies? Are we, you know what I mean? that There's that aspect of like, can we learn from their research? And if we were doing things the product's way and it's opinionated way, we're actually more, well we'll be better off.
01:03:37
Speaker
It's a really fair question. think it comes back to something that I feel like is true the further and further I get in my career too. is It's way harder to change culture and change people's patterns and ways of working than it is to just acquiesce and and change software. like That's just easier to do for some reason. People really do not like to change the way that they are doing things. So to answer your question, like are we just doing it the wrong way?
01:04:01
Speaker
Probably in a lot of cases, but that doesn't mean that organizations are going to accept that and change what they're doing. So sometimes it's easier to spend millions of dollars on custom development than to change the way you're working.
01:04:16
Speaker
Man, I feel like that's a blog post right there. Maybe it is. It's a good one. I like that. and And it's true. Yeah, I've i've been in organizations where you try it. But maybe you can use the excuse.
01:04:29
Speaker
Hey, the software doesn't work that way. Sorry, you got to change what you're doing. People love that. um Yeah, they love it. Let's move on to the lightning round.
01:04:46
Speaker
It's time for the lightning round. Rapid
01:04:51
Speaker
fire, don't slow down. Hands up quick it count. In this game, out. It's for the lightning round. We will ask some very important questions. this is this is you know All of the stuff we've done so far is is fluff.
01:05:06
Speaker
that's that's the This is kind of the thing. If it ended up on the cutting room floor, We're fine with that here at the forward slash. This is the meat of the conversation, the lightning round. This is where we get down to what what everyone wants to hear about.
01:05:20
Speaker
This is what the world revolves around. this This is the important part. So pay attention, listen to the questions very carefully and and choose your answers wisely because there are correct answers and there are not so correct answers to these questions because they're they're hard hitting.
01:05:36
Speaker
Are you ready? That was a lot of buildup, but yeah, I think I'm ready. Do you need a moment to compose yourself to be ready? I'm all right. Let's go. Okay. Giving presents or getting presents?
01:05:52
Speaker
Geez. Can I neither? i hate the whole gift giving stuff. Like I said, utilitarian earlier. i don't Getting them feels like an obligation. Giving them...
01:06:05
Speaker
i feel i I'm not good at that. I don't like doing that. I feel like I'm going to buy the wrong thing. I'd rather just give someone money. What if it's just a gift card? that That seems silly. I'd rather just give them cash.
01:06:17
Speaker
Are you saying receiving a gift card? but Well, I mean, if if you're giving someone something, it's less of a hassle. But then you got to choose which gift card do I give them. Do I go to this place or that place? cash Cash works everywhere.
01:06:30
Speaker
i guess get I guess getting if I have to choose. and If I'm being honest, i mean and I should, i do like getting presents more than I like giving. i mean I know is you're supposed to say giving is the the yeah benefit, but I do do like getting presents.
01:06:48
Speaker
Anyway, have you ever seen a kangaroo in person? Yeah, Cincinnati Zoo has them. They're cool. you get to walk right through. Really?
01:06:59
Speaker
Yeah, they hop right across your path while you're walking. It's fun. Do allow you to box them like you see in those videos online? Is that a thing? They're actually ah very small. So I don't know if these are baby kangaroos or what, but they only come up to, I don't know, maybe chest height or so. They're not huge.
01:07:16
Speaker
So you think you could take Yeah, think I think one. More than one, I don't know. Yeah. and You saw the like the videos of the ones that look like jacked. you know they're like They've been on roids for six years or something.
01:07:28
Speaker
i think I've seen them, but not up close. i don't know if I've ever been up close to a kangaroo.
01:07:34
Speaker
Someday. Someday. I put it on my bucket list. Head of the zoo. It's not far from you. That's a good point. I got stuff to do. ah Would you rather lose all your hair or gain 50% more hair?
01:07:50
Speaker
Where am I gaining the 50% more hair? i'll see you now that's I'll leave that as an exercise to the Well, if it's on my head since I'm losing my hair anyway, I guess I'll gain 50% more hair.
01:08:01
Speaker
So you would rather not look like me with no hair? and That would be the preference. Well, I feel like that's probably a popular opinion. When
01:08:14
Speaker
was last time you stayed up past four in the morning? Oh, probably sometime. Oh, you know what? I went and visited a buddy in Chicago that I hadn't seen in a while and we stayed out late. So that was probably three years ago.
