AI can pretty decently write code. I have created entire social network sites using nothing but AI and they all work well. Sure there were issues and bugs but the AI was able to find the bugs and fix them. Once in a while I had go through and find the errors my self and prompt the AI on how to fix them but the majority of the time the AI could do it all on it's own.
This works occasionally, when the code is uncommon enough that there aren't many examples, but the algorithm is not too difficult to understand. I learned a couple of algorithms by fixing terrible hallucinations, but those are rare instances. Most of the time the problem is either just boilerplate and it's not too different from copying from stack overflow (back when it was good), or the other extreme where the code is so bad it's not enough to fix it.
I agree, there's some truth and value to your comment, as seen with my son's experiences: 1. he spent a lot of time fixing the code of a Chat GPT using friend whose Roblox code (Lua scripting) wasn't working correctly (he's done this multiple times). His debugging skills have gotten so much better because of doing this. 2. His refactoring skills improved quickly when he helped a different group of friends when debugging their code to make it understandable, maintainable and to work correctly. He understands my "coder rants" now since he's gone through this firsthand. :P
If you want to become a programmer, there is no way around: despair, feeling stupid, wtf is means that error message, despair, more despair, thinking that you are the stupidest person in the world, and the fu**** tutorial hell... .but then things get better and that shit starts to become really fun. So for anyone with the same level of resilience as me...keep going...this shit is worth it ❤
As a new programmer…I’ve been hearing a lot about the tutorial he’ll, can you elaborate on what that really is so I don’t fall into the trap.. right now I have an app idea but I’m currently taking course on OOP done with regular procedural. I believe once I’m done with course I might try to follow up videos that build precisely what I want to build because I would understand the syntax and concepts… but does that post graduation study approach fall under tutorial hell as well?
@@potatochipappi In my opinion people are in tutorial hell, because they dont belive that they are good enough to build something on your own. Ive been programming for 2 years and honestly never felt stucking in tutorial hell. Since I remember Ive been modyfying code which I wrote following tutorials, adding new features, looking dive into the code. Imo reading docs its much better than asking AI and every begginer should begin doing it as soon as possible.
@@potatochipappiThis is cope, you can't just take a course and then build something great. You MUST build smaller things. You will fail in the smaller things, there will be bugs, the html will be sloppy, the css will make it look trash and so on. Then, you do it again and slowly increase difficulty, then only after a few months of this, can you even think about beginning the total package you wanted to originally build. That's just how it works. The whole idea about tutorial hell is that ideas are embedded in working memory and then replaced after 5 minutes or 5 hours or 5 days. When you're watching the tutorial, or a video, or reading the documentation, you feel like you know something, even though you don't actually, then when you go to build, you realize "omg, I can't even write the meta tag", then you go back to the tutorial, and because the building part was super hard, you spend 5 h in tutorial instead of 1 minute that you needed, then you go to build again and repeat. What you need is concepts in long-term memory, for which you need to fail and tinker again and again and again. It took me hours over months to figure out how setInterval() really works for example. If I 'studied' it one day, then I still wouldn't know how it worked. But now, I can use it anywhere, anyplace, perfectly. The key thing is that you need to understand the fundamental structure of how things work REALLY WELL: What's the event loop? What is synchronicity? What's a 'this' binding? What's lexical scoping? What's the Box-Model? What's Margin-Collapse? etc. You need to know hundreds of the one line concepts REALLY WELL. The second key thing is that you need to be solution oriented not information oriented, if you don't need to know something(that's on the applicability side and not on the structure), don't go and look it up, lookup what you need, implement it, move on and repeat until you get good. If it works then it works, move on. This is not only a practical consideration or from a fear of learning , but rather true learning is application and not information. The knowledge will automatically come to you over months and years of figuring one thing out, your brain is not a computer that it will instantly input everything about one thing. Information is NOT knowledge, knowledge comes from building with it, using it, not reading or watching. It is a 10x better use of time to build something yourself than it is to watch a video of someone doing it. I put up front-end related things here due to the channel but it applies to all of programming. Advice from me: endeavour to make the course/watching/reading time be less than 10% of the total time spent, 90% needs to be in you fixing bugs, banging your head, asking chatgpt about how something works(NOT THE CODE TO COPY PASTE!!!), reapplying things again and again til it's good enough and building. Also, don't forget the thinking stage, spend a solid amount of time just thinking what the problem is, what the structure will be, what you need to do etc. Don't underestimate a pen and a paper, that's more important than the code. If you don't flesh out the problem and the solution, then after you build, you will realize that that's not what you wanted or that your architecture is spaghetti and that will waste 10x more time than the 10 minuts it would take you on paper.
I worked for AWS in Madrid as a Professional Cloud Solution Architect. I felt dumb 90% of the time solving problems, but when I succeeded, I felt like Einstein.
Issue is overcomplicating stuff, find an idea on what you trully want to make, and you will find a way how to guarantee. You can ask chatgpt how to make it but dont use it for code, but roadmap
You should absolutely use it for code, just don't let it be creative (unless you're brainstorming). I find it's much faster to give a thorough english-language description of an algorithm and have chatGPT (or claude, which is more reliable) write the actual code than to write the code myself. It's just as good at getting you to think through the problem, and you don't need to get hung up on remembering all the syntax and function names. LLMs are pretty good at following clear, unambiguous instructions; it's the following of vague instructions that'll lead you to where you don't want to go.
Before, I only asked AI for a roadmap and didn’t rely on it for coding. I was doing well with other languages and even built my own projects without AI. But everything changed when I reached my third year in college, where I had to learn PHP entirely on my own, without any help from professors. PHP really put a damper on my passion for programming.
1) use Ai to be ambitious and do difficult projects 2) fuck up 3) learn to understand code 4) debug successful 5) **feels the fear of not writing code yourself 6) forget the fear and do harder projects 7) do this until you are used to it Repeat step 1 to 6, and if your conviction is hard and you are too stubborn to give up, you can actually be dependent on Ai while able to program stuff
I needed to hear this with the project i am building. Gonna change my approach and try to write the code myself. Break it down first and lay out the steps, then build it myself.
Let me give you another PoV. Part of the problem is the web development itself. It has ALWAYS been about A) Quick solutions B) Ready-made solutions. Always, since 90s where you can inspect the source of the page - something unheard in "normal" programming. This has pros and cons, pros being the insanely rapid development it brought, cons, well, hacky code, mind-numbing resource consumption, compered to the actual work done, etc. When this is the norm, no wonder the AI will first conquer web development. And again, before AI there were the frameworks, the templates, the CMS, arguably even the UML code generators - all tools which take development away from the developer. Noone writes a page from scratch for 25 years and the AI is yet another of those tools.
Breaking down the problem is what helped me to get away from tutorial hell. I was telling my self what is the simplest app that i could build something that i will not need anything for just documentation and that's it. It was bit painful but when i finished my first app it clicked to me and i stopped using tutorials and started building my own stuff which as you said provided that dopamine hit. When it comes to AI it's the opposite i tried it and when it shown me a result i didn't feel like continuing the project anymore because it wasn't mine anymore. So i deleted it and rewrote it. I will never use AI for coding it almost killed all the excitement from it for me.
I use co-pilot, which I use mostly to learn new stuff. For example, I recently started working on Ruby on Rails and Co-pilot is quite good at telling you different ways to write a piece of code you just wrote, using language-specific features you were not aware about.
chatgpt is honetsly stupid at coding. it uses weird functions and arrangement, starts hallucinating a new method that literally does not exist or just can't seem to debug anything by itself. However, it is useful to talk to and send YOUR code to check if its efficiecny and ensure the logic is correct, and you cna put in it's memory "dont send me code"
Also never fall into the trap of thinking that things should be done only always in one perfect way. It's helpful to see fresh perspectives (ways that are possibly easier or faster to write) on areas of coding which you already rather familiar with.
