AI Coding Makes Development Easier - So Why Are We Burning Out?

[00:00:00]

Matt: AI coding has come onto the scene, burst onto the scene, and it's a hot topic. Today, we're gonna be talking though about the human element of AI coding. More specifically, we're gonna be talking about the effects of it because it's kind of weird, right? We go to start a project, we're like, "Hey, I'm gonna go start this golf app," or, "I'm gonna go start this little system utility for myself."

And in the past, we'd be googling and checking into things, and we'd engineer it ourselves for the most part, unless a piece of software already existed. But we just kind of prompt it into existence. We more or less say to a- almost a digital employee really, "Hey, I want this thing. This is how I want it."

Maybe there's some back and forth, and then boom, that thing exists. And I know that I'm obviously oversimplifying, but that's largely what is happening, and that's gonna have a kind of a weird human effect. Are we gonna like that? Not like it? Are we gonna burn out 'cause of it? What's gonna happen? So in this episode, we're gonna be talking about AI coding, and we're gonna be [00:01:00] talking about that human effect of AI coding. So if this sounds interesting to you and you wanna support the show, you can go and check us out on that Patreon, leave a review or rating on your podcast app, join us in our Discord server, or share this with your friends. Now, Mike, you're actually the one that wrote up this episode, so I am going to pass it along to you

Mikhail: Absolutely. So, I've kind of been, uh, uh, hinting at this probably for a little while now, uh, on, on the podcast. I know Matt and I have had many conversations with this, and as well I've had many conversations with other developers on this kind of idea that AI has kind of two effects. One is, like, draining you of a lot of things and, like, leading to burnout, and then another one is actually, like, maybe an exciting part, right?

Like, I... There's this rollercoaster of emotions where, like, you'd sometimes do something with AI that you weren't able to do, and you're, like, excited that, "Oh, this is really cool." And then all of a sudden you go down and you're like, "Well, if I can do this, then Jim can do this down the street," and you know, then [00:02:00] it starts to get kind of doom- doomy and then, like, happy.

I don't know. It, it's a really big rollercoaster. And what I've noticed is that now that's starting to perpetuate into the higher levels of developers. It's not just random developers like me talking about it. We're talking about, like, a principal engineer from Cloudflare, right? Like, that's someone that's worked at Cloudflare for a really long time at this point, knows a lot about development.

He's now starting to feel a very similar way. So I'm gonna read this, uh, tweet that, that kind of inspired me to write this episode and, uh, go down this path, because I do wanna talk about it. I think that people listening to this podcast have probably felt this way, so I just wanna give that voice to them, and then I do have some recommendations at the end as well about, like, what I think you should do to at least balance yourself out a little bit.

Because if you keep going down the drain, the AI as a drain, then you're gonna burn out 100%. I feel like that, that's 100% how my mind is working right now, so I need to [00:03:00] myself work on that. So it's almost a therapeutic episode for my- for me. So without fur- further ado, I'm just gonna quickly read this tweet that Dylan Mulroy, again from Cloudflare, uh, I'll add, uh, I have a link in the show notes to the tweet, as well as a quote tweet that I'll talk about in a second too.

Uh, "Y'all, I'm not gonna lie, it's way harder to get joy and satisfaction out of building with AI than it was before. Constantly straddling burnout before far le- being far less immersed in hard problems, constant context switching. I'm tired of the uncertainty of where this is going and how to do it well."

Now, that again, that's a quote from Dylan Murphy. That's from the tweet. And, or sorry, Dylan Mulroy. And I do strongly resonate with this, with this idea of, like, not getting satisfaction out of what you're building especially. I used to sit down, solve a hard problem, and feel a reward at the end when it actually worked.

Now, I would be frustrated in between, and sometimes that frustration was pretty [00:04:00] high. And, like, I would stand up and swear and stuff like that, and yes, that was bad. But I would also have this peak of reward of like, "Oh, I fricking solved that hard problem." And I was able to focus on it for a while, and I was able to solve it, and that led to a, um, kind of like a, a, a peak of endorphins essentially.

What I'm seeing with AI more rapidly is that I'm able to solve the hard problems, and I don't have as much of the frustration element to it. Some- I still sometimes do, honestly, because AI's not perfect, and sometimes I, I approach problems in the wrong way even with AI. But when I solve it, it's not me that's solving it.

In my head, it's always like AI is solving this, right? So I'm always questioning my ability to do anything at this point. So that is really the, the just the, where this person is coming from, where Dylan is coming from is like you're never really satisfied. You're never really, like, happy [00:05:00] with the solution that you've come up with because is it really you that's come up with it?

And that is the thing that always, like, balanced the burnout for me is that satisfaction. And now it's just slowly c- like, slowly gain... The burnout slowly gets there, gets there, gets there, and there's no, like, way to drain it at the end. There's nothing at the end that's really helping me stop it from getting there because it's just always, it's either frustration or nothingness.

Frustration or nothingness

Matt: Well, I, I, I can totally see this. I mean, we, we recently had a, an interview episode with Matt Diamante, while the episode wasn't about this, one of the questions that we discussed very briefly was, if an AI does it, did you really do it? And there is the i- idea that because it was your idea, because you prompted it into existence, obviously you did it. you could argue that point from tech that's pre the AI we have today, which, which is you're [00:06:00] using tools. So I mean, if I take a, WordPress installation and then I put a theme in there and I do a whole bunch of content and changes and things, is that my website? I'd think that it's pretty well undisputed that yeah, that is my website, even though I didn't set up a lot of the visuals 'cause the theme handled that.

I didn't set up the database and things 'cause WordPress more or less handled that But you would still say it's, it's yours. But argument to, even against that is that as a person that goes in and up a blank database and st- tells WordPress where it is and gets WordPress files ready to go and sets up the, does the installer more or less, does the installer and gets everything going, or even if you have, say, a cPanel and you use a one-click installer, you're still logging into the cPanel, getting the hosting set up. You're, you're still very hands-on, you're not just to the server. [00:07:00] You're not just talking to the cPanel, whether that be via text or You're not just talking to it and say, "Hey, go install that. Hey, go do this. Hey, go do that for me." 'Cause, uh, 'cause the AI is, is doing that. That's why I made the parallels with almost like an employee doing it. But what's weird is it's almost like every single employee has an employee. More or less, every single employee has an assistant, and that's strange because it's not like you need an assistant 'cause you were necessarily overworked. people are still in the same position, meaning their same job, that they were before all this AI stuff took off, and the place was run well, you weren't completely overwhelmed. And hopefully that's the case today, but yet now we have, we all have assistants. And it is, it is weird. mean, uh, we're talking about, Mike, you and I are talking about, uh, this, I have a golf app idea. I had one before, but I've, like, kind of refined it [00:08:00] now, and we're talking about how to do it. there's a lot of talk about, while they're, on the business side of things, you want that app to be as good as possible, as quick as possible.

