one concern is that using AI as a junior dev could lead you to a local maximum in your learning, where you will become very good at using the AI, but you won't be able to write better code than the limit of what the ai can write, while a senior dev who didn't learn with AI would be able to.
I think the bigger problem will just be that junior developers won't be able to think for themselves. They will just describe the problem and won't try to solve it, they will just copy it from AI. Compared to reaching local maximum in learning, I think the bigger issue will be inability to think about a problem and think about solutions.
Is that really a problem though? Everyone always talks about this but if AI is only going to get better and better, most people are going to continue using it in excess. I think eventually, there will be different tiers of engineers, some who can write code from scratch and some that need to leverage AI. I compare this to a developer who relies on wordpress to build pages vs one that builds them from scratch.
@@rand0mtv660 Just noticed this a while back, since I'm only a programmer for 2 years ( 1 year in front and now somewhat in the backend) I was mindlessly using chatgpt thinking that it solves everything and it did upon stumbled that some code does not make sense so what I'm doing now is just trying to solve the problem, write the code and then try to reduce complexity and redundancy using chat gpt. It keeps functionalities the same but way cleaner.
I don't know man, you have a good point but this is how tech works, every day developing basic things will be easier and faster and you need to leverage these features to be above other devs. Finally, companies don't matter how you build the product, they only want to get a good product, if you are a junior and you are able to resolve problems and make scalable, secure, and good-performance web pages, they don't matter how you built it. Socrates thought books or any kind of writing was going to make our brains weaker because we wouldn't think. We can apply the same principle here, It's just another tool we can leverage. But this is just my opinion, I can be wrong.
what I've noticed is that with these ai tools, I start to think less and less about how to solve a task, even a sometimes trivial one. I mean, for example, how to merge two data arrays into one, depending on some conditions. I think for about 10 minutes, write code, fail, open chatgpt. I don't like it. I think I'm getting dumber. I no longer feel the joy of solving problems because I'm leaving it to AI. If I let AI write my entire code base, it's "prompt engineering", not programming.
The better the dev is, the better the prompts are, the more efficient the tool is. You still need to understand, modify, fix, handle edge cases, tests. And think about architecture too. Chances are junior devs starting with that will produce horrendous codebases.
As a student, I'm kinda torn about using AI for coding. I get that people who learned pre-GPT can use it no problem (and I would too), but for me, it's tricky. I try to understand every line of AI generated code, but I worry I'm cheating myself and turning into just a code reviewer and copy-paster? At what point do I consider myself good enough at having learnt in the "old way" to start using these tools regularly without worrying about handicapping my learning? Right now I'm holding back a little because what I make with AI is way better than what I can just by myself. With the thought that if you get a taste of the good stuff you wouldn't wanna go back
Just going to chime in here and say to read/learn/watch (not just whichever is your favorite learning method, I mean do ALL of these) about programming paradigms, patterns and structures. Find topics relevant to the language/stack you're working in, but also knowing other languages and technology stacks helps too. The aim is to have a more general or abstract knowledge about what you can do when you are faced with a problem and need to construct a solution. And just to be clear, I'm not saying you need to spend hours every day learning, but if you spend even just 15-30 minutes learning something new not even every day but when you have a little spare time and mental energy you'll be better off.
Here's my take. If "writing code" is your job, you probably should be replaced. As TRUE developers, our job isn't to "write code." I mean who here actually enjoys the grunt work of literally writing out code? It absolutely sucks sometimes. Our job is to solve problems using the tools available to us. This is yet another tool for us to use. I use Cursor, and it makes me 10x more productive (this is not a sponsored comment lol). It's honestly refreshing not having to manually write out everything by hand all the time. Of course, I still have to understand what it's outputting so I can correct it if I need to. Consider that people who know how to read/write COBOL in a world where its been abstracted away get paid much much more than your average developer. Today's JavaScript could be the future's COBOL.
don't worry, most devs don't even know how to communicate with other humans. Do we really think they'll be able to figure out how to write well thought out prompts for AI?
And clients are not very tech friendly either. And they’re usually lazy. They don’t understand how to fix their basic email issues, do you think they’ll replace you by building their own apps with cursor or ChatGPT? Nah….
Agree with this comment. People that can only write code but can't solve complex business problems will be pushed out of the industry. Eventually I think the only engineers that will remain will be the architecture masters that have to peice things together and are really good at debugging
Ai can starting to solve the problem by themselves. Within next 5 years, there will be no web developers with high paycheck. Time to learn ML and Math from now on.
Personally, I feel like %99 of the reason I used to use copilot, was because it just wrote boilerplate/completed obvious statements/declarations etc. I bet you could get the same boost if you had a really cracked amount of code snippets and a really well configured text editor. I ended up stopping my sub to copilot because it would send me into autopilot mode (no pun intended), as it just keeps recommending more code than desired, and I would just "LGTM!" and accept it. Partial skill issue on my part, but I do think I've improved in skill and enjoyment of programming after I ended my sub. Though, I think this stuff is definitely useful for topics which you thoroughly understand, and learn nothing by doing manually.
I found copilot useful enough for typescript types and that type of boilerplate. For $10 if it saves you even 30 minutes that’s worth it from an engineering salary.
Nothing is slower in my day to day than reviewing a junior’s code. As such, until tools like this get *really* good, to the point of needing minimal review, im faster just writing it myself.
Personally, I stopped using the in-editor stuff. Every time I used it, I spent more time fixing the "small bugs" then it would take me to write the stuff myself with some research.