Asking AI for something else than generating input data based on schema or reformatting a file or piece of text when it comes to programming is pure madness
Thank you. I am currently struggling with my JS learning path and I was kind of giving up to AI as I felt that I never would be able to understand how to use and combine different things, but your video was right on time ! Thanks a lot
I work as an electrician learning to code takes too long so fuck all that noise I’ll take my microsaas generator and make it work focusing on smooth UI, security, and meeting my intended functionality needs. Get with the times. I learned how to use fire base while making my first flutter app. And while securing it found out a lot about different potential vulnerabilities in databases and code. While trying to deploy ran into apple specific hurdles I had to stumble through and learn from like how to setup Xcode. Yes I use ai to code but I’m not an idiot and I am learning a bunch of skills that are going to help me in the next project
Even when using AI to generate the code, it's still a pain in the ass during the design process. Because gpt, as a machine, cannot generate a complete product on its own, it can't just magically finish your coding assignment in one or two prompts. After multiple times of doom-prompting (ask gpt to generate the code and spamming prompts until it works, most of the time it doesn't), begging the AI to write the code that works as a whole (dumb shjt ik), I've realized that I can break the process into multiple smaller problems, ask the gpt to write the simplest codes possible, then I apply that logic into what I'm trying to build. Sometimes it works, sometimes it doesn't and I have to write codes on my own :/
Reading documentation or how-tos for very unfamiliar things is very good and is never bad. Even more, being able to tackle the new challenges fast is excellent. Whoever can build a decently sized app without looking up anything is a literal god.
I am going in other ways I am sick that everything is in different programming languages, so I just ask ai to implement small blocks of code and slap them together maybe making minor adjustments to see weather it works, like if you are trying to fuck around with shaders and stuff and other stuff (especially if you wanna run something perfomant in WebGl), you care more about getting math and concepts working correctly, without caring about which programming language you are using (ps, try using c++ for 2-3 years all other programming languages will be feeling like using a library or something)
This is random video for me and I do not code anything but it reminded me of university, where we learned Matlab. We had no prior knowledge of it or any coding but teacher gave is separate tstks that we had to do over few weeks, googling everything that is needed but the most important part was that we had to write comment for each line to explain what it does (so even stolen code will be understandable) I got task to create billiard game where I have table, balls and by entering angle and power of shot it had to draw animation result of that hit. Unfortunately I had to steal few lines for code from internet because when ball hitting walls it had to go in other direction and I was struggling to flip coordinate system (basically my ball received direction after collision mirrored of what it's supposed to be). It was a fun project, I enjoyed it
Asking ChatGPT to write code is like asking a biologist to build an human. They can probably create something that would fool a someone with minimal knowledge on how the thing works but the result probably won't function correctly.
The reality is AI IS going to take over programming and Web Development. THATS the bottom line. We might as well adapt and get with the program. Figure out how to get AI to work for us instead of trying to fight a losing battle. When that day comes that AI does completely take over would you rather be the one who fought it tooth and nail and ends up out on the street or would you rather be the one who has years of experience getting AI to work for you so your skills are very much in demand?
Great video..I'm going to share this with my Dev team. AI is truly bad at doing any complex coding and what you're left with is an application you don't understand which makes it hard to find/fix issues. As bad as it is at creating good code...try asking it to refactor code and enjoy all the breaking changes it introduces. AI doesn't think, it regurgitates.
The problem with using word generators is the model needs to be trained on what the new learner specifically knows to generate code right on the cusp of a new horizon. Unfortunately most new learners do not know how to qualify what they know let alone know how to tell a chatbot to make sure the chatbot doesn't just tell them what they want to hear when they ask the wrong questions.
all you say is true me as a self thought dev I relate on that aloot 😭😭😭 the first thing I do on programing is a simple emulator for my custom assembly and when it works I am so proud of my self and quickly do more complex projects like solving a repetitive school work with codes, writing brainfk emulator, making a simple terminal that you can chat with other people, writing a hamming code and making a simple ai without any ML libs and I am so proud of those yes I use help from ai but AI still suck at logical tinking that will end up me fixing it's code that will never work I love to share this on my classmates (this is our first year IT class 😅) that just copy from chatgpt and never fix it's error
Before we had framework programmers. Now we have AI programmers. Most people aren't going to follow your advice but the ones that do are going to be able to keep delivering features and keep production online.
Another tip that I would like to follow up is to question everything you do when following a tutorial. And be OBSESSED with the little things. For example lets say you wanna write a function to handle form submissions by writing: function handleSubmit(event) { event.preventDefault(); } Now question yourself to a point where you have almost no question about it: Why do I need event.preventDefault()? What does this line do, and what would happen if I left it out? How can I implement this in anohter sample project? Can I modify these even further? What I usually like to do next after understanding a simple concept is building a quick small form submission website using what I just understood to deeply test your understanding. After truly understanding that concept inside out, resume back your tutorial and rinse and repeat for every new concept that you encounter.
When it comes to JS stuff like that I like to track from which object the method is called, so I do a quick read on the documentation and try to figure out what it's doing under the hood.
yeah I personally use AI for this. I throw the entire documentation for a particular library and ask it to help me whenever I want to create a particular thing without having to search for it in the documentation, and I ask it "why do I need to write this specific line?" "what does this function do?" "what happens if i remove this?". if it hallucinates a random answer then i go search for it on the web. The advantage when using AI is that it just follows the context, when in the web it can be a totally different concept. I learned how to use ThreeJS just by reading the documentation and asking chatgpt to help for a project.
I use AI all the time for googling, I can't remember every single thing in the scripting sister language of Java, made in 1999. AI is indeed bad at writing the exact code you need, or sometimes it will write code that doesn't even work (literally throws). It's nice when used to get a general idea of things I might have missed or never considered trying.
chatgpt is good for learning in my opinion but too many people just take its code and keep throwing prompts at it until it fixes its own half assed code. But chatgpt is great if you give it for example a code snippet that you dont understand and ask for an explanation or if you let it write code and then ask for an explanation. That way you will learn a lot about programming and also gpt isnt good for bigger projects as they often require more specific tools that gpt isnt too familiar with.
I think that development is a slow process, people want to build as many projects as possible, its like reading a book, if you read 10 pages a day and grasp what the author want to say, isn't like reading a book in one day, The benefit will never be the same.
Can the next video be on the ui/ux and designing steps ,like I am good at doing logical stuff but it becomes hard to write html and css and where to get unique ideas from
The problem with new developers is that people thinks that you can learn to do programming by just looking at a couple tutorials, and that's not how it works. I remember when I was 14 and had no internet connection whatsoever at the time, all I had was a couple of books about C++ and C#. And as much as I followed them and learn through them, at the end I couldn't do anything either because I didn't understand many basic concepts of computer science that are required to understand how to do what you want to do, that are not explained in those books. Just imagine trying to write anything just after watching a couple videos telling you how to write for loops. It really takes more than just f*cking around or watching tutorials, or even using GPT (it won't work, I tried it already and it's unreliable).
