()=>{}

May was the second anniversary of my personal journey with AI, or generative AI to be precise. A few weeks ago before that moment, we had a quick chat on Zoom with some guys from all over the world, from Switzerland to New Zealand. So, the news was like, “You’ve heard that? One guy used something called ChatGPT and won a coding competition.” So, there was this chatbot, right? And it had this new tech that could understand natural English and respond to it. And get this – it could even code in almost any programming language. Wow!

So as soon as I got there, I created an account and asked for a Basic v2 code doing something like bubble sort. And it worked! I soon found out that ChatGPT even knew my native language, Polish. I don’t know, it’s a bit ironic, isn’t it? Until fairly recently, computers were unable to display Polish diacritics, so we had to type them out using plain latin characters leading to mysteries like “zolw” instead of “żółw” (a turtle in Polish). And what exactly did I do two years ago? I took my private OSS project, Retro Build Tool, and asked ChatGPT to extend its functionality. The result is still kept there (I’ll write a separate post about it, it’s really worth it). Using chat AI app wasn’t as easy as using some modern coding tool like GitHub Copilot, Windsurf or Cursor AI (you need to copy a lot of stuff back and forth), but I was still amazed by the capability of something I couldn’t name. In the meantime I’ve learned it is named “LLM”, or a Large Language Model. I made a pitch about this story with this perhaps not so correct title “How do I stop coding”, which was actually quite successful. My narrative at that time were kind of clear - soon we will stop writing code.

I have to admit, the technology was a bit of a mystery to me. I knew the basics of neural networks, and I knew that this technology was important to ChatGPT, but I still couldn’t quite understand how it worked. I’ve found that using a separate, web-based app has made me realise how important context is. I had to copy bits of my existing code to make use of the generated content, which I then pasted back into my IDE. I’ve started having these weird “hallucinations”, but I’m not sure why. I made it pretty clear that things will change, and not just in IT.

Generative AI is a technology that’s all the rage right now, and it’s getting a lot of hype. It’s also spreading like wildfire. I’d say there’s nothing that’s purely software-based that’s been so quickly adopted by the masses (not even the Internet). It’s no surprise that even the first LLMs were really good at coding. I reckon the tech giants splashed out a fair bit of cash to get rid of those programmers they were paying way too well. But the language model is much more than that. It can process and comprehend any text (or image or any other kind of media). Then something strange happened. Some random people were trying out different scenarios using some simple prompts, and one of them had a hallucination in response. It was just to show the public that it wasn’t intelligent. It’s just plain stupid!” My favourite example is here (sorry, in Polish, you’ll have to translate that yourself).

Up to now, technological revolutions have mostly been bad news for blue-collar, physical workers. People in the middle and upper middle classes felt pretty secure, job-wise and financially speaking. This time, though, it’s the other way around. Generative AI is having a big impact on the skills of people who use it. It’s funny, because normally such revolutions are considered natural, people need to adhere, reeducate and find a new job. This time, you know: it will be a disaster, people will be starving, societes will fall appart, blah blah blah.

It’s no surprise that work on the AI Act has started super quickly, and everyone’s started talking about ethics and all that. But, as with any tech revolution, it’s already too late. Investment has been made and progress has been made. There will always be someone in the world who will carry on the work.

It’s funny, but it’s actually GenAI that might hit non-IT jobs even harder. IT jobs, especially the senior ones, will probably stay pretty much the same. When I first started out, I was pretty sure that the coding business was a thing of the past. We’ve figured that maybe “AI won’t get rid of the coding guys, but the ones who can master AI can get rid of the ones who can’t”. Then, when I started checking out tools called GenAI Coding Assistants, I noticed they’re not as simple to use as they could be, especially with existing, large and legacy code bases (something we call brownfield projects, in contrast to greenfield projects, where AI is most useful). We’ve noticed that GenAI doesn’t actually help junior programmers code like seniors, which is different from what we expected. GenAI mostly benefits the more senior members of the team and means we don’t need to keep on hiring new juniors (but how do we grow the more senior members of the team if not from the juniors?). Coding assistants were pretty basic when they first came out – they could only do simple tasks and they didn’t have much context. But the rise of coding agents is once again shaking things up. Now, the tool can plan the work, think for itself, reveal context itself by traversing project files, and implement new or change existing features with a pretty high degree of autonomy. It still makes mistakes (and I think it always will), but in the end, it produces loads of code that needs to be reviewed. Who’s going to do that? Another AI? I doubt it.

Back to my field, nobody can really predict the future, so let’s assume for time being: coding isn’t dead yet, it’s just being redefined.