I'm Not In Love With Development After AI Showed Up

Transcript is auto-generated and may contain errors.

Hey folks, we're going to go to the experience dev subreddit. This one it is AI related um but more I guess focused on uh engagement interest career focus and less on like the specifics of like how people are I don't know like using specific AI stacks and models and things like that. So um this developer was saying or asking I guess like how how others are kind of looking at keeping their interest and engagement in in software development with all of the AI things coming about. But primarily because I don't I don't think their focus was so much like oh because I think I'm about to be replaced and more that like they're saying as they're using LLMs you know they're able to do more and more complex things and so they're kind of like well trying to figure out where they fit in the process and like their their kind of love and interest of being able to build things is like it's kind of waning I suppose.

So, I wanted to talk about this because I guess it's just, you know, some different some different looks at similar or familiar topics that we've been going over. And I think it's interesting because I think this is going to be very very much like dependent on the person. And so if I were if I were talking directly to the person who wrote this post, I think I would start by asking them more questions than I would um like sort of jumping in to say, "Oh, like here's here's the best way to use AI or here's how you keep your engagement or whatever because I I actually don't know." And I think I would tailor my responses to them based on trying to understand them a little bit better. And so let me explain what I mean.

um when if I like take a step back and I I don't have a crystal ball, but if I were to think about the trajectory of AI usage, how that fits into things, um I don't know even even I don't know a year ago uh was already at the point where I could sort of like outsource to AI some uh non-trivial features. Right. And the success rate was kind of in between, right? Like it could build things and they would work and there'd be other things where um you know I I would be nervous to hand it off to AI cuz I'm like this is just absolutely going to be a waste even though it's really not that complicated. Uh just kind of figured like you know the amount of time and effort that I'm going to have to spend going back and forth with AI on this is almost not going to be worth it.

So, let me get AI to focus on some other things while I nail this part down. And then later on, maybe AI can chip away at, you know, some of the remaining work or whatever. Uh, and in a way that I feel like I can basically tell the LLM, go do it, walk away, come back, and be like, cool. Like, I'm I'm not really concerned. And over the past year, I've noticed that, you know, the how things are going, especially around the models, the tooling, um just all of this stuff is is making uh the the complexity that these these models can take on um you know, much more uh much greater. They can take on way more challenging work. Uh much more ambiguous. um we have more things like uh you know skills and instruction files that can really tailor the focus of what the LLM is doing.

So I just think that if I look at the trajectory that it will continue to move in this way where AI is able to do more and more sort of as a I don't like using one shot as the as the metric but in that direction where we can have more confidence that it's going to do uh produce something that we are happy with with less uh explicit guidance. And maybe just to quickly elaborate on that, I don't imagine a world where if I want to use AI effectively that I am manually handing it a 400page spec to go do it. I think if I'm writing like a an absurdly long spec for AI to go do something, then like that's that's just a different problem.

And the reason I say that is because at least in my experience in software development, uh it's not that there are no specs or no design documents, but if everything I had to do with engineers was here's this very detailed spec of everything, we would get nothing done. And I don't care that AI can help write the spec faster. Like that's less of my point. My point is that I don't I don't want to have to think about everything I'm doing in terms of a super detailed spec. I think that's a stepping stone. I think that's a building block that we have to work with. But I just don't imagine that the world in the future is you better be handing over a detailed spec or else. That's my take though. Um, and so if we're moving in that direction, my my view of this is that we're we're sort of like optimizing throughput almost.

And the reason I'm talking about this is because if this person's talking about their enjoyment, right, their interest, their love of like programming and building things, um my thought here is like I I suspect that the world is moving in that direction of throughput. And so you may find yourself in a world that is like everything around you is like how how are you trying to optimize the throughput. So my thought here is like well where do you where do you fit into that? And I don't mean that in like a a doom and gloom kind of way. I I mean it genuinely in terms of like your interest. If we're if everything around us is moving in the direction of optimizing for throughput if that is a you know if my hypothesis is is true or it's roughly in that direction where do you fit in as in where do you want to fit in?