Okay, from what I'm seeing in terms of hiring trends, companies focus on short-term profits. Same as always. If they have a 10-man team handling a project, they're now saying "We're going to fire 5 and have the other five use AI tools to compensate." Are tools like this harming the job market? Yes. Can we stop it? No. Should we? Definitely not. Tools like Cursor will prevent companies from hiring juniors in the future, because grunt work isn't needed anymore. This will also stop students from actually gaining any real experience. To be honest, I see this getting worse and worse before it gets better and employers realize that most of their seniors are retiring and there's nobody to replace them with.
I've yet to meet an actual professional that claims AI is good enough to replace actual programmers. I am not saying job market is screwed right now in USA but I feel like AI is not the culprit. I could be wrong tho, but I work at an AI company in EU and no one is making claims like this. I have also used cursor before and it's okay, but hard capped by chatgpt4 performance.
@@erkoxify I'm in the EU as well. You've got a point, but you gotta keep in mind that investors don't really understand what AI is and what its current limitations are. They only see that insane amounts of money are funneled into this technology so they're willing to bet everything on it. AI will probably become much better than humans, but it's not there yet. We don't have the models nor the inference capabilities to replace seniors. However, shareholders only want to replace juniors, to lower the expenses of companies. They're too blinded by what people like Sam Altman are saying to even consider that AI is being blown out of proportion. We've barely scratched the surface of AI, but tech CEOs sell it as if it's a mature technology
@@erkoxify AI makes human developers more productive. Meaning fewer man hours are required to complete a task. Fewer man hours = fewer men required. AI is reducing team sizes. Eventually it will replace the teams entirely.
I work as a gamedev and never found chatgpt usefull. It usually gives incorrect solutions, makes shit up and when it does have a correct answer it's the most unoptimized and beginner friendly solution that probably won't even go to a prototype. It is fun at fundamental level to copy paste code from it and have a working solution which makes you happy but it won't take you anywhere in the professional industry. Specially outside of web development.
I'm assuming that it's because most game code is not open source so AI's don't have much game code to train off of. Keep in mind that even copilot struggles a lot and copilot has been trained on very large amounts of data.
As a fullstack web dev of nearly 20 years I can say that the majority of of the job is writing the same boilerplate code. It's basically fetching and writing stuff to a backend and displaying stuff on a viewport. AI will replace web developers within 5 years.
Gotta be honest, that is very impressive. For the part of senior not being keen adopting new tech, I don't fully agree. The fact is that there is plenty of new tech everyday and try and understand which one is valuable or solid and which one isn't, is a hard endevour. Juniors are less used to it so they tend to only see the benefits, but experience teaches us to actually ponder over things that might be good, but in the long term won't be. I could say, as a senior, that using AI to generate your code might be beneficial in the short term, helping writing many lines in a matter of minutes. BUT that is all code, that you are introducing, that you might not even understand fully and that most likely will lead to bugs and problems of maintainability in the long term. Who is right? I guess time will tell
AI does something fantastic that is to prevent going into road blocks, nothing takes you more out of "the zone" than forgetting how some method was called or not knowing 100% a framework api. Is not just that I code faster with AI, I can be coding for many more hours "in the zone"
Pretty cool I’ll try it. Your take on junior devs running circles around senior devs made laugh though. Only an intermediate/senior dev is going to be able to think critically of the output. And btw, I’ve been doing web dev for 20 years and everyone I know uses A.I. for coding now. We aren’t at all getting “left behind”
This is more for mid-level devs. Juniors will attempt to use these tools to skip the domain knowledge. That is currently the gate... these tools are not good for juniors, due to the need for humans to edit the results. This is all tech for mid- and senior- level devs to create MVP's but the extending and refining will always be a mid- to senior- level role Juniors always want to skip the basics and thats always where the most important knowledge is stored.
Making a web page responsive is a solved and well-documented problem, but it is also a common problem, so that makes this tool useful. Still, I wonder how well this would be able to handle something more novel and difficult, e.g. fixing a bug in a CRDT library or writing a test for a greedy graph algorithm.
at my company, we cannot even access websites (even for learning purposes) without company permissions and we for sure cannot copy and paste code into any LLM so yeah i feel like learning to debug code and read documentations are still valuable skills because we cannot always rely on them.
Sorry, but can confirm it. For simple tasks you can use the ai generated stuff but for a more complex use case its not possible to use any of the suggestions the ai is giving me.
@@WebDevCody Our customer needs a time tracking service where you can track project hours, work breaks, and vacations. There are many edge cases to consider, such as correct timezone calculations, public holidays, overlapping time entries, etc. I tried using Copilot for the integration tests, but it failed most of the time.
I think it's safe to say that while these tools will definitely not replace your job, they will make timeline expectations waaay shorter from employers because more and more programmers are going to start using AI and will start delivering products in shorter amount of time, so if you decide not to you're going to be left behind. That being said I do genuinely think that the constant AI usage will drop the quality of code even more (which is an accomplishment considering how bad most react code looks already)
Again the idea of "If you don't you IA to code you will lost the race." A developer that has that much xp talking shit like that will make people fall for wrong paths. A Jr that code little and use AI to code all the time will get to maximum leaning very soon. For me all I can see are big companies trying to sell there product saying that AI will change the game but little had been improved to new AI. They do the same and do go to far from it other. Learn how to code, and you will be fine. Be dependent to this tool and you will be in a conner that maybe you don't like.
Does anyone else have issues with the extensions in Cursor IDE? I mean, if I switch profiles, it keeps the same extensions as the previous one (even tho I remove, or add new extensions).
