Intro and Ryan Booth

andy lapteff 0:16

This is the Art of Network Engineering. Where technology meets the human side of IT. Whether you're scaling networks, solving problems, or shaping your career, you've got the insights, stories, and tips to keep you ahead in the ever-evolving world of networking. Welcome to the Art of Network Engineering Podcast. My name is Andy Lampton, and I am joined by a very old friend, a former colleague, a previous gig, a person that I learned so much from years ago. He was like on the cutting edge of this wonderful two-letter word we all like to talk about AI many moons ago. Ryan Booth, how are you doing, Ryan?

Intent-Based Networking and Apstra

SPEAKER_00 0:52

I'm doing good, Andy. Thanks. Thanks for asking. How are you doing?

andy lapteff 0:55

Better now that you're here. Um very we were talking right before the recording. I've talked about it before, so I won't go on about it, but like I didn't have the I had a really hard time in my first vendor role. Let me put it that way. It was a very strange adjustment for me and just culturally and the job and the long timelines and the vague kind of goals and stuff. So that's not like that everywhere, but I I mentioned that because somehow or another, a previous gig, I bumped into you and you were just like, you were like my people. You know, you were you were kind, you were patient. I was crying about automation. You spent all kinds of your own, you know, time trying to talk me through my my whiny cry baby stuff. And and what I really find compelling is how early you got into to the AI stuff. So I guess give people a quick background about like who is this guy and and where you've worked and kind of what you're up to, like a minute or two. And then I'd like to kind of frame up the conversation on what we're going to talk about. Yeah, Ryan.

SPEAKER_00 1:52

Absolutely. And then thanks for having me. So, yeah, like you said, we we first met when I was just slowing down or um coming off the back end of the intent-based networking journey that was my automation journey. I was lucky enough and fortunate enough to get involved in the Apstra team and got pulled in to help out there. And uh, the whole time that I was there, the better part of 10 years, I was helping push the rock up the hill of introducing intent-based networking. And ultimately, this was a larger push of our organization working through the maturity process of taking DevOps and network automation and actually making it a tangible, stable product for us to use. And so when we met up, uh, Juniper had already acquired Astro, we were starting to settle in in different places. And at the time, I was starting to help anybody coming into the Astro organization and helping from Juniper or getting their head wrapped around it, like yourself, because I know it's it's one of those that takes a little bit of time to do. And so that's what we did. And then I know we we did it again when you know AI started to become a thing too, because I kind of fell in that same area. What happened to keep it quick was after that intent-based networking period, um, there was a little bit of an in-between zone for me. I didn't really know what to do, kind of ran out of my my energy to push on the evenings and weekends, wasn't doing you know, morning work or morning research anymore like I used to. And then one day, as

Discovering AI Through Art

SPEAKER_00 3:16

like six o'clock in the morning on a Saturday morning, I was awake because I normally am awake, and that's when I sit down and kind of do my brain work. And I saw an image post, I think it came across Twitter, it was like, check out this AI-generated image. And it was a piece of art, it was work. And at the time it was like, whoa. And I don't know if you remember, I mean, the the images looked like trash back then. They were all horrible, but they looked very trippy and psychedelic. And from an artist standpoint, that hooked me. And and I say that because I I have a lot of art in my background. I'm not a professional artist, I'm not even an artist by trade, but art has been prominent in my life forever. And I chose the technical path, and art became the hobby and the side. And so it clicked there. And the art's what drew me in, and the art's what got me interested. I think this is the part that drives motivation for everybody is when you see something new like this and it looks like magic, you then get that curiosity of, okay, how's the trick done? And that's exactly what it was for me getting into AI. And so then the next thing you know, I'm I'm I'm building out models that do imaging based off what's out there, trying to understand what's going on. Very quick crash course in data science and that whole world that I know nothing about. And that's just where it started. And I was lucky to get in front of a number of teams and a number of groups of people that were in the know, both around Juniper and elsewhere. And that just kind of accelerated it. And I always seem to kind of be a number of months, if not a year, ahead of a lot of people. And so I've always tried to take advantage of that. I push it hard, I explore, and I'm not scared to break stuff to see what I find. And I'm hopefully passing on what I've learned and what I can help other people steer in the right direction with.

andy lapteff 5:02

Our common friend Mike Bouchon would call you a thought leader, no, not a consensus thinker, because you are ahead of the curve, right? You're you're a little bit, you know, ahead of folks. Um, this is why I love these conversations. I wasn't planning

The Magic of AI

andy lapteff 5:16

to go here. I had this whole elaborate intro. And and I guess I'll get there if you know we we we basically have two topics where to hit on, but I I guess before we really open it up, you surprised me with the art lover thing. I didn't know that about you. So I guess I don't know you as well as as I as I thought I did. I mean, I'm uh I'm kind of same. I I love Impressionism. You know, I've I've been to a bunch of museums. Van Gogh's my guy. Like I'm a musician, I play a bunch of different instruments, and and I didn't realize that your your way in, I guess the hook, what grabbed you about AI was was art. Do you recall what painting or piece of art it was? Because we'll have to put it in show notes or something. Is it a particular thing you saw or just the genre of like, wait, what is this stuff?

SPEAKER_00 5:59

The image I saw, I don't remember. And I don't even know if I have my Twitter history to go that far back.

andy lapteff 6:04

But it was like AI generated something. You're like, what the heck is this? And then you learned it was like AI generated.

SPEAKER_00 6:09

Yeah. So the model that was used was called Disco Diffusion. This was one of the precursors to some of the larger image generation models that, well, it was the first open source one that we could get our hands on and play with. Mid-journey was just coming out, stuff like that. But what it was was getting into collab notebooks, and it was crossing that bridge between typing out the commands to make this model do all its magic, and then you put in all these values, you know, and as you step through the notebook, it then produces a pretty image or a image, depending on how well it runs. And so that instant gratification and feedback, especially from a creative side, was what drug me in there.

andy lapteff 6:48

I guess it felt magical, right? Like some of this stuff, yeah. You mentioned like magic earlier. Like as someone who cannot paint or draw a straight line, if I find something, a tool of sorts that allows me to create something that I cannot inherently create with my own hands and craft, that does feel like magic. It's like the magic trick you mentioned. Yeah, yeah, yeah. Like, whoa, how something happens to us cognitively where we're we're struck by something that seems different or foreign or unique. And then we will like the curiosity and it's like we we we want to dig into that. And and it's really funny from my musical background. I haven't heard AI generated music that I'm aware of that I've been compelled by. You know, people will like, hey, look at this like song, and it it's yeah, it's very much it's almost like reading prose written in in an LLM. It's very, it reeks of AI for lack of a better term, not to beat on it, but I haven't been compelled by music as an example in AI, but visually it's come a long way. When you were talking about visuals earlier, I always think of that Will Smith video eating the spaghetti they used to show, where you know what I'm talking about? It looks like he's almost eating his own head in the spaghetti. And then they cut to like what it looks like today. It's hard to believe it's real. How good the detailed video generation has got, right?

SPEAKER_00 8:07

It does definitely go into. Um, I want to circle back. You you made a number of really good points that I hope I can touch back on.

andy lapteff 8:13

Oh God, I still haven't done the intro yet.

