# Is Software Engineering Finally Dead? ## Метаданные - **Канал:** Cole Medin - **YouTube:** https://www.youtube.com/watch?v=tjBpm91ZQM0 - **Дата:** 05.03.2026 - **Длительность:** 17:21 - **Просмотры:** 7,607 - **Источник:** https://ekstraktznaniy.ru/video/11248 ## Описание A couple weeks ago, Boris Cherny - the creator of Claude Code - went on a couple podcasts and said "coding is practically solved" and the Anthropic CEO (Dario Amodei) said that the software engineer role will go extinct this year. Fortune ran a headline: "Software engineers may not exist by year end." That was the last straw for me. Because every six months, like clockwork, someone new makes this claim. Jensen Huang said nobody should learn to code. Dario Amodei predicted AI would write 90% of code by September 2025 - six months later it didn't happen, so he made the same prediction again later. But I'm not here to say AI coding is useless - far from it. I use AI coding assistants every single day, and they legitimately 10x my productivity. The truth lives in the gray area, and that's what I cover in this video. Software engineering isn't dying, but it is certainly transforming. ~~~~~~~~~~~~~~~~~~~~~~~~~~ - If you want to learn how to build reliable and repeatable systems for AI co ## Транскрипт ### The "Death of Coding" Narrative [] Every six months, software engineering, or at least the coding part of it, is six months from being dead. I am sure you've heard this narrative too many times before. It's a pattern that has repeated itself the last couple of years. Like right here, we have Daario, the CEO of Anthropic, who I do respect a lot, by the way. But he went on the record March of last year saying that AI would be writing 90% of code in 6 months. And then come 6 months later in September, of course, it is not even close. And then he goes on the record literally just a week ago again saying that software engineers could go extinct this year in 2026 and it's going to be painful for a lot of people. So he said things like this before, but this time guys it's actually going to happen, right? Well, we'll see. We'll see what happens this year. And Daario is far from alone. All of the leaders in the tech industry are saying things like this. Like Jensen Hong saying kids shouldn't even learn to code all the way back in 2024. The Amazon cloud CEO saying most engineers won't even be coding. Of course, we have Sam Alman saying mastering AI tools is the new learn to code. The Microsoft AI CEO even going so far as to say this year that within a year and a half all white collar tasks will be automated. So crazy claims that honestly I think are pretty exaggerated and a lot of these tech leaders are very biased in what they're saying and honestly I think living in a bubble. So right now I just want to have a candid conversation with you. Let's look at the data and think about why are these tech leaders so biased and living in a bubble? What does this mean for us? What's the true power of AI coding assistance? Because there is a lot of stock in what they're saying. But there's just a massive difference between the capabilities of AI coding assistance and true adoption, especially at enterprise level. We'll cover that as well. And I never explicitly ask for comments on a YouTube video. But right now, I am cuz I really want to hear from you what you think about all this and if you think I'm wrong on any account here, like if I'm misinterpreting any data, if your experience is different from mine, cuz we're going to look at the data. I'll talk about my experiences, put that all together, and talk about what that means for us going forward as well. Now ### Why Software Engineering Won't Die [1:59] here's the thing. I am not saying that coding agents aren't powerful enough to replace most coding. I've done that for myself. I've built the system. So, coding agents are powerful enough. There is a lot of stock in what these tech leaders are saying, but there are two huge caveats that I feel like they're dismissing way too much. First of all, I've seen firsthand how low the adoption of these tools is at an enterprise level. You would be surprised because I teach both individuals and teams at larger companies how to leverage AI coding assistance effectively. And there is so much resistance in these larger companies that I just think there's no way even within five years most companies will be using AI coding assistance to write 90% of code. And then the other big caveat here is there's so much more to software engineering than just coding. And these skills I think are never going to go away. Maybe you've heard the new term intent engineering showing up on the internet recently. It's really just a lot of software engineering principles put into a package to apply to AI. And people are recognizing that we still need these skills even as coding agents are getting more and more powerful. So I could have made this video even a year ago, but what really put me over the edge to make it right now is a couple of comments that Boris Churnney made in some podcasts that people are taking the wrong way. So he was on the Y Combinator and Lenny podcast. And in both he said that coding is largely solved. Now if you watch both these podcasts, which they are great, and you really get into what Boris is