I've been using Cursor for more than 6 months. It's really good. Great for first draft when I'm super tired. And when I'm a little lost at what this bulk of code does. It also reads from other files in your codebase. Awesome companion. But as usual it does get wrong once a while or with unoptimised code.
That’s exactly why I’m finding it amazing. After worker I’m tired and it’s harder to think, and I just don’t want to code. Cursor, although it isn’t perfect, helps me push through even when I’m tired 😆
At this point, I think people are not using the tool. The tool are just using people to improve themselves. They generate the code and we solve some problem for their code and they learn how to solve the problem from us until they no longer need our solving skills. Anyway, there will always be the new opportunities for the jobs.
Idk I feel like anything with even a moderate level of complexity I want to be familiar with at least somewhat. If it is all auto-generated for me it honestly feels more of a pain sifting through it myself to tweak or even to just understand it for liability sake. Cause you know, even if it is auto-generated I am the one including that code and "approving" it hopefully with some level of scrutiny. With things like this its like, yeah sure that is useful. But like templates exist already right? Like template sites you can purchase just once exist. I have a license for a template that I use constantly if I just want to have a quick responsive UI site that I can just make tweaks to here and there depending on the project. And most frameworks worth using nowadays imo lend themselves to templating. I have so much code I already copy and paste myself from. The main difference from this method and using AI is that I am more familiar with the code. Don't get me wrong, I use copilot. But I'd rather have it give me suggestions on the fly like an autocomplete than have it just dump a whole chunk of code on me. Again, mostly for anything with even a moderate level of complexity.
at the end of the day it's simple code; it's React Andy AI. Just because it can do it quickly (obviously it can do it quicker than humans) doesn't mean that it's good at novel or challenging problems. Making something responsive or making a couple of toasts in a next.js app is not a job description either. AI really flops on things that are not well represented in its training data/textbooks, it's not adaptable in that way. Find your niche, learn it deeply & specialise, and you're pretty much guaranteed not to lose your job to AI
AI is good for frontend, infact faster to use as who wants to write bunch of tailwind, but i doubt it will ever replace actual problem solving part of engineering. Unless someone makes AGI and just destroys the entire world lmao, not just programming Btw is it available in NVIM? I have been using nvim just to avoid using copilot (ik you can set it up, but it takes some work), but i am thinking to start using cursor for frontend. Seems better than copilot
I believe it should be a mixture, a balance. Something that is repetitive, and your mind goes ahead of your fingers, and you are not able to type fast enough - those cases using ai to generate code is a must. There are other cases, where it can be used to find a better, or alternative ways to skin a cat. So, generate, learn, practice several times, be done. But I've seen some videos already where people are abandoning code generating AI's because they started feeling they are becoming sluggish. You need to practice typing and coding the "hard" way, ie do it yourself. So, as I said earlier - it's a balance.
Tried it on a medium sized backend app written in Kotlin and tbh it's pretty much useless. I asked it to generate simple unit tests for a validator class and the three test cases I got had 15 compilation errors and generally didn't make any sense. After few prompts I had to redo everything myself. It's great for explaining code to the newcomers but to someone who worked on the project for 2+ years its not very useful. Canceled the subscription the next day...
@@bossrabbit Ah yes, leetcode, the exact material they train this shit on. That's the dumbest benchmark to perform at this point. AI still breaks fantastically trying to solve real world issues or debugging real world code. o1 and Claude generally work well, but they require A LOT of coaching on production Python and C# code. Not so rarely, they make shit worse than it was before.
@@bossrabbit alright AI bro. Soon there won’t be that many humans feeding new things into the programming/plagiarism machine and the curve will flatten further (or the power required for something valuable or novel will increase) and for many people who work with code, programming is not their full workload, or not even the majority. It’s a predictive model, not AGI and in its current form, most AI companies are taking losses and can’t sustain in the long term. But keep circlejerking. I can almost guarantee you were/are an nft/web3 bro before this.
Somehow this makes me sad. At some point AI will become so good, first, we wont need to write code anymore and second, we wont write code anymore. At last, we wont need to think anymore...
Spreadsheets replaced the calculator, AI will replace developers. It's not a question of "if" it's a question of "when". If you're under 50, you seriously need to consider re-skilling unless you want to end up spending your days on the couch, smoking pot and collecting government cheques. Speaking as a full stack web dev of nearly 20 years.
Alright. This is pretty awesome. I just tried it for 2h and it's amazing. If you give it a lot of details and the files it needs to look into, it's pretty good. I just finished some boring tasks.
exactly, i'm in exactly the same boat. It increases my productivity A LOT, it's very good at doing the tedious simple things so I can focus on the actually hard things. I'm using Cody, it's not as autonomous as this, but it's pretty good for what I need. I'll definitely need to give Cursor a try though. How much time would I have spent making that simple functionality? An hour? Yeah not a lot, but that's an hour saved on my end. Did it build something very complex? No. Did it need to? No.
i think you can have different .cursorrules files in different directories? Often i wont want e.g. typescript specific prompts in the backend part thats in python for example so yea just some writing hints, also if you make prompt files you can just at the file which might be simple if that different directory thing doesnt work
update that they did workspaces that you can at - its like a prompt with some files that you can then at so gives you more control similar to what i was talking about...