You want it to release as quick as possible. the way to do that is with AI. But then at the same time I'm thinking to myself, "Well, I've never coded in React Native." Let's just say, for example, React Native. We haven't chosen a stack, but we choose React Native. I really done it? Isn't that strange? And it, it, I think it is strange. It i- it is a good question, and it is, uh, a complicated part almost, I guess, of being a human, where you kinda didn't do it. And I've heard one more argument and, and it's my final example before I let you take over. But one other argument I've heard is that while people who are managers, they don't necessarily know how to code, or maybe they do know how to code and they just don't code.

But they are [00:09:00] the, you know, the IT manager, the tech lead, whatever your company structure is, and they guide They guide your company. Like, they guide your team, and through that guidance, an app appears or a web app appears and, and things. And some people would say that that manager is not the creator of that thing. If you're talking about the individual code, sure, but that manager is still putting in the sweat equity. They're managing the people. They're doing the meetings. They're holding back the brass if they're, if it's a company with many tiers of management. They're managing the budgets potentially of each individual team.

They're still in the trenches. They're just doing the, not the HR, but the administrative, the human administrative part of the job, which is required when, especially when the team is large, and even more so when the company's large. Because you don't want the s- the CEO to suddenly be like, "Let's crush that project." Your, your manager, you want that interface where your manager says, [00:10:00] "Hey, this is actually gonna be a huge revenue source, and we're 95% complete. Do you really wanna crush this?" He's advocating for your team. And so manager is still in the trenches doing something, but that manager would probably feel weird instead of talking to the CEO, he had some sort of bot do it for him. That's strange, and we're living in a weird time

Mikhail: I, I agree. I, I, I agree from couple different levels actually. So the manager stuff is, is a good point, and the other part of the manager argument that I've seen people say is like, "Oh, it's like managing people. You get the satisfaction of them doing things." I, I don't see it exactly that way either because when you're managing people and they accomplish something, and especially if something that you mentored them to get to that point, you do feel satisfaction.

You do feel good about that situation. In this sp- specific situation, y- you're mentoring an AI or you're telling an AI what to do. That's not really making it better. That's not [00:11:00] really contributing anything to that person's career, like that AI's career obviously. You're just getting the work done, and it...

With that, you know, great, you're getting work done, but again, it's, it's back to that like what, what is the thing that staves off burnout? And that is receiving some sort of level of satisfaction at the end of some sort of project. And the m- the less of that we have, the faster the burnout kind of builds up.

It's not an immediate thing. That's why I think when you're first starting to use these AI tools, you get excited, and that's why I think like you, you s- you do have like a, a boost in yourself of like, "Oh, look what I could do." But it builds up over time if you're not careful with balancing it out with something else that will lead to a g- you know, more satisfaction.

Uh, and that's how I'm starting to feel right now is like there is this element of you need to understand your own limits on the burnout side, and if you're not careful, it doesn't... AI or not, you could start [00:12:00] reaching those limits and burn out completely. And we've had I don't know how many episodes on burnout at this point.

I think it's an important topic. We've probably talked about it too much on the podcast, but I think we're gonna continue to talk about it because I've seen people literally get burnt out to the point where they bounce out of a career, and I really, really believe this is happening en masse right now, uh, for many different reasons, but one of them being AI, one of them being layoffs, one of the...

Like the other thing. Like I think that a, an exorbitant amount of people are gonna start bouncing out of the coding career partly because of AI and not being able to realize that it's possible to manage, and that's why I wanna work through a little bit, right? Like this isn't gonna be... I'm not gonna have the answer in this episode.

I wanna make s- I wanna make that clear. I'm not gonna be like, "This is exactly how you should do this, and you're gonna be fine." But as long as you can start to realize the situation, I think you can start to figure out your own answers, and I'll give some recommendations. But you really need to understand where your brain is at to be able to start solving it, right?

Like understand that there is a problem to solve the [00:13:00] problem, and that, that's partly where I see The AI burnout conversation happening. I do wanna quickly go into the, like, the positives that I'm seeing as well, like wh- where I do see, like, almost the benefit, like not almost the benefit, but the, the, the high benefit of how to like not have burnout in AI quickly, and then I'll talk a little bit about the drain.

But yeah, you can absolutely interject first. Yeah.

Matt: a- ask a question? So like

Mikhail: Mm-hmm.

Matt: the reason why you had those highs, uh, when you're like, "Hey, like, you know, I, I've solved this difficult problem," and you had that like high like, "Yeah, this is amazing. I'm doing, doing, I'm doing so good, you know, I've solved this issue." Part of the reason why you had that sort of like uptick in mood is because there was a struggle, and you mentioned the struggle without saying

Mikhail: Yep

Matt: You mentioned the struggle. Do you think, and I know we're gonna get into it, but just as a sort of almost like preface before we get into the specifics, do you think that [00:14:00] the burnout is happening just over time, and that don't have any relief 'cause there's no struggle? there a relationship there at all between like what I'm saying and you're thinking? I- is, is more or less my question. Like, is it because we don't have the struggle period, we just have the droning on in the office?

Mikhail: Well, I, I think, I think partly yes. Like, I think the struggle helped the high be higher, right? Like the struggle when I completed something that was a struggle, I got a, a big boost and a big drop in burnout. Let's say that, right? Right now, there's less struggle, and there's less, like, friction points when you're building something.

But there is sometimes a struggle. But the problem is that even with that struggle, that doesn't lead to any sort of high. Like, that doesn't lead to any boost for me. Because at the end of the day, even if I struggle through and I figure something out with AI and, like, go back and forth with it, and I solve it, I still don't feel the [00:15:00] sense of accomplishment of solving that the same way I did before.

