Transcription
Yeah. So, I've been seeing a lot of YouTube videos kind of pop up of people, developers talking about their experience with AI and how the current state of things is just not that enjoyable and people are actually kind of stressed out about it. I'm going to go through a couple of videos and then also just give my opinions on the current state of AI in this industry because I do think it has changed a lot and I have my own opinions on like is this even enjoyable anymore? Like is this a fun career to do where you're just basically prompting Claude Code all day or you're prompting Codeex to have it do all the work for you and review the work. I'm not going to wash through these and do reactions. I'm just going to kind of give a quick like overview.
This guy basically took a break, a sabbatical for like a month from his job because he just felt like he was just spending so much time multitasking and context switching by managing all these agents trying to implement the code and have another agent basically do code reviews and another one do dependency updates. And I think his main point was like, okay, like the way that we used to code where you actually like go heads down and like try to figure out a hard bug and solve it. It's been distilled to like prompting Cloud Code to basically do it for you and then like you're waiting for Claude Code to finish and in the meantime you have to find other things to kick off to fill in that void of space which I'll put all these videos in the description below if you guys want to check them out.
But I I think I'm seeing a lot of these like this is uh Maximillion. I also watch him as well. He's a really good creator. He makes a ton of tutorials. his his title is it was more fun before AI and I think a lot of people are also kind of feeling this way like if you've been in the industry for a while which I have been in since 2013 professionally which is what almost 13 years now it was a lot more enjoyable before AI for me like there was a period where I was actually excited about this like we could move much faster we could be more productive but now the AI and the models have gotten so good that it just feels like you're just prompting and you're not doing any work yourself you're just reviewing the code that the AI generates and to not use the AI is also silly too because like I've seen the performance benefits and gains uh the efficiency of using these tools to write code like even at work it's a project that's like 600,000 lines of code or more and I had to do a huge refactoring which would probably have taken me like weeks to do by hand and I just prompted cloud code did a did a giant plan and 250 files it changed them all it refactored everything all the tests pass and you know I had to come through and review the code and do some additional changes and reprompt and it did miss some stuff. But we're in an age now where like you can basically completely change your codebase from one prompt and you could just sip co coffee as you like watch the uh the agents kind of go through and after you do all this you there's no sense of like reward before like you would spend all this time trying to learn how these algorithms work, how to write code and now it's just like you prompting your way to a solution. And I guess there there used to be a dopamine hit that you get by doing this, which I just don't feel anymore. Like on all my side projects, I don't feel like I'm actually making them because AI is just basically coding it all for me. And I I get more joy going out and mowing the lawn. Like I feel more accomplishment doing something that's actually physical, like mowing the lawn or I don't know, cleaning up something like, you know what I mean? Compared to like just prompting an agent to kind of do the work.
Okay. This is another one. This guy's basically talking about how he felt like he, you know, AI's basically made him useless because it can basically do everything that he used to do at his job. Okay, this guy got a bunch of views on his, like 166,000 views. This guy Garrett Ross, uh, he's basically talking about how challenging it is to get into this industry right now. Like, he got laid off after working for a while and he's been trying to find a job for like nine months and the interview process is terrible. Uh, some interviews are asking you to use AI. Others are asking you not to use AI. They're going to ask you like to do a leak code problem, which if you take a zoom out a little bit and think about it, cla can solve most of these problems that you're like doing in an interview. So like why are we being asked how to do leak code questions that are kind of silly, right? Like fsbuzz, all these other things. like yeah they do test your knowledge in coding but the the industry is changing and I think it's going to take a while for you know coders to actually start realizing like what's actually important because knowing how to do some of this implementation detail is honestly just kind of become commoditized that it's not really important anymore. I mean this guy even blew a blood vessel in his eye he's so stressed trying to find a job. Cody codes another one worst job market in 26 years.
Yeah. So what are my opinions on what's kind of going on? I'm going to play a little bit of Runescape as I kind of talk because I found this is actually pretty cathartic. I used to play this game back when I was a kid, like in like when I was nine years old or 10 years old. And uh I picked it back up recently. It's been kind of fun just kind of uh mindlessly clicking on things and watching my character level up, right? I get more accomplishment watching a virtual character level up than I do prompting Claude Code to add features to my codebase. Uh that's the unfortunate truth.