2:58 i would rather write code than explain everything to an AI in plain english. and also this task seems easy. but as we increase the complexity i afraid that this is going to be a tedious job
yea, congrats cody discovered ai, will go through through the typical phases of amusement, amazed, then later disappointment, then later regret and anger then it will go through the whole cycle of grief, then finally say AI sucks and no one has to worry about it.
i think only reason AI is not competent writing complex code is because it uses snapshots of existing codes on the internet to train itself rather than learning how to do human level abstract thinking. it doesn't go further than codes. we go further as humans. we think of what is the use case, what is the scalability, what is the reuse potential, what is the audience of this product, how can we make it faster without compromising anything etc. AI only knows how to respond to the prompts.
@@jurassicthunder after 50+ years, people still don't understand what ML/AI is. It has always been the same principle, just that we now have means to process terabytes of data very quickly and efficiently. and you are right, unless something fundamentally changed, no matter how many years has passed AI will always be the same.
Solving non-technical-consumer problems =/= solving engineering problems. Coders/devs often solve consumer problems; engineers often solve technically hard business problems with a slow ramp up time. The former is really easy to automate away, such as with wordpress for small businesses. The latter is really hard to automate away. If your entire skillset is gluing together library APIs, I’ll be a gate keeper and say, you’re likely not much of an engineer. You’re often not *solving problems* you’re implementing an *already made solution* for the problem. You often didnt solve anything. Someone else did and youre just gluing it together. Im not trying to devalue the importance of coders, but just making it clear whether or not one should be worried about automation and AI. You’ll have better job security in the latter. Theres nothing wrong working on an app for people to schedule meetings. But just know, your businesses point of differentiation isnt a technical one. Its likely operations, logistics, distribution, etc.. If your business is writing a better PDF parser for other businesses to use to extract data they need, your point of differentiation is a technical one. Which one do you think gives the person writing the code better job security?
You're making it seem like the point of differentiation is categorically different, but it really is just a matter of difficulty from a technical point of view. Basically, PDF parser - hard / Scheduling app - easy. Although, honestly, a full-featured scheduling app is no joke. I worked on one and the collision detection algorithm is not something that you can ask an AI to develop for you, at least not without a mountain of prompts and patience.
@@parlor3115i probably didnt use the best examples, but i think were in the same idea. The point is to position yourself to solve engineering problems, not apply premade libraries for your day job. As for point of differentiation, i think it us categorically different. If your company views your job as a programmer as a cost sink (i.e. maintaining a web app for say a real-estate company) verses views your job as a value generator (i.e. maintaining a proprietary, custom app which is your value proposition to the customer), your job security will be very different. You want to be what makes the company money. Not what is just overhead on a company balance sheet.
@@hamm8934Fair enough, but trust me, current AI is capable of solving very specific business problems. It's just the overhead of providing the adequate prompts to get it to do so does not justify the end result. Basically it's a situation of "I'll just do it myself, instead of hand holding through the entire process". However, when AI evolves to be more mature / the right AI startup creates one purely for the sake of engineering software (already the case just needs more work and polishing), it will actively ask businesses and technical question and make proposals and act on them autonomously. Personally, I feel the only way we can be safe as SE is if AI development reaches a dead end and they can't develop more sophisticated AI for a few more decades or at least until a point where the hype and investment dies out or slows down considerably.
i get what you are saying but if ai can do all the work where is the joy of writing code anymore it's boring . Unless you are a sadistic and you like obedient ai
I love tailwind and how easy it to style with it, but OMGG the time i need to spend to make the pages responsive is crazzy. If you say it is that easy then my life would get soo much better. I feel like it's with every ai tool, it helps u to be more productive, would it replace me completely... well.... that's an heated topic 😅
You said it in the beginning. I created a live stream, and it's catching lots of views... You are a content creator, aspiring for content engagement, so here we are.
It just feels wrong to do everything with AI, you're learning nothing. I use codeium, which is very minimal assistance and also free. Sure, you're slower but at least you understand what your codebase does.
at this point, I can identify what problems are worth solving by hand, and what I would consider a "boiler plate problem", meaning something I could find on stack overflow and doesn't require much heavy thinking to implement. AI helps write the boilerplate and I'm ok with that.
If the ultimate goal of a dev is to solve coding problems all by your brain, don't use AI. But if the goal is to deliver apps that solve end users' problems, use whatever tools help you reach your goal! I just feel like conservative devs try to romanticise and pride themselves too much.
and this a typical coper. this wont replace devs. but it will replace most. only few will be hired to write the prompts and know enough to say it works or not.
one concern is that using AI as a junior dev could lead you to a local maximum in your learning, where you will become very good at using the AI, but you won't be able to write better code than the limit of what the ai can write, while a senior dev who didn't learn with AI would be able to.
This is so true, it's a big risk to coming juniors. Even bring a seasoned developer, once you get used to this stuff, it's VERY hard to go back...
I think the bigger problem will just be that junior developers won't be able to think for themselves. They will just describe the problem and won't try to solve it, they will just copy it from AI. Compared to reaching local maximum in learning, I think the bigger issue will be inability to think about a problem and think about solutions.
Is that really a problem though? Everyone always talks about this but if AI is only going to get better and better, most people are going to continue using it in excess. I think eventually, there will be different tiers of engineers, some who can write code from scratch and some that need to leverage AI. I compare this to a developer who relies on wordpress to build pages vs one that builds them from scratch.
@@rand0mtv660 Just noticed this a while back, since I'm only a programmer for 2 years ( 1 year in front and now somewhat in the backend) I was mindlessly using chatgpt thinking that it solves everything and it did upon stumbled that some code does not make sense so what I'm doing now is just trying to solve the problem, write the code and then try to reduce complexity and redundancy using chat gpt. It keeps functionalities the same but way cleaner.