So there's less struggle now, right? And that is good, but it's also bad because there is no relief at the end, at the end of the struggle. There is nothing that stops it. So that, that is absolutely a time thing where it just builds up and builds up and builds up. And the more you do that, and the more you go down that path and, like, not be able to find a way to have a sense of accomplishment from the tasks that you're completing, the worse it's gonna get.

Absolutely. Um, and there-- Yeah. For, for me, for a while there, like, I didn't have that sense. Like I, I just like, you know, went, went into my cl- Claude Code, figured stuff out in Claude Code and went about my day, and I was like, "Oh, this is cool." Like, one of the things that I have here as AI as a relief is it does make the job easier, right?

Like, that's how I felt initially, right? Like I, I did feel that, and it is true. It hap-- 100% makes the job of a developer easier. [00:16:00] There is-- You can't really refute that. Uh, if you're, if you're a developer, you know what I'm talking about, like being able to not have to write a bunch of, you know, troubleshooting code or not go, have to go down and, you know, fill out an entire database with seeded data and, like, not have to do a bunch of scaffolding and not have to, like...

There's a lot of things that we don't have to do anymore because of AI, and that makes the job easier. And in some ways, that could be seen as like, "Hey, maybe we can focus on the harder problems now." Right? Like that's what we, what people have always said is like, as soon as we, you know, remove a layer of busy work, we can start using that capital to focus on harder and harder, harder problems.

But we're still needing to do that busy work. Let me be clear, like we still need to input that into AI, wait for the AI to finish or, you know, start another thread, have that AI do that thing in the other thread, and have to go to another thread to work on it. So like it's still being done, it's just not being done by us anymore, for the most part, and we're still, [00:17:00] like reviewing it as well.

But it does make, in general, work easier for us as developers. There's no-- I, I don't think you can really re, re- refute that.

Matt: Well,

Mikhail: Yeah. Mm-hmm

Matt: myself, I mean, uh, w- I, I don't have necessarily the AI burnout right now 'cause I'm still doing a lot of those things, setting up the WordPresses and setting up the CMS and things. But 100%, if I, if there's something new to me or something that I've only done two, three times, handful of times in my career, the AI is able to either do it for me or I'll tell it, "Hey, I want a quick refresher on this," and it's able to do that, and then I'll give it the context of what the specific problem I'm having, and then it'll solve it. But I'm still getting the satisfaction of, okay, good, that, that WordPress is working or whatever. But there aren't, those aren't as h- like, those aren't as high highs. You know what I mean? It's still like, I mean, how many times have I set up [00:18:00] WordPress? Like sure, it's like, yeah, an- a- another one is up and running, you know?

It's not that crazy. Um, but I mean, there's more problems than that. I'll, I'll have like a snippet of code that I can't get working inside of a CMS, and it's, you know, trying to interface it back and forth, and then I'll, I'll get those working, and AI is generally not that great at doing that still 'cause it gets all the version numbers wrong itself, even if you tell it, in my experience. But 100%, like the, the very, very difficult little plugins, let's say, that I make for WordPress that aren't, aren't on the marketplace or anything, just for individual clients, uh, I don't get as much satisfaction out of making them anymore

Mikhail: Mm-hmm. And one thing I, I, uh, one-- another kind of positive thing on that note, actually, is that, like, I, I will still get satisfaction from figuring in, figuring out a new AI workflow, and let me explain a little bit. Like, for example, like you were just saying, like, uh, you know, the WordPress plugin. The first time I figure, I would figure [00:19:00] out how to create that WordPress plugin and have it all automated with AI and have it work, I feel like that still gives me some, like a, a good rush, and, like, I, I still feel pretty happy about that because, like, oh, cool, a whole new unlock for me, right?

Like, I can do all this with AI now. The problem is, is that the second time I do it, I get nothing because I've already figured it out and it's great, and you, you kind of... Again, it's o-only like a battle within your own head where you're oversimplifying the fact that you figured out that part, and now you're like, "Oh, everyone can do it, and so this, this no longer matters."

And that's how it fee- And that, that's why I think, like, this builds over time with AI use because I started figuring out workflows and that was cool, and initially I was super excited because I'm figuring out all these cool different things. Like, I figured out recently how to, like, set up the SSH connection so that the, uh, so that the AI can SSH into a new server.

So I'll, like, spin up a DigitalOcean droplet and have the AI SSH into it and then set up everything other than, like, DNS. Also it'll set up DNS. And then the AI can literally work in the DigitalOcean droplet and set up an entire [00:20:00] Docker stack of, like, a full stack application in there with the database and everything, and now it's like an isolated, you know, netw-networked, uh, container for me to work with that's fairly safe and then I'm, I'm also able to access on the internet.

And, like, I found that workflow super cool. I'm able to do really complex things with it, and now I did it again recently and I'm like, "Oh, this, uh, it lost its appeal." Like, it's great. Like, I can do it, and now, now I can use that and I do-- I have used it probably, like, I don't know, six times since I figured it out.

But I really only got satisfaction from figuring out the first time 'cause the next times felt automated to me, right? And I just, I found that that stuff is happening less and less because I figured out the workflows already. Like, I figured out what works for my specific applications that I'm building.

So mostly, like, you know, the skills that I use already, the, the, you know, the prompts that I'm prompting, all of them have been the same for the last, like, two months. Like, it's all worked and it, it works fine, and that's why I'm not changing it that much. And so, like, for two months the, [00:21:00] the features that I'm developing, I have no real interest in them, like, other than the fact that I need to complete them.

Like, they're just being done.

Matt: the journey is so short

Mikhail: Yeah

Matt: Before, I mean, if you're, if you're setting up a, a complex WordPress setup, like a big WordPress stack, you're manually doing that, even though it might take, let's say, an hour or two hours if you're experienced, you're still in it.

You're still effectively kind of working with your hands, so to speak, for two hours. Whereas how long has it taken you to, to do that work- like that workflow, Mike? I mean, y- you don't... I- if the bot takes 15 minutes but your prompt was five minutes, your journey's five minutes, and then you're

Mikhail: Correct. Yeah. You're either waiting or you're context switching. And like, so you're waiting, y- which is even worse, right? Like, like for, especially for me, like I'm, I'm a focused person. Like I like to work on one thing at a time. But with a lot of this stuff, like you're saying, I, I do context switch. Like I'll work on three different things at a time.

