Coding has turned into a slop factory

You know, I really didn't want to make such a negative video. I'm usually a much more positive person, but I just hate what's happened to our industry, right? I just hate what's happened to the coding industry right now, and I need to talk about it. And the reason I'm even making this video, and the reason which sparked the whole idea in my head, is this tweet over here from the VP of Engineering over at Spotify, where he's basically saying that they're doing 4,500 production deployments every single day, and this is something that he's bragging about right now. My only question is, how is this something to brag about? Like, you're doing 4,500 deployments every single day, all for the end user to not even notice it at all. Like, there is no value in this for the end user. Like, I use Spotify a lot, right? I pay for the premium. I listen to it all the time. I'm more addicted to Spotify Premium than most of you guys are to the hub, right? I use Spotify a lot, and I have not noticed any new features, any bug fixes, new features, faster response times, faster performance, any of that stuff. So where exactly are these 4,500 deployments coming from, and where are they going, right? Especially if the end user, especially if a heavy user like me, doesn't even notice them happening, right? Like, what exactly are the changes you're making if they're not some sort of bug fix or new feature or faster performance or something like that, right? And doesn't all of that just make AI seem completely and utterly pointless from a business value perspective? Like, my only question to the VP of Engineering over at Spotify is, what exactly is the end result that your users are getting? Like, what exactly is the value that you're getting if they don't even notice these 4,500 changes, right? If they don't even notice a positive change in the app, forget positive change, they don't even notice any change. So where exactly are these deployments coming from? Like, I'm willing to bet Spotify is spending millions and millions of dollars in LLM tokens right now, and the end user doesn't even notice them. There are no faster response times. There's no new features. So what exactly is the point over here? And I mean, I've covered token maxing way too many times on this channel, right? We've covered Uber burning 3.4 billion dollars in just four months. We've covered Meta spending, I believe it was 900 million dollars just on Claude tokens. We've covered Microsoft, we've covered Amazon with their token usage leaderboard and everything. And it looks like Spotify is going to be the next victim because what the hell is going on over there, right? And it's not just enterprises and big companies that have fallen for this whole AI psychosis. It's also individuals. And a lot of them are just AI grifters that you see on Twitter. People like, and I hate to say this because I don't want to spread negativity for no reason, but people like Peter Steinberger, the guy who made OpenClaw. Like, trust me, it's not the guys who are using Character AI and AI companions and girlfriends and boyfriends or whatever. Like, those guys are pretty fucked up in their own way, right? They're pretty darn creepy, but even more so than them, it's people like Peter Steinberger who say shit like, "Oh, you should be designing loops that prompt your agents." Like, what does that even mean, right? And if you don't have the context for this, Peter recently tweeted something like, "Oh, you should be designing loops that prompt your agents, and you shouldn't be writing prompts anymore." None of that makes sense, obviously. But, you know, according to Peter, you don't even need to write prompts anymore. You just need to write loops, whatever that means. I don't know what that means. None of his audience knows what that means, and I don't think even Peter knows what that means. But yeah, according to him, you should be designing loops that prompt your agents. You don't write prompts anymore because the LLM can just figure out exactly what you want all the time, perfectly, with 100% accuracy, and you don't need to do anything else. And just so you know, this is coming from the same guy who is currently being paid billions of dollars in LLM tokens by OpenAI. All right? So when he tells you that you need to be designing these infinite token-burning agentic loops, maybe you shouldn't be listening to him, right? But just remember, he's being sponsored by Sam Altman, right? He's getting billions of dollars in tokens from OpenAI. He's not some regular dude who's paying for all these tokens out of his own pocket, right? So if he tells you that you need to make all these loops that burn billions of tokens just to produce slop faster, he's probably not the right guy to listen to, especially when the biggest achievement of his career is building some glorified AI agentic wrapper around other models that helps you do basic shit like check the weather or, you know, send a few messages, which is stuff that Google Assistant