I don't know man, you have a good point but this is how tech works, every day developing basic things will be easier and faster and you need to leverage these features to be above other devs. Finally, companies don't matter how you build the product, they only want to get a good product, if you are a junior and you are able to resolve problems and make scalable, secure, and good-performance web pages, they don't matter how you built it. Socrates thought books or any kind of writing was going to make our brains weaker because we wouldn't think. We can apply the same principle here, It's just another tool we can leverage. But this is just my opinion, I can be wrong.
Dang cody discovered chatgpt, we are done
its actually claude ai which is more powerful then chatgpt.
@@siskoDE difference is minimal, i use both, but i prefer chatgpt only because of better UI and the agents, the gpts
@@cslearn3044chatgpt sucks at logic lol wtf are you using it for
@@cslearn3044I wouldn’t say it’s minimal. I’ve been using copilot since it first came out, and this model was the first time I’m actually impressed
@@WebDevCody idk, when i use chatgpt i get very similar stuff, but probably its because of prompts
what I've noticed is that with these ai tools, I start to think less and less about how to solve a task, even a sometimes trivial one. I mean, for example, how to merge two data arrays into one, depending on some conditions. I think for about 10 minutes, write code, fail, open chatgpt. I don't like it. I think I'm getting dumber. I no longer feel the joy of solving problems because I'm leaving it to AI. If I let AI write my entire code base, it's "prompt engineering", not programming.
Was this clickbait? Yes
Is it a good tool to have? Absolutely.
what is the link for this took
The denial is real lol
The better the dev is, the better the prompts are, the more efficient the tool is. You still need to understand, modify, fix, handle edge cases, tests. And think about architecture too. Chances are junior devs starting with that will produce horrendous codebases.
As a student, I'm kinda torn about using AI for coding. I get that people who learned pre-GPT can use it no problem (and I would too), but for me, it's tricky. I try to understand every line of AI generated code, but I worry I'm cheating myself and turning into just a code reviewer and copy-paster?
At what point do I consider myself good enough at having learnt in the "old way" to start using these tools regularly without worrying about handicapping my learning? Right now I'm holding back a little because what I make with AI is way better than what I can just by myself. With the thought that if you get a taste of the good stuff you wouldn't wanna go back
as a student, your goal is always to learn. if you don't feel you're learning, don't use ai
Just going to chime in here and say to read/learn/watch (not just whichever is your favorite learning method, I mean do ALL of these) about programming paradigms, patterns and structures. Find topics relevant to the language/stack you're working in, but also knowing other languages and technology stacks helps too.
The aim is to have a more general or abstract knowledge about what you can do when you are faced with a problem and need to construct a solution.
And just to be clear, I'm not saying you need to spend hours every day learning, but if you spend even just 15-30 minutes learning something new not even every day but when you have a little spare time and mental energy you'll be better off.
Making something responsive is the most soul sucking part of front end dev.
That an implementing theming.
If you don't do it from the start you are gonna be miserable
Yeah honestly most of what I use AI for in my work is UI and especially responsive UI.
I just use some UI Lib.
because you're too lazy to learn css and it's flexbox
Here's my take. If "writing code" is your job, you probably should be replaced. As TRUE developers, our job isn't to "write code." I mean who here actually enjoys the grunt work of literally writing out code? It absolutely sucks sometimes. Our job is to solve problems using the tools available to us. This is yet another tool for us to use. I use Cursor, and it makes me 10x more productive (this is not a sponsored comment lol). It's honestly refreshing not having to manually write out everything by hand all the time. Of course, I still have to understand what it's outputting so I can correct it if I need to. Consider that people who know how to read/write COBOL in a world where its been abstracted away get paid much much more than your average developer. Today's JavaScript could be the future's COBOL.
don't worry, most devs don't even know how to communicate with other humans. Do we really think they'll be able to figure out how to write well thought out prompts for AI?
And clients are not very tech friendly either. And they’re usually lazy. They don’t understand how to fix their basic email issues, do you think they’ll replace you by building their own apps with cursor or ChatGPT? Nah….
Agree with this comment. People that can only write code but can't solve complex business problems will be pushed out of the industry. Eventually I think the only engineers that will remain will be the architecture masters that have to peice things together and are really good at debugging
Ai can starting to solve the problem by themselves. Within next 5 years, there will be no web developers with high paycheck. Time to learn ML and Math from now on.
Ehm, I enjoy writing code, and I bet most programmers do.
Personally, I feel like %99 of the reason I used to use copilot, was because it just wrote boilerplate/completed obvious statements/declarations etc. I bet you could get the same boost if you had a really cracked amount of code snippets and a really well configured text editor.
I ended up stopping my sub to copilot because it would send me into autopilot mode (no pun intended), as it just keeps recommending more code than desired, and I would just "LGTM!" and accept it. Partial skill issue on my part, but I do think I've improved in skill and enjoyment of programming after I ended my sub.
Though, I think this stuff is definitely useful for topics which you thoroughly understand, and learn nothing by doing manually.
I found copilot useful enough for typescript types and that type of boilerplate. For $10 if it saves you even 30 minutes that’s worth it from an engineering salary.
Nothing is slower in my day to day than reviewing a junior’s code. As such, until tools like this get *really* good, to the point of needing minimal review, im faster just writing it myself.
Personally, I stopped using the in-editor stuff. Every time I used it, I spent more time fixing the "small bugs" then it would take me to write the stuff myself with some research.