That's, that's my max. I can't do these like 15, 20, [00:22:00] 30 agent loops that some people do. I can do three. And even then, that's stretching me too far, and that, that leads to burnout as well, is like I'm, I'm trying to balance these things and like going back into this one that finished 15 minutes ago and going to this one.

Like that's not my preferable way of working, but I'm also not in the sense of like I'm just gonna twiddle my thumbs for 15 minutes while this agent is doing its thing. So like that combination also kinda sucks. And like I know I was talking about the relief side of things, but that's definitely a drain.

The context switching sucks. Um, because even if you're just prompting a, a, an agent to do something, if the, if what you're prompting is very different, you have to change the wording, you have to test it. A big thing with, uh, with this AI stuff that I've learned is the testing side is almost more important than reviewing, right?

Like if you're, if you're gonna be, if you're gonna be, uh, like trusting AI to write code, then you need to test the code that it writes. You can't just review it. You need to see that it's actually doing the things that it needs to [00:23:00] do. And you can get automated, like I do automated testing a lot, but like you also need to verify the automated tests.

So having to like even context switch between the testing that you have to do is a drain at, on its own, and then having to write the c- the prompts, and then having to follow what's going on and make sure that the reasoning is correct. Like It's not-- It seems pedantic. I think, I think it does, it does seem like it's like I'm complaining about little things.

But the problem is, is that the little things add up, and over time, without being able to have any satisfaction from the work that you're doing, this does 100% lead to burnout. And that's what I, that's what I'm personally seeing right now, uh, is that, like, I've invested a ton of time learning this stuff, right?

Like, I've learned a lot of the skills. Again, I've learned a lot of the workflows. I've learned a lot of the tools. I'm using Cursor, I'm using Pi, I'm using, uh, Claude Code, I'm using Codex, I'm using OpenCode. Like I- I'm using enough freaking tools. It's enough [00:24:00] tools. And like, usually when I-- Y-

Matt: many. Ha

Mikhail: you know, I'm, I'm for sure using too many, but I'm almost chasing the satisfaction of learning something new, of learning a new tool.

But even that has now collapsed for me because, like, all these tools are almost the same. And so, like, when you pick up one, you pick up another, and it's like, okay, so it's-- now it's the same thing. Like, I figured it out once, and the ne- the, the other times I don't get any satisfaction. So like- It's it's an intrinsic problem with making stuff too accessible, and it sounds like I'm complaining about something that I shouldn't be complaining about, but I really, really think that this is re- this is resonating now with a ton of people.

Like, the reason that I talked about this even from the get-go is that that tweet that I just-- that I talked about the first, uh, in the first part of the show got, like, over 3,000 likes, and that tweet's targeted directly at developers. So we're-- I'm not the only one feeling this, and th- I'm not some... Like, I'm, I'm a scrub.

Like, I, I, you know, I'm not a principal engineer at Cloudflare, let me be clear. [00:25:00] If a principal engineer at Cloudflare is starting to feel this, then I know for sure that this is starting to affect a very large part of the industry from all the way to the top to all the way to the bottom, and we need to figure out a way to somehow balance this out, otherwise there's gonna be a massive loss of talent.

Um, which maybe is what companies want or, like, what, what the natural, the natural way of things in this particular case, but, like, I don't wanna be part of that lo-loss, if that makes sense.

Matt: Well, I was gonna say, like, there's many skills that are now antiquated,

Mikhail: Yeah

Matt: on the table is coding. The jury's still

Mikhail: Mm-hmm.

Matt: but is one of those skills, and it probably won't be something that just goes away overnight. probably won't go away super quick, but it is on the... It is on trial really, uh, as, as to is this gonna be one of these skills that is, that is old and antiquated and, and isn't necessarily needed anymore. [00:26:00] However, w- it's very easy to forget that there are many segments to markets. gonna be people that hand code stuff and will always hand code stuff. I mean, heck, we have products that can be made en masse, and when I say products, I mean everything ranging from, like, wallets to game consoles that are not handmade, and we can mass produce them and, like, maybe there's handmade parts, but many of it's automated, it's assembly line. not considered handmade. But then we have artisan things where, like, people who w- like leather working, that's, like, a skill that not a lot of people have anymore, but it's respected and people buy handmade, like, leather wallets and belts and things because it's a sought-after skill. it doesn't necessarily mean that that skill's gonna completely disappear.

If coding does go the way of, you know, the dinosaur, it's not gonna be completely the way of the dinosaur. [00:27:00] There's still gonna be something where it'll... You know, this is hand, this is hand coded or this is something along those lines. There's so many different segments to a market

Mikhail: Yeah, I, the only way, the only thing about that argument is, like, it's gonna be... It might be such a small subsection eventually. Not right now, obviously. Right now we're still in the transitionary period, but tr- eventually it might go the way of, like, the dinosaur or the Morse code, for example. Like, Morse code is still learned to some degree, but it's not really a useful skill.

It's just something that, like, is a bygone product. People are interested in it, and people still learn it on the side, right? It's not something that's necessary for a lot of different... I, like, I don't, I don't think that ships use Morse code to communicate anymore. I'm pretty sure that that's now not part of the process, even, like, even, like, a, as a fallback.

I'm not 100% sure on that,

Matt: idea how

Mikhail: but I'm pretty sure that that's not used. But I know that, I know that there are Morse code classes. Like, you can still take Morse code, [00:28:00] and I think that's, that's how coding will be, is like, there'll still be, like, coding and traditional coding classes, and there's still gonna be people that learn how to code and maybe even enjoy it.

But during this transitionary period, it just feels so daunting that, like, it's almo- like, when I go to hand-code, which I still do on occasion, very o- very rarely, I almost feel like I'm wasting my time. Well, I, n- sorry, not almost. I do feel like I'm wasting my time, and that's another part. It's another part of the burnout for me is, like, stuff that I was doing six months ago that seemed very efficient and very, uh, very productive is now no longer productive.

So I think that rapid switch, right? Like, if we're ta- talking about this from a, like, a point in time perspective, like right now, I think the rapid switch about the skills that you had are no longer really needed, and now you're learning new skills all of a sudden, and it's like it happened overnight, it seems like.

[00:29:00] That definitely leads to this kind of drain on your brain as well. So there's so many factors working against us as developers. Like, f- first of all, you have to kind of embrace it. You have to embrace the coding or, or the AI stuff no matter what because companies are starting to... Like, CEOs are starting to get AI pilled or not it's even starting.