saying, I can get behind it. But when you take this at face value, coding is largely solved. It's a very exaggerated claim that a lot of people are misinterpreting, thinking that software engineering is dying and that he's saying the same thing as Daario. But he literally clarified later on X that software engineering is still very important. He said, and I quote, "Someone has to prompt the clouds. Talk to the customers. This is very important. Coordinate with other teams. Decide what to build next. Engineering is changing and great engineers are more important than ever. " And so looking at this right now and Boris, he's the creator of Claude Code. So Daario is his CEO saying something that seems very contradictory. Software engineering is going to go extinct in 2026. Software engineering is more important than ever. So how do you wrestle with this? I mean these contradictions even within the same company is why I'm very confident in saying that these people are living in bubbles and they're very biased in what they are saying. And all we have to ### Financial Bias in AI Tech Leadership [4:21] do here is just follow the money. I mean it's a pretty simple argument. Like if we look at Enthropic here, they went from a $61. 5 billion valuation to a whopping $380 billion in less than a year. And so yeah, Daario is benefiting a lot when people are leaning more and more into AI coding assistance instead of writing code themselves and trying to be software engineers. So that you have this whole narrative of like don't even get a computer science degree. That's very helpful for people like Dario, also people like Jensen. So yeah, his net worth was only $3 billion in 2019. And then this is pretty outdated, but even in 2024, it was already at $90 billion. Just insane. And this makes sense. Nvidia is selling the chips, most of the chips for LLM training and inference. So, he is also benefiting a lot. And then, yeah, of course, Sam Alman, just as another good example here, OpenAI is valued at over $730 billion now. And by the way, Sam is in line for a 7% equity stake that is going to be worth over $10 billion. So, he's doing pretty good for himself with this whole AI craze right now. So with all of that being said, the truth right now really seems like yes, software engineering is going to evolve significantly over the next couple of years, but it is not going away. Maybe we will need less engineers, but the role is still going to be very important. And also AI is not going to be writing 90% or more of code for most companies. We'll get there eventually, but not like this, not in the next year. ### My Personal AI Workflow [5:47] Now, allow me to share some of my own personal experience here because I have a lot of it I think will really tie this together because I went from being that engineer that refused to use AI coding tools to now teaching thousands of individuals and teams at large companies how to use them effectively. So, back in 2024, I was one of those people that said no to all AI coding tools. I wrote all of my code by hand still. But finally, November of 2024 is when I installed Windsurf. That was my first AI coding assistant. And man was my brain blown. Like I wish I had used AI coding assistants for the last year. And so I didn't start doing a ton with these AI coding assistants. Like they just help me with writing a function or editing a single file. But over time, like the next 6 months, I built up my full system that I call my piv loop plan, implement, validate. I have a video on my channel where I cover it that I'll link to right here. So, I've gotten to the point now where I delegate all of my coding, at least almost all of my coding to the AI coding assistants because I put myself in the driver seat for the planning and the validation at the end, sandwiching the implementation. So, I'm not vibe coding. I'm still doing everything a software engineer does where I define the architecture and tech stack. I take requirements and translate that into the context for the coding agent. I do a lot of code reviewing and even fixing code by hand still. But I am using AI coding assistance as a primary driver. Like I said, there is a lot of stock in what these tech leaders are saying because the tools definitely are capable enough. That's not the problem. The problem is these exaggerated claims saying that all companies are going to adopt them. AI coding assistants are going to write all code in the next year. That's what I think is really not the truth here. So, there are two more big things that I want to cover right now. Both of which I'm confident you can get behind. The first is that AI coding assistants still require an insane amount of guidance to get the job done reliably. Like Boris said, the role of software engineer that you have to take on is more important than ever. The other thing is you would be surprised what adoption of coding agents looks like at an enterprise level. I talked about this a little bit already, but I'll share more of my experience and a lot of data that I found to back up what I have seen. So, both of these things pointing to the fact that all of these claims on the internet right now are very exaggerated. So let's start with the argument for software engineers. And honestly the ### The Evolution to Intent