Ah, yes, the "tinker until it works" crowd. Real blast from the past, aren’t they? Look, people who say we shouldn't use AI for coding are like those who'd prefer riding a bicycle instead of using a car because, hey, "you get there eventually." Sure, they’ll arrive, but they'll be exhausted, sore, and probably a decade behind. Back in the day, coding in binary wasn’t just tedious; it was madness. Imagine writing one and zero for hours just to make a light blink. Then assembly came along, then high-level languages, then interpreters, compilers... all to save people from reinventing the wheel every single time. AI is just the next logical leap. It's like the Jarvis to your everyday code-the assistant who speeds up the grunt work, leaves you time for creativity, and catches things you might miss. So, tell the tinkerers to keep tinkering, and while they’re doing that, we’ll be the ones building the next billion-dollar company with the time we saved.
Did you just compare the current LLMs to Jarvis and Cars. When did I say don't use AI at all. How the hell you learn something if not by tinkering, I am genuinely asking.
@whosajid If the goal is to Understand the code and actively take part in building your project, you can still ask AI to write your code especially if you have time constraints. Do you agree?
@whosajid LLMs are constantly growing and improving it's just a matter of time before they get to Jarvis level especially with quantum computing, also Jensen Huang said the goal is to get AI to write a code where a person without much skill just gives some prompts in English and Voila. People are actually building full apps with cursor AI in a short amount of time.
meanwhile me spent a 3 days to fix just a simple problem and after that i laughed so hard and that was been really good like to see that i've done smth :)
i agree even if i made a small toggle button i am wayy wayy more satisfied and motivated to do more complex projects then if i would have just asked gpt or copying youtube code
kind of off topic to the video but it took me a while to get out of the tutorial hell part of programming. i tried to learn lua and python and wasn't really able to learn anything and i just memorized certain functions but i didn't know the problem solving aspect of it all and i eventually forgot everything i learned. it wasn't until i decided to learn JavaScript that i fixed this. if i wanted to learn about a specific topic (lets take if statements) instead of looking it up, i would watch a video of someone explaining the general format and syntax of if statements (if i couldn't find a video, i would read documentation) and then i would get some questions off the internet relating to if statements and i would take what i learned and fix the problem. after all of this i would document it as if i was writing a book to publish so if i ever forgot how to do something, i could look at my document and get a quick refresher. but every since i started doing that, i noticed i dont really need tutorials, if i need to fix something, ill use documentation and perhaps even StackOverflow if it really comes down to it(AI is a last resort if i ask a question on multiple different forums and cant get an answer as AI usually just gives you the answer and doesn't explain it that well), and if i need to create something from scratch but i dont know how to do something, ill look up documentation for values or properties that allow me to do what i want to do. this might not work for everyone but it definitely helped me get out of tutorial hell.
I agree but there are AI models that are better than others so using chatgpt will certainly result in messy and buggy code, but if you use DeepAI or Claude 3.5 sonnet it will be pretty awesome I use both and give them code and ask the same question so these AIs give each other AI code and ask what happened in the code and compare it to the old, theirs and the other AI's. DeepAI is often better than Claude (in pure programming and optimization, and it's more up-to-date).
I think its really just a matter of using AI for a new topic a couple times then doing it on your own from there on, after all its best use is as a teacher that teaches you something new in the best way you learn
F*ck around - sure, when writing your own projects you can. But if you're programming as a job nobody will pay you for "f*cking around". You need to use all available tools to meet the requirements and do it in given time. If asking AI for code and then correcting it is faster than writing it on your own from scratch, then that is the tool you're expected to use, most of the time.
Exactly. This video is so stupid. AI can’t write code, but it’s capable of guiding you and explaining niche features. You’d realize this after seeing AI fail miserably at writing your code for you. A majority of tutorials or forum posts are filled with unnecessary fluff to the point where you can’t find what you’re looking for. I don’t need a technical explanation of everything. Oftentimes I’m just looking for a quick refresher and it becomes a 30min search session for what could’ve been explained in mere seconds. Even documentation suffers from the same problem.
For those of us who are pretty confident developers, do you have any advice for monetizing the skill? I've tried everything from e-commerce, web games, crypto projects, live streaming widgets but still nothing as lucrative as I would have hoped.
Great question: Few ideas: 1. Start a UA-cam channel Teach or build in public. It's not hard, trust me. But you do need luck. 2. Build something that people want(like really really want. Pay you to use the product - want). You are going to need very strong marketing skills and a lot of good luck. 3. Show this confidence in an interview with a startup founder(or tech lead). The easiest and most realistic route. Follow a lot of small tech founders on Twitter and maybe LinkedIn as well. See if you can jump in for a part of the pie. Or, If you are good in python or ML, jump on the AI wagon rn.
i made a project that was totally ai and i understood nothing. but recently i have started to make my own project using the same technology as the ai project but this time without any ai. it has been much more difficult but it is a lot more rewarding
Instead of blindly copying the code in the beginning you should watch the video and then code whatever feels good to you . See the approach not necessarily the code.... Like if the code is in react try to do it in vanilla js or if he used scss for styling use tailwind ...... A good way to keep yourself in check is that when watching a tutorial make sure when the utuber makes a mistake in video you don't make the same one , instead write the correct code before utuber goes ahead and correct his mistake
Yeah, I’m only willing to do this if it’s C++, the language that introduced me to the phenomenon of dreaming in code and allowed me to easily ai my way through js and especially python at my current job without consequences, so far lol 😅.
Hey guys I understand what he is saying. But I think it’s good to write your own code by yourself if you are trying to learn the basics of your tech stack really well, loops, if statements, functions, how to debug. After that, I think using AI to write your code is good because the complex stuff is all bunch of basics lumped together. Also if you understand the basics, you can just spend more time reviewing and optimizing the A.I code. Also you will be able to build big projects and make money from them. You get to do other things either that just building projects like actually making an impact with your projects. The goal is to become good enough to talk to the A.I and review it’s code. While you review its code you can ask questions about the code and add comments in your code to explain the code to your self.
Code writing is the PERFECT use case for AI. You don't write code for the sake of writing code. It's not the goal, it's a burden. I'm a programmer. I'm already using AI as an assistant. It makes translating code from one language to another extremely efficient. Since languages are very similar but still different, it's easy for the human mind to mix them up, but the AI doesn't. It's also rare to know literally _all_ programming languages, but an AI can indeed know them all. It can also rearrange the lines automatically (we don't need AI for that, but it still does it), explain a code, _expand_ a code, read a full library and give you the codes from that library...
If you need to know every programming language (really, more than 2 or 3) to do your job, something has gone very wrong; you've created an imaginary problem for AI to solve. AI can write trivially easy code (even then, you are likely to get nonsensical results), so it can save on typing, but it can not create new ideas: it simply regurgitates code that has already been written. There is no innovation, so you will always be limited in the problems that you can solve.
@@azure-hawk In my first internship, I had two languages to use, C++ and Python. In my first job, I had to learn C#. In my spare time, I work on games ; RPG Maker required me to learn Ruby, and Minecraft required me to learn JavaScript. I also came across software using lua language. As an engineer, I also regularly use MatLab, which uses a _modified_ version of Python that has its own syntax rules. It's dumb to state that programmers don't have to learn several languages. Most likely, you aren't one and obviously don't know better. AI doesn't regurgitate things ; that's what a _regular program_ does. When building a neural network, we create a mathematical representation of a dataset. This mathematical function is then use to create NEW data. Literally what scientists do whenever they want a mathematical law, except we do it automatically through a process we called training. It's _interpolation_ of data, not regurgitation.
@@azure-hawk In my first internship, I had two languages to use, C++ and Python. In my first job, I had to learn C#. In my spare time, I work on games ; RPG Maker required me to learn Ruby, and Minecraft required me to learn JavaScript. I also came across software using lua language. As an engineer, I also regularly use MatLab, which uses a _modified_ version of Python that has its own syntax rules. It's dumb to state that programmers don't have to learn several languages. Most likely, you aren't one and obviously don't know better. AI doesn't regurgitate things ; that's what a _regular program_ does. When building a neural network, we create a mathematical representation of a dataset. This mathematical function is then use to create NEW data. Literally what scientists do whenever they want a mathematical law, except we do it automatically through a process we called training. It's _interpolation_ of data, not regurgitation.