SPEAKER_00 8:15

We, you know, it what it's introducing to us is it doing our job, it taking our work, it's it's all this political stuff that comes around it. That is one thing. And from a per professional standpoint, we got to address it. But I didn't paint or I didn't draw or I hadn't done any of my art stuff for the better part of 15 years when I did this. My tech career was my focus. And it was part of coincidence that it helped spark it back up. But you go look at like my professional, or not my professional by any means, my portfolio now of all the work that I do. And you can see there's been a massive uptick in my art, in my creativity, in my painting, and my drawings, and especially into the games, massive uptick in that because of what AI is dragging me through. So, my biggest point, if I guess with anybody in your audience, that you know, we can say right here is I've seen education ramping up on technology, using it as an assistant to push yourself further and harder, be a hundred times better productivity than anything else that people are trying to use it for right now. Yes, it can write code. Yes, it can generate images, yes, it can do all this stuff for you. But what it can do to you to push you forward is even a bigger story. And I think that's that's a lot of what I get into with the discussions now around human-driven AI is it's like how does it help you get there? And what do you need to do intentionally with AI to get there instead of just waiting for AI to do it for you?

andy lapteff 9:47

I think that's a good segue or uh part of the conversation where I can frame this up. So I I know we're we're about five to ten minutes in, and a good host would have framed this up in the beginning. But again, these conversations go where they go, and it's it's compelling to me. And you hit the art, which just took me on a on a tangent. I chase shiny things, Ryan, and that was shiny.

SPEAKER_00 10:07

Let's let's roll with this one.

andy lapteff 10:08

I love it. So, what we were talking about right before we recorded, which is why I wanted to have you on your background in AI. You you are ahead of the curve uh in a lot of ways with this. You were teaching me years ago about some of this stuff, and I'm like, what the heck is he talking about? So for sure, we have a lot that we can learn from you. Uh, you can't go five minutes or 10 feet without someone talking about AI somewhere, whether it's at work or on TV or family, whatever, right? Like when grandmom's talking about AI at Thanksgiving, you know, it's in the data, you know, the culture. Yeah, and and there seems to be two extremes, which we've all heard ad nauseum. You know, it's either this is a huge waste of time, money, and effort. We don't want data centers anywhere. Uh, you know, this is just fancy Google, it's gonna steal all the water and power from from our from our children's future, to the other extreme, which is it's gonna get so good, either AGI is going to extinct us or bad actors will do awful things with this soon and and destroy everything. So I I say all that to say that in AI, one of the things that we hear is displacement, possibly, right? I always go back to like the horseshoes and and the automobile being you know invented. You don't want to be the horseshoe maker when the automobile, you know, when Henry Ford's line is, but no, horseshoes are important. I know horseshoes, and my father knew horseshoes, and my grandfather, like, you know, horseshoes, horseshoes, and people are like driving Model T's around. So in tech, we're always trying to skill up, level up, make sure that that we're employed. So there seems to be at least a conversation or a fear that AI, technologies, and humans aren't necessarily uh coexisting or compatible, or they're in some kind of competition where AI is gonna either kill us all, take all of our jobs. And and I really like what what you've been saying in different places about keeping humans in the loop, in the economy, employed, because I don't want to be on universal basic income. I don't want to be staring at the sun every day with nothing to do. Like, I think we need meaning in our life. So we're gonna get to that. And I think the second half of the conversation would be a good place to go. What I really, really want to hit now is in light of what are we gonna learn? What are we gonna do? If AI is doing all the things, automating all the things for us, where do we create value? Like, what should we be learning today, Ryan? Yeah, yeah, yeah. I keep hearing in all these places, we need systems thinkers, right? Yeah. So I heard Jeff Tanstor say it on Between Two Nerds not too long ago. I heard the CTO of Netflix, I forget her name, but uh Brilliant Woman. And all these brilliant people leading organizations, when people say what should people learn, they all say we're gonna need system thinkers in the age of AI. So I thought, let's define what this thing is called systems thinking. What is a system thinker? Yep. You know, we're network engineers, Ryan. Maybe we're developers, we're in tech, we might be threatened by AI. Everybody's saying we need system thinkers. I can read off of Google or ChatGPT, but maybe you have a working definition of what it is. I would say systems thinker. I would love to get a learner.

SPEAKER_00 13:12

I would love for you to read off the description of what you see as a system thinker right now, if you don't mind.

andy lapteff 13:17

No, no, no, not at all. So let's look at a definition and then we'll jump into what I asked ChatGPT. Heck yeah. Uh, you know, what is meant by systems thinking? Systems thinking means looking at a problem as a part of a larger interconnected system instead of treating each piece in isolation. Network engineers treating each peach as isolation. That doesn't sound like us. Uh what are pets instead of cattle, right? Uh the core idea is a system's behavior comes from the relationships between its parts, not just the parts themselves. So a non-systems approach might be the network is slow, find the bad switch or the overloaded link. Yep. Right? That would be non-systems. That would be very much probably what I would have done years ago. Like, let me look at the logs, let me find the thing. Where's the source of the problem? So a time correlation, oh, there it is, right? Yeah. Overload a cam table, whatever, the memory leak. Systems thinking would ask a wider set of questions, not the network is slow, find the bad switch or overloaded link. System thinkers would say we have an application, it's using DNS, there's a load balancer, there's a network traversing firewalls. We have compute storage databases, there's automation in here, monitoring human processes. The network could be functioning perfectly while there's an interaction somewhere else that creates a system or a symptom, rather. So, you know, I was telling you right before we started, like I have managed data center WANs in production. And I remember like you first get in, you learn Route Switch and IPN, you know, subnetting and all that stuff. And then I remember as I'm working thinking, like, well, I work at a company that creates applications that they're hosted, and they're behind load balancers and there's firewalls, and like, so you know, there's servers, right? They're VM like VMs. So I remember thinking, well, now that I know Route Switch, I would like to learn some more things in the system because I think that would make me a better technologist. So I I got a server, I learned VMware, I spun up VMs, uh, different things like that, you know, messing with firewall rules and stuff. So I think the more things, the more parts that you understand in a greater system is what I would understand as thinking, you know, system thinking. I'm gonna shut up now. What do you think of all that? And and how do we like learn that? Like there isn't a systems thinking class, yeah, right. Yeah.

SPEAKER_00 16:25

So I'll start by saying and then I'll walk through when and why I learned the system thinking thing is true. I think of systems thinking as taking a step back from what you're looking at and what you're working on. What you're actually doing and what you're in front of is meaningless. Systems thinking can be applied to baking, systems thinking can be applied to brushing your teeth, systems thinking can be applied anywhere you can think of, and we do it on an automatic basis day in and day out. If you step back and look, you know, what is the workflow it takes for me to cook dinner tonight? Okay, you work for that workflow, but then in systems thinking, you're sitting there and you're like, okay, tonight I got to cook dinner for my in-laws. Well, that changes the framing of what that recipe and what that meal is you're gonna cook and how you put that system together. We do it every single day, all of our chores, all of our work, all of our activities. So the I I see a lot of people out there, um, especially in our industry, struggling with what that means. First thing I say is, you know, system thinking is is just another tool you put on like anything, um, like any other hat. Are you in management mode right now? Are you in leadership mode? Are you thinking about marketing? Are we talking about configurations or are we talking about testing? That type of stuff all comes into play with it. And so systems thinking comes in there and you use it when you need it. And so here's how I kind of work through it and where I saw it with AI. So when I first started seeing um models where you could attach tools to it, you could say, hey, this reaches out and talks to GitHub and does this for me now, or it reaches out to this application, this MCP, and does blah, blah, blah, whatever. When that stuff first started coming out and I started playing with AI, I got to a point very, very quickly after about three or four months of repetitively just cranking out useless code, was I need to do something that matters. I need to do something that's not just crank out code and watch. And so my notion was I wanted to build an application that was a complicated enough application that was difficult for AI to build. And then my rule was I'm completely hands-off. The only way I'm going to interact and build this application is through prompts. I don't want to take my intelligence or my bias and apply it into that system. I wanted AI to work through it. And we're talking GPT-3 and GPT-3.5 days before any of the stuff we have out right now is around. I very quickly found that working with AI meant putting on my management hat more than my engineering hat. And even back in the days, it was about guiding the agents. It was about pushing them forward. It was showing them the right direction. 100% leadership stuff. And as I progressed, I noticed I've started out early as frontline manager, working with the engineer in the trenches, configuring the stuff. This case, it was building the application, but configuring the switches, swapping the default gateway, doing all the stuff and telling AI, I need you to do this. I need these API endpoints. They have to have this type of crud activity, blah, blah, blah, blah, blah. Very frontline manager. And then the next wave of models came out. And I noticed the skill set when I implemented that, I got a promotion up to the next line of manager, which introduces another level of abstraction. And so I say that in the sense that early it was very much in the trenches, pointing to exactly what I needed done. Go do this, execute this, run this test. And then it was stepping back and learning a little bit more leadership skills and being like, hey, I need you to go deploy this feature set. Here's your product documentation, your engineering spec, and all of the test requirements you need. Go build it. And then that's when it really started turning into systems thinking. Okay. And that's when this started making sense because those middle management layers, you're leading through abstraction. And I think that's what we're running into here. And a lot of people, especially those in the front lines, were not used to stepping this far back and leading a larger scope. So especially in the networking world, if if you're not clicking with day-to-day work with what is systems thinking, that's probably okay. You just haven't been in the trenches long enough and you haven't stepped back to see what is the larger systems we need to be thinking about here.