01:08:31
Speaker
As my grandfather used to say, ain't nothing good going on at 2 a.m. So um that's not my usual ML. Yeah. Yeah. just to remind you, ah the statute of limitations, uh, for theft, depending on how much is, believe seven years in Ohio. So that just and keep that as context for this question. Have you ever stolen anything?
01:08:57
Speaker
No, I don't think so. Is that because you're afraid of being prosecuted or is that ah that the real answer? No, I really just don't think I have. That's good. That's a good thing.
01:09:10
Speaker
I won't answer that. What's your middle name? Matthew. Matthew. Do you like that middle name or not? It's fine. It's middle name.
01:09:22
Speaker
It's a middle name. It's utilitarian. It exists. Exactly. it It serves its purpose. Don't have strong opinions about it. Nothing nothings strong. um Cake or pie?
01:09:35
Speaker
Cake. If it's good cake. Can I say that? Can I qualify? You can. You can. It's not like there's a right or wrong answer to these things. They're just stupid questions. It's not what you told me earlier.
01:09:48
Speaker
we got to get people to stay to the last of the show. You know what i mean? Like, so if we so build it up, then like, oh, i got to hear this. Right. But now they're like disappointed. Um, what's your favorite number? 42.
01:10:00
Speaker
forty 42. Okay. two forty two okay That's a, that's a fair one. That's, that's a very geeky answer. Real original. Um,
01:10:10
Speaker
what is the last song you downloaded? Oh, I don't think I have an answer to that.
01:10:20
Speaker
I don't know. I introduced my kids recently to Bohemian Rhapsody, but we just play that on YouTube. That's probably the best answer I've got. Bohemian Rhapsody. Okay.
01:10:31
Speaker
I will accept that.
01:10:35
Speaker
Who was your first celebrity crush?
01:10:40
Speaker
Emma Stone. Emma Stone. from what what where did that How did that come to be? What was that zombie movie? don't remember.
01:10:50
Speaker
She in a zombie movie. I think so.
01:10:56
Speaker
ah don't remember. We'll have to check that. We'll see if that answer is valid. I think mine had to be. Zombieland. There it is. Was she in Zombieland? Okay. yeah that's good movie. That's your first crush Zombieland? When did that come out? Like three weeks ago? No.
01:11:10
Speaker
but How old are you? Golly. Younger than you. Yes, apparently. My first celebrity crush, I think probably is either Daisy Duke, like the OG, whatever that lady's name was from the series back in the day, or Heather Thomas. I think that was her name from the Fall Guy, the TV show.
01:11:34
Speaker
One of those two. i can't I don't know which one was first, but that's who that would be. Okay, well, that wraps up the lightning round here at the forward slash. Anything you'd want to share in closing? Anything your're what you're ah promoting these days and you're getting into? you're writing books? if you Have you starred in a movie or anything lately that's coming out? Anything like that?
01:11:58
Speaker
Uh, if you're into cloud native and you're in Cincinnati, we just started up a Cincinnati cloud native group. We're doing a summer series right now. We're actually doing our second series on Kubernetes virtualization here soon. And we've got another one on open telemetry. So check that out if you're interested.
01:12:16
Speaker
is that available remote as well? It is not available. So that's for Cincinnati people only. Full in person. okay Yep. And check out my medium too. I post there somewhat regularly, just stuff I'm working on and some interesting proof of concepts I'm doing.
01:12:34
Speaker
right. Any he words of wisdom that you'd share when it comes to platform engineering, any sort of sage advice you'd share with them? Yeah, I'll circle back on the product piece, like treat it like a product. Make sure you've got a cross-functional team. Make sure you're actually talking to your users.
01:12:50
Speaker
um On the AI front, I'll say like first principles, like think back to, it's not as scary as people think it is. For the most part, it's stuff that a lot of us are already familiar with, especially when you think about the agentic flows. like It's really just a lot of, at the end of the day, the agents are more or less making API calls to accomplish these different tasks, and there's just an LLM component to it. So it's not as scary as you might think.
01:13:14
Speaker
So don't be don't be overwhelmed by it.
01:13:20
Speaker
All right. Fantastic. Thank you for appearing on the podcast with us, Dylan. Yeah. Thanks for having me it was a lot of fun. Yeah. We always have fun on the on the forward slash here. If you'd like to get in touch, drop us a line at the forward slash at Caliberty.com.
01:13:37
Speaker
C-A-L-L-I-B-R-I-T-Y. See you next time. The forward slash podcast is created by Caliberty. Our director is Dylan Quartz, producer Ryan Wilson, with editing by John Corey. Marketing support comes from Taylor Blessing.
01:13:52
Speaker
I'm your host, James Carman. Thanks for listening.