Trying again: In my first internship, I had two languages to use, C++ and Python. In my first job, I had to learn C#. In my spare time, I work on games ; RPG Maker required me to learn Ruby, and Minecraft required me to learn JavaScript. I also came across software using lua language. As an engineer, I also regularly use MatLab, which uses a _modified_ version of Python that has its own syntax rules. It's dumb to state that programmers don't have to learn several languages. Most likely, you aren't one and obviously don't know better. AI doesn't regurgitate things ; that's what a _regular program_ does. When building a neural network, we create a mathematical representation of a dataset. This mathematical function is then use to create NEW data. Literally what scientists do whenever they want a mathematical law, except we do it automatically through a process we called training. It's _interpolation_ of data, not regurgitation.
which projects do you recommend doing? Can you give a list for junior lvl (Todo), intermediate, etc. I don't know which type of project I should start building.
I wasn't planning to censor but did it anyway. Here's why- I know it ruins the flow and feel of the moment. But if YT decides to push down the reach just because of some weird advertising policies, all the hard work will go to waste. Put yourself in my shoes. Edit: Thanks for your feedback, I really appreciate it 🙂
You have a sunk cost fallacy. Because most of us invested enormous time in figuring out how to code, plus layers of different languages and frameworks, now we justify why coding with Ai and relying more an more on it is wrong and "replacing yourself." reality is and we all know it, that we have to be more of architects than coders.
For the purposes of learning you should actually learn. In the end of the day the AI is still outputting code for you and one day the code might not work and you'll have to debug. If you have never done it with your own code, you'll have a very hard time debugging the AI's code. And the problem might be too complex that you won't be able to debug the prompting either.
@RedactedBrainwaves2 in the immediate future probably. I'm of the opinion that coding will be like a switchboard operator, obsolete. And you will have to take more strategic approach to building. But I might be too optimistic (I am a technical startup founder)
@@helLf1nGerTA still, even if the AI does the coding and runs everything and you get the fully compiled application, someone needs to know how it all works because we can't trust an AI to do everything. We no longer have phone operators but phone companies still hire technicians who understand how the system works for routing phone calls to fix any issue that might occur with the automated systems. We still need people to know the low level aspects of programming languages even though js and python exists, and programming languages are more reliable in their mechanics than AI.
I would love to agree with you. Unfortunately I do not. It's like saying that you need to understand perfectly what now it's considered to be low level stuff (like memory management) to write working programs. The reality is that you do not need that, you take it for granted in 90% of the cases. Tomorrow natural language will be just a new level of abstraction and high level programming lanfuages will be the new low level staff. For sure, ij some application you will still need a complete understanding, but these are exception, not normality. The future software engineering will mainly be requirement engineering.
Unfortunately writing the complex stuff with AI simply does not work at the moment because the AI doesn't understand anything. I've tried numerous times to write complicated functions with AI, explaining it in detail the input parameters and required output, and it just fails and fails. And when asked to fix its issues, it consistently claims that it has found and fixed the problems yet makes zero changes to the broken code. At least right now, it is much easier and consistent to just write the code yourself.
@twnyeneyme6058 Yes, I know it is still not able to write complicated stuff, but the entry barrier for writing code is much much lower now. I also know that progress is not linear but for sure in the future LLMs will not be worse than now and I bet the tools that natively integrate LLM (as Cursor) will be much better even without newer and/or more powerful language models
Disagree with the line of thought. Instead avoiding the use of AI while coding, make the project so large that ai alone can't keep track of everything at once
Junior Engineer (1968): When I'm writing memory management routines I'm not able to do some of the math in my head without help from one of these new calculators. Stupid Senior Engineer (1968): It's okay to use it when you encounter complicated math but ideally you want to practice enough so that over time you won't need to use a calculator as often. In 2022 I had 6 senior and 1 junior engineer working for me in a lab in Seattle ($150k per month). Now I get more done on my own with AI ($400 per month). If you find this offensive, you're scared. If this helps you understand what's happening on the ground so you can plan for the future, all power to you
Why would I find it offensive. I use AI all the time. Not just for writing code. Because tbh it's not good at solving real problems. But yeah it helps me save time with smaller chunks of the project.
Ask AI to write your code, confused why your code doesn't work, then actually learn by fixing the AI's hallucinated code
by asking the AI how to fix it
Practice for the future xD
AI can pretty decently write code. I have created entire social network sites using nothing but AI and they all work well. Sure there were issues and bugs but the AI was able to find the bugs and fix them. Once in a while I had go through and find the errors my self and prompt the AI on how to fix them but the majority of the time the AI could do it all on it's own.
This works occasionally, when the code is uncommon enough that there aren't many examples, but the algorithm is not too difficult to understand. I learned a couple of algorithms by fixing terrible hallucinations, but those are rare instances. Most of the time the problem is either just boilerplate and it's not too different from copying from stack overflow (back when it was good), or the other extreme where the code is so bad it's not enough to fix it.
I agree, there's some truth and value to your comment, as seen with my son's experiences:
1. he spent a lot of time fixing the code of a Chat GPT using friend whose Roblox code (Lua scripting) wasn't working correctly (he's done this multiple times). His debugging skills have gotten so much better because of doing this.
2. His refactoring skills improved quickly when he helped a different group of friends when debugging their code to make it understandable, maintainable and to work correctly.
He understands my "coder rants" now since he's gone through this firsthand. :P
"Spending an hour fixing a feature that'll take users 2 seconds to use" so true
How many times I've thought to myself "Huh, that's all of them?" When explaining the features of my app I spent weeks building.
Lol bold of you to assume they're even using it
If you want to become a programmer, there is no way around: despair, feeling stupid, wtf is means that error message, despair, more despair, thinking that you are the stupidest person in the world, and the fu**** tutorial hell... .but then things get better and that shit starts to become really fun. So for anyone with the same level of resilience as me...keep going...this shit is worth it ❤
Agreed although I’m not that hard on myself. Programming is just fundamentally difficult. People need to understand that
As a new programmer…I’ve been hearing a lot about the tutorial he’ll, can you elaborate on what that really is so I don’t fall into the trap.. right now I have an app idea but I’m currently taking course on OOP done with regular procedural. I believe once I’m done with course I might try to follow up videos that build precisely what I want to build because I would understand the syntax and concepts… but does that post graduation study approach fall under tutorial hell as well?
@@potatochipappi In my opinion people are in tutorial hell, because they dont belive that they are good enough to build something on your own. Ive been programming for 2 years and honestly never felt stucking in tutorial hell. Since I remember Ive been modyfying code which I wrote following tutorials, adding new features, looking dive into the code. Imo reading docs its much better than asking AI and every begginer should begin doing it as soon as possible.
@@potatochipappiThis is cope, you can't just take a course and then build something great.
You MUST build smaller things. You will fail in the smaller things, there will be bugs, the html will be sloppy, the css will make it look trash and so on. Then, you do it again and slowly increase difficulty, then only after a few months of this, can you even think about beginning the total package you wanted to originally build. That's just how it works.
The whole idea about tutorial hell is that ideas are embedded in working memory and then replaced after 5 minutes or 5 hours or 5 days. When you're watching the tutorial, or a video, or reading the documentation, you feel like you know something, even though you don't actually, then when you go to build, you realize "omg, I can't even write the meta tag", then you go back to the tutorial, and because the building part was super hard, you spend 5 h in tutorial instead of 1 minute that you needed, then you go to build again and repeat.
