Zuckerberg DROPS AI BOMBSHELL: The End Of Software Engineers

Probably in 2025, we at Meta, as well as the other companies that are basically working on this, are going to have an AI that can effectively be a sort of mid level engineer that you have at your company that can write code. And once you have that, then in the beginning it'll be really expensive to run and you can get it to be more efficient. And then over time we'll get to the point where a lot of the code in our apps and including the AI that we generate is actually going to be built by AI engineers instead of people engineers.

So that clip has been doing the rounds in the AI community. And in this video I'm actually going to take a different perspective than the current ones that you're seeing, because I don't think most people are understanding the true dynamics of exactly what's being said here, even for those just interpreting the clip at a face value. So it's really important to understand the nuances of exactly what is being said here and that's what this video is going to dive into. So this is of course Mark Zuckerberg, the CEO of Meta, and he's talking about the fact that in 2025 AI is on ramp to essentially change in terms of how they get their code done. Now, I agree with him to an extent, but when we actually take a look at things, things might not be as they seem. So in the first part of this video, I'm actually going to go with the majority here and say that, you know, it is a potential scenario where the majority of code being written is done by people engineers. Now this is not the only person at a top billion dollar company that has said this. But later on in the video I'll get into why most people are missing the most critical part about this statement and why things might not be as they seem. But as I was saying, okay, this is not the only person at a billion dollar company in the higher ups that has said this Amazon earlier. Well, not earlier this year because it's 2025, but last year Amazon said in 2024 in a leaked conversation, meaning that we weren't even supposed to know this, but they said the Amazon cloud chief actually told employees that most developers could stop coding as soon as AI takes over. And they shared their thoughts during an internal fireside chat in June. And we actually got this knowledge because Business Insider obtained a recording of the meeting. Now this was something that was pretty crazy. I covered this in a video and it was largely one of the most viewed videos on my channel in terms of the post AGI Economics series where I talk about how technology is developing in a way that will change the economy and, of course, the current roles that we currently do have. And the reason I want to bring this back into the context of this video is because this is a statement that we've seen time and time again. Now, on the surface, of course, it does sound bleak, and there were more statements that were, you know, a little bit in more detail. So this is a screenshot from that previous video. But you can see right here, he says if you go forward 24 months from now, which is two years, and that would be June 2026 or some amount of time, he said, I can't predict where it is. It's possible that most developers aren't coding. And that, of course, you know, is the guy who became Amazon CEO in June. And he said that, you know, coding is simply the language that we talk to computers in. It's not necessarily skill, you know, of itself. The skill is how do I innovate, how do I build something that's interesting for my end users to use? Now, I personally do agree with this to some extent, because what I've actually done is I've actually looked at the numbers and at the very specific details. But another thing to note here is that he does state that, you know, most developers aren't coding. It doesn't say that they're not working, just simply that the role of developers may change. Now, like I said before, let's ignore the actual facts of what was said, but look at the numbers in where these systems are in terms of their benchmarks, because that is largely going to give us a quantitative view of where these systems are, regardless of what this CEO said or that CEO said about the specific thing. So four to five months ago, in that same video, we can see that these benchmarks that we were looking at here, and this is the SWE bench, a software engineering bench. And you can see that, you know, in 2024, at the start of 2024, it was around 7%, and during that time it went up to around 40 to 38%. And I'm doing this so I can show you where we are today. So that was five months ago. The highest score that we got was around 38.8%. I do think that the cosine genie one, I'm not sure, but it's not on the Leaderbo anymore. So maybe that record was taken down due to an error or something like that. But let's say that four to five months ago, the highest record that we got on the SWE bench, which is, of course, in this video, how they're talking about. People are not going to be coding because AI will fulfill that role. So we can look at that time five months ago and see where the improvements are now. And we can see that now where we are, we're actually at a 62% on the SWE bench verified. So it's almost a double in terms of the actual performance on these models. Right, right here you can see that very recently, towards the end of 2024, code story midwidth agent +swe search managed to get 62% of the swe bench. And this bench is a pretty, pretty hard bench to crack because, you know, it's not in the training data. It's not something they've seen. This was actually a pretty decent benchmark. And this is something that you can see slowly but surely every single month it's around, I would say, 2% or so that we increase or maybe like 5% or so. So it's quite likely that in two years this benchmark might be cracked. Now, what's crazy about all of this is that if we actually take a look at the most recent model, we can see something that is crazy because remember how I said it was almost, almost a jump? We can see that right here on the O3 benchmark, and I do apologize for the quality of this, but we can see that on the SWE bench, the verified benchmark, as we spoke about, we can see that O3 gets 71.7%. And I know you can barely see that, but it is 71.7%. So in just a year, the benchmarks have essentially doubled, which means that if we look forward 18 months by now, is it quite likely that we are going to have systems that could potentially get close to 100% or that level? Now, I do think that that is possible, but in that time, are we going to have models that are cost effective enough for them to be that useful? That is going to have to be something that we're going to have to discuss because, of course, cost is something that is really expensive. And we do know that right now, O3 is ridiculous in terms of the price. Now, of course, as you do know, as time goes on, these models will continue to get cheaper and cheaper. But take a look at what they've said about the O3 model in terms of its coding performance. We can see that O2 in terms of, you know, well, it's O3, but, like, this is the second iteration of their Model we can see that on competition code forces. This is a model that is scoring higher and higher. We can see that O3 is in like the top 1% of coders, which is something that's absolutely crazy. And it's only their second iteration of the model, whereas the first iteration 01 preview, we can see it wasn't that great. And like I said before, it's quite likely that 04, which will actually be the third iteration of searching models, is going to be released sometime in 2025. So it's quite likely that this level is going to get even better. So take a look at this short video because I think it's rather fascinating to, you know, see what they talk about. But I'm actually going to circle back on this because it's pretty crazy what OpenAI are talking about when they talk about software engineers. And they actually talk about that in a moment. But take a look at what they say about O3.

