Originally from my website, I'd appreciate some support to grow it!
A little bit of history...
Around 2020 - 2021, Github announced Copilot - a generative coding assistant powered by AI. It was in technical preview and was available for a limited group of developers for beta testing. They advertised it as a tool to help writing boilerplate code, backend server codes, test cases, etc. while keeping security in mind.
According to many users, it wasn't that great at the time, but there was hope for sure
Around the same time, a project called DALL-E Mini happened as a submission for a Google hosted hackathon. It generated weird images. However, it was fun. People had nothing to do, because of COVID-19, which made them generate more and more of those weird images.
This marked the beginning of the new AI era. Lots of people got awareness about AI and generative AI - thanks to the weird image generating service.
On 2022, OpenAI announced ChatGPT, and nothing was the same ever since.
ChatGPT has been trained with vast amount of data existing in the internet. And it uses a sophisticated algorithm to predict text streams and keep the conversations flowing and interesting. However, due to this specific way of implementation, it was never a great tool for coding. Everybody knew that, including OpenAI - which resulting in them releasing better and improved models.
But the problems still prevail!
Problems with these generative models
- Lack of context: These tools often doesn't know the existing methods, components, practices, and standards. Sure, latest tools have better context but it is never perfect. And you have to really understand the project and be a proficient programmer provide all these contexts.
- Over complexity: Going along with the above point, since it doesn't really know the context, it tries to be perfect by adding all the possible scenarios. Sometimes you don't need all that. An experienced programmer can always do it better and shorter.
- Hallucination
- ZERO UNIQUENESS: I've seen my pals releasing their portfolios since the release of tools like cursor, etc. Funny enough, they all look the same. Same with their blogs, everything looks the same.
I'm 100% sure in few more months, they will release way better models addressing all these issues. But that's not the real deal.
So why am I ranting?
When these tools were released we knew how useful they can be, and how we could have used them to help our work. Everyone knew that these are not going to give you the perfect output. And people with actual experience often joked when they heard things like AI is going to take over your job - because it can never replicate your creativity. It can never be you or replicate your emotions.
However, all these noobie coders, with 0 skill and 0 knowledge, happened to discover these tools, and now have made it their personality. They call themselves skilled for some reason. My brother in Christ, you are not skilled; it's just some other person doing something for you. Let's get to this point later.
It has become a crisis now most of the entire industry is being filled up with these noobs, hiding under AI. Funny to hear arguments like it's going to save time. Don't get me wrong, it does save your time, and you should use them for that exact reason. But there's a way to do that. Use AI to complete repetitive tasks (like writing test cases), or research new stuff.
Don't just give a prompt, and call it real skill.
Do you realize,
You are just an intermediate between the requirements and the actual product. Your only job is passing the requirements to the tool and waiting until the outcome. Do you realize that a 12 years old with a decent background can do that too? Do you realize if someone designed some tool that would identify the requirements by itself and finished the product by itself, you become completely useless?
My friend, it's your job is in danger, not mine!
Let me quickly re-iterate why crap that are entirely built with AI are useless.
- No uniqueness, no emotion. All it has is perfect structure or perfect grammar.
- Lacks context, and sometimes do things out of context
- Long ass codes (it's laughable to see 10000+ lines of code for a simple website)
- Shit performance, and no security, unless you are specific about it.
- You know nothing about the final product
I'd like to talk more about that last point. I made a solution to the N-Queens problem 7 years ago. I was so proud about that and I still remember how I did that to this date. I'll be honest, the performance was garbage because I was brute-forcing the thing. But I know how to improve it because I know exactly where it went wrong.
I talked with some guy that has came up with an algorithm for something. I'm not going to say who or what. They mentioned they finished that project in last month. I asked how they implemented it. They just spat out some stuff that are not relevant to my question. They said they did it in last month so they don't really remember how they did it... I'm pretty sure they didn't even see that code until this point.
So we both revised the code and explained what was going on. I asked if they have any idea on how to improve the algorithm. They had no idea because they had 0 skills and 0 clue of the algorithm. On the other hand, I saw so many areas for improvement.
Your ChatGPT is not going to help you improve something if you don't know what to fix. And to know that you have to actually understand the work. For that
- either you have to write the entire thing by yourself
- or just be so good that you can look at the thing and understand what's going on.
Now your ignorant ass might argue,
"cAn'T yOu jUsT gEneRaTe wItH Ai aNd iMpRovE iT lAtEr?"
- To be actually good you have to do things by yourself. So I'd rather keep on practicing that.
- It's harder to nitpick things one by one. I'd rather write the thing by myself because I actually know how to do the work.
