AI Didn’t Kill Developer Jobs — It Changed The Work

Added: 2026-05-29

[00:00:00]Yes, certain entrylevel jobs are disappearing. Just like when Action Script uh fell out of favor, those entry-level jobs disappeared. I'll go over here. When VB6 fell out of favor, those jobs, those entry level jobs, those jobs in total disappeared. You know, you getting the idea. When Pearl CGI fell out of favor, if you're looking to get a Pearl CGI job, those jobs disappeared.

[00:00:29]Same thing is happening now. Yes, if you're looking backwards, you're looking at writing raw code or re or using React or something directly. Uh yes, those entry-level jobs are going away. But that's just normal. It's part for the course.

[00:00:47]Uh in the software game, again, I've seen this over 30 years. There's plenty of entry- level jobs if you leverage AI tools in the software development in every game. So I see bookkeepers who know AI and know how to use AI effectively are going to get jobs right away. Accountants who do that, lawyers who do that will do it properly because right now people are making mistakes, right? Because AI is far from perfect, makes all kinds of mistakes. So let me address the uh problems that AI has which are many. AI hallucinates, makes bad assumptions. It can give you answers and it seems like uh people will say AI is lying to you when it gives you wrong answers. It's not lying. It just does.

[00:01:27]It AI does not think. It's a giant associative array. Now, we think that it's thinking because most of the time we're not thinking either, by the way.

[00:01:35]Most of the time humans use the system one, the associative system because we don't use system two which is the logical system very often because it's a very heavy uh process meaning it takes a lot of energy. So our brain is constantly always trying to conserve energy because it's still things where hunter gatherers doesn't realize we can go to McDonald's and get energy anytime we want. So unfortunately that's the situation we're in anyhow. So yeah, AI makes a lot of mistakes. uh AI has no logical capacity. There's nothing new out of AI. It's all derivative, right?

[00:02:10]It's all derivative. To prove my point, we just have to look at the Apple study.

[00:02:15]So Apple basically tested all these advanced AI models against a brand new set of tests that were very close to the standardized test that uh all the benchmarks that the AI companies use.

[00:02:28]Now the thing is the AI benchmarks uh from what I've been told that they are gamed to a certain extent. Basically they know that they have to I'm an AI vendor. I got a model. I want to train it against a particular uh benchmarks like the classic one for video is uh Will Smith eating spaghetti. Maybe they should add in Will Smith slapping out people. So the AI companies will train their models, their AIs, their large language models against these predefined tests and then they'll show the results.

[00:03:00]Look, it's 5% better on this and 2% better on that and 10% on better on that. They go, "Oh, look, we're getting smarter." So what Apple did, Apple smartly did is they basically created a new set of tests that mirrored the predefined tests, but they're brand new.

[00:03:15]And then they tested the AI against these tests at the similar level, and they just failed miserably. They didn't they weren't like 20% worse. They were like they just failed like 99% of the time or 98% of the time just destroyed.

[00:03:27]So proves that the AI can't think because if they were thinkers they would be able to see oh this test is kind of like that test so we can you know extrapolate blah blah as humans can do.

[00:03:36]AI can't do that. So because AI cannot think it has no capacity for logic like the the Apple study shows. You can go look up the paper. Um that's why it seems to hallucinate. That's why it seems to lie. It's not because it doesn't have that capacity to override a bad association or an erroneous association tree if you will that uh leads it to a stupid answer. So that's why humans have to inter interject themselves into the process where they um they verify what AI is doing. You have to think of AI as a giant associative array that is weighted, right? Uh, that's how developers look at it. If you're a non-developer, what does that mean? You have to look at as the world's fastest total junior who needs to have everything that they do checked over and over again. That's what an that's what an AI is essentially. It's very powerful, but it's not this artificial general intelligence around the corner adjacent uh thing that's going to destroy the world.