What you need is concepts in long-term memory, for which you need to fail and tinker again and again and again. It took me hours over months to figure out how setInterval() really works for example. If I 'studied' it one day, then I still wouldn't know how it worked. But now, I can use it anywhere, anyplace, perfectly.
The key thing is that you need to understand the fundamental structure of how things work REALLY WELL: What's the event loop? What is synchronicity? What's a 'this' binding? What's lexical scoping? What's the Box-Model? What's Margin-Collapse? etc. You need to know hundreds of the one line concepts REALLY WELL.
The second key thing is that you need to be solution oriented not information oriented, if you don't need to know something(that's on the applicability side and not on the structure), don't go and look it up, lookup what you need, implement it, move on and repeat until you get good. If it works then it works, move on. This is not only a practical consideration or from a fear of learning , but rather true learning is application and not information. The knowledge will automatically come to you over months and years of figuring one thing out, your brain is not a computer that it will instantly input everything about one thing.
Information is NOT knowledge, knowledge comes from building with it, using it, not reading or watching. It is a 10x better use of time to build something yourself than it is to watch a video of someone doing it.
I put up front-end related things here due to the channel but it applies to all of programming.
Advice from me: endeavour to make the course/watching/reading time be less than 10% of the total time spent, 90% needs to be in you fixing bugs, banging your head, asking chatgpt about how something works(NOT THE CODE TO COPY PASTE!!!), reapplying things again and again til it's good enough and building. Also, don't forget the thinking stage, spend a solid amount of time just thinking what the problem is, what the structure will be, what you need to do etc. Don't underestimate a pen and a paper, that's more important than the code. If you don't flesh out the problem and the solution, then after you build, you will realize that that's not what you wanted or that your architecture is spaghetti and that will waste 10x more time than the 10 minuts it would take you on paper.
> wtf is means that error message
almost a javascript moment, one might say
I worked for AWS in Madrid as a Professional Cloud Solution Architect. I felt dumb 90% of the time solving problems, but when I succeeded, I felt like Einstein.
Issue is overcomplicating stuff, find an idea on what you trully want to make, and you will find a way how to guarantee. You can ask chatgpt how to make it but dont use it for code, but roadmap
You should absolutely use it for code, just don't let it be creative (unless you're brainstorming). I find it's much faster to give a thorough english-language description of an algorithm and have chatGPT (or claude, which is more reliable) write the actual code than to write the code myself. It's just as good at getting you to think through the problem, and you don't need to get hung up on remembering all the syntax and function names. LLMs are pretty good at following clear, unambiguous instructions; it's the following of vague instructions that'll lead you to where you don't want to go.
I usually use AI to check if the code I wrote can be simplified and less redundant, to check if any part/line is unnecessary.
Before, I only asked AI for a roadmap and didn’t rely on it for coding. I was doing well with other languages and even built my own projects without AI. But everything changed when I reached my third year in college, where I had to learn PHP entirely on my own, without any help from professors. PHP really put a damper on my passion for programming.
A damper?
Why did you have to learn php... its terrible.
Where can i learn php? Any suggestions?
@@felixluna4184 what everyone just hates on that programming language for no exact reason
Same story for me too lmao
1) use Ai to be ambitious and do difficult projects
2) fuck up
3) learn to understand code
4) debug successful
5) **feels the fear of not writing code yourself
6) forget the fear and do harder projects
7) do this until you are used to it
Repeat step 1 to 6, and if your conviction is hard and you are too stubborn to give up, you can actually be dependent on Ai while able to program stuff
Actually, a really usefull project is integration
Application as well
I needed to hear this with the project i am building. Gonna change my approach and try to write the code myself. Break it down first and lay out the steps, then build it myself.
Let me give you another PoV. Part of the problem is the web development itself. It has ALWAYS been about A) Quick solutions B) Ready-made solutions. Always, since 90s where you can inspect the source of the page - something unheard in "normal" programming. This has pros and cons, pros being the insanely rapid development it brought, cons, well, hacky code, mind-numbing resource consumption, compered to the actual work done, etc. When this is the norm, no wonder the AI will first conquer web development. And again, before AI there were the frameworks, the templates, the CMS, arguably even the UML code generators - all tools which take development away from the developer. Noone writes a page from scratch for 25 years and the AI is yet another of those tools.
Yep, So true. Sadly I also just do web dev. You sound like a pro from C, C++ or Java background.
Breaking down the problem is what helped me to get away from tutorial hell. I was telling my self what is the simplest app that i could build something that i will not need anything for just documentation and that's it. It was bit painful but when i finished my first app it clicked to me and i stopped using tutorials and started building my own stuff which as you said provided that dopamine hit. When it comes to AI it's the opposite i tried it and when it shown me a result i didn't feel like continuing the project anymore because it wasn't mine anymore. So i deleted it and rewrote it. I will never use AI for coding it almost killed all the excitement from it for me.
I use co-pilot, which I use mostly to learn new stuff. For example, I recently started working on Ruby on Rails and Co-pilot is quite good at telling you different ways to write a piece of code you just wrote, using language-specific features you were not aware about.
this is why ai prompters infuriate me, ai is not here to replace, but to help, yet they want it to do everything for them
I agree.
if you are just using ai your code will reach a limit anyway x)
chatgpt is honetsly stupid at coding.
it uses weird functions and arrangement, starts hallucinating a new method that literally does not exist or just can't seem to debug anything by itself.
However, it is useful to talk to and send YOUR code to check if its efficiecny and ensure the logic is correct, and you cna put in it's memory "dont send me code"
Also never fall into the trap of thinking that things should be done only always in one perfect way. It's helpful to see fresh perspectives (ways that are possibly easier or faster to write) on areas of coding which you already rather familiar with.
As an enthusiast hobby coder who wrote only vanilla javascript for 6 years, for browser and node, I second this video.
This video made my want to start a project in python, I haven’t coded in a little while but this perspective is the GOLDEN NUGGET
Regarding 'misuse' of AI we all have to learn what works and what doesn't work in our own frame at our own time.
1:07 lmao showing purple guy as an example
If you dance I’ll dance
This ai era is like the white coke of every developer u get addicted then despair
Asking AI for something else than generating input data based on schema or reformatting a file or piece of text when it comes to programming is pure madness
Thank you. I am currently struggling with my JS learning path and I was kind of giving up to AI as I felt that I never would be able to understand how to use and combine different things, but your video was right on time ! Thanks a lot
I work as an electrician learning to code takes too long so fuck all that noise I’ll take my microsaas generator and make it work focusing on smooth UI, security, and meeting my intended functionality needs. Get with the times. I learned how to use fire base while making my first flutter app. And while securing it found out a lot about different potential vulnerabilities in databases and code. While trying to deploy ran into apple specific hurdles I had to stumble through and learn from like how to setup Xcode. Yes I use ai to code but I’m not an idiot and I am learning a bunch of skills that are going to help me in the next project
@@thruhazeleyes my point exactly! The ciders are a bit jealous that other can now just jump in to their exclusive club without,"paying the dues"
Even when using AI to generate the code, it's still a pain in the ass during the design process. Because gpt, as a machine, cannot generate a complete product on its own, it can't just magically finish your coding assignment in one or two prompts.
After multiple times of doom-prompting (ask gpt to generate the code and spamming prompts until it works, most of the time it doesn't), begging the AI to write the code that works as a whole (dumb shjt ik), I've realized that I can break the process into multiple smaller problems, ask the gpt to write the simplest codes possible, then I apply that logic into what I'm trying to build.