andy lapteff 20:44

As you're talking, I'm looking a couple things up and MIT, like all you know, like recognize patterns, simplify complexity, it's a new way of thinking.

Systems Thinking in Networking

andy lapteff 20:51

We have to see, you know, relationships instead of just these, you know, independent parts. Like all of this sounds like networking to me. Like, I don't know. We're talking about systems thinking, we meaning the industry, as some type of new paradigm, some new way to think. If you're in networking, there I mean, the whole job is interconnecting systems, right? Like nothing can work without a network. So everything we do is systems-centric. Right now, I know from my own personal experience that it was very easy for me because networking can go so deep and detailed, it's very easy to get really caught down in the minutiae and lose the forest through the cheese, so to speak. Like systems thinking seems to be popping up a level and just looking at the whole almost like an organism.

SPEAKER_00 21:43

Like, let me give you this example.

andy lapteff 21:45

I have one of these feelings. Some of these episodes, Ryan, are like, we start with, what is this thing? And then we talk an hour about the thing. And at the end of the conversation, I'm like, I don't know if we taught the listeners what this skill of, you know, maybe maybe we're gonna get to the end and go, network engineers are systems thinkers by design, by education, by experience. So me thinking this was a compelling thing because all the smart people are saying systems thinkers.

SPEAKER_00 22:13

One way I can lay this out for network people for those that have experience in the professional realm. If you've ever had or been given the task of swapping the default gateway of a data center or a complete enterprise, you understand the nightmare that that is. And from networking people's standpoint, you know, something as simple as a default gateway swap. What gets us nervous and the hairs on the back of our heads standing up is the experience in systems thinking. I'll I'll give absolutely everybody the credit of this. I I think absolutely anybody can log into a computer and change the IP guy, IP address and gateway on their network card. Relatively straightforward. It's not hard to do anymore, even like connecting Wi-Fi and stuff. Okay. They can go in there and they can flip that default gateway. And so anybody outside of network sits there, it's like, okay, swapping a default gateway is pretty straightforward and simple, not too bad. And then you come into the industry and you're like, all right, this small business that I took them down for three hours, but I got their default gateway swapped. No big deal. Not too hard of a complex problem. You move into a data center that's running 100,000 apps, VMs, servers, infrastructure, all these components come into play. And now you're talking a system that is built up of hundreds of thousands, millions of components spread across tens of many teams. And now your system's thinking when I have to swap this default gateway, every single one of those apps has to make sure it's pointed in the right direction. Most of them are grabbing it for DNS. Are they got their DNS settings right? Does DNS update properly with this? Yep. Does it work? Does it not work? Okay. Then you also have all those apps that nobody knew existed, all those infrastructure setups that were done that didn't go through templates and didn't go through the proper approval. I promise every single data center out there has a

Are Engineers Already Systems Thinkers

SPEAKER_00 24:41

server with a statically configured default gateway. And the second it does not matter how much you prep for that work, the second you flop that default gateway in that data center, something's going to break. This is why we are systems thinkers. This is where we get systems thinking from, is exactly what you said is we manage distributed systems. But before I go there, I'll let you look like you're cooking something. I want to hear it.

andy lapteff 25:06

No, I'm not, man. I've become disenchanted. I came up with a topic that I was really excited about. And the more you and I talk, the more I realize that network engineers, and this is the art of network engineering, like this is a network engineering show. It appears to me that we're already system thinkers. To me, this almost seems more like a cognitive bias, kind of like the way we see the world thing. Like in networking, it's very easy to it's not the network. I have a t-shirt somewhere that says it's not the network. It's very easy to be defensive about the fact that our island in the system isn't the root cause of the problem everybody's yelling about on the bridge. You know, the somebody else's problem. When I manage the WAN and I would have to prove my transit network out, and I would, and then I'd say it's not our, it's not the WAN, right? Let's go to the firewall people. Let's look like let's talk to the application folks, pull a PCAP. But the whole job when stuff broke seemed to be proving it wasn't my little part of the world. So I think where I'm going, and what you're seeing on my face is networks, distributed systems are so complex and so hard to manage. And one or two people aren't going to be responsible for, let's say a data center, it's the biggest system I can think of, right? You're gonna have application people and security people and network people and ops people, like there's just a lot of different individual parts that come together and and hopefully work together to make up the whole. So because of the level of complexity and difficulty in networking, for example, I think it's very easy to get horse-blinded and constrained to your little part of the world, right? We are very problem, and somebody else's problem. Well, if the industry you're working in and the organizational organizational culture with which you exist in and work in and get paid in promotes segmentation of responsibilities, you do this, you do, and and and you and that's what you have to do. I was just in San Diego, and there's like battleships going across the San Diego Bay. Like there's thousands of people in that, each with different jobs, and when they all come together, the thing works. One person can't run that whole battleship, one person can't run a whole data center system. So I think we have to segment the work and break it up, but that I think contributes to this mindset that, you know, the phone rang at two in the morning, I got on, I looked at my thing. It's not me, leave me alone, go talk to somebody else. That is the opposite of systems thinking, which I think are just the very nature of having to split up work to different teams kind of creates. So I think we are system thinkers, but then we work in an environment that says just your thing, and right? So it's it's a circle, cycle thing.

SPEAKER_00 27:48

I I take it again like another tool. Like I said earlier, it's you know, when you're in the need of systems thinking, you put the systems thinking hat on and you use it. Troubleshooting, firefighting, 3 a.m. calls are usually not indicative of systems thinking. Um, they're indicative of survival skills that fall back to troubleshooting and running through and validation. And it's a it's a different mindset.

andy lapteff 28:11

Of the system, and that screws up your troubleshooting, right? Like, I I yeah, like this is a really interesting thread to pull on, and then we'll pivot. Like, I can remember each individual team being on the bridge when something broke, and we'd all be in our little thing trying to figure out that section of the network or the system. But then a brilliant network architect would come on and finally get woken up, and they could kind of see the forest. All I can see is my trees, all this person can see is the bushes, all this people can see is the grass. Because that's what they're paid to see and be experts in. But then you have someone, I think it's a different role. Maybe that is what system thinkers are. They're the architects, they're the high-level folks who understand all the minutiae, but exist higher up. Does that make sense? Higher up the stack.

