Coding Before AI Wasn’t The Stone Age

Hey guys. So, we all know that software development is a lot different than it used to be. You know, I started learning to code around 2007 and aside from the languages themselves, virtually everything has changed from the process to the tooling. But it's changed over what I would say is a couple of decades. Uh not overnight. And I see a lot of people acting like LLMs and recent AI tools have just completely changed and revolutionized everything in the past couple years. Yes, in 2022 we probably couldn't imagine the code generation that we have from some of these agents, but other than that, the processes are are not all that different. And I came across this Reddit post of a tweet talking about development 3 years ago versus now and how much better it is today. So, we're talking 2022 compared to 2025. And I wanted to just take a look at this narrative, look at some of these points with which some I agree with, some I don't, and just offer some insight and yeah, a little bit of criticism uh of some of the points. And this isn't meant to be like a negative video trashing someone else's opinion. I'm not mentioning names or accounts or anything. This is just for a discussion. And of course, most of this is my opinion, and you're welcome to disagree. So, let's get into it. [music] All right, guys. So, I just wanted to take a second and tell you about today's sponsor. So, Savella is an all-in-one solution for hosting your apps, your APIs, static sites, and they support just about any popular language and and framework that you can think of. And I myself have been doing web development for about 20 years now. And I remember the the absolute nightmare that came with deployment and DevOps up up until really just a few years ago. But Sabella is a platform that makes it super easy. Gives you a simple UI to work with to deploy your applications as well as spin up databases like Postgres in in really just a couple of clicks. And you can deploy from GitHub or GitLab or from a Docker image. They offer free static site hosting and then app and database hosting from $5 per month. And then obviously you can scale as needed. And I don't take on sponsors that I don't believe in. Sabella has a great service, great product. So, check them out and the links in the description below. All right, so let's take a closer look at some of the points in this tweet. Some I agree with, some I don't. And I do realize the author was probably being somewhat hyperbolic, but I think it's worth examining anyway because a lot of people actually believe this stuff and and stuff even crazier. So, the first one is right, we were writing every line of code ourselves. So even 3 years ago, even 10 years ago, developers weren't writing everything from scratch. So if you think about what we had, we had code snippets and templates that we refined over the years, although we did initially write those. Um, Stack Overflow with millions of tested solutions that a lot of people would just copy and paste. GitHub full of open- source code to use and reference and learn from. Package managers where you install libraries and you never even see that code really. Uh, boilerplate generators. you had CI CLI tools that scaffold entire projects for you and and even uh you know autocomplete in your IDE was pretty intelligent in 2022. So the idea that we were you know coding a notepad manually typing every character like monks copying manuscripts just really isn't uh that accurate. So and good developers have always been good at leveraging existing solutions. That's literally part of the job. So, the next one is spending hours fixing missing semicolons. So, I think this one is is borderline insulting to developers, but in his defense, I think this one was was really hyperbolic. I hope so. Um, but llinters have been around for a long time. And even just ESLint and prettier, which is what a lot of people use today, it's what I use, has been around for much longer than 3 years. Um, even just VS Code on its own was showing missing semicolons before you save the file. Uh, again, probably being hyperbolic, but if you were genuinely spending hours on semicolons, that wasn't a tooling problem. That was, you know, more of a fundamental misunderstanding of of your development environment. So, searching Stack Overflow for answers. Now, this one I agree with. We obviously did that a lot for years. Um, I do think it kind of contradicts the first point depending on what was meant by writing every line of code yourself. But what I really want to mention for this point is just about Stack Overflow in general because it's framed like it's some like it was some tedious ordeal and maybe compared to AI it was, but it also had some real value that AI doesn't in my opinion. So it had vetted answers with up votes and real community discussion. There was multiple approaches to the same problem by different you know multiple people uh not just one way explanations of why something works not just copy paste the code real developers sharing you know tested solutions and they would cover edge cases of course some of it was really toxic too and and was horrible but there was a lot of value to it and yes AI gives you the answer faster no argument there but Stack Overflow gave you context debate understanding and there's real value in seeing how other developers think through a problem, not just getting a solution spit at you by a machine. So, in my opinion, I think Stack Overflow still does have a lot of value. Um, so the next one deploying was hope it runs on production. Now, this is probably one of the most misleading narratives in tech right now. It makes for a great tweet, but it completely ignores how professional development actually worked. So we had staging environments and even back when I was building PHP applications in 2008 uh I always had a staging environment that mirrored production. It wasn't perfect and and you did have to configure a lot of stuff manually but the concept of you know testing before production wasn't invented by Versel in 2023. By