Engineering [8:02] best way to quickly explain why we need to be a software engineer so much still is to look at the evolution of prompt engineering to context engineering then to the new intent engineering. So prompt engineering, you've probably heard of this before. It's been around for a few years now. It's really just the art of wording things in the best way possible for the LLM to get the best single output. And so this doesn't really look like software engineering at all. It's really just wording things in the right way to LLMs. But then last year, this evolved to context engineering. So this is sort of a direct evolution of prompt engineering because we're going from single inputs to get single outputs to creating this entire system of context for our AI agents, especially AI coding assistants, giving them all of the context they need to plausibly solve the task at hand. So it's a lot more detailed and this is starting to look more like software engineering because a big part of the context ecosystem that we provide is going to be things like the architecture, the tech stack, the key decisions that we made around the feature or app that we want to build starting to feel like everything except for the coding is our responsibility because we are the ones doing the context engineering for the agents. And then more recently, this has evolved to intent engineering. Now, I'm not just trying to throw buzzwords at you here. I'm trying to make a point of how we are evolving more and more over time how we work these coding agents to sit more as that senior engineer role. So, the idea with intent engineering and Nate B. Jones covers this on his YouTube channel and his podcast. It's really good stuff. But the idea with intent engineering is not only do we want to give good context around our codebase and things like that, we also want to be very clear on things like the success criteria, how the agent can validate its own work, making sure we're super aligned on what we're building, so that not only is the code correct, but that the application or feature is actually what we want. And there's a lot of effort that goes into that. But now we're really getting to the point where it sounds like we're just a software engineer doing everything but the coding because we're defining the tech stack. We're asking the right questions. We're translating stakeholder requirements into the context ecosystem for our coding agent. There's a lot of work that we have to put into this. It feels like we're doing a lot of handholding because this really is what it takes for our coding agents to be reliable and creating a system that can be repeatable as well. Not to mention how important the validation strategy is for context engineering as well. Us doing the code reviews, defining workflows for the agent to review its own code. Again, so much of software engineering principles being built into how we use these coding agents that people are pretty universally considering the best practice right now for working with these tools. Which by the way, all these best practices around context engineering and intent engineering, getting good results with AI coding assistance, I teach all of that in the Aenta coding course in the Dynamis community. And so if you want to learn how to build a system to get reliable and repeatable results with AI coding assistants, definitely check out the course and the community. This teaches you how to be the software engineer that we're saying you need to be right now. And you don't even have to have a technical background to get started with this. But anyway, everything that I firmly believe in and teach and what we're talking about here with this evolution and really becoming the software engineer, this is why I just can't get behind the statement here that software engineers are going to go extinct this year. There's just no way that it's going to happen. Now, the ### Challenges in Enterprise Adoption [11:34] second big thing that I wanted to talk about is the enterprise adoption. So, like I said, I do a lot of trainings for teams at larger companies helping them adopt AI coding assistance. And there is a very wide range of people on these teams. There are some that are super gung-ho about coding agents. They just want to learn everything. And there are a lot of, especially senior engineers that are very resistant to these tools because they know the pitfalls. Large language models hallucinate. That's never going away. AI coding assistants introduce a lot of security issues actually in your codebase. if you don't have a good system and so if you don't know how to make that system and you know the concerns it's very hard to adopt these tools especially for production enterprisegrade software which by the way that's why I love doing these trainings because I teach the system that enables these larger teams to use AI coding assistance but yeah I've also seen how crazy slow the adoption is at enterprise level because there's so much corporate red tape there's procurement processes security reviews engineers that don't trust the tools in the first place like I said and so all this leads to a very slow adoption. And the other thing to think about is if you were to go and take the Agentic coding course in Dynamis, you would be immediately equipped