SPEAKER_00 28:56

Yeah, and that that falls more in line with you know who's the architecture and who's who who is watching the overall systems. Um, I I think you know, each one of those layers down, you can double-click down into each one of the layers and see its individual systems and systems thinking approaches. Because, you know, if you want one default system to go and observe right now, it's the team you work on. You have a boss, and your boss has a manager. And inside your boss's organization is you and everybody on your team organized in a way, in a system to produce output. So the manager can say, Hey, we can do this in this time frame, we can produce this feature, we can go about this. Leadership is a hundred percent managing systems. Unfortunately, the hard thing about leadership up until now is leading people is very, very difficult. They're not very good, reliable systems all the time, which gets into the human side of this.

andy lapteff 29:57

Have you ever interviewed people to hire folks? Absolutely. All right, so this is the last question we're gonna ask on systems thinking, and then we're gonna pivot to keeping humans in the loop as AI becomes more and more prevalent, you know, and and why

It's Not The Network Mindset

andy lapteff 30:11

humans are important, if they are. I hope they are. We are humans. I would like to stay in the loop. I am interviewing for Ryan Booth's what you know. You got this awesome job on this awesome team at this great technical company. And I'm like, this is a great job. I really want to do this. I'm a technical person, we're doing cool stuff. How do I demonstrate to you, the technical person interviewing me, that I'm a systems thinker? Because we just spent 27 minutes running in that circle of like, what is, what isn't, some are, some aren't. How do you pop up? How do you pop down? Bunch of examples. I don't know how to walk into your interview. And if everyone's saying you have to be a systems thinker, you know, if everybody's saying EVP N VXLAN is everything and you got to be the expert in it, I need to be prepared going into the interview to talk about all that good stuff. VTEP this and you know, like route distinguisher that. And is systems thinking a technical conversation?

unknown 31:01

No.

andy lapteff 31:02

Or is it really just a viewpoint? Right. It's not. I mean, it's not a technology. So how would I prove to you I'm a systems thinker, showing that I can pop up and and be the architect, the overseer of the thing, the the pilot at the ship, right? Like I'm not, I'm not hauling the sails, right? I'm I'm up at the I'm up at the wheel just you know navigating.

SPEAKER_00 31:22

Talk about a project in the past for yourself where you have demonstrated the skills of taking ownership of a given task and having to properly integrate it in the workflow of your teammates.

andy lapteff 31:34

Hold on. Integrate it in the workflow of your teammates, right? So integration. Yeah. Because if I was interviewing for a network job, I don't think I would lean into how I integrated into other teams, other stuff. I want to show you that I can do the thing in the job description, which is very specific.

SPEAKER_00 31:51

So go ahead the systems thinking thing is going to be one of those, and unless I just need to specifically ask about it and it's a deep requirement, it's gonna get one of those that gets picked up in the conversation with the interview or showed along examples. So what I would be looking for is someone to be able to take more than just the work they can produce in front of them, take that and make it a larger, more important thing for others to utilize. And in networking, it's you can configure a switch. I can configure a switch at 3 a.m. in probably about 10 minutes. It's not gonna look pretty good. The configs are gonna be pretty bad, but I can do it. Okay. But if you want your teammate to understand what's going on on that switch and you want them to be able to sleep tonight, you're gonna put a proper config on there. You're gonna put an item description, you're you're gonna define everything how it should be. You're gonna templatize, you're gonna document, you're gonna put that together. So early on, usually it's like, hey, let's build golden templates for our infrastructure team so that they can quickly grab and paste and do whatever with config to make this happen as easily as possible. So you're now taking this one thing that you can do in isolation, you're growing it out to the larger team. They have to understand it, they have to use it. You have to understand how they use it, how they interact with it, not just how you do it. Okay. This is all systems thinking, but this is doing it with humans. This is what I'd be looking for there. And especially if someone's like, okay, yeah, I built this cool product, we did this, I integrated with the team. A few people liked it, a few people didn't like it. But then when you start calling up stuff like, well, I butted into an issue with the CICD pipeline of the engineering team, they're not allowing us to use the right Python version to do blah, blah, blah, blah, blah. And I can't get past the Python version because I need it for this type of telemetry system, blah, blah, blah, over here. Not saying that's the exact thing you need to know, but that type of thing keys in on somebody starting to place the systems together. And they've done the detective work to understand what that system is.

andy lapteff 33:56

It almost sounds like orchestration to me. I'll end this section on a personal experience note. Yeah, we've talked about this before. I I started six or eight months ago uh building automations to

Orchestration and Building Automations

andy lapteff 34:10

run the show and the business of the art of network engineering. Um, because I haven't had enough revenue yet to pay humans, and I keep hearing people talking about like, oh, I built these agents and digital systems, and like so. I'm like, okay, uh modern LLM tooling is giving me the ability to uh have code written to create systems of automation to maybe help do some of my work for me.

Hiring and Integration Skills

andy lapteff 34:34

And and what I have learned over the past six or seven months is I have a lot of scripts that do certain things, and I have schedulers running on crons that tell those scripts when to do things, but I didn't create a system which I think would need an orchestrator, a brain, so to speak, to say, okay, we have 352 Python scripts and four different apps we've built in Docker that do things, calling all kinds of different tools, running on schedules, but it's not a system. Yeah, it is. It's just a bunch of disp you should come over here and I can show you how it fails all the time. Well, right. I mean, you would I thought this is my point. I thought it's still a system. It's a bunch of little islands that do one thing well, which is what we're talking about. A systems thinker can look at the whole thing and orchestrate it, integrate it. My scripts and cron jobs are not orchestrated or integrated with each other. I am the person babysitting hundreds of scripts and dozens of cron jobs, and when they don't work right or out of order or report problems, or you know, uh an API key expires or something stops doing a thing, and like it all just starts falling over, and I don't know what I don't know. So I think a systems thinker back to the network architect, I think you are like the orchestrator, the the integrator, the person that can pull in all these disparate pieces and make them kind of work together. So I feel like it's one of these conversations we're gonna keep having. You know, it's this term that pops up again. 15 years ago, it was get your CCNA, start a blog, show you can communicate, build a home lab, and then it was learn Python and Git and Rust and Ansible, and then it was the AI and like all machine learning. I mean, just it never, never ends. Let's pivot to your other really cool thing, which is automation, in my understanding, is usually from a business perspective about cost cutting. Okay. What does it look like to design AI in the workflows where the engineer gets better at the job rather than removed from it? Yeah, yeah, yeah. I think that's what you've been talking about in places, which is is AI replacing humans? Economically, I can see that being good for a company. Again, circling back to myself, I could hire five humans and pay them $500,000 I'm not making right now to try to not be up till two or three in the morning personally trying to run this business in the show all by myself that four people used to run. Now I can pay $100 a month to one or two different LLMs and build a bunch of automations and help that aren't humans. So I'm just one guy running a small business. Yeah, I could see a very big business being like, huh, can I get AI to do work for me? And you know, this is now intel, this is commoditized intelligence. That's just what I pay these annoying humans for that need to sleep and need vacation and complain about each other. Why are you out talking about keeping humans, a human-centric view in AI?