What is the thing that you actually enjoy doing? So this person was kind of hinting at like, hey, like I think what I'm going to start doing is instead of just, you know, starting with AI, I'm actually like, especially on more complex things, I'm going to start by doing a pass myself, right? I'm going to go through this. I'm going to try putting some things together and then, you know, afterwards bring in AI. And like I I think when I hear that um to me it sounds like where the interest is and where the engagement is is like kind of think being able to think through these hard problems, right? And like and and go through that problem solving experience. And again, this is why I would want to start by asking them more questions versus just prescribing anything. That's kind of what I hear when they're saying that, right?

I that's they're trying to say I want to be able to do this part in particular and then I would bring in AI. Okay. So going back to what I was saying about, you know, if the world around us is optimizing for throughput and this is going to be sort of this expectation on us, what does it look like to be able to to meet that expectation of high throughput and still kind of fit in to the part that you want to enjoy. And I'm not saying I know the answer to this, right? I maybe that was a disclaimer to start with. What I'm not about to do is lead us down a path where we solve this. Um, I'm trying to lead us down a path where we can ask ourselves the right questions because you're going to need to answer it for yourself, right?

So, if you want to spend time doing more of the problem solving part, um, and we have to think about throughput. Okay. Well, in order for LLMs and AI to be effective working on these problems after, what do we need to be able to give them? Right? Because again if other people are are you know offloading even this part to AI are they having more throughput? I don't know the answer but like is that if that's the direction everything's going and things are being optimized that way you know are you going to find yourself basically if you think about your throughput being bottlenecked on this part specifically um com you know compared to other people and that's going to be the thing that holds you back even though you enjoy it. So, I'm not saying that's right or wrong. I'm just saying that I think it's a balancing act because you may feel pressure, right?

You might feel this pressure that if you're focusing your time or doing your sort of a version of software development with AI particular way, you might feel this pressure on you like, you know, the next time you're writing into Reddit, it's like, hey, everyone else around me is doing X, I'm doing Y. Um, and my manager is not happy with me. They're telling me to change. So, I'm just saying there's probably going to be this external pressure on you. And so, what is the right balance? Like, what what does make sense here? Um, I'm going to I'll use part of this drive to kind of talk about how my own sort of personal perspective has evolved a little bit from AI usage. But I think before I do that, um, my point here is like at least for this person, it sounds like they have part of it that they're interested in, right, that they want to keep.

It sounds like it's actually going through the steps of problem solving. What I don't know is like, is that actually more on like the design of things? And this is where like if you're hearing this and you're like, well, I I don't care necessarily about the problem solving or maybe you do. like what are the things that you care about genuinely because it will be different for everyone. Um I do I post a lot of stuff on social media, right? So I' I've seen different people from across different parts of the world saying different things back to me where conversation might be like hey like no one likes legacy code bases and you'll have people jumping into the conversation being like actually I really do I do really like the challenge of working in a legacy codebase and trying to move things around without breaking like for them it's like there's so much um what's the word like so many constraints so many difficult challenges that those types of challenges are very interesting to them.

So I don't know like for example if the evolution that AI is going down is like hey look you can just point an LLM at you know codebase. And in 30 minutes it's rewritten your entire codebase in uh assembly and it's entirely optimized for robots. Like that person would be like that sucks. That's the part I like to do. So, um, what is the part that you like to do? Right. I I know that for me, I I've spent a lot of time over the years like I like building um what seems like I don't know, like a reusable framework or something, right? I'm like, "Hey, I've seen this problem a couple times. I should probably go build a ridiculously overly complicated library for it that no one will ever use. I might not even use it, but it's fun putting it together. I like doing that.

Um, it's the same thing when I've done like video game development. I like playing with the patterns and like just building things. I like I like the sounds very backwards from production code, but I like exploring the complexity. like if we go build in this direction, what interesting complexity can we add like systems and whatever extensibility that we definitely don't need but like would be fun to try and explore. So uh I've seen patterns of that in myself. Um for some people they might hate that. Maybe some people are like hey look you know um the part of software development that I love is is really just like highlevel architecture system design. Like I love stepping back and putting the building blocks together. Maybe that's at different uh like I usually call this like different zoom levels of like um if we're talking about a distributed system like maybe you don't care about the particular implementation of some service.