Okay, from what I'm seeing in terms of hiring trends, companies focus on short-term profits. Same as always. If they have a 10-man team handling a project, they're now saying "We're going to fire 5 and have the other five use AI tools to compensate." Are tools like this harming the job market? Yes. Can we stop it? No. Should we? Definitely not.
Tools like Cursor will prevent companies from hiring juniors in the future, because grunt work isn't needed anymore. This will also stop students from actually gaining any real experience. To be honest, I see this getting worse and worse before it gets better and employers realize that most of their seniors are retiring and there's nobody to replace them with.
I've yet to meet an actual professional that claims AI is good enough to replace actual programmers. I am not saying job market is screwed right now in USA but I feel like AI is not the culprit. I could be wrong tho, but I work at an AI company in EU and no one is making claims like this. I have also used cursor before and it's okay, but hard capped by chatgpt4 performance.
@@erkoxify I'm in the EU as well. You've got a point, but you gotta keep in mind that investors don't really understand what AI is and what its current limitations are. They only see that insane amounts of money are funneled into this technology so they're willing to bet everything on it. AI will probably become much better than humans, but it's not there yet. We don't have the models nor the inference capabilities to replace seniors. However, shareholders only want to replace juniors, to lower the expenses of companies.
They're too blinded by what people like Sam Altman are saying to even consider that AI is being blown out of proportion. We've barely scratched the surface of AI, but tech CEOs sell it as if it's a mature technology
@@erkoxify AI makes human developers more productive. Meaning fewer man hours are required to complete a task. Fewer man hours = fewer men required. AI is reducing team sizes. Eventually it will replace the teams entirely.
I work as a gamedev and never found chatgpt usefull. It usually gives incorrect solutions, makes shit up and when it does have a correct answer it's the most unoptimized and beginner friendly solution that probably won't even go to a prototype. It is fun at fundamental level to copy paste code from it and have a working solution which makes you happy but it won't take you anywhere in the professional industry. Specially outside of web development.
I'm assuming that it's because most game code is not open source so AI's don't have much game code to train off of. Keep in mind that even copilot struggles a lot and copilot has been trained on very large amounts of data.
As a fullstack web dev of nearly 20 years I can say that the majority of of the job is writing the same boilerplate code. It's basically fetching and writing stuff to a backend and displaying stuff on a viewport. AI will replace web developers within 5 years.
Gotta be honest, that is very impressive. For the part of senior not being keen adopting new tech, I don't fully agree. The fact is that there is plenty of new tech everyday and try and understand which one is valuable or solid and which one isn't, is a hard endevour. Juniors are less used to it so they tend to only see the benefits, but experience teaches us to actually ponder over things that might be good, but in the long term won't be. I could say, as a senior, that using AI to generate your code might be beneficial in the short term, helping writing many lines in a matter of minutes. BUT that is all code, that you are introducing, that you might not even understand fully and that most likely will lead to bugs and problems of maintainability in the long term. Who is right? I guess time will tell
fair take
AI does something fantastic that is to prevent going into road blocks, nothing takes you more out of "the zone" than forgetting how some method was called or not knowing 100% a framework api. Is not just that I code faster with AI, I can be coding for many more hours "in the zone"
Pretty cool I’ll try it. Your take on junior devs running circles around senior devs made laugh though. Only an intermediate/senior dev is going to be able to think critically of the output. And btw, I’ve been doing web dev for 20 years and everyone I know uses A.I. for coding now. We aren’t at all getting “left behind”
This is more for mid-level devs. Juniors will attempt to use these tools to skip the domain knowledge. That is currently the gate... these tools are not good for juniors, due to the need for humans to edit the results. This is all tech for mid- and senior- level devs to create MVP's but the extending and refining will always be a mid- to senior- level role
Juniors always want to skip the basics and thats always where the most important knowledge is stored.
Making a web page responsive is a solved and well-documented problem, but it is also a common problem, so that makes this tool useful. Still, I wonder how well this would be able to handle something more novel and difficult, e.g. fixing a bug in a CRDT library or writing a test for a greedy graph algorithm.
Does your company allow you to give these tools complete access to codebase or is this something for personal projects.
personal projects only
at my company, we cannot even access websites (even for learning purposes) without company permissions and we for sure cannot copy and paste code into any LLM so yeah i feel like learning to debug code and read documentations are still valuable skills because we cannot always rely on them.
Cursor is great. Another reason to use it - ZERO Microsoft tracking and telemetry.
Sorry, but can confirm it. For simple tasks you can use the ai generated stuff but for a more complex use case its not possible to use any of the suggestions the ai is giving me.
walk me through what your idea of a complex use case is
@@WebDevCody Our customer needs a time tracking service where you can track project hours, work breaks, and vacations. There are many edge cases to consider, such as correct timezone calculations, public holidays, overlapping time entries, etc. I tried using Copilot for the integration tests, but it failed most of the time.
@@WebDevCodyi guess big tech, finance companies, ecommerce companies etc.
I think it's safe to say that while these tools will definitely not replace your job, they will make timeline expectations waaay shorter from employers because more and more programmers are going to start using AI and will start delivering products in shorter amount of time, so if you decide not to you're going to be left behind. That being said I do genuinely think that the constant AI usage will drop the quality of code even more (which is an accomplishment considering how bad most react code looks already)
Again the idea of "If you don't you IA to code you will lost the race." A developer that has that much xp talking shit like that will make people fall for wrong paths. A Jr that code little and use AI to code all the time will get to maximum leaning very soon. For me all I can see are big companies trying to sell there product saying that AI will change the game but little had been improved to new AI. They do the same and do go to far from it other. Learn how to code, and you will be fine. Be dependent to this tool and you will be in a conner that maybe you don't like.