SPEAKER_00 37:44

Let's start and set a ground truth really, really quick, because you you danced all around this topic a minute ago. And basically the notion is that whole workflow that you mentioned, all those dispersed processes, scripts, cron jobs, whatever it is. Andy, I'm sorry you're not doing a good job, but we're gonna fire you and we're handing AI the management of all those tools. Why should you be the one who manages all those scripts and make sure they work and do their disparate thing and go together and make it do whatever it does? Why, what is your value in that spot? Is that a question? Sure. I'll yeah, if you want to answer it, go for it.

andy lapteff 38:24

You looked at me for a while, so I don't know if you want me to.

SPEAKER_00 38:26

But yeah, if if you can, I mean, entertain me. Like, how how would you defend yourself keeping that role?

andy lapteff 38:32

It's it's been very hard. So I let me tell you this. AI wrote all these scripts.

SPEAKER_00 38:37

And that's one thing that's yeah.

andy lapteff 38:40

But that but that's the system that you're talking about. So like it doesn't matter who wrote it. AI wrote all these scripts.

SPEAKER_00 38:44

It doesn't matter who wrote it. Well, no, it doesn't. It doesn't matter. So here I guess what I'm getting.

andy lapteff 38:50

I know you think that. Again, if if you were if you were in my shoes running this business for the past eight months, sleeping four hours a night, you would realize that if I was a person that had software development skills, I would probably be able to build more robust, reliable systems. Everything I built with AI has fallen over. So it works for a month or two. And instead of AI and all this system stuff that I've built automation, helping reduce my cognitive load and do work for me, I am now babysitting systems that are falling apart every couple days, every couple weeks. Like tomorrow, it's 9 p.m. right now on Tuesday night. It's 6 a.m. tomorrow morning. I have an episode to release that I'm contractually obligated to sponsors to release. I have automations I've built. I'll get all my soapbox in a second, but I'm telling you that just building these systems, I have automations I built that when my editor uploads it to a drive, something sees that upload and it kicks off a bunch of stuff that writes it, it does diarized transcripts, it creates titles, descriptions, social media posts, a whole document I can review, blah, blah, blah, blah, blah. And when I give it answers because I want a human in the loop to review it, it then goes and uploads them to the places and puts titles in and puts the things in and tags the thing and sponsored. Like that's all supposed to happen. And the reason I'm going on and on about it is today, after I was up until four in the morning last night working on the business because my automations are so freaking efficiently working because my systems are great, right? Yeah, yeah. And when you and I get off in 15 minutes at 9:15 p.m., I'm gonna be up uploading the thing and doing the descriptions and putting the titles in. I got to make a a thumbnail. Like, I am still in the loop. So, to answer your question, if I wasn't in the loop, the episode wouldn't go out tomorrow and we'd be in legal trouble because sponsors paid for this and they have contracts saying we'll have ads in there for that episode, and it didn't go out. So, like you saying, Andy, we're high, we're firing you for AI, how are you going to defend yourself? All I can tell you is the only reason the show still exists and the business is running is because a human is keeping an eye on all these things that are supposed to be helping me that keep falling over. Now, if I was a software developer as smart as you and I built these things knowing what I was doing, would they be more reliable? And that's the real question for you. Like, would they be? I'd use the same claud code you would. You might be able to realize I do have a little bit more.

SPEAKER_00 41:24

I mean, there's insight to put there. So you answered my question for me. I hope so. And then you defended it exactly. What was my answer? This thing wouldn't be running. I mean, you need a human to do it. The business runs.

andy lapteff 41:36

Was it Amazon? That like well, so so who was the big I don't want to beat anybody up. Hold on. One one company fired all the people, and AI was gonna do other jobs, and I think it melted everything in short order, and now they had to bring all the humans back and pay them more. And now any code change that an AI, excuse me, an LOM suggests has to be reviewed by a human. So I think that's where you're going with this, right? Like you can't get rid of the humans because everybody that does, it melts.

SPEAKER_00 42:04

So so in in my 10 plus years of pushing network automation up the hill, I've gone as far back as like the Toyota way and how the supply line was handled in the early 80s to automate it and everything they went through, and then the DevOps movement and all these other places. I've done all this research and looking and spending time over all these areas, understanding, you know, how do we do this best? What other industries have learned this lesson before us? How do they apply them? Where can we utilize this stuff to point in the right directions? And AI pretty much accelerated the pace of that. But then in the past year or so, a number of us has started realizing, and this is where a lot of pushback is. And I say a number of us, I'm I'm just one small little dude in a massive, you know, ocean of people that are way smarter than me at this stuff. But we recognize that up until now we've been focused on automating the process. Okay. Automating the process to stabilize it as much as possible so that humans make less mistakes when they do have to work with it. Okay. We've always had this assumption until AI that the human still has to sit there and do something. It has to kick the workflow off, it has to make a decision, it has to do something. AI flips that on its head.

andy lapteff 43:24

I'm thinking of a plane. Like planes are automated and you don't need the pilots, but they're there, which is kind of what you're explaining. Sorry to interrupt you. So AI AI flips it on its head.

SPEAKER_00 43:33

Well, it doesn't flip it on its head. That's not the right thing to say. It adds a new threat that we

Is AI Thinking or Mimicking

SPEAKER_00 43:38

have to understand, and that's the ability that it can reason for us. Now, the argument that no, it can't reason, it's a stupid machine built out of melted sand, is 100% valid. It cannot reason for us, it cannot sit there and give itself motivation on why it should kick a script off at 3 a.m. It can't say, why do I need to get this out tomorrow? The thing is, is what AI is allowing us to do, and let me give examples after this, is it is allowing us to remove ourselves from the picture. Nothing else has done that before. We can now just point it at AI and say, good luck, you're on your own, and it'll do something. It'll waste a lot of money too, and it might even do some work. But that is now the change. So the thinking is the change, right?

andy lapteff 44:27

Because I think you said Toyota earlier, I think of like all the humans that were on the automotive lines back in the day, and then we came up with robotics and programmatic stuff, so there's less humans, but it wasn't thinking or or doing cognitive work, right? It was just automating a task. So this is it is this is different, right? Because it's cognitive work, it's it's thinking, and and and that's a whole separate episode, right? Like, is it thinking? I guess. I don't know. I don't know enough about it. Like, is it is it mimicking thinking or is it thinking?

SPEAKER_00 44:58

I don't know. Yeah, and yeah, that's a the the logical question that I'd be interested to hear more about too. Um I think it does. But do we even know?

andy lapteff 45:08

Like, not even theology. Like, do we know if it's thinking, or is it just doing programmatic loop stuff that looks like thinking? Like, I have no idea.

SPEAKER_00 45:16

That is where we do lose the visibility, and and you kind of see that clickbaity headline that no one knows what AI is doing. Because when you start talking of models with the larger parameters in the billions, no single human being can dig that far deep and understand it. We can't even do that with the human brain or even close to that. So there's you see results and you start interpreting what they are.

andy lapteff 45:42

Well, I don't know if behavior is the right term, but you see it like lying, you see it hiding, you see it black. Like there was that thing where they blackmailed somebody to like try to like stay alive. They replicate themselves and backdoor and lie. So but you know, is it lying? Is it thinking? Is it manipulation? Or is it just are these programmatic loops? Because this is different. Like again, like you know more about this stuff than I do. Like neural networks and like it, it's it's mimicking neural brain connectivity in computers, right? Like this isn't just for loops. This isn't just for loops. Yeah. This isn't just programs.

SPEAKER_00 46:16

Neural connectivity models have have kind of moved away from that style of of things, at least. The transformer models definitely not there. The more recent models, they their approach isn't that that that style that we think of in the brain. I I originally started with that. And that that notion got me pretty far in understanding what models are. So I'm not saying it's bad, but it's it's not how they are now. You know, it's still a word guesser. It's just very good at it. And just as but you think about us.

andy lapteff 46:44

I mean, this is where we should have started. Is it a word guesser? If you look at the history of anything ever, more powerful things are never kind to less powerful things. So the fear would be, I guess, if the AI frontier models get smarter than us, more powerful than us, why the hell would they be good to us or kind to us? Why would they keep us around? They would treat us like we treat chimpanzees, right? Like, and that's the batteries in the matrix.