Sometimes it works, sometimes it doesn't and I have to write codes on my own :/
Reading documentation or how-tos for very unfamiliar things is very good and is never bad. Even more, being able to tackle the new challenges fast is excellent.
Whoever can build a decently sized app without looking up anything is a literal god.
I am going in other ways I am sick that everything is in different programming languages, so I just ask ai to implement small blocks of code and slap them together maybe making minor adjustments to see weather it works, like if you are trying to fuck around with shaders and stuff and other stuff (especially if you wanna run something perfomant in WebGl), you care more about getting math and concepts working correctly, without caring about which programming language you are using (ps, try using c++ for 2-3 years all other programming languages will be feeling like using a library or something)
Oh sweet summer child, building toy apps and not knowing how the language works.
This is random video for me and I do not code anything but it reminded me of university, where we learned Matlab. We had no prior knowledge of it or any coding but teacher gave is separate tstks that we had to do over few weeks, googling everything that is needed but the most important part was that we had to write comment for each line to explain what it does (so even stolen code will be understandable)
I got task to create billiard game where I have table, balls and by entering angle and power of shot it had to draw animation result of that hit. Unfortunately I had to steal few lines for code from internet because when ball hitting walls it had to go in other direction and I was struggling to flip coordinate system (basically my ball received direction after collision mirrored of what it's supposed to be).
It was a fun project, I enjoyed it
Wow, thanks for sharing.
Asking ChatGPT to write code is like asking a biologist to build an human.
They can probably create something that would fool a someone with minimal knowledge on how the thing works but the result probably won't function correctly.
This ☝️
The reality is AI IS going to take over programming and Web Development. THATS the bottom line. We might as well adapt and get with the program. Figure out how to get AI to work for us instead of trying to fight a losing battle. When that day comes that AI does completely take over would you rather be the one who fought it tooth and nail and ends up out on the street or would you rather be the one who has years of experience getting AI to work for you so your skills are very much in demand?
Great video..I'm going to share this with my Dev team. AI is truly bad at doing any complex coding and what you're left with is an application you don't understand which makes it hard to find/fix issues. As bad as it is at creating good code...try asking it to refactor code and enjoy all the breaking changes it introduces. AI doesn't think, it regurgitates.
It's just autocomplete on steroids
If you are using ai to generate your code, you never liked programming to begin with.
The problem with using word generators is the model needs to be trained on what the new learner specifically knows to generate code right on the cusp of a new horizon.
Unfortunately most new learners do not know how to qualify what they know let alone know how to tell a chatbot to make sure the chatbot doesn't just tell them what they want to hear when they ask the wrong questions.
Such a based take 🙏
Also, the first code is always the worst one. Build, rebuild, and rebuild again.
all you say is true me as a self thought dev I relate on that aloot 😭😭😭 the first thing I do on programing is a simple emulator for my custom assembly and when it works I am so proud of my self and quickly do more complex projects like solving a repetitive school work with codes, writing brainfk emulator, making a simple terminal that you can chat with other people, writing a hamming code and making a simple ai without any ML libs and I am so proud of those yes I use help from ai but AI still suck at logical tinking that will end up me fixing it's code that will never work
I love to share this on my classmates (this is our first year IT class 😅) that just copy from chatgpt and never fix it's error
Before we had framework programmers. Now we have AI programmers. Most people aren't going to follow your advice but the ones that do are going to be able to keep delivering features and keep production online.
Another tip that I would like to follow up is to question everything you do when following a tutorial. And be OBSESSED with the little things.
For example lets say you wanna write a function to handle form submissions by writing:
function handleSubmit(event) { event.preventDefault(); }
Now question yourself to a point where you have almost no question about it:
Why do I need event.preventDefault()? What does this line do, and what would happen if I left it out? How can I implement this in anohter sample project? Can I modify these even further?
What I usually like to do next after understanding a simple concept is building a quick small form submission website using what I just understood to deeply test your understanding. After truly understanding that concept inside out, resume back your tutorial and rinse and repeat for every new concept that you encounter.
When it comes to JS stuff like that I like to track from which object the method is called, so I do a quick read on the documentation and try to figure out what it's doing under the hood.
yeah I personally use AI for this. I throw the entire documentation for a particular library and ask it to help me whenever I want to create a particular thing without having to search for it in the documentation, and I ask it "why do I need to write this specific line?" "what does this function do?" "what happens if i remove this?". if it hallucinates a random answer then i go search for it on the web. The advantage when using AI is that it just follows the context, when in the web it can be a totally different concept. I learned how to use ThreeJS just by reading the documentation and asking chatgpt to help for a project.
best thumnail ever ... i knew i did the right thing when i subed this channel
Beginner Python user here: hell yeah I'm printing every step!
I use AI all the time for googling, I can't remember every single thing in the scripting sister language of Java, made in 1999.
AI is indeed bad at writing the exact code you need, or sometimes it will write code that doesn't even work (literally throws). It's nice when used to get a general idea of things I might have missed or never considered trying.
2:15 The had us on the first half 😆, but they lost us with the "ask the ai for guidance"
chatgpt is good for learning in my opinion but too many people just take its code and keep throwing prompts at it until it fixes its own half assed code. But chatgpt is great if you give it for example a code snippet that you dont understand and ask for an explanation or if you let it write code and then ask for an explanation. That way you will learn a lot about programming and also gpt isnt good for bigger projects as they often require more specific tools that gpt isnt too familiar with.
This ☝️
I think that development is a slow process, people want to build as many projects as possible, its like reading a book, if you read 10 pages a day and grasp what the author want to say, isn't like reading a book in one day,
The benefit will never be the same.
I don't. I ask it a question, and it gives me an entire code.
Can the next video be on the ui/ux and designing steps ,like I am good at doing logical stuff but it becomes hard to write html and css and where to get unique ideas from
Very soon. Like in a few days.
The problem with new developers is that people thinks that you can learn to do programming by just looking at a couple tutorials, and that's not how it works. I remember when I was 14 and had no internet connection whatsoever at the time, all I had was a couple of books about C++ and C#. And as much as I followed them and learn through them, at the end I couldn't do anything either because I didn't understand many basic concepts of computer science that are required to understand how to do what you want to do, that are not explained in those books. Just imagine trying to write anything just after watching a couple videos telling you how to write for loops. It really takes more than just f*cking around or watching tutorials, or even using GPT (it won't work, I tried it already and it's unreliable).
Lmao, I am Sir Toothy and you just made my day - you're most welcome for the inspiration
I was going to DM you on Reddit, but I thought it’d be better if you stumbled upon this video on your own. So happy to see you did 🤎
the thumbnail is lore accurate
insane video, full agreement 🤝🏻
Ah, yes, the "tinker until it works" crowd. Real blast from the past, aren’t they? Look, people who say we shouldn't use AI for coding are like those who'd prefer riding a bicycle instead of using a car because, hey, "you get there eventually." Sure, they’ll arrive, but they'll be exhausted, sore, and probably a decade behind.
Back in the day, coding in binary wasn’t just tedious; it was madness. Imagine writing one and zero for hours just to make a light blink. Then assembly came along, then high-level languages, then interpreters, compilers... all to save people from reinventing the wheel every single time. AI is just the next logical leap. It's like the Jarvis to your everyday code-the assistant who speeds up the grunt work, leaves you time for creativity, and catches things you might miss.
So, tell the tinkerers to keep tinkering, and while they’re doing that, we’ll be the ones building the next billion-dollar company with the time we saved.
Hey I'm just saying, you can either leverage the future and new tech or remain in the past and perish in the process.
Used AI for that one
Did you just compare the current LLMs to Jarvis and Cars.
