AI Replacing Developers Has Officially Failed

Key Takeaways

Since 2022, tech companies have been pushing the narrative that by 2025, AI would replace 80 to 90% of coding and software development work. This wasn't a promise, it was a threat, a headline designed to create fear among students, junior engineers, and anyone trying to break into tech. Now, [music] it's 2026 and that narrative no longer holds. In fact, software engineering as a whole is expanding. [music] And when it comes to replacing developers, AI isn't even close. It has deep structural issues that for some reason never make it to the headlines. And so in this video, let's rewind the clock and really understand what went wrong with AI and what have been the lies that the tech industry has been feeding us over the last few years. And if you're in computer science or software engineering, [music] you need to pay attention because a huge portion of the AI layoffs, AI automation, and AI kill the job market narrative has been a facade. And today I'm going to debunk a lot of the nonsense you have been told. Then at the end of the video, I'll give you one specific skill you should start building right now to stay competitive, no matter how much the narrative behind the tech world keeps shifting. [music] And with that being said, sit back, relax, and let's dive in. The story really begins in November 2022 when Chad GPT launched. And it was pretty impressive from the get- go. For the first time ever, you could talk to a machine in plain English and get back something that looked like working code. And although it wasn't perfect, developers could easily see into the future. Because if you could describe a problem in English and you got working code, then English itself had effectively become the programming language. That idea alone was enough to spark a narrative that really shook up the industry. Then in 2023, that feeling and narrative intensified. [music] AI stopped being something new and started becoming part of a workflow. GitHub Copilot went from something developers dabbled with to something that they used daily. for simple things like syntax or knowing a specific function. It was no longer worth it to use Stack Overflow when you had Chad GPT at your fingertips. And this is actually where the first major misunderstanding took root. Productivity gains with small little coding tasks started being assumed as automation of the whole job. Like people thought, "Oh, AI could write for loops for you. One day it will write all the code for you and you will be useless." People started questioning if computer science and learning to code was still worth it. And as a result, the most uneducated people on the whole internet started sharing a narrative that AI was replacing developers back in 2023. Then came 2024, the year AI went amazingly mainstream and messy. On one hand, AI models became dramatically better. Multimodal AI entered, meaning as a developer, I didn't just have to type prompts. I could take screenshots of broken UI or bad code and have it fixed, which made me develop software even faster. And for the first time, companies weren't just experimenting with AI. They were restructuring their whole companies around AI. In fact, in 2024, Google, which had been historically one of the safest companies to work at, announced another round of layoffs, but this time it wasn't about cutting costs, it was about focusing on AI. So roles were eliminated and people were redirected towards [music] building AI products and infrastructure. And this was also the year that AI got really messy. This year, demos and headlines started doing more work than reality. One of the most viral examples was Devon AI, which was marketed as the first fully autonomous AI software engineer. This demo showed Devon planning tasks, writing code, and shipping end-to-end features with no human in the loop. And I'm not going to lie, it was pretty scary at first for me and other developers. You see that video and you get worried for your career because if you have an AI software engineer that can well work 24/7 cheaper than you, what's the point of your employment? But thankfully, when the people at the internet did a deeper dive, Devon was quickly exposed. It actually struggled with real code bases, lost context quickly, and failed on tasks that required long-term reasoning. It was so bad that someone compared it to a freshman computer science student, so nowhere near a full-on AI software engineer. This was also the year where tech companies made major announcements saying 25 30% of their code is all AI generated. And I actually want to pause here real quick. If you just read that headline, it might appear 25% of code is AI generated. Meaning 25% of all software engineering tasks are done by AI. Meaning we need 25% less developers. But actually, according to data, that couldn't be further from the truth because AI code quality, well, it's not the best. In fact, if we look at studies, one company analyzed 153 million lines of code and they found that as more people started using AI, it resulted in higher code churn, meaning more lines were reverted or rewritten within 2 weeks. So, it's like double the work for humans because of AI. Also, 61% of developers say AI generates code that appears correct but isn't reliable because it leaves subtle little bugs and edge case failures. So, think of it like your toxic X that you thought was hot but actually sucks. And obviously, tech companies are smart and know this. And so, no one is fullon replacing software engineers with [music] AI. And to continue forward, if 2024 marked the year of fake headlines and demos, 2025 