So, let's talk about my opinions on like where where the current state of things are. I mean, I do have a course which I talk about how to use agentic coding and how to use these tools. So, I do think that it is super important to learn how to use these tools, but unfortunately, I think when you've been in the industry for so long and you're so used to doing things a certain way, when AI first started coming out, like I was super excited about it. I'm like, "Oh, this is really cool. I can move a lot faster. I can implement stuff. I could probably work on many different side projects at the same time." But recently, I think at the end of the last year, like I just kind of reached a breaking point where I'm like, "This is no longer fun." like it's no longer fun to prompt uh these agents to do all the work for me. Like I'm not feeling a sense of accomplishment anymore like I used to. I will say that if you're like new coming into the industry, I'm guessing that you don't feel this way because now you have like this tool that can help you be a lot more productive and efficient that you didn't even know like how to code beforehand, but now you can actually ship features. But me, I've been doing this for 13 years and I'm like looking at it from a different angle where the sense of reward that I used to get from coding is gone. Like I used to spend a lot of time trying to become really good at typing. Like my words per minute was like 150. I was actually training that to become really good at coding fast. I got really good at using my IDE, opening files, diving in the codebase, refactoring stuff. Um, I got really good at understanding code and, you know, web development at least. I'm not like a super awesome coder, but I spent a ton of time like learning all these different frameworks and languages and libraries and how to become really efficient at just coding in general. And I feel like overnight, I mean, realistically is over a course of like two years, these models have gone from not even usable to completely relate replacing my need to code. Like I don't even code by hand anymore. Um, like I I use cursor and clawed code. I basically prompt my way to add features and fix bugs. Even at work, I do the same thing. Like I'll create a a plan and cloud code. I'll kick it off. I'll do some refinement. Maybe I modify some files here and there using my IDE. But for the most part, some of these editors, like, you know, Cursor 3 came out with this editor mode where it's focused more on agentic coding where you're letting the agents do all the code edits for you. Um, and the more you embrace that, the faster, more efficient you're going to be because even though it does still mess up sometimes, it's it's reached this like threshold where it's more accurate more often than not. And so it makes sense if you want to create like a Docker file, you could just ask Opus to add one for you. And it's going to analyze your codebase and try to figure out what exactly it needs to add into the Docker file. And if you do have some knowledge about how some of the things work, you're going to be more successful with agentic coding. But I don't know, it's just like something something just it's not as fun as it used to be.
Now, to kind of hit on some other points that were in those videos I mentioned, one of the points was mentioned that like there's this level of anxiety and stress now that you feel like you have to manage all these agents and you have to learn how to set up skills and you have to learn how to do like agent teams or sub agents and how to have all these things running and like you feel like you you're just missing out on something. And although I did feel that for a little bit, I have found that with larger realistic projects, you're not like kicking off all these concurrent agents. Usually, you have the mental capacity to focus on one thing at a time, like one story, maybe two stories, and you have to prompt it and then by the time it's done implementing its changes, you come back and you have to manually test it. There's some stuff that it misses. You have to go and reprompt it. And so realistically, like I don't see how someone could be, you know, concurrently kicking off tons of agents unless it's a brand new project. I have found that these tools work best when it's a brand new green field project, but on an actual like brownfield project that already has a ton of business logic, has a ton of complexity. Concurrent agents isn't just like a viable thing in my opinion. So I found the sweet spot of agent coding is I typically just have one terminal open and I will focus on one thing at a time. If that makes me a slower developer, so be it. I'm at least happier. I can focus on just one uh feature at a time, one bug at a time. I don't have the context switch. I can actually pay 100% attention to the logs that Claude Code prints out as it's trying to plan out or implement stuff. And I can catch little things along the way that it messes up on. So, I can go and reprompt it to fix it. Now, often you can kick off some other sub aents like to fix little things here and there, but I have found that like every time you make a code change, there's a higher chance that you broke something that's completely unrelated to what you're trying to fix and you just end up wasting a bunch more time, right?