You just know that some service facilitates something and how these things connect together. Maybe some people that level is not interesting. It's like hey if I'm building this service I want to talk about the building blocks of it. Like that's really interesting to me. So I think I think everyone's gonna have a different take on on the parts that they they love and they hate, right? Um so I think that's important to understand and I'm not saying that you have to have an answer, but I think it's important to explore because um number one, forget AI. I think that's really important to understand in general because if you and this applies I think to things outside of just software development by the way. If you spend too much time without knowing or trying to understand the things that you really enjoy, you might realize that over time you're spending a lot of that time doing things you don't enjoy.

So um I think trying to get more in tune with that's very beneficial in other parts of life but especially like this is a channel on software engineering and like career stuff right so especially in your career as uh as someone working in software engineering I think that's important to understand so that if those opportunities come up you can try moving more towards them or if you know there's things that you dislike and like the thing that you're spending time on is primarily that cool. Like you have an understanding. It's not just like, "Wow, I hate my job." It's like, "No, it's maybe not my job. It maybe you, you know, love the manager you work with, love the company you work for, but the stuff you're doing just for some reason sucks." And you're like, "Well, I hate my job." Well, maybe there's something else you could be doing at work, but you've never actually like kind of challenged that thought.

So, I think it's important to understand these things. So for this person, they're taking one stance on it. I would try to understand more about that and then go through this thought exercise of like this is again my sort of belief where we're going with AI and especially because of that that's all speculation. I don't I don't know that for sure. So again, if you're doing this thought exercise, not only is it important for you to try and figure out like what are the things that are engaging for you and disengaging, but also if you have a different take on where we're heading with with AI usage, then like I think that needs to be factored in too. I kind of gave you mine. Um I know maybe to elaborate a little bit further, like I I don't think that this is a matter of like, okay, well, we don't need humans ever.

I've been, you know, when more of this stuff was coming out and I like was pretty early on in code commute, I was saying like I don't think that that's going to happen. I know some companies have been trying it. Let's get rid of the software engineers and replace them with AI and then oops, let's hire them back. Uh I also know that people see a lot of layoffs and then they go oh that must be because people are being replaced by AI. Um, aside from those some companies that are doing that uh stupidly, I think there's other reasons that you're seeing layoffs that aren't people replaced by AI. Um, but I I don't the the future that I envision is not people go away. It's just AI. Uh I think it's you have people and they are you know much more ampl like their their capabilities are much more amplified because they have AI and whatever other tooling we end up you know creating.

That's my take though. Um, and I say that as someone like to to kind of back that up a little bit, right? While it's only been a couple of years, like my my own development has has kind of moved in that direction. And so, like at work, I'm a engineering manager, so I actually don't spend time actually writing software. Um, participate in other parts of software development, but I'm not actually writing code at home. Like I just I live streamed last night for for around two hours and was showing people like not not to say like hey look how I'm building and it's the best way to build but more like with this kind of stuff like I can go on all of these like side quests for myself that in the past would have been like why would I bother with that like cool idea but like I got a million other things to do now I can do those million things at once.

probably not the best uh but uh but like I can explore those and that's very like exciting for me. I'm going come back to this in just a moment. So um I think that you kind of have to balance like where you think this stuff is going and I think that's going to be a big part of your your decision-m. So, um, my my meta point with trying to answer this question is like understand, um, where you fit into that process. And that's not a matter of like just cuz AI can do it doesn't mean that it should. It's more like what part do you love to do because I genuinely feel if you love to do something, more often than not, you're going to get very, very good at it. Why? Because you love to do it. You love learning about it. You love getting better at it.