and Siri could do back in the day and with a lot fewer hallucinations. Like, seriously. And if you guys can't tell, I'm talking about OpenClaw, which is this AI assistant app that Peter vibe-coded. Not even built, vibe-coded. It's a completely useless app. Nobody's using it for any kind of serious work. And if you just step outside of tech Twitter and the echo chambers of AI for just a minute, you'll see nobody even knows that it exists. And my only question to Peter is, if it's such a revolutionary app, why hasn't it gone completely viral yet? Why hasn't it just blown up? Because apparently it's good enough, according to Peter at least, it's good enough that you can just automate every single task on a computer, right? You can just offload all of your computer work to an AI agent. If it's that good, why isn't everybody using it? That sounds like something every single human on the planet should be using, right? Why aren't they? Maybe because it's not as good as you seem to think. Maybe because it's not as good as the marketing seems to tell you, right? Like, if it was that good, it would have just blown up already, obviously. Because if you have that good of an app, everybody should be using it. But instead, nobody's using it, but everybody's hyping it up. Why? Because tech Twitter, especially with AI now, is just a big fucking echo chamber where you just repeat what everyone else is saying without even trying the things out. The whole industry has just drank the AI slop Kool-Aid so fucking deep into their throats that they don't even see that they're not even building good software anymore. Like, they're just producing slop. Nobody's trying to create good software anymore. Everybody's just creating slop. I've yet to see any good consumer apps built with AI. Like, there's a few out there, but there's not a lot. People are building better tools, better models, better harnesses, better skills, all that stuff, better MCP servers, but they're not using those things to build real consumer apps. Like, they just keep chasing the next big tool instead of using the tools they have to create something real with them. First, it was Markdown files, like you had to build this huge knowledge base of Markdown files that the agent could then derive context from. Right. Then it was Ralph Wiggum loops or whatever, which I still don't know what that means. And I'm completely fine with that because it got outdated like three days later. Then it was OpenClaw, which we already discussed is completely useless, and now it's infinite token-burning agent loops, which, again, I don't know what that means. Probably Peter doesn't even know what that means, and he's the guy who came up with the entire concept. It's like, okay, I'll give you a really intelligent frontier model. Think something like Fable 5 or even Opus 4.8, and I'll give you a really great harness, right? Like something like Codex, which is a really good harness. And we'll just assume that Codex can work with a cloud model, right? Just for the purposes of this little experiment, thought experiment, if you will. I'll give you a really intelligent frontier model, and I'll give you a really great harness. What are you going to build with that? Oh, I'm going to go and train a new model that's even better than the previous one. Or I'm going to build a new AI chat app, my own chat app, and that's going to help me write code even faster. And then I'll create another chat app using that chat app that's going to help me write code even faster than that one. And then I'll create another one, then I'll create another one, and it just keeps going on forever. You're not using these tools to build something real. You're just building the next version of the tool just to produce slop faster. The point is not to write code as fast as possible. We've already solved that. These AI agents can produce millions of lines of code in hours, days, and yet there's no good quality software out there. Like, we're not increasing the level of, we're not increasing the amount of good software in the world. It's like you keep sharpening your axe until hell freezes over, basically. You need to get out of that sharpening-your-axe mentality. It's like college students who pursue higher education, like some PhD or something, just to delay their unemployment. Like, it's going to happen sooner or later. You may as well just step into it, right? Like, you can't stay stuck in the preparation phase forever. You gotta, like, at some point, you gotta just get out there, use the tools you've already got, and go build something real with them. And I mean, this reminds me a lot about how programmers back in the day, before LLMs, used to build tons of side projects, and then they would struggle to find users for it. Why? Because none of these people have some sort of business sense. And we're seeing something very similar happen now because none of these AI grifters have any business sense. They don't know what they're doing. They don't know that they actually need to create good