[00:04:41]This doom porn, this doom uh narrative is a constant tool is a is a tool that's used constantly. There we go. By power by the powers of it be to influence people. So I think the tech bros decided to leverage fear. Maybe some of them believe in it. I don't know. I believe they said this is a good tool to um to leverage to to to because fear is the biggest motivator. So this is a good tool to get people interested in our tech and so that they'll invest hundreds of billions of dollars, right? Nothing motivates humans like fear. So just like when Excel came out, people had to learn to how to use Excel. Bookkeepers had to learn Excel, right, to become bookkeepers. That's, you know, I'm sure when you're a bookkeeper 1980, you didn't need to learn how to use uh Excel and home computers. But then when you know Windows started hitting the scene and Mac OS started hitting the scene in the early 90s in a big way then to get that entry level job yes you had to learn Excel you had to learn how to use Word you had to learn how to use Windows or Mac OS you had to learn these things right uh today it's now uh AI right AI is the new Excel if you're a software developer once upon a time to get a job as a web developer you had to understand HTML you had to understand how to slice up images and put them into tables. You had to understand a little bit of Pearl CGI if you wanted to build apps. You want to do Pearl CGI or CC CGI and later and so forth. Like for me, I was did the bulk of my coding in the Java days. It's a whole bunch of Java stuff like JSPs and serlets and XML uh config files, all this kind of stuff.

[00:06:26]Deployment on re on servers and server configs. These are all antiquated now.

[00:06:32]Like there are some people are still using these things. You got a lot of legacy out there, a lot of legacy systems, basically old systems that still work, but they're they're using an old stack, an old technology. If it's no, you're no longer going to get a job.

[00:06:46]Like if I was stuck in that time frame where I was just doing JSPs and serlets and POJOs and XML descriptors and so forth and Apache Tomcat and Kyo resin and uh you know that kind of stuff, I would I wouldn't get a job. Those those jobs are gone. So AI is just doing that not just in the coding space it's doing across the board. So there are going to be plenty of entry level jobs but there just going to be you know traditional skills plus AI you got to learn your languages got to learn your JavaScript I would recommend the web stack you got to know know HTML CSS JavaScript understand the basic constructs of the web the foundations request response cycle servers clients that kind of stuff you got to know this so you can direct the age AI and override it when it does stupid stuff. So the next issue I want to address is the productivity um gain from AI which I think in the end is going to be 25 to 35% overall which is huge but people are saying well if AI is so productive we're going to start losing all these jobs all these junior all these jobs are going to disappear because they won't need people because AI will do it no here's the thing it's going to create a whole new set of jobs it's going to create a whole new set of opportunities you got to understand software companies are always trying to keep up to get the software the new versions out to improve their their code base to improve their operating systems. So instead of the next version of iOS coming out a year from now it will come out three months four months from now instead of the next version of Windows coming out a year from now. Same thing, right? There's it's just going to speed up the cycle.

[00:08:23]So I think companies that are firing people now they're not firing uh competent people they're firing the people that were marginal I believe also there's a lot of firing because they just overhired during co uh they need people trained in AI so job security entry level jobs are people who are whatever you do plus AI accounting plus AI bookkeeping plus AI legal research plus AI software development plus AI I you know foundations plus AI and and then there's a whole set of AI skills. Now there's a whole idea you're just going to use AI and and nothing will happen and then you you know without any training without structuring it properly without using it as an advanced user you're going to get all kinds of bad results. So I talked about this in a recent vlog. I had a friend of mine and said oh look I'm use Claude code the best out there apparently and we're going to develop a website. So yeah we're gonna develop a website. It was just a branding site, nothing to get excited about. So they, you know, it generates some spits out some initial templates. Looks pretty good. But then the real problem hit that 80% wall I talk about. And people say, "What is this 80% wall? Either works or doesn't work." No, you don't know software. But the 80% wall is when you get something that seems to work fairly well and it covers 80% of the processes that you need that you need, right? It looks 80% is 80% finished the website or the applications 80% working all the functions are working but you have still things you got to deal with security some subtleties in the behavior that make all the difference whether or not this is production ready or not and then [snorts] it falls apart if you don't use the AI initially and properly right you have to use it initially and properly you have to use it properly initially so you can't have somebody who doesn't know what they're doing build up something to the 80% point, then it starts falling apart and then have somebody else come in and fix it because it's already a mess behind the scenes. Basically, for developers out there, what what happens in that situ? You get a bunch of spaghetti code, a terrible code. So, at the inception of the project, when you're using these tools, if you want to build production ready software, um you have to structure your um your processes properly. So you got to know what you're doing. You got to know what you're doing as a developer even if you're using AI quite a bit. So in terms of jobs, yeah, they were seeing layoffs now because there's a readjusting just like when flash action script fell away, they started firing all these action script developers, but eventually they they morphed that into uh HTML 5based application development, that kind of development or react development, eventually angular so forth. So, so there is going to be a transitional period where people are going to well, they're getting laid off now because these are people who are not aligned with the new um they don't have the AI skills, right?

