The world of software engineering is evolving fast. AI-powered tools are revolutionizing how code is written, and more people are jumping into tech...
Some comments have been hidden by the post's author - find out more
For further actions, you may consider blocking this person and/or reporting abuse
I'm not looking forward to a future where we have to fix all the AI mess created because companies thought they don't need experienced people anymore...
I am. It's going to be a hiring frenzy.
AI is expontentially improving.
Whatever you think it is, it is gonna be better, not worse.
So, you can wait.. I wouldn't recommend it.
I tend to disagree. We are already seeing diminishing results. It does not look very 'exponential' to me, but rather the typical s-slope. Sure, bet on AI. I'll bet on my skills.
agreed with you @syeo66
@syeo66 Well, for us, who were working with sofware for more than 20 years, we don't need to bet on our skills, we already did.
Now, what it is not being exponential, is the developers knowledge. But AI is growing.
Itโs really scary just thinking about that!!!
NAILED IT.
I just primarily think that this whole vibe coding concept is a marketing strategy because I personally don't trust AI with the code it generates for me. It still hallucinates a lot.
Just as you wrote.
Real engineers will soon be in big demand companies begin to realise that AI generated code has scalability and security issues
Real engineers are being hired to work for AI development. Not for fixing some kids stuff. haha
That's a valid point, but it's not necessary for every single individual within the category of "Real Engineer" to go for AI development. But one thing is always be true, wherever they work, they'll be the savior and the superior till AI gets in a state of ruling the human race! ๐คฃ
yes, the real are the saviors
While I agree with the core message, there's also a potential for AI to democratize development for those who might not have had traditional CS education. The key is responsible and continuous learning.
Development is already democratized. Anyone can get an education online through free resources.
Yes, that's correct @ac89 , I myself a self learner, though I've also pursuing degree on CSE. Most of the time, or even before getting into university I started learning by myself with the help of online free resources, but the net maturity of learning is, weather we self learn or getting a degree on any decipline, we have to ensure learning the core, considering it a mandatory thing to adopt within own self.
Absolutely! AI is a powerful tool that opens doors for many who may not have had access to traditional CS education. The key, as you said, is responsible and continuous learning. AI can accelerate growth, but relying on it without building a strong foundation can be risky. The best developers will be those who use AI to enhance their skills, not replace the learning process. Appreciate your insight!
I think it always "vibe coders" will always from come time to time with different appearance.
last time, people who only can code with third party library, plugin. Only mastering the framework but not the language itself.
Currently its time for people who only code with AI help.
It's our job to make sure we understand what is valuable if we doing by ourself or by AI.
Totally! "Vibe coders" keep changing. First it was relying on frameworks, now itโs AI help. Our job is to know whatโs valuableโdoing it ourselves or with AI.
Nice article, I don't have any numbers at hand, but I can guess quite a bit of these "vibe" coders are actually SaaS founders, and this is not necessarily a bad thing.
When you just starting a business, if you cannot code, it's really expensive to hire a dev team and this will ultimately shelve some great ideas.
So if these founders can "vibe" code their way through an MVP, eventually a percentage of them will succeed and this will grow the startup base, hence specialists will be needed later on to come in and "cleanup" or scale that product. Also, being able to use AI to save weeks or even months of work could mean they are proffitable much sooner.
As for junior developers, common sense should prevail. If you rely solely on AI to write code for you, without understanding the core concepts and want to be a professional in the industry, you may get away for entry-level jobs but there's going to come a time in your career when you need to apply your mind and do something complex, then you're in trouble!
Well, getting profitable on garbage-code product isn't something I can get behind.
Yes, creating some prototype or tool for your own usage can be done with LLMs (if you don't mind all the ethical and ecological questions), so you can validate your idea with prototype before spending money on real dev. But maybe you can create prototype in Figma or other tools as easily and more reliably. Or pay some freelancer a little money to do that for you.