So on the left hand side we show the coding evals. So it's like code forces yellow which measures how good a programmer is and the higher is better. So as we can see on the plot with more thinking time, O3 mini is able to have like increasing evil all outperforming O1 mini and with like median thinking time is able to measure even better than 01.

Yes.

It's like for an order of magnitude more speed and cost we can deliver the same code performance on this evaluate or even better experience. Right. So although it's like the O3 mini high is still like a couple hundred points away from Mark, it's not far that's better than me. Probably just an incredible sort of cost to performance gain over what we've been able to offer with 01. And we think people will really love this. Yeah, I hope so. So on the right hand plot we show the estimated cost versus code versus yield trade off. So it's pretty clear that ostreet mini defines like a new cost efficient reasoning frontier on coding. So it's achieved like better performance compared to better performance than okay with a fraction of cost.

Amazing.

With that being said now like I.

Said before, we're actually going to take a look at what they've been saying about software engineers at OpenAI. So this guy that you're seeing right here, this is Noam Brown to this guy right here, this is the guy that actually worked on reasoning at OpenAI. So you know the O1 series, that is his bread and butter. He was essentially the main person around that reasoning idea and how searching for new possibilities could actually lead to improvements. And in this short clip, in this interview, he actually talks about how O1 actually acts already like a software engineer, which could be a glimpse at the future of the roles of AI agents and AI systems within these large companies. And remember, I've said that, you know, this is something that is quite expensive, but when we look at the average salary for a software engineer, provided that it is somewhat cheaper than the $250,000 that you're paying these people per year, it's quite likely that as long as it can become cheaper than that, it's quite likely that they will find value in these AI systems.

O1 does feel, at least to me and I think other people on our team like a better coding partner than the other model. I think it's already authored a couple of PRs in our repo and so in some ways it is acting like a software engineer, because I think software engineering is another one of these STEM domains that benefits from longer reasoning. I don't know, I think that the kinds of rollouts that we're seeing from the model are thinking for a few minutes at the time, I think the kinds of software engineering job that I do when I, when I go and write code, I think for more than a few minutes at a time. And so maybe as we start to scale these things further, as we start to follow this Trend line and let 01 think for longer and longer, it'll be able to do more and more of those tasks and we'll see, you'll.