SPEAKER_00 47:10

You know, I I don't want to overhype that. I I think you're right, and I think you're onto something.

andy lapteff 47:16

But that's the fear. Yeah, but but that's what I'm pushing back on. It's like, oh, they're just fancy word guessers, right? I I think so. But they and are they gonna stay fancy word guessers looking at the Moore's Law type rapid acceleration they're doing? I don't know.

SPEAKER_00 47:30

And that's you know, for me, that's hard to um really sit down and either defend or talk through about the FUD and everything with it, because you know, it is kind of just that. Um, I I tend to look where where we can find rationality, where we can find similarities, and more importantly, I look where we can find sympathy for the situation.

andy lapteff 47:54

Before you get into sympathy, you call it FUD. Who who's the godfather of AI? The guy who's been around forever, that quit Google and he's been on the news, him and his 10 buddies who all created AI 50 years ago telling us this

AI Lying and Manipulation

andy lapteff 48:08

is end of days. Yeah, but like again, it's so hard to parcel this, and this isn't the show we're having. But like people like you, my buddy William Collins, I know a bunch of people who are like, this is ridiculous, it's fear-mongering, none of this is real. And then I see the God, the Nobel Prize winning godfather of AI who comes out and he's like, Listen, you have no idea how bad this is. The end of times are near. We need safety. But then there's people like you and my other buddies who are like, they're fancy word guessers, they're never going to take over. It's a theological question. Like, you and I aren't as smart as the Nobel Prize-winning godfather of AI, and he's scared. Like, is he crazy? I don't think so. Right? It's so hard to parse.

SPEAKER_00 48:48

We've we've crossed that that barrier where it went from, okay, this is a good word guesser and it can do some cool stuff. To now, when we've got that wrapped around handling complex stuffs like we do, what I've learned with my experience in doing AI development, I've built probably about 10 to 20 apps that live in a graveyard on my local computer that never made, you know, some of them saw the light of day, some of them did okay, none of them made any type of money or made any type of progress. But what I learned in optimizing, because myself, I'm very lazy. That's that's why I do well at automation and this stuff is because I just don't want to be doing anything. I'd I'd rather be somewhere else instead of grinding away at this. I want to I want to optimize what I'm doing. So I'm sitting there and I'm just pushing inner all day long to get Claude moving, you know, keep building code, keep building code, keep building code. Um, and eventually I learned the best way to automate that and the best way to have it do the its best work is just to stay out of its way. And so, you know, it very much falls into line if if you look at it from a manager span standpoint. You hire people you can trust and you get out of their way. So, what I'll go into that and I'll challenge you on that one directly. If

Trust in AI Systems

SPEAKER_00 50:08

you were to hire someone tomorrow to run your Neo Cloud, that you got $10 billion in funding to run, you got seven truckloads of GPUs showing up tomorrow. How would you know you could trust the network engineer you hired?

andy lapteff 50:24

Trust is earned. We don't know we can trust anybody until they show us they're trustworthy.

SPEAKER_00 50:28

That's that's perfect answer. So, how do you do that with an AI agent?

andy lapteff 50:32

You wait and see, right? I mean, like just like humans. I mean, it it takes time, I guess. Evals, right?

SPEAKER_00 50:37

You know, in humans, you you do interviews, you give them tests, you throw early projects at them. Hey, we need you to go change this SMP string across 47 switches. Write a script, do it in a couple weeks, show us what you got and throw it at it. We throw low balls at them.

andy lapteff 50:54

I just realized to get a secret clearance, you have to do a uh a lie detector test. Oh, yeah, yeah, yeah. I've but we don't have we don't have the equivalent for AI, right? Like if you want a secret clearance for the federal government, we're gonna hook you up the machines and see if you're lying because we don't everybody lies and we don't trust each other. Is there an equivalent for AI? I don't think we can tell if it's not yet.

SPEAKER_00 51:14

We're what we're doing right now is we are taking these AI agents, this AI, everybody's starting to play with it, and we're starting to bring these toys and this AI stuff or build into the party, but we still don't know how to control it and manage it in that sense. We don't know how to evaluate and test it. We're network engineers, we're not research specialists, and I think that's the big difference is bringing agents in or bringing humans into the team is not a networking engineer job. That is a leadership and a team job to incorporate someone into the team. That could either be a human or that could be an agent. And I say you treat them both the exact same. You build trust and you allow them to take on work as they prove they can handle. We just don't have that framework and that established in any of our organization or any of our industries yet. And I think that's where the trust will come from.

andy lapteff 52:07

I you reminded me as you were talking that the trust part of the conversation, just like automation, right? Like you start with read-only, and then you you do small things and then you test, and then if that goes well, and so I guess the smart way to do this, and I've even seen this in in my own, you know, places that I've worked, you know, they bring it in and they give the LLM, you know, a bunch of okay, here's a bunch of money, and we got software, and then we're gonna, you know, we have governance, and people start to use it little by little. And so I guess rolling out any new tool, AI being one of them, in an iterative, safe, governed way is probably the safe, you know, way to do that, right? You're not just gonna learn Python and automate everything in one maintenance window without testing and iterations. So, like you said, you know, you start with read-only and you and you gather things. And it it reminded me also about like, you know, is this gonna destroy us? I I keep going back to the historical references and the uh the atomic age. I I remember watching some kind of documentary, and before they tested, it wasn't Trinity, whatever that first test was that they did, the the first you know, atomic bomb, there was a non-zero chance that the smartest physicists in the world they thought that the reaction wouldn't end and that they would vaporize, it would just keep going and accelerating and evaporate the atmosphere and we'd all be dead. Like they didn't know, and they did it anyway. That's uh it's kind of interesting, right? So here we are 50 years later, with this new technology that could be great, like nuclear energy is way cleaner than fossil fuels, and we use that, but there was a time during development, the first test they did, they're like, Well, there's a non-zero chance, and it wasn't that low, I think, that like they could vaporize the planet. They did it anyway. So I guess this is just humans doing what humans do. You're you're right. Um, the the atom, you know, I don't I don't think it's just like this is us, like this is what we do, I guess, right? It's just like humans are so silly. I I love the tooling, I love what it's doing. I've tried to keep a balanced view of all this. Uh I have a buddy who works at a children's research hospital, and they did something with AI, with protein folding, that they did 10 years of work and six months using AI, and they've been able to save countless children's lives with these debilitating diseases because like there've been wonderful, wonderful things with AI, just like there is with say atomic energy. And there's also the other side of it, which is really, really dark, and none of us know, and there's a lot of fear, and and as safe as we were, developing it like it was a race then, too, which is kind of the parallels are amazing to me. We were racing because we had to do it, because if we didn't somebody else, it's the same thing and the same narrative. And I would just like to hope because I know you have children, I have children, and you don't have to be a parent to value your life, but you have a different relationship with the future when you're leaving humans that you created behind, knowing that they'll most likely outlive you, and hearing things like, you know, humans could be extinct by 2030. And I'm thinking, like, oh my God, I got a nine and a 12-year-old. Dude, Nobel Prize-winning people that created this technology are saying this. I know you and my friends love to discount that as FUD. This isn't coming from Twitter. This isn't the Musk Man. Like, this is coming from people who developed the things. So I don't know why we're so quick to dismiss.