They're start- they're, they're really AI pilled at this point. Um, not everyone, but, like, a very large chunk.

Matt: We have people that comment on our clips that are

Mikhail: Yeah

Matt: saying that, you know, there, there's, like, a large subset of people that are not

Mikhail: Yeah.

Matt: pilled, CEOs or otherwise

Mikhail: I, I think yes. Th- th- there's definitely, there's definitely going to be exceptions, and I think there's a, there's still a large chunk of companies that haven't embraced it in, in a large way. But regardless, in the zeitgeist and the ones that I'm seeing around me, like my own perspective of, you know, maybe seven or eight different companies that my friends work at, all of them, like literally all of them, are [00:30:00] AI-pilled.

All the CEOs have gone fully AI crazy, and like this is like I- AI psychosis, whatever you wanna call it. I don't know if they're all in the same group chats, but it, it seems like if I talk to this guy that works at company A and this guy that works at company C that have nothing to do with each other, have the exact same messaging.

So like they've-- It's either maybe the CEOs are actually talking to ChatGPT and getting the same response. I thought maybe that as a re- as a reasoning, or they are some sort of a secret group chat for CEOs. But it's just a weird message coming down that is very coherent. V- Like it, it's sa- the same across the board, where like you're either in or you're out kinda thing.

Like you either use AI and only AI or you're no longer part of this company's future. Like that's the messaging that people are getting. Um, so that in itself is a scary thought, right? Like n- not a great, not a great feeling, and then all the other stuff that comes with that of like product managers putting PRs in and, uh, [00:31:00] CEOs putting PRs into your code base that you were so like guarded against for so long.

Now you're having to like lift that restriction It's just kind of a losing, it's a losing battle. It's a losing battle on, on that side. So you, we need to learn some ways to figure out how to benefit from this for ourselves. One being make the work easier, right? Like that is a thing and that is a good thing.

Like you're not having to solve these ridiculous problems sometimes, you let AI solve them, right? Like you need to find ways of alleviating this pressure

Matt: Well, when you, when you were saying though that all the CEOs are in a group chat, I mean, I, I'm sure su- some of them are within the same circles, especially if they, they attend the same conferences and things like that. But

Mikhail: Maybe

Matt: think, I think one thing that they all share in common is they're in the same segment of business.

And what I mean by that is segment of the internal business, which is they're trying to get most amount of productivity [00:32:00] through their company to make money. Just like how we mentioned that the manager is doing the meetings and is the human interface to the higher-ups and things like that, th- this, like CEOs are in that same business segment within their business, and that goes for people in tech, people in vacuum cleaners, right?

Everywhere. Those CEOs are trying to make the best vacuum cleaners, the cheapest they can, have the best profit margins. The same, the same goals apply clothing manufacturers, where they're trying to sell the best clothes at the best prices with the best profit margins. The same goes for anything, is my point.

Is it's... That's a commonality across pretty well every single CEO, short of maybe not-for-profits and certain other niche cases, artisan companies that are there, designer companies even to an extent, where they're there for the name and they have, like, a, a, a reputation to uphold kind of thing. Um, there's a [00:33:00] fly buzzing around as I'm trying to talk.

Good Lord.

Mikhail: Of course. Yeah

Matt: Uh, that's, that's, like, where their commonality is, and so you are gonna have some of those, those common conclusions, and that's why it kind of seems like they're in a group chat, 'cause they're not in a group chat, but they're in the same... Like, one CEO of a scooter company have probably a bunch to say, have a great conversation with the CEO like, a sneaker company,

Mikhail: Mm-hmm.

Matt: have a great conversation with the CEO of a person that makes, like, Game Boys or something.

You know what I mean? Like, they can all, they can all chat if they were in the same room 'cause they're all in the same segments of their businesses. They're all in the same sort of position with the same goals. Their game is the same, or

Mikhail: Yep

Matt: at the very least

Mikhail: Yeah. And I, I think, I think the, the end, the end point of that discussion is, like, we can't change their game. We can't go in and we can't really affect it. So what we have to do is play the game [00:34:00] that they're, that they're telling us to play and find ways to make it work for us. One of those ways, again, make the work easier.

Some other ways that I, that I found is, hey, AI is a pretty good tool, um Try to use it for something that you, you want to use it for. Like, something that w- has worked for me recently is I've been wanting to make these little games. Like, I've always, I've always been kind of like a, you know, I, I'm a gamer at heart.

Like, I don't play that much anymore, but I've always, I've always played games ever since I was like four, three or four. I had a Sega Genesis. And like, it o- it's always intrigued me making games. Matt and I made a game, uh, with an artist one time, and that was probably my best experience in programming still.

Like, this was like, I don't know, a decade ago now. Um, that was the most fun I've had in programming. So I know that that's like something that I've always wanted to do. I don't think I, I ever considered it as like a career path because I just don't particularly like the industry that much maybe, like the game making [00:35:00] industry.

I've always looked at it as like a, a grind and like that's what it's always seemed to me. Um, but I've always wanted to make games. So AI comes around and hey, maybe I can try it, try it out. And I did. I made like this little elevator game, uh, that like super random. Uh, but I just thought like, hey, those old elevator operators had to press the button to go to a specific floor, and they were controlling the algorithm, right?

Like, there was no algorithm. They were the algorithm, and they could do it efficiently or they could do it inefficiently, but like realistically they just have to get the person from point A to point B. And I made a game around that concept of like you're that elevator operator and you have to do it as efficiently as possible, and the more people you get from point A to point B without losing patience wins.

So random little thing, but

Matt: the way

Mikhail: yeah, it's like like i- I thought it was a pretty interesting game. Like, I didn't think it would actually work out the way it did, but like it did. And I would not have done that probably, uh, without AI. And that's the framing that I'm putting into my own [00:36:00] head, and I did really enjoy the process of making that game.

It was a new thing. Uh, you know, I made a game des- uh, like a GDD, a game design document similar to what Matt and I did for making the game before. I did a bunch of play loop testing before I even worked on graphics or anything like that. You know, I, I thought of the music that I would want for the game and got the music in the game.