You will just continue to get better. Does that mean that AI will never be better than you? No. Um I can't make that claim. But to me, that does suggest that if you're if you're focused on that kind of stuff, you might you might find different ways that it's not that like AI can't help you at all with that, but that might be some part of the process that you keep and then you find different ways to augment yourself using AI even in that part of the process. And then for all the other that you don't want to do, right? because you're like for example I hate writing tests or I hate writing observability into my code or I hate doing security audits whatever it is there's a whole bunch of in software development where you might be like I just don't find that interesting

right like if you don't then if that's a part of what you have to do in your role how do you make sure that you're optimizing the hell out of handing that over to AI and letting it do a good Because the other thing is like just because you can hand it over to AI doesn't mean it's going to do a good job. So it's not just oh this you know this looks like X like give it to AI I want to get rid of it. It's this looks like X I want to give this to AI and like based on everything I've been doing how do I you know provide that to the LLM in a way that is going to do a good job so that I don't have to spend more and more time babysitting AI doing that I don't like. So, um yeah, I think it'll look different for everyone.

In this person's case, I would say um maybe that means like prototyping and exploring and trying out some challenges. Uh, but I don't know if it's more design versus just exploring the problem space, but then try to to take on some of that and figure out like what do you need to put together so that you can hand that off to AI to go do the other important parts that you don't want to do. Okay. So, for myself, uh, I kind of gave you a hint as to like historically things that I've really liked to do. Um, and so it's kind of a very high level like hey putting together frameworks and things like that even if they're not used like I I once wrote I don't know if people remember this before Unity there was something called XNA um and so it was you

could write like basically it was like having Unity but uh more basic so you could write C code um and build games and so was not um not like Unreal Engine or anything like that, but you could you could write games with it. And so I started building um like basically like a UI framework uh for I think I had it in XNA and then I had it in Unity and like why I don't like I don't really know. Um I think it's just cuz I enjoyed doing it. Did it like did I ship it to like millions of people? No, I didn't. I probably used it like a tiny bit myself and then like just gave up cuz I was like cool, like this was fun and now now I don't want to do it. So I I have found that's a pattern where I end up making things like that.

But if I think about that more, it's actually a lot less about like do I love the um I don't know like the mechanical parts of like stitching these things together like writing out the code and like I really don't but that's been a necessary evil for me for for a very long time. So if I want to do something like this, I have to go write out the boilerplate code and whatever else and find ways, you know, with scripts and anything else to to reduce that that noise. Right now I can use AI for that. Um, I was saying this on stream last night, but from building out things like this, there's almost for me almost always this inevitable point in time where um, whatever I'm building is kind of like, I don't know, deviated into something else or ballooned into something that's kind of messy.

And and I have this realization where I'm like, one second, I got to pass this bus. um right into all this traffic. Um I have this realization where I'm like, okay, like I've probably made some design decisions here that I need to undo or we got to backtrack or anything else. And so historically what that's meant is that I end up having to uh like basically kill it. Like I give up and that just means that like I have something that I've been spending time on that just disintegrates, right? Like I literally give up on it and it's because it's like too much mental effort to kind of like refactor it, bring it back to life, do the, you know, do the right thing. Like it's just too much. So it just stopped. And so now again, like that's not a part that I enjoy when that kind of thing happens clearly or else I would spend time doing it.

And now with AI, I have the ability to kind of offload that. So when I think about the parts that I enjoy, it's it's really more like I I like I like seeing things like my ideas come to life and my ideas are generally not um like visual. So I don't like, you know, visualize something and I'm like, "Oh, that would be really cool to draw or to like um you know, to have a 3D model of that." like I I don't feel like I'm visually creative that way. Um same thing with music. I don't feel like I'm creative that way. I don't like maybe I am for those things and I haven't unlocked it or something, but that's just not how I perceive myself. But for for like software, like I I think it's different where I have ideas and I'm like that would be something neat to see come to life.

like whether or not it's like super useful or is a you know billiond dollar business or whatever like completely not relevant right completely not relevant it's more like is this do I think that this is something interesting that I would like to see come to life and so for me that's that's the interesting part like how can I describe this thing how can I shape it how can I talk about what it's supposed to do and and and then it comes to life. So again, historically that means I have to go piece these things together. I have to go write the code. Um and I'm finding that again with AI, I have less of that, you know, I have less of that to do, which is great cuz then I can focus on the parts that I really enjoy. And I think one of the nuances there is like it's not that I don't care about any of the the technical details.