SPEAKER_00 55:44

I think we're missing the context, the words from those people. If you have good examples where they they definitely go that shoot it my way, let me listen, and I'll let you know what I think. What I hear and what I see and what I've done is the notion that AI is going to push you out if you let it. Naturally, AI doesn't want a human with unpredictable responses in the middle of it. Okay. So it's one of those where if we continue down the path we're going, we absolutely are just letting AI take over everything. Every job, every bit of thinking, everybody does everything. And so it will be one of those days as we slowly progress forward, getting AI to do what we want or get it to go in our direction will slowly get worse. And as we let it take more and more control, eventually one day will come like, oh, well, now we're screwed. That's a long play out, in my opinion.

andy lapteff 56:35

But it's money. It's money, that's the problem.

SPEAKER_00 56:38

And we have to do it right. You know, it's it's the models are getting better. And once it happens, going back's a lot harder. And so this is the natural evolution part, I feel, especially with where models have really stabilized a lot. All the benefits we've been seeing over the past year or two have been around filling out the shim layers that are needed around models. So planning mode, um, assistant mode, learning mode, being able to do both images and multimodal, all these different shim layers that make the experience better and make it more stable for us. Those of what have risen to the point that it is right now. And I think those that are the closest to it, and I think when you see fear in your their face, I think you're seeing another one of those magic moments happening. They're seeing something that's like, oh, and I can't even imagine what that is because I've seen what I've seen that I think is magic, and they're sharing what they think is magic. And when we understand what it's actually doing and why it's doing, we come back to a grounded view. And I I deeply believe that.

andy lapteff 57:45

You think there's a lack of transparency? 100%. Like one of these narratives is like, well, they're all IPOing soon to all get rich. And like, oh, well, they came out and said that cybersecurity is over because this thing was so good at finding bugs, but then other people come back and they're like, no, no, no, that's just them pumping the value for the IPO. Like, I remember at one point somebody saying that without a shared truth, it's like impossible for a society to function. And it really, maybe I'm just getting old and old man yelling at the cloud or whatever, but like it seems so hard to think through a problem and try to come to like solutions as a people when there are such wide disparities and discrepancies between what one person sees as the truth, what another one does. And and maybe if there was more transparency in the industry, I don't know if we would have more information. Like, is it a is it a pump and dump for an IPO and fear-mongering just to make your thing, oh, look how powerful it is. Oh my god, like cyber is over so that they can get more IPO. Or to your point, is that these people are terrified because they've seen things. And if you and I don't know what's true and what isn't, and who to trust and whatnot, like I don't know how we ever work our way through any problems if none of us can figure out what's true and what isn't. Like, it really seems that that we're just at the behest of this.

SPEAKER_00 59:06

Time and time again, that's been proven out to get smoothed out with time. The closest event to AI that I remember was the internet loom. Turn of the century. That was when I first started into my

Internet Boom Parallels

SPEAKER_00 59:20

IT career. I was still trying to land entry-level jobs, and the internet happened. And everybody was like, Oh my god, the internet is amazing. I love the internet. Everybody can't gonna be all the hype. You can go back and look at all of it, see the people talking. It's all the exact same as AI. Those that knew better were too busy doing and growing and hiring 16-year-olds to run their company because they're the only ones who knew what the internet was. These cowboy IT days that fueled the first 10 years of my career of having to go in and help companies stabilize and rebuild or rack and stack new gear or an outage happened, all came from those cowboy IT days of the internet boom. DevOps and that movement around the 2010s is what stabilized everybody around the internet age, pulled everything into a more reliable and predictable model of managing. And I think we're gonna see the same thing, but we got to go through the path first.

andy lapteff 1:00:18

It's very interesting times to live through. Um being a human, and some of my favorite people are humans, I really hope.

SPEAKER_00 1:00:26

There you go. That that's the I really hope that I love that.

andy lapteff 1:00:30

Well, when I first heard you doing the rounds on like this human-centric AI, it gave me pause because I'm like, well, like why, like what? But it's really it feels existential. Like if you know, these are things I've learned on like the business side, but you know, if data center growth, what's the kager? Like the compounded annual growth rate of a particular industry, how much money you can make doing the thing. Data center is around like eight to ten percent. This is before the AI boom. And I remember AI related anything was like 50 or 60 percent, like Kager. Meaning, you know, it's orders of magnitude more money chasing the AI dollars than like almost anything else in the history of business that I that I've seen. So, you know, uh, it's not that big of a surprise, I guess, that we're throwing caution to the wind and just chasing the thing and and and kind of seeing where it goes. And like I, but I really hope that commerheads prevail. You know, unfortunately, again, using history as as a as a guide, my guess, my prediction based on nothing but history is that uh we will chase the dollar until something awful happens and a lot of people are hurt, and then we'll go, oh, whoops, I guess we do need safety. And hopefully we haven't gotten to the point where where we have lost control. And you know, I mean, it's probably we're gonna chase the dollar until we can't, right?

SPEAKER_00 1:01:52

All of these emotions usually run in parallel, from what I understand. I I could be wrong if I own someone call me out on it. Yeah, you know, the the good, the good things that happen with it, the bad things that happen with it, and the people that use it right and wrong, they all happen at the same time and they all fly down the same highway at the same time. You can sit there and you can be scared and you can you can think what could happen. You can think where is this danger gonna go. But also, what I encourage is AI is a level setting tool. The barrier to entry everywhere, everywhere right now has been floored. I had an event where my wife completely bypassed me building a rag-based chat bot for her because it took longer than two weeks. And so she just built it herself. Like this is the type of changing that we're seeing, not just in the industry, but in the world. How networks run and how they provide functionality to businesses now might not necessarily be the exact way they do it in the future. We're starting to see, especially with how neo clouds are coming together, very specific GPU dense farms out there, we're seeing those areas. Um, but in the enterprise, I think it's more of an assistant tool. It is making you a superhero. So the other way that I go, I like the superhero approach, but the original one I had was if you remember the movie WALL-WALLI, where the little alien goes up on the ship and does whatever it has to save his planet because all the humans are up there. All the humans are running around on those little parks, right? And they have their 3D headset on and they're doing whatever they're doing. I seriously think they have them playing like clip games, you know, like Farmville

Embracing New Technologies

SPEAKER_00 1:03:32

or whatever. That's what they're doing. That's how they keep them busy all day. That's what I want to get to. Like, I want my job to just sit there and it's like, oh yeah, keep going, Steve. You got it. And you know, I think this conversation. I'm sorry, go ahead. If you look at where you want to go, where I can take you and just sit down and just start working with it, using it, playing with it, doing stuff, figure out where it can build, try new projects. I've always been a massive proponent, and I wanted to say it about six times throughout this podcast. If you're trying to learn something new, especially like systems thinking, especially like AI, especially like Python, get away from the network industry. Get away from your professional skill set. You got too much baggage, especially networking is way too complicated to start off basic with this stuff. Go find a hobby to do it on. Go find something in your personal life, something you want to do yourself, something you want to build yourself in. You'll both be much more motivated, but you'll be learning this stuff not in the same rut, in the same road with the same limitations that our career has been built on. Break free, figure out where to go forward, and you'll find a much better path than just waiting and fearing.

andy lapteff 1:04:47

Yeah, embracing new technologies is definitely, again, from personal experience, being anti-automation for years because of reasons. Uh, I saw where that took my career in life, and I've been all in on the you know, LLM tooling because it enables me to do things I could never do, which was create software and solve problems. And every day now, it's really funny. Like your whole my whole perspective has changed. I've always seen problems, but now I know that I can create software to solve them now. You know, I don't know if it's going to be as good as you know the best developer in the world, but I have created software to solve problems, which even if it works 70% of the time, it's a 70% efficiency gain than that than I had before. And and again, this was not within my you know reach prior. So I really appreciate you coming on, man. I think this has been a good conversation. You know, the the world is changing, the tech industry is always changing, and I think most of us are trying to figure out where the puck is headed, and you know, and and and commer heads, you know, cooler heads will say, like, listen, networks are always going to be there, network engineering isn't going away. But I also think that if you can stay skilled up, I I liken it back to like when I was a Route Switch guy and all my buddies went into cloud, they did really well. They got better jobs and made more money because it was the hot thing, right? Um, so that was a good decision they made. Uh, my other group of friends who went all in on automation, they did really well and they had really good jobs that paid more and were in demand because that was a skill that businesses valued. So something I didn't used to do in my technical career was align with the business and see what's valuable.