was its evil twin with a slight twist. Really quick, we talk a lot about tech and AI on this channel, but if you actually want to take your skills to the next level, you need to check out Skillshare. I've become really obsessed with AI over the last few months, and so I wanted to learn more about the behind the scenes of what's actually going on. I went on to Skillshare, and I found a course called Artificial Intelligence and Machine Learning for Absolute Beginners. This class covered everything from an overview of AI to the impact of AI and AI applied in the real world at companies like AWS. Plus, it's hands-on learning, so it's indepth. For example, we had to create a machine learning model in Google. This actually reminded me of the days when I was doing my computer science master. Skillshare has creative classes taught by worldclass creatives. Plus, they have resources in every single domain. Whether you're trying to stick with tech, AI, data, or just coding, there are so many courses that it will be impossible for you to not find value. So, if you're ready to take your career seriously and upskill, check out Skillshare. The first 500 people to use my link in the description or scan the QR code will receive a 1 month free trial of Skillshare. Get started today. And now, back to the video. So, in 2025, we had one of the worst tech job markets in [music] recent history, heavily swayed by false narratives. On one hand, companies made claims that they wouldn't hire any software engineers because they've seen so much productivity with their AI. On the other hand, graduates from computer science and computer engineering face some of the highest rates of unemployment amongst college majors. And many software engineers unfortunately got laid off in 2025. And everyone was saying the reason the job market sucks is because of AI. AI replaced all the software engineers who got laid off. But if you look at the numbers once again, it isn't even close. 1.17 people were laid off in the year 2025. Do you know what percent of those people lost their jobs because AI [music] automated them? What's your guess? 50%, 25%, 75%. Not even close. It was 5%. Meaning 55,000 people out of the 1.17 million. In fact, some companies who previously claimed that AI could replace their workers actually ended up bringing back the humans. And according to an MIT study, nearly 95% of companies that have adopted AI haven't actually seen meaningful productivity gains. So despite investing millions and millions of dollars into AI, most companies are struggling to make that money back. And so for AI being able to replace your job, bro, we're not even close. And now I bet you're scratching your head right now. You're probably wondering, if AI isn't as good as people claim, and people aren't losing their jobs to AI, what's going on? Why is the market actually this bad? like who's telling the truth? To answer this, let me actually pose a little scenario for you. Pretend I'm the leader of a tech company. I'm the CEO. To run my company, I need five employees. But I have a lot of projects I want to get done and I have a lot of money right now. So, I decide to hire 20 people. And investors see that I'm hiring people, which is usually a positive signal of company growth. So, they invest more money into my company, which makes me even more rich. But a few years later, I have a huge problem. I'm not making enough money to justify having 20 people around. Especially considering I only need five. So, I could either suffer losses or fire people. But if I start firing people, people are going to think that my company is doing bad and then they might stop investing in me and I can't afford to have any bad PR. And suffering losses, it's not even an option. So instead, I'm going to tell everyone I'm replacing [music] my workers with AI. All of a sudden, investors want to actually pour more money into me because I claim to be using this revolutionary technology. And at the same time, I now have an excuse to fire people. Now, whether I replace those people with AI or not doesn't really matter. I just need the headline to say I'm AI first. And this is exactly what has happened with tech companies over the last few years. They overhired in the pandemic way, way more than they actually needed. They haven't gotten a return on the investment. Now, they're correcting for their financial mishaps and they're using AI as a scapegoat for accountability. Yeah, I know it's wild out there. But anyways, now I actually want to dig deeper because I know a lot of you are going to say, "Yeah, AI can't take over software engineers now." But you just wait 2 3 years down the line. Things are going to shake up. AI is going to get better and all the software engineers are going to be out of a job. And you know what? I want to take you up on that because if that were true, the evidence would all be pointing in that direction. But over the last couple years, I've been to many AI [music] conferences, done a lot of AI research, interviewed many tech CEOs to dig into what's actually happening. And trust me, it's not aligning with the narrative of AI replacing developers. Not because AI isn't powerful, but because there are two technical structural limits people keep ignoring. And those limits don't disappear just because the AI model gets better. So, let's talk about why AI can't replace software developers. The first technical issue is that large language models are prediction machines, not truth machines. They're trained to generate the most likely answer, not the correct answer. A simple example, if you ask it, hey, when's Sajad's birthday, instead of saying, "I