Should've used AI to proofread this comment
@@AS-fv1tz You speak English because it’s the only language you understand. I speak English because it’s the only language you understand.
@@felipelauton812Nope
@@felipelauton812 nah I speak other languages too
@@AS-fv1tz you was already mogged brother
"Ok so it shows you the changes so you can review them...aight imma just accept all lmao"
😆
Why cant this be a extension?
Does anyone else have issues with the extensions in Cursor IDE? I mean, if I switch profiles, it keeps the same extensions as the previous one (even tho I remove, or add new extensions).
I've been using Cursor for more than 6 months. It's really good. Great for first draft when I'm super tired. And when I'm a little lost at what this bulk of code does. It also reads from other files in your codebase. Awesome companion. But as usual it does get wrong once a while or with unoptimised code.
That’s exactly why I’m finding it amazing. After worker I’m tired and it’s harder to think, and I just don’t want to code. Cursor, although it isn’t perfect, helps me push through even when I’m tired 😆
At this point, I think people are not using the tool. The tool are just using people to improve themselves. They generate the code and we solve some problem for their code and they learn how to solve the problem from us until they no longer need our solving skills. Anyway, there will always be the new opportunities for the jobs.
Is this a separate IDE or a vs code extension?
Ide which is a fork of vscode
So do you still plan on using Supermaven for auto-completion and Cursor for larger additions, especially ones that span multiple files?
nah cursor seems to be just as fast imo. no need for supermaven at this point
@@WebDevCody dang, alright imma check it out
Hey Cody, Does Cursor autocomplete code when you start typing like CoPîlot does ? Cause if it does it might be something I would change for!
Idk I feel like anything with even a moderate level of complexity I want to be familiar with at least somewhat. If it is all auto-generated for me it honestly feels more of a pain sifting through it myself to tweak or even to just understand it for liability sake. Cause you know, even if it is auto-generated I am the one including that code and "approving" it hopefully with some level of scrutiny.
With things like this its like, yeah sure that is useful. But like templates exist already right? Like template sites you can purchase just once exist. I have a license for a template that I use constantly if I just want to have a quick responsive UI site that I can just make tweaks to here and there depending on the project. And most frameworks worth using nowadays imo lend themselves to templating. I have so much code I already copy and paste myself from.
The main difference from this method and using AI is that I am more familiar with the code. Don't get me wrong, I use copilot. But I'd rather have it give me suggestions on the fly like an autocomplete than have it just dump a whole chunk of code on me. Again, mostly for anything with even a moderate level of complexity.
This is cool indeed, I'm gonna try it. I still don't see myself building features with it but for writing test...omg super cool.
Incredible tool. Is it possible to find something similar but with locally hosted models (which also can look into all project files)?
For sure it's game changing and for sure it's super powerful and useful but we are not cooked (yet)
at the end of the day it's simple code; it's React Andy AI. Just because it can do it quickly (obviously it can do it quicker than humans) doesn't mean that it's good at novel or challenging problems. Making something responsive or making a couple of toasts in a next.js app is not a job description either. AI really flops on things that are not well represented in its training data/textbooks, it's not adaptable in that way. Find your niche, learn it deeply & specialise, and you're pretty much guaranteed not to lose your job to AI
AI is good for frontend, infact faster to use as who wants to write bunch of tailwind, but i doubt it will ever replace actual problem solving part of engineering. Unless someone makes AGI and just destroys the entire world lmao, not just programming
Btw is it available in NVIM? I have been using nvim just to avoid using copilot (ik you can set it up, but it takes some work), but i am thinking to start using cursor for frontend. Seems better than copilot
I believe it should be a mixture, a balance. Something that is repetitive, and your mind goes ahead of your fingers, and you are not able to type fast enough - those cases using ai to generate code is a must. There are other cases, where it can be used to find a better, or alternative ways to skin a cat. So, generate, learn, practice several times, be done. But I've seen some videos already where people are abandoning code generating AI's because they started feeling they are becoming sluggish. You need to practice typing and coding the "hard" way, ie do it yourself. So, as I said earlier - it's a balance.
What's the reason you don't wanna use interfaces and prefer types 6:43
No good reason, just keeping it simple.
@@WebDevCodyoh! 😅 I thought I missed something in TS.
I dunno about 'hours' of work saved, maybe an hour based on the ui you showed, still cool as!
Tried it on a medium sized backend app written in Kotlin and tbh it's pretty much useless. I asked it to generate simple unit tests for a validator class and the three test cases I got had 15 compilation errors and generally didn't make any sense. After few prompts I had to redo everything myself. It's great for explaining code to the newcomers but to someone who worked on the project for 2+ years its not very useful. Canceled the subscription the next day...
When AI can become creative by itself and not use human data to formulate content, then we are in trouble. Even LLM creator says LLMs are NOT IT.
Every couple of years we need to reinvent same BS solution to teach newcomers, why it sucks... and this will never change.
As long as I have to coach it 6 times to make a working string replacement script, I’m not concerned for my job :’)
Bro, AI is breaking Leetcode records left and right. Don't get complacent, AI is coming for your lunch!
@@bossrabbit Ah yes, leetcode, the exact material they train this shit on. That's the dumbest benchmark to perform at this point. AI still breaks fantastically trying to solve real world issues or debugging real world code. o1 and Claude generally work well, but they require A LOT of coaching on production Python and C# code. Not so rarely, they make shit worse than it was before.