When did I say don't use AI at all.
How the hell you learn something if not by tinkering, I am genuinely asking.
@whosajid If the goal is to Understand the code and actively take part in building your project, you can still ask AI to write your code especially if you have time constraints. Do you agree?
@whosajid LLMs are constantly growing and improving it's just a matter of time before they get to Jarvis level especially with quantum computing, also Jensen Huang said the goal is to get AI to write a code where a person without much skill just gives some prompts in English and Voila. People are actually building full apps with cursor AI in a short amount of time.
thank you for that vid homie
You can't even use AI generator if you don't know what are you doing nor you don't know what is your end result.
meanwhile me spent a 3 days to fix just a simple problem and after that i laughed so hard and that was been really good like to see that i've done smth :)
i agree even if i made a small toggle button i am wayy wayy more satisfied and motivated to do more complex projects then if i would have just asked gpt or copying youtube code
chatgpt is robot slave use it or people will use it and get ahead of you. it's very powerful
Finally a new Sajid video
Also in my experience, AI just makes code up. Like code that doesn't even exist. So of course tou try to implement it and it doesn't work.
Yeah happens all the time.
kind of off topic to the video but it took me a while to get out of the tutorial hell part of programming. i tried to learn lua and python and wasn't really able to learn anything and i just memorized certain functions but i didn't know the problem solving aspect of it all and i eventually forgot everything i learned. it wasn't until i decided to learn JavaScript that i fixed this. if i wanted to learn about a specific topic (lets take if statements) instead of looking it up, i would watch a video of someone explaining the general format and syntax of if statements (if i couldn't find a video, i would read documentation) and then i would get some questions off the internet relating to if statements and i would take what i learned and fix the problem. after all of this i would document it as if i was writing a book to publish so if i ever forgot how to do something, i could look at my document and get a quick refresher. but every since i started doing that, i noticed i dont really need tutorials, if i need to fix something, ill use documentation and perhaps even StackOverflow if it really comes down to it(AI is a last resort if i ask a question on multiple different forums and cant get an answer as AI usually just gives you the answer and doesn't explain it that well), and if i need to create something from scratch but i dont know how to do something, ill look up documentation for values or properties that allow me to do what i want to do. this might not work for everyone but it definitely helped me get out of tutorial hell.
Very on topic and thanks for sharing with us. I did the same thing 🤝
I agree but there are AI models that are better than others so using chatgpt will certainly result in messy and buggy code, but if you use DeepAI or Claude 3.5 sonnet it will be pretty awesome I use both and give them code and ask the same question so these AIs give each other AI code and ask what happened in the code and compare it to the old, theirs and the other AI's. DeepAI is often better than Claude (in pure programming and optimization, and it's more up-to-date).
Yeah Claude 3.5 > ChatGPT 3.5, and v0 is also nice.
I think its really just a matter of using AI for a new topic a couple times then doing it on your own from there on, after all its best use is as a teacher that teaches you something new in the best way you learn
F*ck around - sure, when writing your own projects you can. But if you're programming as a job nobody will pay you for "f*cking around". You need to use all available tools to meet the requirements and do it in given time. If asking AI for code and then correcting it is faster than writing it on your own from scratch, then that is the tool you're expected to use, most of the time.
Exactly.
This video is so stupid. AI can’t write code, but it’s capable of guiding you and explaining niche features. You’d realize this after seeing AI fail miserably at writing your code for you.
A majority of tutorials or forum posts are filled with unnecessary fluff to the point where you can’t find what you’re looking for.
I don’t need a technical explanation of everything. Oftentimes I’m just looking for a quick refresher and it becomes a 30min search session for what could’ve been explained in mere seconds. Even documentation suffers from the same problem.
@@modestmii bruh
@@modestmii This video is message for beginners.
Im very new with coding. I want to build my oown webshop. What is the best way to start so i can understand this?
Great video btw!
Learn the basics of HTML, CSS and JavaScript. For the next 6 months don't do anything else.
Meet me after 6 months.
For those of us who are pretty confident developers, do you have any advice for monetizing the skill? I've tried everything from e-commerce, web games, crypto projects, live streaming widgets but still nothing as lucrative as I would have hoped.
Great question:
Few ideas:
1. Start a UA-cam channel
Teach or build in public. It's not hard, trust me. But you do need luck.
2. Build something that people want(like really really want. Pay you to use the product - want). You are going to need very strong marketing skills and a lot of good luck.
3. Show this confidence in an interview with a startup founder(or tech lead). The easiest and most realistic route. Follow a lot of small tech founders on Twitter and maybe LinkedIn as well. See if you can jump in for a part of the pie.
Or,
If you are good in python or ML, jump on the AI wagon rn.
i made a project that was totally ai and i understood nothing. but recently i have started to make my own project using the same technology as the ai project but this time without any ai. it has been much more difficult but it is a lot more rewarding
😂😂😂
@@p3ter408 whats funny
Good for you, bro 🔥🔥🔥
@@Efandr thanks bro
Always love your content....
I'm learning while Ai is coding for me. I understand more with AI than any tutorial or course in a short time
Instead of blindly copying the code in the beginning you should watch the video and then code whatever feels good to you . See the approach not necessarily the code.... Like if the code is in react try to do it in vanilla js or if he used scss for styling use tailwind ...... A good way to keep yourself in check is that when watching a tutorial make sure when the utuber makes a mistake in video you don't make the same one , instead write the correct code before utuber goes ahead and correct his mistake
People will do anything but read a book
Thanks for all the good advices!
I wonder why the title of this video is not "f*cking around until it works" 🤔
Great video, thanks!
Yeah, I’m only willing to do this if it’s C++, the language that introduced me to the phenomenon of dreaming in code and allowed me to easily ai my way through js and especially python at my current job without consequences, so far lol 😅.
I wish I could code, let alone dream in C++. Massive respect 🙏
idk but when i use ai it fcking hallucinates and then i started having hallucinations too
Hey guys I understand what he is saying. But I think it’s good to write your own code by yourself if you are trying to learn the basics of your tech stack really well, loops, if statements, functions, how to debug. After that, I think using AI to write your code is good because the complex stuff is all bunch of basics lumped together. Also if you understand the basics, you can just spend more time reviewing and optimizing the A.I code. Also you will be able to build big projects and make money from them. You get to do other things either that just building projects like actually making an impact with your projects.
The goal is to become good enough to talk to the A.I and review it’s code. While you review its code you can ask questions about the code and add comments in your code to explain the code to your self.
unbelievably large L take. doing the basics over and over for years are the only way you can ever be good at any skill in life
Thank u alternate me for this uncommon video
I think you missed the point. AI should be used like a pair programmer.
Yeah exactly. Sorry if the video didn't make that clear.
AI for me just helps me complete tasks faster. Always work smart.
Always work smart.
But also,
Always think long term, especially if you are trying to build a career.
I was just about to start another tutorial before i saw this video
Code writing is the PERFECT use case for AI. You don't write code for the sake of writing code. It's not the goal, it's a burden.
I'm a programmer. I'm already using AI as an assistant. It makes translating code from one language to another extremely efficient. Since languages are very similar but still different, it's easy for the human mind to mix them up, but the AI doesn't.
It's also rare to know literally _all_ programming languages, but an AI can indeed know them all.
It can also rearrange the lines automatically (we don't need AI for that, but it still does it), explain a code, _expand_ a code, read a full library and give you the codes from that library...
If you need to know every programming language (really, more than 2 or 3) to do your job, something has gone very wrong; you've created an imaginary problem for AI to solve. AI can write trivially easy code (even then, you are likely to get nonsensical results), so it can save on typing, but it can not create new ideas: it simply regurgitates code that has already been written. There is no innovation, so you will always be limited in the problems that you can solve.