Um but one again if I were to use my zoom level analogy certainly I don't care about like what the if statement says or if someone used a for loop versus a while loop like I don't care. Um, there are there are some patterns in code that I have preferences for and would not die on a hill if someone used a different pattern. Like that's fine, but more like if I see this perpetuated, um, there are some patterns I've seen in code bases where I'm like, hey, I think if we stick to this, we're probably going to run into some headaches. So like my suggestion is to lean into you know this other pattern instead. So I like guiding things in that way. This is the same thing that uh I would do as someone writing in a code base for a team or as someone if uh it's less of the case now because as a manager where I am, you know, I I'm not actively in the code bases.

So there's probably stuff I would see in there where I'm like I I don't agree that that's the right thing to do and someone, you know, maybe there's people on the team that are like, we also agree and like the code is just, you know, at some age where they're like, cool, yeah, we could go clean this up, but what's the value? Doesn't matter. Um point is that I I like having influence on the direction of things. um and and architecture like I I like that. So it's just the higher level technical parts I think that I'm more interested in spending time in but that's from enjoyment. So I think that's the the other important thing that I want to acknowledge here is like what I'm not trying to convey to you is that I think those are the more valuable parts. I'm not trying to say, "Hey, be like me." And I'm listing off the things that like are the best.

I'm just saying these are my opinion. Like that's where I like to spend time. That's what I enjoy. And like I know I enjoy these things because when I'm building software outside of work, like I don't get paid for all the software I build. I like to do it, right? Like I I built Brand Ghost originally, which is something I sell, but I built it originally because I am a content creator. I need a tool that does what I need. I I have brand ghost, but like that's a combination of something that was like need cuz I need a tool like that want because I'm a software developer that likes to build things and then also happens to be something that I can productize. So it's I think like a a trifecta, right? Kind of kind of worked out that way. But there's certainly other things like um I talk about needler a lot on on this on this channel, right?

So I have some open-source assembly scanning source generation kind of thing that I I built that is how I like to to structure a lot of my C applications. And like that was did I need to do that? No, because I I already am doing that kind of stuff manually. It's fine. But I think that was beneficial. So maybe could argue need um and like also just very interested in it. That fits very much into what I was saying before around like building frameworks, right? Like that was a fun thing to put together and it was more fun because I could use AI and I could say like hey this would be a cool feature if needed could do it. Um would I use it? Like yes I could see some scenarios where I might do I need it? No. So I could go ask AI go build this feature and for me it was fun to be able to to do that.

So yeah, I think there's just a handful of things that I can refer to that are outside of work where I'm like, I put this together because because I was interested and I enjoyed doing it. But if you told me, Nick, you have to go write every line of code for that, I would say, well, it's just simply not worth it. Not worth it because one, I'm not getting paid for it. Two, it's going to take a ton of time. And like compared to the part that I enjoy, there's going to be way more time spent on the parts that I really don't enjoy. So why would I? Am I capable of it? I think that's another interesting question to ask. Um, for a lot of this stuff, I would say yes. But, uh, AI's also allowed me to kind of venture into some some areas of like I'm not skilled at, right?

I was joking around on on my stream last night that like I had, you know, a PR that went up from an agent and it was doing some frontend code and I kind of joke saying like I'm going to go review the backend stuff cuz the front end I just don't care about and I actually don't really because it's just like a private dashboard. But regardless, the real reason is because like I I don't have enough experience to say like these are patterns and things that don't make sense to me, right? like change to something that's more valuable. I don't have that experience. So, like when I'm reading the front-end code, I'm like, yeah, like I can see how this makes sense. I can see that these tests are actually using real code instead of just uh cuz I've seen LLMs basically the system under test is completely fake or mocked.