But most of it was, was done by AI, right? Like most of the actual implementation of what I'm saying was done by AI. I was just kind of the manager of the situation and the play tester and the designer of the, uh, uh, of the, of the game, and I got satisfaction out of it. So like I got-- I felt good. It was a nice little project, and I got a little bit of my meter refilled.

And that's where, again, part of the episode came to me is like we have to find that way of doing it. Like we have to find something. It might not be a game for you. It might be music. Like you might be obsessed with music. Maybe you can use AI to help you learn guitar or help you make music or whatever, like and that could be your thing.

Or [00:37:00] try to find a way to benefit from this and not have it only be a doom- a doom machine, right? Like if you're, if you hate your work but you have to use AI, there's ways you could get AI to do things that could make you like it. That's really the point of that conversation

Matt: Well, I, I mean, like it's good to have... 'Cause I know we were, we've mentioned a couple times we wanna talk about some positives, and there's a lot of negatives, unfortunately, with this AI stuff. Or maybe, maybe, I don't e- I don't- maybe the scale is more balanced. Maybe there's just as many pros as there are cons, but we're

Mikhail: Mm-hmm.

Matt: the negatives right now, having sat with AI for quite a while. what's, what sucks, like not to bring it back to negative, but it's great that like you built that elevator game in that, and like it works great. The game is cool. You know, it's cool. But like if you go to patch that, are you getting the satisfaction, or are you more frustrated at the AI being like, "No, no, like patch it like this," and it's not gonna quite [00:38:00] do it the way you want?

Or like, because you've kinda solved the elevator game, your own sort of comments earlier in the episode, you mentioned that once you've solved something, the next time you go through it, it's not as satisfying. Do you see that happening even with the elevator game? Where if I, if I go to you and I say, "You know what? It'd be cool to have themes where it's not an old-timey elevator, it's a space elevator or some..." You know, whatever. Or

Mikhail: Yeah

Matt: what the heck they have that, it's not an elevator, but it's the thing in Star Trek where you get into like the elevator cart and you go anywhere in the ship.

Mikhail: Mm-hmm.

Matt: what if you work through the ages where you like went from the original, like old school 1920s elevator operator all the way through the whatever the future elevator is from Star Trek."

I can't remember the name of it now. like would that excite you, or would you go like, "Oh, great, now I gotta prompt this AI"? Because the reason why I ask this is 'cause of the comments you made earlier, but also one other thing. I think you've mentioned this on a tweet though, or on X or [00:39:00] whatever you want to...

An X post, uh, where- When, you start becoming frustrated at the speed of your work when the AI isn't acting as quickly as possible. Because, because we've sped up work so much, when the AI does hit a hitch, when you do have to go in there and manually do something, I don't know if it's entitlement, I don't know if that's, uh, the right word to use, but it's almost like we feel entitled we feel frustrated at the fact that now the work has slowed down. And so, like, we're not inter- you know, we're not going as fast as possible because when we make a new concept like your elevator game, coming up with a whole new concept and a whole new working game and everything, at least to me, feels like a big jump in progress, even if it, like, line per line code is not. It's like, whoa, like, I made this cool new game. You know what I mean? But then when you get into the weeds and someone's like, "I want..." Go back to websites. "I want my slider to actually push fade in. I don't [00:40:00] want it to just fade in." And you're like, "Push..." You know, go to the ChatGPT and you're like, "Push fade."

And it's like, "Okay." And then it pushes the whole page down and it's, you know, messing around, and you're like, "No." And then you don't wanna do it 'cause it's gonna take you four hours. You think that you could get it done in an hour with the ChatGPT 'cause the whole slider only took 30 minutes, there's that, you know?

Mikhail: I, yeah, I, I see what you're saying, and the reality is I don't have a good answer for that yet because this is a new thing for me. Like, this is literally, like, you know, a week ago thing that I created this game. And I, I have done some patches and they haven't bothered me, um, but I haven't made a second game.

So I'm, I'm worried that maybe that second game is gonna have that same feeling that the second, you know, app that, uh, learning a second app is gonna have. Now, maybe not, because I am starting to get excited about another game that I'm, like, thinking of, and now that I have this one under my belt, it's kind of like almost, you [00:41:00] know, snowballing into excitement for me.

Um, but u- until I actually do it and get down in the nitty-gritty, like, I won't know 'cause it's too early. I, I just... I do think that you have to find that passion thing. And the other part of the passion thing, and something that maybe is an indicator for me as well, is like you have to find the people that use these things to enjoy them as well.

So I've been having trouble, like, w- 'cause a lot of the stuff that I do at my, like, uh, job is a lot of internal things and, like, beta things, and, like, I just don't have a lot of feedback from them, so I'm just making stuff into the void.

Matt: and stuff.

Mikhail: Correct.

Matt: stuff behind closed doors

Mikhail: I'm doing a lot of stuff behind closed doors, and a lot of that stuff remains behind closed doors other than used by, like, handfuls of people.

And, like, I get some positive feedback, you get some negative feedback, and that's fine. Like, it is what it is. But I, I don't have, like, a ton of passion for that because there's just not a lot of people using it. If the game takes off, I feel like that would change things for me. Like if, if, you know, if 10 people play the game [00:42:00] and all of them like it and give me these ideas, and that's kind of what's been happening for the most part, I, I do get some satisfaction from that of being like, "Okay, I was right on this one."

Um, so just having something that has some impact does help, uh, even if you're using AI to create it. Like, it, it, it's especially difficult when you're using AI to create things that no one uses, I think. Or, like, it's just an intranet thing or, like, a small... Like, you're, you're just solving tickets. I think that's where a lot of the burnout will come from.

But if you're making something and all of a sudden it becomes good and people use it, I do think that that would solve some of the issues. Um, I just, yeah, I don't have a lot of experience with that yet. Obviously, the game had the little, like, you know, a few people playing it and that's it, but yeah. It's...

I don't know.

Matt: I, I

Mikhail: TLDR, Matt, I don't know

Matt: fun

Mikhail: It, it, yeah, I, I will release it, actually. I, I have most everything done. It's the-- By the time this blog, this episode is out, there will be a blog post on the HTML Things website. I have it ready to go, [00:43:00] um, with the, with, like, talking about why I created the game and stuff like that, and, and it'll have a link to the game as well.