I guess another thing I want to kind of pinpoint on is that we have these tools now where people who have no coding experience can basically build full stack applications and get them deployed out and have no idea what they're doing. Like there's tools like Lovable that can deploy out full applications and they just prompt features after features and for the most part they work. Are they riddled with like performance issues and security issues? Probably. And unless you are a coder, you're not going to be able to read through the code and understand those. Uh you kind of have to embrace and trust these tools that they do everything perfectly, which is not the case. And I don't think they're going to reach that case unless you just want to spend a bunch of tokens of having various different models, implement the code, review the code, refactor the code, reanalyze the code, and then at that point you spend like two hours watching these AI models turn over the same feature. And uh you know, probably being a human in the loop is going to give you better outcomes at the end of the day. And to be a human in the loop, you have to know how to code. And so I guess my point is that like although these tools can help you move faster, if you want to ship a quality product, you still have to know what you're doing and you still have to review the code. And if you're like on a startup, yeah, then maybe you can have some flexibility of just like blindly merging the stuff that Cloud Code is implementing and you can just let it write the test and like write the implementation using those tests and then have another model review it and like you could fully automate all this stuff. I'm just not really certain that at a certain point it just starts to break. I I think at a certain point it starts to fall apart. Um, I could be wrong. I'd be very curious if someone's working on a project where they are not actually in the driver's seat watching these agents and commanding these agents. I I do think at some point maybe we can have fully autonomous systems building out entire applications. Like I know there's Gas Town and other things that are reach like critical mass of like 300,000 lines of code that just kind of build on themselves. I think autoclaw was like another thing that kind of you know got huge in scale which uh was basically just like you know let AI just keep on adding features to it. Honestly, I'm just rambling at this point but I think sometimes I just question like what is the purpose of the career that I have spent so much time trying to get into and is it time for a pivot? I know like from what I've seen webdev might be the lowhanging fruit that AI replaces first. It's really AI is good at front end is good at backend. I've also seen AI be very terrible at CSS and I have to go in and actually fix the CSS myself. So, it's still not a a silver bullet. Like AI can't fix every problem and you still have to have someone kind of directing it when it can't get it done. Um, sometimes just throwing a bunch of tokens and money at the problem doesn't actually get it working. But at the same time, like I'm trying to figure out like I don't want to just be there babysitting an an agent the whole time. But I'm guessing there's other niches in software engineering like different industries like maybe you're doing lower level coding like with C and you're building up performant embedded systems and maybe AI doesn't work as well in that. And then you also need to like focus more on like double-checking memory management. Make sure that's good. Make sure that the the CPU cycles are actually super fast and not wasteful, which you probably have to be more experienced. I'm I'm kind of questioning like where is my pivot? What should my pivot be? Should I dive more into security? Can AI do security just as well as like security expert? Um, do I need to dive more into like being a project product manager or a product owner? Someone who actually like commands the agents and has a higher vision of how the product should work? I don't really know. But I will say that on a larger team and a larger project, like there's still a bunch of human process and red tape and bureaucracy, which slows everything down. Um, if you are in a system that's already in production, you can't just let AI make a pull request and click the merge button and cross your fingers and hopes hope it doesn't break anything, right? There's still tiers of testing. You need to make sure that what you just added doesn't break everything. Often the agent will fix 95% or 90% of what you want, but then there's the other 10% you have to keep on going through and prompting it and fixing it and refining it and reprompting it. Sometimes as a human like you just prompt poorly. something you add into the prompt just makes it output some terrible results and you go back through you re reread the prompt and you realize oh there's probably a better way I should have said this. So it's those type of issues that you run into with agent decoding that I think still keeps you safe as a software engineer. You still have to be in the loop. You still have to be kind of like babysitting this stuff.
But yeah, I don't know. I'm just trying to figure out like what what do I do next? Like I could try to build my own SAS product and then at that point I know AI I could probably do it really well and I probably have to just focus more on leaning into marketing. That's another niche I could try to go into. I got my YouTube channel. I'm decent at making educational videos. I made a course that's doing pretty successful. I guess all I know is that like since AI has came out, I hyped it up on my channel. I was talking about it a lot because I did see it being the future and the writing was on the wall at the time. But now that it's here and like it's super good and it's efficient, I I just don't get that dopamine hit of coding. Like I just don't feel like I'm having fun doing it, but I also need to pay my bills and I have to like, you know, I have a family to provide for. So I'm trying to figure out the balance of like it's extremely hard to pivot, especially in this current uh economy. Like it's going to be so hard for people to even find a job. So I I'm not going to be able to pivot to another niche. Like imagine me trying to become a game developer. like that's not going to happen. I have zero game dev experience. No one's going to hire me to code up a game. Um, so I'm kind of like stuck in this web development niche where that's probably the only niche I could try to like find