Be able to tell that we've achieved AGI internally when we take down all the job listings. And now this is the part of the video that I'm actually going to flip back on what I've been saying for the entire start of the video, because I think that whilst yes, the people like Mark Zuckerberg and the internals at Amazon saying that, look, of course we are probably going to have a situation where we don't have any human coders. That is quite the statement. I think that even if we get into a scenario where they're not hiring as many software engineers, I still think that if you are currently a software engineer, you're probably going to be valuable more now than ever than you probably ever were. And that is not just me stating that to make you feel good. I've actually been doing some deep dives on this and the research is actually quite surprising. Like it even surprised me as I was researching for this video. So take a look at this because this is where you want to pay attention if we actually take a look at what the WEF are saying. The World Economic Forum, they actually spoke about the future of jobs report on 2025. Now I'm not going to dive into the entire report because this is going to be something that is, you know, a 20 minute, 30 minute video because there's just so much to talk about. But take a look at this right here. And number four in the top fastest growing job section, we have software and application developers. Okay, so software developers. Software and application developers. That is one of the top fastest growing jobs by 2030. Now remember, the World Economics Forum's job is essentially to predict and prepare the world for exactly the future that we're heading into. And so I don't think there's another organization that is going to be better suited to predicting the roles of the future. So the fact that they saying that, you know, the fourth fastest growing job role in this AI society in this post AGI economy is going to be software and application developers, I think that tells us something about the role of software developers in the future world. And I don't think it is one that where software developers are completely useless. I actually think it's the opposite. And that's because of a few reasons. So the main reason of this, and you might be thinking, well, if you know these AI systems are going to be, you know, software developers, what the hell are software developers going to be doing? So let me introduce you to a paradox that has been spoken about, and I know no one else is talking about this, but let's get into it. So the Jevons Paradox is an economic principle that basically states that when a resource becomes more efficient to use, demand increases, leading to more resource use. Okay, think about that for a moment. When a resource becomes more efficient to use, the demand increases, leading to more resource use. So if a resource becomes more efficient, it becomes cheaper to use when, which makes it more attractive to consumers and industries and the increased affordability leads to higher consumption. So let me explain to you guys exactly what this means for software developers and why the industry is probably going to expand. In my genuine opinion, stage one is the low barrier to entry. If we now have a scenario where it is a very low barrier to entry for anyone to be able to code, that essentially means that it's quicker to develop applications and this lowers the cost to creating software. And as a result, more companies and individuals can afford to develop software for their specific needs, even for niche or experimental products that were too expensive to justify, which means that if more people now have access to, you know, software because it's cheaper, that now means that we have number two, increased software use cases. With AI enabling faster and cheaper development, businesses can launch more startups, they can launch more apps, more platforms, more products, and this leads to a broader demand for software. And as software becomes more embedded in more industries, what this means now is that it means that we actually need more software engineers because we're going to need software engineers to supervise, we're going to need them to debug the code, we're going to need them to ensure that the AI generated code is correct. We're going to need them to design the architecture and the purpose of the software. We're going to need them to handle very specific edge cases, right? Complex logic that AI might struggle with. We're going to need them to create and refine AI models themselves, which is of course a growing demand. So all of this is to say that whilst, yes, it might seem like on the surface the software developers are completely done and I guess you could say that AI could supervise, AI could debug and AI could design the architecture, whilst yes, that is still possible, it doesn't discount the fact that now we have a low barrier to entry, which means right now we have so many people developing apps that if you want to actually fix your app or take it to that next level, you're going to actually need a software developer who knows the complete backend, who knows how to code, who knows how to connect this up. And that's something that I've seen firsthand with individuals who are in this space. There's actually an increased demand. Like I've been building certain things, certain automations and certain pieces of software for things within my businesses and I've oftentimes had to drop on a consulting call to with a software developer for an extreme rare use or edge case that an AI system simply cannot fathom. So it's something that I personally do believe that whilst yes, you might get to the fully autonomous SWS because of course companies are going to be rapidly trying to drive down that cost, I still think the entire ecosystem as a whole is probably going to be putting those individuals at a significant demand. Now remember what I spoke about, like I said before, the fact that this isn't something that. I'm just saying, if we look at Apple's recent job posting for a software engineer, Apple actually recently spoke about how they need experience with AI agent based frameworks like Crewai and LangChain. So like I said before, because there's all these things like Crewai, there's like LangChain companies are actually hiring people to become specialists in these roles. And the crazy thing is that if you are a software developer or if you have any experience related to that, a lot of that stuff is very similar to using agentic workflows. So. So you're actually more valuable now than you were back then. So that's genuinely my honest opinion. Because if Apple is now hiring and on the resume, they're looking for people who actually have experience with AI agent based frameworks. And that's of course something that is on these software developer jobs. It's an indication that the industry is of course expanding now. We also can take a look at what was said during this conversation on the YC YouTube channel and they actually speak about how software developers could potentially actually have more leverage now.