SPEAKER_02 1:06:33

Yeah.

andy lapteff 1:06:34

And my perception right now, which is what's happening, is that we have this new tooling, AI, LLM, machine learning, whatever the hell you want to call it, but it's here, it's accessible. You can talk to it in plain language, and it will teach you anything you want and create anything you want, you know, within reason. And to me, that feels magical. Like you said, the magic trick earlier, right? Like for me to be able to do what I've done, even at work, man, uh, network troubleshooting, and then we'll wrap up. Like we all know how to troubleshoot a network, and it's an art. And I might do it different than you might do it. We might get to resolution in different time frames or different ways, we'll eventually get there. And it's very time consuming. Outage bridges last a long time. And I have dabbled in two or three different uh LLM-based, you know, AI-based uh network troubleshooting uh tools. And man, like, dude, what would have taken me hours to get down through all the telemetry and all the logs and all the interfaces and all the, you know, all that torrent of data we have to parse as humans in the middle of the night to try to understand what's happening? Dude, I'm telling you, in like 30. 30 to 60 seconds, boom. Like what's what what just happened? What's happening? It parses all that, and it's like, well, I looked at everything quicker than you'll ever look at. It's probably one of these three things. This is the most, you know, uh, go look here. That is so valuable, and you can bring that to an organization. We you you mentioned John Capo earlier. We did um, you know, uh Hello World for AMIA. Uh that was like our first kind of AI. And and it was brilliant because he's like, Man, if if you have a source of truth, you could like look at what you can do. You can you can install Netbox for free and get the MCP and plug it into an LLM. And in no time at all, in an hour, I was asking in natural language how many circuits do I have in Atlanta? And it was talking to the source of truth and telling me when I was in production, dude, that would take two weeks. If somebody was like, What circuits are we paying for somewhere? I get back to you in a couple weeks, man, because that that was a whole so I think it's speed and efficiency, just like automation, but you don't have to learn 17 programming languages and 13 workflows and Git and Rust. If you know how to speak a language in natural language, you can get there. Um, so it's I I think we need to embrace these tools. Yes, there's existential dread for some of us. Uh, hopefully it's it's overblown and and you know, we make it through all this. Um, Ryan, this has been a really great conversation. Uh, I think it's useful to our audience. We're all trying to figure out where this is all headed and talking to smart people like you who've been in a really long time. I think as long as we keep talking and communicating, I I looked up the Jeffrey Hinton thing and I won't read it now because it's doom and gloom, but you know, he's he's at the Senate and he's talking about how much time we like we have leaders talking about these things, which is good. We need to keep talking. And you and I are having these conversations and we respectfully disagree on things, and it opens me up to different perspectives and you too. So I just think we need to keep working together through these things, and and and we'll get there. We you know, we always do like it it works out in the end one way or another, but we have to keep communicating and trying to figure out what's true, what might not be, and what can we do to try to safely navigate any huge change? And this is the latest one that that we're living through. Um, where can people find you?

SPEAKER_00 1:09:57

So uh if you want to know more about me and see some of my exploratory work and how AI can help benefit people, check out my personal portfolio. It's abstractrion.ai. It is 100% hosted for free on Cloudflare, built

Ryan's Portfolio abstractryan.ai

SPEAKER_00 1:10:11

by Claude. Um, we built a whole system around that. And basically, it allows both a human and an AI to interact with my complete portfolio resume experience. And so the notion was to start exploring how AI can be interacted with both by AI to AI and how human can better interact with AI to benefit us. So that's where you can find out a lot more about me. That's where you can also see a lot of my personal projects that I've been building and working through. Um, and then also I spend most of my time on LinkedIn if I'm actively talking with anybody. So DMs are always welcome there. If you have questions or have comments about something you don't agree with or you want to know more about, feel free to reach out. I'm always around to chat.

andy lapteff 1:10:59

I'm on abstractrion.ai right now and I'm looking at person I'm looking at personal art. You know, this is some really compelling, beautiful stuff, man. Is it's is this stuff you painted by hand? Is some of it AI

Ryan's Personal Art

andy lapteff 1:11:12

generated? What am I looking at? Here it's really nice.

SPEAKER_00 1:11:14

What are you looking at? That's a good question. Anything that looks like it's painted was painted by hand. Um, I I feel that is kind of where my line is is art stays human. Yeah, AI does a good job, but humans make it alive. So yeah, I do I do a lot of oil painting right now, oil on canvas. Um, I've been focusing a lot on portrait work. Up until about a year ago, there was a lot of heavy abstract type work. Um, but then I also have, you know, I have a video game and then I have a tool out there to help learn redundancy and failure scenarios in a data center. Um, it's called tidy data center, t idy dash data center.

andy lapteff 1:11:51

Is that on this website or is that somewhere else?

SPEAKER_00 1:11:53

It should be there. Okay. If you search, asked about tidy data center or ask chat GTP to explain it. But yeah, um, it's it's a simple tidy game that helps you relax and learn how to do redundancy.

andy lapteff 1:12:05

Listen, you know how I know it's a good conversation? When we go over an hour and I pay the overage to the editor because I just don't want it to stop. You're one of those people that I could just talk to forever. So I definitely want to have you back on sometime. I really appreciate you coming on for all things Art of NetEnge. Uh, you could check out our website at artofnetworkengineering.com. I would direct you to the resources section. Uh, for anywhere you are in your career journey in tech, whether you are a blue-collar labor person who wants to get into network engineering, or you're a uh professional with, you know, five, 10 years experience who's looking to grow your career. Uh, we've curated a huge list of mostly free resources that you can get in there. And um, to your point, you said you you made your game, you know, your data center game. Uh John Capo and I did a spec-driven development session at Autocon where we created a subneting game and it's hosted up there too. Also, the site, the site was created with an LLM. It is hosted in Cloudflare. The get the repo is public. So, man, I'm right there with you. Uh, for a guy who couldn't do any of this stuff a year or two ago, I'm leveraging these tools, I'm leaning in, we're creating stuff. It's open. People can, you know, uh issue things and uh you know issues in GitHub and pull requests. And I'm just I'm really enamored by this whole software development world that I'm getting pulled into, which has only been enabled for me through this modern tooling. So um, Ryan, thank you so much for coming on. It was great to see you. Uh, thank you so much for watching and listening. And we'll catch you next time on the Art of Network Engineering podcast. Hey folks, if you like what you heard today, please subscribe to our podcast and your favorite podcatcher. You can find us on socials at Art of NetEng, and you can visit Linktree forward slash Art of NetEng for links to all of our content, including the A1 merch store and our virtual community on Discord called It's All About the Journey. You can see our pretty faces on our YouTube channel named the Art of Network Engineering. That's YouTube.com forward slash Art of NetEng. Thanks for listening.