with all of the rules and commands and skills, everything that you could apply to building your own software in just like one day. It does not take long to take my system and use it for yourself. But when I come in and do these larger trainings for companies, four hours, even six-hour workshops I've done before, I teach the entire system, but they have to now take that and create a standard for their entire team. They can't really use the tools effectively unless everyone is using them, at least in a pretty similar way. And so it can take days, weeks, even months to create this full system after the training. So there's just such a delay. That's why I'm saying that the adoption is not going to be that fast for most of these larger companies. And of course, I have a ton of data to back up the things that I've seen firsthand as well. So, first ### AI Usage Data and Statistics [13:31] of all, there was a survey done in 2025. Out of all engineering teams, 90% of them now use AI in their workflow. So, it seems like a very high percentage, but this number can mean a lot of different things because there are a ton of companies out there that they want to be AI enabled or AI native, whatever they call it. So they buy a bunch of licenses for cloud code or GitHub copilot or whatever, but it's not like their teams are using them on a daily basis, know how to actually leverage them. So 90% now use AI in their workflows, quote unquote. Uh but then if we look at the Stack Overflow survey from last year, only 51% of professional developers are actually using AI tools daily. So again, this huge disconnect that we have. And just another stat for this here, if I haven't given you enough already, 76% of executives believe their teams have embraced AI, while only 52% of engineers would say the same thing. The license is available. The company is AI enabled. But these engineers know there's so much more they could be doing with these tools. They're not using them daily. So they say they're not AI enabled, even though their leadership says they are. And just one more here. 21% of AI licenses remained underutilized. So just continuing to go with this narrative here where the tools are becoming available but they're really not adopted end mass yet and it's going to be so long until it is the case and so yeah maybe at some point when we reach AGI whatever that is whenever that is software engineering will be dead but I think that generally the role of software engineer is just as protected as a lot of the other roles that people are scared AI will replace but it's just going to take an insane amount of time for any of that to be a real concern. And I think that's the last hot take that I have for you here. So definitely let me know in the comments now that you've gotten through everything. What do you think about all the data that I showed you? My general thoughts and experiences. I'm really curious if you've seen anything else as well. And ### How to Thrive with AI [15:23] so the last question to answer really quickly here is what should we do about all this? Like what do I practically do given all of this information? Well, the first thing I'd recommend is never vive anything into production. We have to be the software engineer. Coding agents are capable, very capable, only with the right system that we engineer. That also does mean that it's important to keep your technical skills sharp. I disagree with Jensen Huang that kids should not learn how to code because learning how to code is very important to be a good engineer overall, even the skills outside of coding. And so you can review the AI generated code because validation is also not going away anytime soon. And then also just to help you maintain your code bases. It's good to use your AI coding assistant as an educator, not just the coder. If you can work with it constantly to understand all the changes, how your codebase is evolving, that prevents you from getting to the point where you're relying on the coding agent so much because you really don't understand all the decisions that it's made along the way building feature by feature. You don't want to lose control of your codebase. And even if you're not vibe coding, you're trying to do a lot of planning and validating, if you're not asking the right questions to understand what it's doing, you're going to get to the point where you're just way too reliant on your coding agent. The people who will thrive are the engineers that use AI coding assistance as a force multiplier while still maintaining a deep understanding of their craft and their code bases. And the people who struggle are the ones that have given up agency completely and they just crumble when they are no longer using AI to help them with anything. Because when AI makes a mistake, they have no fall back. So those are just a couple of points. I hope that helps you just think about things you should do right now, how you should maintain your craft, even using AI coding assistance to help you be more technical if you aren't right now. And so if you appreciate this video, you're looking forward to more content on agentic engineering concepts around context engineering and intent engineering, I would really appreciate a like and a subscribe. And with that, I will see you in the next video.