quality software to, you know, get users and get paid. Right. I mean, just look at Copilot and Windows, right? It's completely useless. Nobody likes it. Nobody enjoys using it. Nobody is even using it. People are trying to disable it and get rid of it from their system. And yet Microsoft will keep shoving it down their users' throats. Why? Because they're just so deep into their AI psychosis, right? And even Microsoft, a trillion-dollar company, can't find a good use case for AI for normal everyday users. At least not one that's just utter slop, right? And that's something very interesting, and I think you should pay attention to that. But yeah, before we get sidetracked, it's like there's a new way to write slop faster every single week at this point. Markdown files, Ralph Wiggum loops, prompt engineering, agenting, you know, coding agents, whatever. And now it's loop engineering. And I mean, I'm all for chasing efficiency and getting better tools and everything, but what exactly are we achieving with all of these tools? Well, for one, GitHub goes down every single day at this point. GitHub never used to go down back when I was learning the code back in 2019. It just never happened, right? AWS and Cloudflare go down all the time now. Like, I blink my eyes, and there's a new AWS outage, and the entire internet has just gone down. And this is something that never happened before LLMs. And, you know, I could be wrong because I haven't looked at this, looked at the statistics, right? But this is something that never happened before back when I was learning the code, way before LLMs. It's just something that never happened. Like, AWS used to be very stable, and now it seems like every other week I try to go on some random website, and it's just down. And I just go check the internet, and it's like the entire internet has just went down because AWS went down. And this is something that never happened before. And if that wasn't enough, Uber. You didn't think I would forget about Uber, right? Uber has burned 3.4 billion dollars in just four months, right? Their entire 2026 budget, they just burned through it in just four months, spending all of it on LLM tokens, and they have nothing to show for it. And that's not me saying it. You can look at the head of engineering over at Uber. You can look at his tweets. He'll tell you he cannot connect any of those LLM tokens usage to real features or real bug fixes or anything. He doesn't know where all of that money and LLM tokens went. Spotify is doing 4,500 production deployments every single day, and nobody even knows where they're going because there's no new features, there's no faster response times, there's no fewer bugs. They're probably spending millions of dollars in LLM tokens. I haven't checked the official figures. They haven't disclosed it, but they're probably spending millions of dollars in LLM tokens, and their users don't even notice it because nothing is changing in the app or in the web version or any other software that Spotify is shipping. To be honest, I wouldn't be surprised if most of those deployments, if 4,000 of those 4,500 deployments, was just some intern at Spotify, just randomly restarting the Spotify servers again and again and again, just to get us to 4,000 deployments, right? I wouldn't be surprised if Spotify was doing that because I don't see how you can make 4,500 deployments every single day, right? Every single day. And somehow nothing in the app or the website even changes. I don't see how anyone can do that. Like, this whole timeline just sucks, man. The coding industry right now just absolutely sucks. And if you told me back in 2019, when I was learning the code, that this is what the industry would become, I probably wouldn't have learned the code in the first place. I probably wouldn't have started because, like, what are we doing over here, man? We're producing so much slop, nobody's using it, and we're just destroying the internet, destroying society. What? What exactly are we achieving by doing all of this? Fuck all. Nothing. Anyway, before this video gets way too long and I just yap for another 30 minutes, I should probably end it over here. If you resonate with anything that I just said in this video, please, I'm encouraging you, don't indulge in this slop production, right? Don't indulge in this slop production. Produce good software, high-quality, clean software that people genuinely want to use, all right? If you genuinely solve their problems, they're going to use your app, no matter what it's all about, right? Even if it doesn't have AI, even if it doesn't have the fancy tech, right? And please don't listen to these AI grifters on Twitter and everything. They don't know what they're talking about. They are so deep into their own psychosis that they don't know what's right and wrong anymore, all right? Just trust me on that. Anyway, yeah, that's all I want to talk about in this video, and I'll catch you in the next one, hopefully with a much more positive outlook on the entire industry.