a job into. But at the same time, I don't know if it's the most safe job, right? You could probably find some kid in college to vibe code an entire application, which in fact, this is actually happening right now. There's people who have no experience who are in college. They're vibe coding applications, they get it to be a decent size and then they see that there's a bunch of issues with it and they reach out to consultants and then they have to come in and fix the issues that uh you know these these applications were viodated from the get-go and that's something we're seeing in the industry I mean we've even seen this at our job but we have to go fix other people's code that you know the startups will try to move fast they'll try to just add features so they can get customers and then the software system reaches a a critical mass point where it's like okay now we actually need people who know what they're doing but I do think coding is still cave. I just don't think it's as fun as it used to be. Um, and maybe I'm just old and like I'm just it's nostalgia of like spending 8 hours trying to solve a bug that now Clog Code can do in like 5 minutes. Um, you don't really get that hit anymore, you know? Like right now I'm about to level up to level 54 and attack and like it's going to feel good. I don't know why, but it's going to feel good and it's going to feel better than prompting Claude Co to add a feature to my codebase because even though it's just me clicking on these little sand crabs, I feel accomplished leveling up the level 54 for some reason. And by the way, this video is not sponsored by Runescape, but if you are looking for a fun, chill, brainless game where you can just click on things and level up and feel accomplished when you level stuff up, go check it out. I think it's a pretty cool game. Um, but yeah, I don't know if you guys have comments, leave below. if there's anything else you think I should hit on. You know, this video I didn't plan for you rambling, but I'm just seeing a bunch of people always pop up on my feed about, you know, how terrible this industry is, how the the lack of joy, the joy isn't sucked out from the industry. But I also think that this is a great time if you are able to work past this. Like if you're already a professional, you're able to work past this. It's a great time to build a SAS product, get it out there, and then maybe focus on other things like managing a business, marketing, stuff like that. Because I do think right now, like we do kind of have a secret weapon. If you already experienced as a web developer, you can ship out and crank out web applications faster than ever. And now once you have the entire application built, you're just going to be limited basically by marketing it and getting people to use it. But then you're also going to be fighting a bunch of other people who are able to basically recreate what you just created by a couple of prompts as well. So it's a very interesting time and I think um you will have an edge up on other people if you either have a really really deep tech technical knowledge on something like you really understand how SQL database works and you get hired to really manage a SQL database or you just have like a really broad understanding of business and marketing and all this stuff.
All right, I've been rambling for a while in this video, so I'm going to go ahead and stop it. But let me know if you kind of like this vibe where I'm just like playing Runescape but also talking about my thoughts on software engineering. I do think I I I don't really make many videos like this where I actually talk about my opinions, but at this point I just don't really enjoy making content like teaching people how to do stuff in code just because like it's been commoditized like Claude Code basically can build out a ton of stuff. And yeah, I mean if you're stuck and you can't get past something, you don't know how to do something, these agents can do it for you and they can teach you. So it just on my channel I have like 1,200 videos where I just spend a lot of time teaching people how to code and now that also feels like it's basically been stripped away from me right so 13 years of web development uh you know getting good at coding getting good at my ID that feels like that's been taken from me and the joy has been removed from that also you know my YouTube channel teaching people how to code and really getting them excited about learning XJS or learning this library this framework that's also been taken from me because I just feel like you can just ask these agents to do it as well. Um, so I'm trying to like figure out what the heck I'm going to do with my channel, what the heck to do with my career. Luckily, I'm still employed. It still pays the bills, but it's just something about it. It's just not fun anymore. Um, and I I got to figure out how to pivot, right? That's that's what I'm trying to figure out right now. So, it's also why I haven't really been publishing much videos, as you can tell. Like, I maybe published a video like I took like two or three weeks between a video. It's just I I think of an idea and then by the end of the day I'm like I don't want to go record a video. I honestly don't. Like oh a new model came out and I'm like I don't I don't care. I don't want to go make a video about that. Or a new IDE came comes out or a new code editor or a new agentic coding tool like Codeex comes out and I'm like I just I don't care. This stuff it's just not fun. It's not fun making videos on hype updates that come out. It's not fun teaching people how to code when no one cares to watch it. So I don't really know what to do with this channel. I don't know what to do with my career. All I know is that I need to keep making a paycheck. So, I got to try to find ways to pivot, maybe build a successful SAS application, which again, that's a whole another can of worms. And even that's hard. Like even making your own SAS product is hard now because now you're competing with a ton of other people who might not even have as much experience coding as you do, but they're able to vibe code an application because maybe they're 20, 25 in college, don't have any responsibilities and they can just spend their whole time just coding on their product and getting it out there and marketing it versus like I can't compete against that, right? After my 9 to5, I'm tired. I don't want to I don't want to do anything. I just want to play some Runescape, right? Anyway, let me know in the comments what you guys think. Have a good day.