[00:11:24]They don't have the AI skills. So, that's the key. If you're watching this and you're a junior, you're coming out of school, you're wondering what I should do, you learn, learn AI. It's not trivial. Just using chat to do searches is not using AI. It's just like saying that you use you you type in a couple of numbers into an Excel spreadsheet or a Google spreadsheet and somehow you're you're an expert at spreadsheets. You're you are not right. And AI is infinitely more complex. Like I've been using AI for a little while now. And to use it properly, there's all these it's it's complex. understand how to harness AI, how to leverage it, how to integrate it into business workflows. You have to know the different models, the strengths of the different models, you know, have to you have to know how to coordinate, how to create endpoints, the MCP, uh how to uh connect one model to the next, how to stagger, how to write the prompts, how all these things, and it keeps this you got this idea of of skills, MD files. It just keeps going on. It's not trivial. It's not trivial.

[00:12:28]So there's going to be a huge amount of jobs for people who are professionals in a particular area, software, whatever, accounting, but who know AI again. So there's going to be this little transitional dip. Then I think that jobs going to shoot up. That will bring me to my last point about this subject in terms of AI and jobs and its impact on the broader economy. I think we're going to have a huge boom in the number of jobs are going to be created. Why?

[00:12:54]Again, based on my 30 years experience in in uh the technology business, I've seen this over and over again. [snorts] I've seen this over AI is now revealing is starting to reveal brand new use cases, things that we could not do before that we can now do and a lot of companies are going to be interested in that. I'll give you a quick example in terms of the past websites, right?

[00:13:18]Websites are a brand new tech, right?

[00:13:21]And what web apps and websites did, it made magazines and newspapers defunct.

[00:13:26]Like that business got crushed. No question. But it was replaced by a far bigger business, a far more profitable business for people involved. When I say I mean employees and and business owners, but employees as well, like web developers and high-end web designers, they were getting paid major bucks, right? So they replace a lot of the traditional graphic designer jobs, right? And it all went onto the web. And then the web allowed a whole bunch of companies now who would never have hired a graphic designer, right? Never hired a videographer. They're now never hired a photographer. They're hiring all these people now because they have websites for their businesses, right? There's a whole new category of job and demand that just did not exist before because the technology did not exist before.

[00:14:13]We're seeing that in AI already, right?

[00:14:16]So there's I believe for every job lost there might be three four jobs created right but not any old stuff right not any old stuff right [snorts] if in the early 90s you still wanted to work on typewriter repair or you wanted toh make ribbon for the ink for the typewriters yeah you're out of that job no question about it but if you just got into new stuff there's plenty of opportunity for you so that's the key to all this so there's going to be a lot more jobs new jobs we have not seen before. A lot more opportunity. The entry level is just changing, right? The entry level, the entry level developer is no longer React and uh and basic JavaScript and HTML.

[00:14:59]Now the entry level job is the foundations.

[00:15:03]Be aware of what React and Angular and Vue can do, but more on the AI stacks, understanding orchestration, understanding harnessing, understanding all these things about AI. And what happens? You'll be a master at that. And then so you you you'll make the decision, but the AI won't be able to.

[00:15:20]You'll say, "Okay, for your particular task, we're going to use uh React." And then you're going to set up the orchestration. And then next thing you know, boom, boom, boom. Then the AI is writes all the boilerplate code for you.

[00:15:32]I want to dispel one other myth that's coming out I disagree with. I think this whole idea that developers are just going to be spec writers and then they're going to hand this monolithic spec to the AI and then the AI is going to generate all the code, that's total BS as well. We've tried that in the past, right? [snorts] They used to do that kind of design before agile came out in software development back in the uh early 90s. I remember studying I said this sucks. Now I always was agile oriented because I've been I was an entrepreneur outside of tech before I be started writing software. So I applied entrepreneur techniques to my software development which is basically agile. You know you have a light framework of what you're going to do. You set up the the principles and basic parameters and then you start implementing the components that you've laid out. But it's an iterative process, right? So it's lightweight uh with rules and then you iterate that out. That's how AI is going to be developed. It's not going to be like I'm going to sit there for months and write a big giant spec and then give it to the AI and away it goes and it'll come out with a beautiful piece of software. That's nonsense. Whoever's suggesting that to you, turn them off.