But anything where people depend on you? Where they give you their (personal) data for you to manage? If you just give them this black box some LLM generated, you are just irresponsible, and you are making money unethically IMO.
Also, there are great OSS frameworks, libraries and whatnot which make creating your v1 much easier, and you most likely don't need team of devs. Having one technically literate person on your team when you want to start a technical bussiness seems pretty reasonable.
agreed with your presented thought
Thanks for the great points! I agree, SaaS founders using "vibe coding" for an MVP is smart, especially when resources are tight. AI can help speed up development and boost profits. But for junior devs, understanding core concepts is key, AI canโt replace the need for real skills when things get complex!
I have nothing against using AI tools. The problem lies in blindly trusting code that is produced by such tools.
I have the same issue with No-Code tools, in general. I remember in the early 2000s, using Dreamweaver to build websites. The markup was utter garbage! I prefer having the control to keep my markup semantic, add minimal JavaScript, and sprinkle in CSS to make my interface pretty.
Yes, exactly! Anything generated by AI (or any kind of automation) is just garbage!!!
surely is ๐ฌ๐
is something I will file under the "karma" category.
I know this is a hot take in the modern world, but if you have an idea and need AI to build it for you then really you have nothing and you deserve nothing for it. I'm fine with people building them for the benefit of humanity or as a showcase, but selling something you had AI plagiarise for you is not something anyone is entitled to.
totally agree
oscar goes to @leojr94_
RIP!!!
๐คฃ
In the end, we will need a strong fundamental first to use AI to code for us.
If we can represent our fundamental skill with number, let say our fundamental skill is 1, then AI can make it 10x works faster (1 x 10). But, if our fundamental skill is 0, even with 10x AI helps, it will still make 0 works.
Anyway, i also just published an listicle about AI tools that we can use for vibe coding. Have a look if you dont mind :D, thanks.
dev.to/syakirurahman/top-10-ai-fig...
That's the key, understanding the fundamentals allows someone to craft the best prompt, and from the best prompt, you can achieve the most effective results.
Cant wait for people to figure out that not all Vibe devs should have been hired and that's when our knowledge comes in and I'll charge triple the amount I usually would ;)
Every school costs money ;)
;)
Well Written Article. Totally agree with you that Vibe Coders are everywhere. We aimed for hiring a mid or junior for our Team. We see a good code that some parts are working and some parts are not working, which is completely fine, but sometimes the person has no idea why these parts of the software are working. I also understand that with the AI, it is much harder to feel confident and be good at something. I guess at least for the first year of career they should not use AI, they should just chat with AI and not copy paste any code to their IDE.
and the best suggestions will be not using AI, while doing any kind of learning!
I took enough CS classes in school to read and understand code, but I never thought Iโd be able to spin up a benchmarking pipeline for three different LLMs in a dayโuntil I actually did (shoutout to Lastmile AIโs AutoEval, which made it ridiculously easy). Vibe coding is real, but I feel like to really leverage AI-generated code, you still need a solid foundation. Otherwise, youโre just running black-box outputs and praying it works.
The more you know about a topic, the better you can craft prompts and get the most accurate responses from AI. Your knowledge sets the limit for how well you can work with it.
For example, imagine you're coding a sorting program. If you don't know much about sorting algorithms, AI might suggest Bubble Sort as the solution because it's an easy algorithm to implement. However, while itโs simple, itโs not always the most efficient. If you were more knowledgeable about sorting, youโd know that algorithms like Quick Sort or Merge Sort could perform better depending on the data. Without understanding things like time and space complexity, you might miss opportunities to improve the efficiency of your program.
So, itโs not just about codingโwhatever task you ask AI to help with, the more knowledge you have on the topic, the more effective your results will be. If you don't have that background, your ability to explore and improve will always be limited.
Note: I wrote this reply using AI, but not everything, I just used this as a tool to rephrase my own word, as being Bangladeshi I'm not as well to effectively write in English, so I always use AI to fine-tune my public writing, and I am not ashamed to tell this publicly.