And I, I have it hosted and stuff like that. So, like, it's out there right now. I just haven't really socially put it out there, so I'll, I'll post a link on X. It will be released, I have no doubt in my mind. I-- It's a really small game, so I'm not expecting anything from that. Um, just more like, "Hey, I, I'm hoping that, you know, five people play it and enjoy it," because now I'm thinking about the next game.

Because this, this was more for me figuring out, like, is it possible to make small games? Yes. Can I have fun making those, them? Yes. Can I have fun playing them? Yes. And boom, now I've found my thing that can potentially bring me back from burnout. That, that was really the goal of this game, because I was feeling...

Like Matt knows, I've been feeling like shit for a little while now, uh, for multiple different reasons, um, AI being one of them. But, like, yeah, I needed something that would pull me out of it, and that's why, again, [00:44:00] I implore people out there in the same situation to actively go out of their way to find that thing.

'Cause if you're not gonna do that, I know for myself, I would just go down the rabbit hole of, like, "I don't wanna be a developer anymore."

Matt: and

Mikhail: And I

Matt: Yeah

Mikhail: would def- Like, I was going down, down that route. Like, I was, I was legitimately going down that route. I'm a little bit better now. We'll see how, how it goes, but I'm a little bit better now having found this, and I'm hoping others also find something

Matt: That's good. And, and man, I, I don't know if you saw me smirk when you were talking about that negative thing, but the very first thing I wanted to say was, "And you don't wanna know the other reason why he's been down in the dirt," and just leave it, just let that sit, is the first thing that came to my mind.

I was like, hang on, I'm just gonna... It's not gonna do that. Let's just not.

Mikhail: Yeah.

Matt: You don't wanna know the other shit that's

Mikhail: Yeah.

Matt: on. This is, uh, this, that in two weeks we could have a Mike exclusive.

Mikhail: I mean, we might, we might. We'll see. Stay tuned for a Mike exclusive. Um, but yeah, e- everyone's going through their [00:45:00] shit, though. That's the thing, like it's not just me, it's not just you. Like, everyone has their own shit in the background, and you hear people bubbling it up sometimes, and that's, that's natural.

But like, everyone has their own shit, especially right now in my opinion. Like, there's no ideal life out there. And that, like even accepting that is like, is a good thing, like you're not the only one suffering obviously. Um, but yeah, no, it's... This game has been cool. Uh, the other thing that I want, before we end this episode, 'cause we're, we're like, I think I'm at the end of the show notes here.

And there will be some more stuff in the show notes by the way, so just check them out on htmlallthethings.com, and I'll probably have a link to this game or maybe like a side link to the blog post, um, about the game. Uh, but before, before I wrap up my stuff here, one other thing that's helped me, um, this is maybe a controversial one, I even have a star beside it, is trusting AI more.

And I can explain that a little bit. So what I realized was really, really draining me [00:46:00] was reviewing every single line of code. That, I don't know if, like the ratio of reviewing to writing changes as you progress in your career. It always goes up for reviewing and less for writing. That's just how it is for an engineer for the most part.

As you go from junior to mid to senior to principal, you're, you're reviewing more code. With AI on top of that, pretty much all you're doing is reviewing code, and reviewing code is I, I haven't found anyone that loves it, but it's super draining because every line could have like implications and you're trying to figure out the connection between one line and another line.

It is super draining and it's super boring. Um, so I figured out workflows where I don't have to review every line of code and that again, that's partly why I think this is gonna be controversial 'cause there's a whole actual fight. Uh, and I was thinking of doing an episode on this. We might do a web news on it, but there's a whole fight going on right now about what kind-- like [00:47:00] when should we should stop reviewing code.

Like the... I think Theo posted a, a, a message about like how, how, where do you think we are about reviewing code? And the fight broke out of like, how could you possibly think to not review code? And then the other side is like, why are you even wasting your time reviewing code? So you definitely have to review code.

I'm not saying don't review code, but figure out the critical parts, review them, and then find ways to review less code. For example, a lot of times I'll rely on testing like te- writing, writing tests and reviewing tests over reviewing code. If the tests pass in the right way, both security tests and functionality tests, I know that the middle part, the stuff that's happening is working the way it's intended.

Is it working as efficiently as it possibly could? Probably not. Accepting that is another part of it, right? Like accepting that this code might not be the best code you've ever written or you could write. As long as it's doing input output correctly, you know [00:48:00] that you can rely on this code if you have the right test and then manually testing.

So I find myself, I like to manually test things regardless, but it's definitely a little bit easier for me than reviewing code to manually test all the workflows. Like almost putting your mind into a QA aspect versus a reviewer aspect has helped me as well, right? Like just finding that workflow that worked for me for reviewing less code and not being as draining 'cause just, I can't review thousands of lines of code every day.

I can't, I can't do it. Like it's just not, it's not possible. And, and if someone's doing it out there, hats off to you. Great. Awesome for you. Um, but like I just can't... That's not my life. I can't do that. I can review a couple pull requests a day with like a few hundred lines of code and get some good, you know, responses.

But the amount of code that we're writing guy, like it's outrageous. You can't review it all. It's not possible efficiently, at least. Like you can skim it, but you can't review it all. So

Matt: I, I

Mikhail: that's the last point there

Matt: to code [00:49:00] reviews, but I used to do, uh, peer reviews on router configurations, and if y- so if you've never worked on, like, enterprise-level routers, you- you're more or less scripting it. Like, you're going in and you're writing, like, not lines of code, but lines of configuration. Like, open this port, make that port a VLAN, make that port have this protocol, do this, do this. And it can get very, very complex, especially in an existing environment where you're changing something big and you need the traffic to keep flowing. And so we would peer review each other's configuration changes. Like, "Hey, I have a ticket coming up this Saturday. Can you please review it?" Send it to the team on Wednesday, and people peer review it.

And I'm sure that some IT teams have processes in place that where maybe two people need to approve it or something like that. And that also sucked. Like, going through line by line, again, not of code, but of configuration, especially when it gets [00:50:00] convoluted, especially when you don't have full context 'cause you might be working on a ticket in a completely different, completely different data center, completely different networking closet, and you're like, "Wait, what is all this?" You gotta go in, you gotta kinda like pull on your own theory, see what's going on in that building, check out the network diagrams, you know, et cetera, et cetera. You know, and it, it is difficult. And there is a little bit of acknowledgement that every single line is gonna be reviewed by everybody in a very complex configuration.