[00:16:43]They don't know what they're talking about. No, no. It's going to be I define the spec, lightweight spec, lightweight spec. Here are my components. You know, as a developer, I'm gonna say, "Okay, for this, I'm probably going to need this and this library here, maybe this framework here. Okay, then you talk to the AI. You give it a bunch of parameters of how to write the code."

[00:17:01]And then you start developing components with it. That's where you're going to get your five and your 10x productivity with AI working with traditional software tools. In addition to that, there's all these uh AI first development um jobs and opportunities that are already appeared. I have developed a couple of AI first off three AI first no four AI first applications.

[00:17:25]Basically I trained up some models to facilitate some of the things I do during my day saving me hours and hours of time but they make mistakes but I have to review. So instead of something taking me an hour to do with the help of AI I get it done in 10 minutes five minutes but I have to double check because it does make mistakes. So anyway, I just wanted to deal with this whole idea that uh developers are just going to be uh developing monolithic giant specs and handing it to AI and then going have a coffee and a cigarette and a donut while builds the app. That's just silliness and nonsense. So there you go. I'm Uncle Steph. I'm here to dispel the AI myths. I hope uh it's pretty clear. This again, it's based on three decades of software. I keep emphasizing that because you got to pay attention who's talking, right? I see so many talking heads out there, some of the biggest names, they don't know what they're talking about. They haven't written code before. It's like it's crazy, you know, and they have these very strong opinions based on no background. Nothing, nothing. You got to have, you know, it's like when I was a fighter, when I first was learning how to fight, I did a lot of combat sports.

[00:18:35]Did it for 20 years or so. Um, I remember at one point after many years of doing, I had done judo and wrestling.

[00:18:42]I had done taekwondo catch Campbell. I'm a black belt. I remember I go to boxing. I remember I used to make fun of boxing. B just punching. It's easy. Punch is easy. And people were talking about, you know, complexity of throwing good punches.

[00:18:56]Like that's that's easy. You know, it's all in the techniques, I thought. Right.

[00:19:00]But then when I actually went and fought a a boxer, my boxing coach kicked my ass.

[00:19:06]And I learned that I didn't know how to punch very well. And even though he was punching in front of me, I couldn't see the nuances in his body mechanic and his positioning that made his punches hundred times more effective than mine.

[00:19:20]Right? There's all these little nuances and these little mechanics and um punching is fairly simple compared to development. Same thing with software understanding these big systems understanding how software business works how uh you know remember AI is just software so these reporters these people who've never written software who you know or who've never brought product to market they don't get the nuance they don't have it they've never done it it's not an insult to them it's just they never done it you know so um it's complex stuff bro it's complex stuff so these talking heads so many out series, podcasters talking about it. It's just like I just rolled my eyes. I hope you take this to heart. I'm very encouraged by this. I will tell you, listen, I will tell you if I thought this AI stuff was really what it was. Last point. Oh, yeah. I forgot this. Doom porn. Doom porn. Y 2K. Fake fake uh crisis. You know, Y2K was a technology thing. And you know, 1999, 2000, people were freaking out. People were building bunkers. They were storing food. They thought planes were going to fall out of sky. Nuclear wars were going to start.

[00:20:33]It was nuts, you know. And people thought this and it was all fake. It was all Uh same thing with weapons. WMD. Oh my god. Saddam. And it just goes on and on and on. I see it all the time. Don't fall for it. Don't fall for it. I know it's hard because our brains are literally designed to overemphasize potential threats, right? Think of all the times of all the times you had anxiety. Think about all those times you've had anxiety and nothing happened.

[00:21:03]The opportunity is huge. If you're a beginner, I've given you I've laid out some of the things you have to do. AI is going to it's it's actually a big um it it levels the playing field. It levels the playing field. Actually the the the advantage that money had the advantage that wealth and economies e economies of scales had are diminished quite a bit not 100% diminished trust me it still takes work to put out good software