My conversation with AI (ChatGPT), attaching for demonstration purpose:
me: 'reohrase this "The amount of knowledge you have in a topic that you're doing with AI, is the limit of how well and effective prompt you can produce by yourself, and how fine-tuned and standard response you can generate from an AI."'
ChatGPT: .....
me: 'also give an example, imagine in a scenerio, where you are vibe coding a sorting program, generally, i predict that it will produce the program with Buble sort, and you don't know much about sorting, as a result you are being in the dark that according to data you have to sort, it eill better to use quick sort or merge sort or anything whatever, but for you lack of knowlegde you will comsider using that produced bubble sort, where maybe if you had knowlege of space time complexity and knowledge about more sporting algorithm you had enough room to improve and make that program more effective, so it's not about coding, any topic you want to get done by ai, the prerequisites is to keep enough knowledge about that topic within you, otherwise you're level of exploration will be always lkmited'
ChatGPT: ....
Me: 'in using bubble sort also mention that, generally it's the initial choose of ai as being easy algorithm'
ChatGPT: (gives the final version, which i replied you with)
Did you notice the similarities between my final reply and my conversation with AI? Isnโt it just a reflection of my words, only more refined? Of course, it is! I donโt worry about whether my spelling or sentence structure is perfectโI just express my thoughts to AI to bring my ideas to life.
Maybe my prompts arenโt always the most effective or well-structured. Though I consider my prompting the worst it can be, thatโs part of the learning process. Iโm still learning about promptingโnot in a structured or deliberate way, but passively. I simply experiment, refining and reshaping my prompts as I go, trying to align responses with my expectations.
That said, I rarely put much thought into crafting prompts. I just ramble until I get the response I want. Of course, this means I waste a lot of time in the process. I know that using more effective prompts could save me time, but honestly, I donโt care. As an extrovert, I enjoy passing time by writing or rambling in written form. Itโs fun, and maybe it even makes me feel good.
Maybe Iโll always prompt like thisโuntil Iโm under pressure to improve or forced to make my prompts more effective and structured.
Whatever...., I see AI as a tool to refine and shape my imagination, not to replace it. I also make sure to use it only for topics I truly understand. If something doesnโt make sense to me in real life, I donโt blindly rely on AI for it either.
I have a Junior colleague who's completely dependent on AI tools. He's unable to tell what the 3 lines function is doing, without running it first. He's in the company for two years I was forced to put him through the interview again. With just pencil and paper. We'll see how that turns out.
yeh, let's see how that's turns out!
AI-assisted coding is a game-changer, but as you pointed out, it shouldn't replace fundamental problem-solving skills. While AI can speed up development, true expertise comes from understanding how and why things work under the hood. The divide between 'vibe coders' and foundational engineers is real, and those who invest in core CS knowledge will stand out in the long run. Thanks for sharing this perspective!
Exactly! AI can speed things up, but itโs the deep understanding that sets great developers apart. Investing in core CS knowledge is the key to long-term success. Thanks for the thoughtful comment!
iโm not too worried honestly. companies will hopefully realize the amount of time needed to debug and maintain the vibe coded slop far outweighs the actual benefits of hiring real engineers that know whatโs going on
yes you are right
The split between coders and foundational engineers was already a thing. AI only made it much more visible.
correct
What I saw with vibe coders is that, because they don't know the software development cycle, they are making shitty software.
The stuff here is, if you plan your software, create an architecture design document and a technical specification, separate the tasks modularly and the implementation, there's no way AI fail. Seriously, treat it like a Dev, not like a replace.
The problem with these kids that just want to feel they "program", is that, they get a prompt, do all the job with AI and THEN, ONLY THEN, they debug.
No unit testing generated, no regression testing everytime you incorprate a feature.. etc.
That's the problem.
Not how they program, but how they work.
Same stuff, bad developers fails in the same stuff.
AI is a powerful tool, but true expertise comes from understanding the code, not just generating it. The future belongs to those who master both.
Appreciate your understanding.