In a utopian world, in a perfect world, in a f- perfectly responsible world, every line would be accounted for and checked, and everything would be great. You know, w- like every peer review would go amazing, or every peer review would find issues and then have those fixed, right? That's the utopian view. But if you notice, I said Saturday is when the configuration changes, and that's because there's a certain element of pushing [00:51:00] these changes where we go, we are gonna peer review it and possibly find the big issues, but there's an acknowledgement that there may be problems, and we're gonna take a responsible step by putting this configuration in place during a date and time that has the least amount of traffic so that if something goes awry, we are able to roll it back. There's a certain reliance on backups and things like that. again, this is not the responsible practice. Like, if you were learning this in school, they would be like, "Never do this, never do this." But in the real world, when deadlines are there, when you need to get things done by a certain time, uh, while you're also working on something else, like you might be multitasking on top of the deadline Immediately re- like some realism kicks in where you go, "I'm gonna review the big parts of this, but I may not be able to review absolutely everything," and we're taking responsibility.

We're having... Making sure our backups are in place, and we're doing this at a good time. And I, like, while those things are always [00:52:00] great to have, it is almost, almost a little bit of an acknowledgement of, hey, not everything is gonna be peer reviewed to the nth degree because it's gonna slow us down too quick or too much. not gonna be able to hit those deadlines. We're gonna be in the same context for way too long and wasting time most of the time, you know, a quick check over is gonna be like, "Hey, port two, not port three." You know, that'll fix it. Whereas in the, in the, in the moment, you might be like, "Wait, something went down.

Oh, shoot, it was port two, not port three," and you might just be able to fix it. So not the most responsible way to do it, but again, real world, when timeline and money and everything else comes into play, it is gonna... It... You know, you're not gonna be able to code review everything. You're not gonna be able to configuration peer review everything.

It's just, it's just not realistic part of the job, unfortunately, for, for many companies. If it's an extremely well-managed utopian company somehow, then sure, you review absolutely everything, and the person has to be alert, [00:53:00] and the person has to be awake, and they have to be invested in it. You're just not gonna get all those things every single

Mikhail: Well,

Matt: something

Mikhail: it's a massive trade-off, right? Like if you, if, if your intention is to review every line, then you have to push back releases by X amount of time.

Matt: Agreed. Agreed

Mikhail: Like, and then you have to play to the person's capabilities, right? Like again, a, a good p- pull request is usually between 300 and 600 lines of code. I would say 300, like l- closer to 300 so that, uh, the reviewer actually has a chance to catch things.

'Cause after that we lose context. We start to lose context very quickly, and the re- like we're not writing these like succinct little PRs, so if we want that to be the case, then we have to spend more time writing succinct PRs, smaller PRs, have more reviews, and be okay with the time trade-off. And in some industries, you're talking like, you know, healthcare and stuff like that, and life critical s- s- situations, yeah, you, you deal with the trade-offs.

You [00:54:00] deal with the time sinks and like the, "Okay, I can't review all this today. I'll have to wait for tomorrow because I'm drained." And be okay with that response. Like if I said that, if I, you know, if that was said in a small s- tech startup, people would be like, "What the fuck are you doing? Get, get that code in production right now."

That's literally how it, it, it's treated right now, and that's bad. Let me be clear, like that's not a great, that's not a great feeling, but that's

Matt: We see ma- uh, we rely on

Mikhail: Yeah,

Matt: just

Mikhail: break.

Matt: grinded to a halt by the fact

Mikhail: Yeah

Matt: on top of another one, and even going into the console, 'cause we know what we're doing, and like display none-ing the thing on top. Oh, now that back thing's unenabled 'cause there, there's actually more to the story there.

Something so simple as a button being disabled inappropriately can just screw everything over, and that can so easily happen when you're not reviewing code

Mikhail: Yep. That's the thing. Like, so it's just, it's just the reality of the situation. You have to learn ways to optimize that without having to review every line of code, and it sucks. It does. It means that code's gonna break and, uh, but [00:55:00] if you can at least save your sanity and still have decent code Then do that.

And that helps. That does help. Reviewing less code helps my burnout. Period.

Matt: Fair enough.

Mikhail: Yep

Matt: and creating. you're creating things, putting out into the world. That's a, I think that's a Tony Robbins line loosely.

Mikhail: Yeah. I think that does help. Yeah, for, for sure. Like, actually creating something and putting it out and getting feedback is-- it, it feels good. Highly recommend it. Try to find your own thing that works for you. Again, uh, it's gonna be different. And, uh, I don't think these two things are gonna be enough for me either.

I'm gonna need to figure out more ways because it's helping, but I know that I, at the back of my head, I still feel it. So I'll-- Whatever I find, I will talk about, I will post. Uh, so yeah, you can hit me up on the socials at, @htmeleverything on X, @htmlthings on all the other ones. Um, and yeah, we'll, uh, we'll keep you updated.

If you have something as well that y- works for you for [00:56:00] keeping the burnout at bay, please let us know, because that's what we're looking for right now.

Matt: 100%. And, uh, also let us know what you think, you know. Let us know what you think about the, the, the peer review thing, let us know what you think about the whole AI coding thing and the human element of that. Uh, what do you think about all this? I notice a lot of people are very against AI, but some people are very for it, some people are skeptic. It's all over the place, so please let us know in the comments on Spotify or, uh, wherever y- you want on the various social platforms. I mean, we got... We, we post a, a lot of clips and a lot of stuff everywhere, so, uh, you know, lots of ways to reach us, of course. And I think that's it. So if you want to support episodes like this, you can do so on Patreon, patreon.com/htmlallthethings, just like these fine people did. Tim from The Web Hacker on Jason from Geek Life Radio via geekliferadio.com. Garrett Segall, Level Up Financial Planning via www.levelupfinancialplanning.com. [00:57:00] Magnus from Yes Web via yesweb.se. Syntaxify from the HTML All the Things Discord server, and Stacy Mosler from the website swoonworthydesigns.com. remember that we have a Scrimba affiliate link. If you want to save up to 20% off a Scrimba Pro plan, you can click that link in the show description or in the show notes at, on with full details on how it works in those show notes on htmlallthethings.com. And we also have a contributing author on our website, Michael LaRocca.

He's the author of Self-Taught: The X Generation Blog at selftaughttxg.com. Feel free to leave a comment or review on the platform you're listening to this on, we are signing off