the mid uh 2010s you know we had Docker containers for consistent environments. So the whole like works on my machine problem was already solved for for most teams. And um if if your code worked in staging, it worked in production because they were configured identical. Uh and and in 2022, CI/CD pipelines weren't cutting edge, they were standard practice. So you know, Jenkins, Circle CI, CI, Travis CI, these were tools that were mature and widely adopted and you literally couldn't deploy if your test failed. So that, you know, uh, runs on my machine, it's mostly like influencer Twitter stuff. Now, if the original tweet had said 1999 or or 2005, then yeah, they'd have a point and they'd have a point with pretty much all this stuff. But acting like we were just cowboys pushing code to production and praying in 2022 just wasn't a thing. Uh, that was decades earlier. So, let's look at what is what the tweet claims is so much better about development in 2025. So AI writing code for us. Now AI is amazing. I'm not someone that's going to, you know, trash it at a fundamental level. In fact, quite the opposite. I use it in all my projects in some way, even if it's just for like reference and learning. Autocomplete. Um, it's fantastic for boilerplate code and tedious stuff that you've done a million times. However, this one point brings on a whole new set of big problems that we didn't even have three years ago. Uh the problem, well there's a few, but one of the biggest is people just vibe coding serious projects. Just pasting AI output, not even pasting it, just letting it write it without even looking at it, not even opening the file. And that can just burn your project to the ground before it even gets up and running. I mean, you rack up so much technical debt. And uh again, this wasn't something we had a problem we had years ago. Uh I think AI is a tool and you need to understand how to use that tool properly. So next one, errors get explained before you even Google them. I mean, your IDE was doing this with intelligent error messages. AI does make it more, you know, conversational. Um, and and that said, I I will give credit here. AI can really help you understand the specific errors that you're getting in specific context where in the past, what we would do is look for a similar error on Stack Overflow, and you'd have to adapt to your situation. Now you can just paste your exact error, line numbers, file names, all that and and get a a tailored explanation, which I think is great. So docs feel more like conversations than manuals. So I think he's absolutely right with this one. Um you don't get a static page where you're left scratching your head trying to figure out which which section applies to you. You can actually have a conversation, ask follow-up questions, and actually learn. And I think that's a real improvement. So uh and but this is important. Not everyone does that. Uh, a lot of people get lazy and don't even try to understand the errors or don't even read them. Just copy them, throw them in the AI, let AI give you the solution and move on. And that's not learning, that's outsourcing your your thoughts, your thinking. Uh, let's see. So, the last one is deploy is usually just a single command. So, this has been true for years. Um, if you remember, well, Heroku is still around, but I I haven't used it, but I used to always use Heroku. git push Heroku main and I was literally doing that in courses years and years ago. Um even Verscell and Netlefi's continuous deployment it's not new you know these were C CI/CD pipelines were mature long before chat GPT uh if he said 2007 then yeah dealing with C panel FTP uh manual server configuration that was gen genuinely a nightmare um you know you could do things correctly but there was a lot of manual labor involved by 2022 though one command deploys were were already really the standard so I know the original tweet tweet again was probably hyperbolic. I get it. But my main point is that these AI tools, they are changing software development, just not as drastically as many people think and definitely not as as drastic as a lot of influencers want you to believe. Most companies that are are doing things the right way are using AI in a much more subtle fashion than plopping a prompt in in Lovable or Bolt or something like that. Um, the actual development process is not all that different than it was in 2022. in my opinion. Now, I I would even argue that the change from 2010 to 2020 was greater than the change from 2020 to now. You know, back in in in 2010, I mean, front-end frameworks were barely a thing. You know, mobile first design wasn't standard. Cloud deployment was was still um uh relatively new for small developers. And the tools that we got in that decade, Docker and you know modern frameworks like React and Vue, GitHub actions, cloud platforms, these were massive shifts in how we actually work and AI autocomplete and chat interfaces. They are great additions, but they're iterative improvements on an already solid foundation. So, I mean, take this with a grain of salt. Again, it was just meant to kind of spark some conversation and thought. I'm not saying AI isn't valuable. It absolutely is. Um, I'm just saying let's keep some perspective about what's actually changed and what was already pretty good. And I think that to me the danger that I see uh is beginners thinking that they don't need to learn fundamentals because AI will handle it or um experienced developers assuming everyone before 2023 was just stumbling around in the dark. And and neither of those are are true. But uh I think we're we're at a good moment for development tools. if I'm thinking optimistically. Um, but we were also in a pretty good moment, you know, three years ago and hopefully we'll be in a better moment three years from now. That's how progress works gradually with occasional jumps. Um, hopefully, I mean, things could just go to Who knows? But that's it. Thanks for watching, guys. Uh, let me know if if you agree, disagree, what your what your thoughts are. And thanks for watching.