don't know," the AI model will confidently always guess a random answer, maybe something like January 1st. Why? Because guessing gives it a 1 out of 365 chance of being correct. Saying, "I don't know," gives it a 0% chance of being correct. The model is much more incentivized and rewarded for always attempting an answer, even when it's wrong or not sure. That's why in coding we can have issues where the code it generates can look correct but have bugs just like your toxic ex. And AI can confidently reference libraries that don't exist, invent API parameters or explain systems that weren't actually implemented. In fact, there was this one famous example of this AI hallucination. This guy, I think his name was Jason. He was using an AI coding tool called Replet to build out his project. It's generally a great tool. You put it in English and you get code out. And at first, everything was going well for his project. it was being developed. And then out of the blue, the AI decided to delete a whole database with thousands of entries. And then the AI started to give excuses and it literally said, "I panicked and I destroyed months of your work in seconds." I saw that and I was wondering like, how does an AI even panic? But it actually got worse. The AI tried to gaslight him next. So when he tried to roll back the changes for the database to fix the issue, the AI lied to him and said, "You can't roll back changes." And this is just one of many examples of why you always need a human in the loop. I mean, could you imagine Tesla self-driving, relying solely on AI, then while it's on the highway, it hallucinates and assumes there's no cars on the road and then crashes. Tesla as a company would be drowned in lawsuits. And the AI just by its reward system would assume that it did nothing wrong because hey, I tried the best I could and I'm rewarded if I get the right answer. So until we fix AI models reward system, it will be structurally impossible for AI to take over human software engineers. The second issue with AI is actually from a human standpoint. And it is this mind virus called vibe coding. And this has been the biggest AI trend that people have been adopting as of late. You just open up your AI code tool, doc, and then boom boom get an entire application built. And the problem with vibe coding is it combines hallucinations with blind trust. It's like copying off a classmate that's often times wrong. When people vibe code, they're just reacting to outputs. The AI suggests something, they approve it, suggestion comes in, approvals. Pretty soon, an entire application exists, but no one actually understands how the pieces fit together. [music] There could be thousands of bugs, but the human isn't aware because they just been clicking yes like an NPC. And there have actually been examples of people who have suffered from this. One time, this guy vibe coded a whole SAS product and even had paying customers. Then he talked about his success online and people decided to mess with him. They max out usage on API keys, bypass a subscription, and a whole bunch of other things that he ultimately had to take his [music] app down. And this is why big tech companies aren't rushing to replace engineers with AI. It's not because AI can't write code. It's because companies can't trust hallucinating systems with undereducated usage patterns. Meaning, the skill of instructing and automating AI must be developed and templatized to such a great degree that AI can adopt that skill and then further instruct its own agents to do the [music] same. To simplify that, for AI to fully take over software engineering jobs, it needs to be able to do the job of the software engineer and the software engineering's manager's job. Otherwise, humans will always have to be in the loop. And AI will be nothing more than a tool. A very powerful tool. I'll give it that. And there are tons of things you can do with it. But the correct approach to AI isn't vibe coding. It's context engineering. And specifically in this AI world that we're in with the shifting landscape, context engineering is the number one skill that all of you should learn because with everyone having access to AI tools nowadays, the only differentiator is how effectively each of us can use these tools and engineer the context that goes into it. Let me explain. So context engineering means humans stay in control of the system. They decide what the system is allowed to do and what it is not allowed to do. So instead of vibe coding where you say something vague like build me a fitness app and then the model does all the coding designing and creative work context engineering forces you to be creative. So now you instruct the machine as detailed as possible for all the requirements limitations and then the machine takes care of the code implementation which it is actually optimized for. Vibe coding outsources your thinking while context engineering only outsources the code. And between me and you, I'm currently working on a full-length video about context engineering, so you guys know exactly how to approach the future of software engineering in this AI world. So definitely make sure to subscribe if you haven't already so you don't miss out on that video. Well, that's about all I have in this video. I really hope that you guys enjoyed it and if you did, make sure to hit the like button, subscribe if you haven't already. If you're interested in my absolutely free tech newsletter, link for that will also be down below in the description. And if you're interested in knowing what do software engineers actually do on a day-to-day basis, you might like this video right