Like, let's double check for that. Um, so I can rationalize that it's doing the things I want, but I don't feel like I have good guidance on feedback to give it. So, you know, just a a good example of like, cool, I didn't have to go learn a front-end like stack just to be able to to have a dashboard, right? I don't need to do that. Um, if you've heard me talk about Blazer before, like Blazer is really cool because not cuz I think it's the best technology that's ever been invented. Blazer is cool because as a C developer, I can go take all of my C knowledge and build UIs. Um, like web UIs. And so I don't have to use Blazer if I don't want to. Maybe it is a good fit for the thing I'm doing, but at least I can make a decision about that versus just because the language.

maybe the rest of the ecosystem it is a good fit and that's cool and I'll use it otherwise I I can say cool like let me lean on AI and go build this. So that might be a factor for you too. Um kind of go back to some of the earlier discussion, right? For some people um when we talk about interest and things like that, some people are like I just want to learn different things, right? I was telling someone recently that like I wish um I like the result of learning something. I love the the feeling of like coming out of something that I've been focused on being like, "Okay, I feel like I've learned." Um, but what I what I don't like about myself, and I can admit this, is like I'm very like hesitant and resistant to to like new challenges. I don't like that about myself like but that's something I can say out loud, right?

Like I know that about me. If if for example, if I am almost at the office, right? If I walked into the office and my manager was like, "Hey, great news. Um there's going to be a bit of a change, but like um you know, uh we're we're going to I'm totally making this up, right? We're going to move you to this other team that does something completely different." I would be scared. Like even if he was like, "And there's like a pay raise." I'm like, "That's cool, but like I would be scared." Uh, and it's funny because I mean it's maybe that's a lot of imposttor syndrome. I've said this before that any time in my career where I've been put into a new area uh after a little while I'm like cool that was probably the best learning experience I've ever had. Every time it's really good coming out of that once I feel like I've learned.

Um, if I change that scenario, if I walked in and my manager said, "Hey, there's there's this other opportunity. We could just move you to this other team. Assume everything's going to be taken care of." Like, you know, obviously it doesn't work this way, but assume everything is just taken care of. You can go work on X instead. Um, and even if you were to put some like pay incentive, right? I would be uncomfortable. the pay incentive would have to be quite high for me to like overcome that. And I don't like that about myself. I really do wish that I was someone who who saw, you know, challenges like that is like exciting. Um because I think there are plenty of people and I say they're very fortunate, right, who have these different challenges.

Maybe they come up in like regular work that they're doing where they're like, "Cool, I have to go um explore this part or new tech stack or um maybe we have a platform and I have to think about a different domain that people are are using our platform for. Whatever it is, like some people thrive on that." And for me, it makes me anxious. Like I said, I I love coming out of that feeling like hell yeah. Like I feel like I've learned like I work on at Microsoft I work on the routing plane for Microsoft 365, right? So it's a it's a very very big platform and like I came into this arguably knowing very little close to none about routing. Am I an expert? No, I'm certainly not an expert. I would never sit here and say that I'm an expert. Do I work with lots of very smart people?

Yes. Do I think that I'm good at my job? Yes. So, it's fine, right? I've learned a lot and I know if I don't know the answer, I know like how to go find it. So, I can be effective. But when you put me into new situations, I I don't like that feeling. So, you might be someone who really does. And I would say when it comes to AI, instead of like, you know, new new area that you know nothing about, like what can you do? Like you do what I do. You type into your prompt into your claude code or co-pilot or whatever you want to use. like go launch a fleet of researchers and go understand everything about this topic and come back and summarize it for me because I just want enough information that I can go do something with. Uh I don't want to have to be immersed in it because it it like scare is the wrong word.

It makes me uncomfortable and I don't like that about myself. So anyway, hope that is helpful. I'm just getting to the office here. There's a lot more traffic today than I expected, but that's okay. At least I'm not late for my meeting that's in 5 minutes. I was nervous about that. No good parking spots. Come on. Okay, we'll take this one. We'll take this one. Cool. Yeah. So, thanks for watching, folks. If you got questions, leave them below in the comments. Um, otherwise, you can go to codemute.com, submit totally anonymously there. And then, like I mentioned, I do a live stream every Monday at 7:00 p.m. Pacific. I won't be doing one this upcoming Monday because I'll be hosting my in-laws. Uh, by the time I post this, though, that's probably in the past, so don't worry about it. Uh, but yeah, you can check that out at the Dev Leader podcast on YouTube.

And then my main channel is Dev Leader if you want some AI tutorials, C# programming, that kind of stuff. And then otherwise, just find me on social media. I post all the time and happy to engage with you. So, thanks for watching. See you in the next one. Take care.