Even better if I can write it by myself, and use AI to improve that. Because as a human there are things that I might miss. And I actually do that. AI is really helpful in that regards.
My rule of thumb is do the base by yourself, establish structure and rules. And let the slaves fill in the blanks for you.
credits: from hackernoon, https://hackernoon.com/ai-please-wash-my-dishes-let-me-write-a-desperate-plea-for-creative-freedom Where's the AI to do my dishes while I do my art?
I just plugged in a funny meme that is actually talking about something serious. I got it from another blog from hackernoon. Please read it too (link is pasted above). We need AI to do our hard chores so we can focus on our creative work. Not the other way around guys!
My challenge to those AI losers
- try doing something that actually works, it shouldn't break on a random Saturday because your sorry ass committed passwords and secrets to your Github.
- try to actually implement something with limited resources. Storage limitations, poor computation power, runtime limitations. AIs were trained with codes on ideal environments.
- give it some new language, it wouldn't do shit.
- try to work with a private API. And watch AI inventing new methods.
How about the CEOs saying AI is the future?
Let's go through a list
- "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 midlevel engineer that you have at your company that can write code. (Mark Zuckerberg, source)
- "AI is going to eliminate a lot of current jobs, and there will be classes of jobs that totally go away. AI is also going to change the way a lot of current jobs function, and it’s going to create entirely new jobs." (Sam Altman, CEO of OpenAI, source)
You have to realize that these people are CEOs of companies that relies on success of AI. Zuck wants to say Meta AI is the future so more people uses that. Sam Altman wants to promote OpenAI too. So ofcourse they are going to claim statements like this. It's clever advertisement...
Conclusion
AI definitely has it's place in the future, and I'm never denying the fact it is a really useful tool. It just has to go to the right hands. AI is never going to replace real skill or creativity. So keep on what you have been doing over the past years. AI chaps can continue doing their thing and make it easier to get more jobs for us.
There's more concerns with AI such as environmental, legal, and ethical issues - but we don't even have to get to that.
Just use it responsibly, thanks!
RE: My challenge to those AI losers
✋ You rang? Cool. Pull up a chair. 🪑 Let’s talk.
Perfect! I’ll send you a personal invite to my very first “vibe coded” app as soon as our work OS gods approve it (ETA: maybe this decade ⏳). Was it flawless? Nope. I had never touched AI before. We took some messy loops until I learned how to tell it what I wanted. Then it started tracking itself. Bit by bit, I learned exactly what AI could — and couldn’t — do, and how to use both to my advantage. ⚡
Oh, you mean like JVM-OS-layer deadlocks that birth immortal zombie threads? 🧟♂️ The kind even GC refuses to touch? I'm talking years of survivors, mocking every fresh deploy. Add in the kind of trace dump that’s pure nightmare fuel: Spring + Hikari + JDBC. That “unsolvable” mess? AI helped me track it down, errors and all — in under three days. Stabilization fix included. (No exorcist required… though I seriously considered it. 😅)
Sure. Go? Rust? COBOL? Klingon? Elvish? Morse code? 📡 Whatever. I stick with post-VB because I value my sanity, but code is code. Nobody’s lightning-fast in a language they haven’t touched in years or ever. That’s why I make AI do the heavy lifting: comb through the official docs, organize them into neat little input/output rules, and then I set the standard. You know when something smells wrong — then you check it, prove it, and often find a better way.
If you’re sitting there watching AI hallucinate without stopping it, that’s not an AI problem. That’s a you problem. The second it goes off-script, I slam the brakes — hard — fix whatever led it astray,
/clear
, re-prompt. Done.Letting AI free-range through production code is like giving a raccoon the keys to your kitchen and acting shocked when it’s eating spaghetti on the counter. 🍝🦝🤷♀️
In just the last two sprints, AI has helped me:
This summer, it pinch-hit as teacher/mentor/test giver for my intern when I got pulled into a last-minute war room. (I wasn’t entirely sure that would work... until he quizzed me later. We’re even.)
Outside of work? That mystery song stuck in my head for days? I tossed my nonsense clues at AI — “2014? white mountain… feet? fur?” — and it nailed it in under three seconds. (Dirty Paws, Of Monsters and Men. Second guess. I wasn’t even 40% right. 🎶) Seriously — Google that and see how long it takes without AI.
So am I an “AI loser”? Fine — I'll take it. But I’m also twice as productive, documenting 5× better, and killing bad ideas before they waste time. AI’s not a magic button, and it shouldn’t replace you. But refusing to touch it? That’s how you stay on dial-up while the rest of us have already moved on.
I’ll take my chances with the DeLorean. 🚘🚀