@@azure-hawk
In my first internship, I had two languages to use, C++ and Python.
In my first job, I had to learn C#.
In my spare time, I work on games ; RPG Maker required me to learn Ruby, and Minecraft required me to learn JavaScript.
I also came across software using lua language.
As an engineer, I also regularly use MatLab, which uses a _modified_ version of Python that has its own syntax rules.
It's dumb to state that programmers don't have to learn several languages. Most likely, you aren't one and obviously don't know better.
AI doesn't regurgitate things ; that's what a _regular program_ does.
When building a neural network, we create a mathematical representation of a dataset.
This mathematical function is then use to create NEW data.
Literally what scientists do whenever they want a mathematical law, except we do it automatically through a process we called training.
It's _interpolation_ of data, not regurgitation.
@@azure-hawk In my first internship, I had two languages to use, C++ and Python.
In my first job, I had to learn C#.
In my spare time, I work on games ; RPG Maker required me to learn Ruby, and Minecraft required me to learn JavaScript.
I also came across software using lua language.
As an engineer, I also regularly use MatLab, which uses a _modified_ version of Python that has its own syntax rules.
It's dumb to state that programmers don't have to learn several languages. Most likely, you aren't one and obviously don't know better.
AI doesn't regurgitate things ; that's what a _regular program_ does.
When building a neural network, we create a mathematical representation of a dataset.
This mathematical function is then use to create NEW data.
Literally what scientists do whenever they want a mathematical law, except we do it automatically through a process we called training.
It's _interpolation_ of data, not regurgitation.
@@azure-hawk You're clearly not a programmer. Don't speak when you don't know better.
(I'd love to expand, but Baby UA-cam won't let my comment pass.)
Trying again:
In my first internship, I had two languages to use, C++ and Python.
In my first job, I had to learn C#.
In my spare time, I work on games ; RPG Maker required me to learn Ruby, and Minecraft required me to learn JavaScript.
I also came across software using lua language.
As an engineer, I also regularly use MatLab, which uses a _modified_ version of Python that has its own syntax rules.
It's dumb to state that programmers don't have to learn several languages. Most likely, you aren't one and obviously don't know better.
AI doesn't regurgitate things ; that's what a _regular program_ does.
When building a neural network, we create a mathematical representation of a dataset.
This mathematical function is then use to create NEW data.
Literally what scientists do whenever they want a mathematical law, except we do it automatically through a process we called training.
It's _interpolation_ of data, not regurgitation.
And all that hard learning to be replaced by AI and become obsolete of course without a job
bruh😭
Critical role spotted.
remember bois Pythagoras stared at tiles.
I use claude to declare a variable at my job 💀
💀
Chatgpt can't write javascript code always make mistakes
which projects do you recommend doing? Can you give a list for junior lvl (Todo), intermediate, etc. I don't know which type of project I should start building.
I think AI can help you here.
Either censor your script or don't censor your audio. It completely ruins the video otherwise.
I wasn't planning to censor but did it anyway. Here's why- I know it ruins the flow and feel of the moment. But if YT decides to push down the reach just because of some weird advertising policies, all the hard work will go to waste.
Put yourself in my shoes.
Edit: Thanks for your feedback, I really appreciate it 🙂
AI can teach you in 5 minutes what a 1hour tutorial/course can/or not,
Can't argue with that.
You have a sunk cost fallacy. Because most of us invested enormous time in figuring out how to code, plus layers of different languages and frameworks, now we justify why coding with Ai and relying more an more on it is wrong and "replacing yourself." reality is and we all know it, that we have to be more of architects than coders.
Not really. You still have to problem solve with ai, and you won't be able to because you've never had to
@nickwoodward819 we have to agree on the definition of problem solving before having this debate.
For the purposes of learning you should actually learn. In the end of the day the AI is still outputting code for you and one day the code might not work and you'll have to debug. If you have never done it with your own code, you'll have a very hard time debugging the AI's code. And the problem might be too complex that you won't be able to debug the prompting either.
@RedactedBrainwaves2 in the immediate future probably.
I'm of the opinion that coding will be like a switchboard operator, obsolete.
And you will have to take more strategic approach to building. But I might be too optimistic (I am a technical startup founder)
@@helLf1nGerTA still, even if the AI does the coding and runs everything and you get the fully compiled application, someone needs to know how it all works because we can't trust an AI to do everything. We no longer have phone operators but phone companies still hire technicians who understand how the system works for routing phone calls to fix any issue that might occur with the automated systems. We still need people to know the low level aspects of programming languages even though js and python exists, and programming languages are more reliable in their mechanics than AI.
What is the AI app name you used to generate code? 😊
v0 by vercel
I would love to agree with you. Unfortunately I do not. It's like saying that you need to understand perfectly what now it's considered to be low level stuff (like memory management) to write working programs. The reality is that you do not need that, you take it for granted in 90% of the cases.
Tomorrow natural language will be just a new level of abstraction and high level programming lanfuages will be the new low level staff.
For sure, ij some application you will still need a complete understanding, but these are exception, not normality.
The future software engineering will mainly be requirement engineering.
Unfortunately writing the complex stuff with AI simply does not work at the moment because the AI doesn't understand anything. I've tried numerous times to write complicated functions with AI, explaining it in detail the input parameters and required output, and it just fails and fails. And when asked to fix its issues, it consistently claims that it has found and fixed the problems yet makes zero changes to the broken code. At least right now, it is much easier and consistent to just write the code yourself.
@twnyeneyme6058 Yes, I know it is still not able to write complicated stuff, but the entry barrier for writing code is much much lower now. I also know that progress is not linear but for sure in the future LLMs will not be worse than now and I bet the tools that natively integrate LLM (as Cursor) will be much better even without newer and/or more powerful language models
Who agreed, because I agreed with him💪💪💪 what a good video
love from india bro ..
He is also from India....
@@Coder.tahsin really ?
This needs less views so that more of my peers would get stuck in tutorial/GPT hell, hence making me a little bit more ahead of everyone
Damn you plotting 😭🙏
Disagree with the line of thought. Instead avoiding the use of AI while coding, make the project so large that ai alone can't keep track of everything at once
💀
@whosajid probably should rephrase, a good project should be complex enough that an AI cannot code the whole thing without human help.
@@Iron_spider99 Got that the first time only. Nice one.
any books you recommend to become a developer job. To tackle these projects without spending weeks on a same thing
The Odin Project.
Thanks ❤
Same with flutter code, i can write functional code than Ai.
Eh. I enjoy bug fixing the code others have worked on.
You like suffering.
You are a programmer.
2:45 - Brilliant
Learn how to properly use AI
NA .. Ask AI... it will be much better at coding anyway.... no point in learning something that a maschine does 100000 times better
Yeah 100000% 😂😂😂
Junior Engineer (1968): When I'm writing memory management routines I'm not able to do some of the math in my head without help from one of these new calculators.
Stupid Senior Engineer (1968): It's okay to use it when you encounter complicated math but ideally you want to practice enough so that over time you won't need to use a calculator as often.
In 2022 I had 6 senior and 1 junior engineer working for me in a lab in Seattle ($150k per month). Now I get more done on my own with AI ($400 per month).
If you find this offensive, you're scared. If this helps you understand what's happening on the ground so you can plan for the future, all power to you
source: trust me bro
@@jvmgang source: taxi driver
Why would I find it offensive. I use AI all the time. Not just for writing code. Because tbh it's not good at solving real problems. But yeah it helps me save time with smaller chunks of the project.