Now it is actually more leverage. It is far more leverage to know how to code than ever before. And it's actually even more important and it will make you way more powerful. Like you don't have to be all the way in the weeds on everything. You actually are going to be able to like orchestrate this giant army of agents. And I think of Mickey Mouse and Fantasia just like, you know, like learning this new magical sort of ability. And like, you know, suddenly all the brooms are like, you know, walking and talking and dancing and it's this incredible menagerie of being able to build whatever the heck you want whenever you want. Just like, like literally from any computer, from any web browser.

Yeah, I try to come up with like a Moore's Law type type thing where it's like the return on. On learning code is like doubling every six months or something like that. So learning code a little bit in, you know, 2020, you know, was not that useful because you would still you will, you get blocked, you wouldn't how to deploy something, you wouldn't know how to configure something. Let's go to 2023 with chat GPT, learn to code just a little bit. We'll get you Fairly far because ChatGPT can help you. And then 2024, learning to code a little bit is a massive leverage because we have agents like this and others and there's a lot of really cool tools out there like cursor and others that will get you super far by just like having a little bit of coding and just extend that forward like six months later you're gonna have even more power. So programmers are just on this massive trajectory of increased power.

So this video was kind of like a 5050 on one side you have the CEO saying that look, there are going to be no more programmers that are writing our code. Maybe, maybe those programmers are not writing the code. But I can guarantee you that where we're headed is to an area where, where it's quite likely the software engineers are probably not going to be engineering the software anymore. It's quite likely that you software engineers are going to be orchestrating multiple AI agent systems. It's quite likely that you're going to be orchestrating these AI agent swarms, putting them in scenarios where they're going to be able to get 10 times the work done than a traditional software engineer. And that is where I think things are going. And we can see this by the fact that number one, we have the Apple's job posting showing that for software developers, they now require experience with AI agent based frameworks like Crewai or LangChain. And the World Economic Forum is actually predicting that the fastest growing job sector number four is software applications and developers. So in theory, this is why I said it's a nuanced point that most people did miss. Most people like, oh, software developers, yada yada yada. Trust me guys, as long as you can learn AI agent workflows and that kind of thing, you're going to be having the most leverage possible because you're going to be essentially able to manage digital employees. They're going to get more done than before. So that is my honest opinion, source developers might not be writing code, they will be of course orchestrating these AI agents, which is where we're going to be going to next. And of course hopefully this video shed some light on the scenario because I saw so many videos that simply have the wrong opinion and clearly missed this entire nuanced point of view because they didn't see these certain points that currently were out there. So with that being said, if you guys enjoyed this video, hopefully this provided some value to you and I'll see you in the next one.

Hope you enjoyed.