@@bossrabbity’all AI bros are tiresome
@@rdvansloten you’ll have plenty of time to contemplate that thought when you’re on the unemployment line friend 😉
@@bossrabbit alright AI bro. Soon there won’t be that many humans feeding new things into the programming/plagiarism machine and the curve will flatten further (or the power required for something valuable or novel will increase) and for many people who work with code, programming is not their full workload, or not even the majority. It’s a predictive model, not AGI and in its current form, most AI companies are taking losses and can’t sustain in the long term. But keep circlejerking. I can almost guarantee you were/are an nft/web3 bro before this.
Why is my composer view not fullscreen like yours, did you have to change a setting?
I type cmd shift i
Somehow this makes me sad.
At some point AI will become so good, first, we wont need to write code anymore and second, we wont write code anymore. At last, we wont need to think anymore...
Cant we just cut the chase and ask it to make a $10k mmr app
I wish
You always find amazing tools, great video! Keep it up Cody 💪
I'm surprised by the amount of dislike (hate?) when we mention AI..
I couldn't figure out how to give my chatgpt key then I learned the API key from the plus plan doesn't work.
I guess we just subscribe.
Spreadsheets replaced the calculator, AI will replace developers. It's not a question of "if" it's a question of "when". If you're under 50, you seriously need to consider re-skilling unless you want to end up spending your days on the couch, smoking pot and collecting government cheques. Speaking as a full stack web dev of nearly 20 years.
spending my days on a couch doesn't sound too bad now that I think about it
@@WebDevCody same lol. lets lay in one together and watch some non-sense on the telly :D
Very kind of you to say please to the AI overlords
Alright. This is pretty awesome. I just tried it for 2h and it's amazing. If you give it a lot of details and the files it needs to look into, it's pretty good. I just finished some boring tasks.
command + shift + I ? I can't open it, anyone could help?
it is good for easy changes, now try asking it to generate a faster than Clickhouse NoSQL analytics database
I don't even know what that is
@@WebDevCody thats the exact point cody. You should know enough about a subject to make the thing work. Classic Mathews Principle (Matthew 25:29)
I'm both happy and sad you made me take another look at cursor :o
I can't wait for someone to build this sort of llm integration into neovim with some plugin.
exactly, i'm in exactly the same boat. It increases my productivity A LOT, it's very good at doing the tedious simple things so I can focus on the actually hard things. I'm using Cody, it's not as autonomous as this, but it's pretty good for what I need. I'll definitely need to give Cursor a try though. How much time would I have spent making that simple functionality? An hour? Yeah not a lot, but that's an hour saved on my end. Did it build something very complex? No. Did it need to? No.
Thanks for showing cursor's functionality, its better than just code completion 👍
Now I started feels like Dev jobs are in trouble specially for newbies
Agreed that CoPilot is crap now, we must assume they're doing a major update. Codeium is my go to now but will give Cursor a try
i think you can have different .cursorrules files in different directories? Often i wont want e.g. typescript specific prompts in the backend part thats in python for example so yea just some writing hints, also if you make prompt files you can just at the file which might be simple if that different directory thing doesnt work
update that they did workspaces that you can at - its like a prompt with some files that you can then at so gives you more control similar to what i was talking about...
Mate, have you heard a tale about a boy and the wolfs?
frontend finnicking is the most tedious part of fullstack dev. Like i just wanna write business logic xD
Hopefully cursor has an affiliate program cause I am about to go try it.
2:58 i would rather write code than explain everything to an AI in plain english. and also this task seems easy. but as we increase the complexity i afraid that this is going to be a tedious job
yea, congrats cody discovered ai, will go through through the typical phases of amusement, amazed, then later disappointment, then later regret and anger then it will go through the whole cycle of grief, then finally say AI sucks and no one has to worry about it.
you know me so well.. ai sucka video, in how about 12 days sound good?
@@WebDevCody take your time
i think only reason AI is not competent writing complex code is because it uses snapshots of existing codes on the internet to train itself rather than learning how to do human level abstract thinking. it doesn't go further than codes. we go further as humans. we think of what is the use case, what is the scalability, what is the reuse potential, what is the audience of this product, how can we make it faster without compromising anything etc. AI only knows how to respond to the prompts.
@@jurassicthunder after 50+ years, people still don't understand what ML/AI is. It has always been the same principle, just that we now have means to process terabytes of data very quickly and efficiently. and you are right, unless something fundamentally changed, no matter how many years has passed AI will always be the same.
Hahaha
Same, I cancelled my Anthrope subscription and paid Cursor instead
i really like the theme you are using can you share please ?
Noooo don't go down the dark clickbait road! You don't need it!
I needed to reach my monthly view quota
@@WebDevCody Honestly, fair enough 👍 Keep up the good work
demn cursor server is now officially overload
hug of death - hopefully they used serverless
Yes, AI tools are really phenomenal
Solving non-technical-consumer problems =/= solving engineering problems.
Coders/devs often solve consumer problems; engineers often solve technically hard business problems with a slow ramp up time. The former is really easy to automate away, such as with wordpress for small businesses. The latter is really hard to automate away.
If your entire skillset is gluing together library APIs, I’ll be a gate keeper and say, you’re likely not much of an engineer. You’re often not *solving problems* you’re implementing an *already made solution* for the problem. You often didnt solve anything. Someone else did and youre just gluing it together. Im not trying to devalue the importance of coders, but just making it clear whether or not one should be worried about automation and AI. You’ll have better job security in the latter.
Theres nothing wrong working on an app for people to schedule meetings. But just know, your businesses point of differentiation isnt a technical one. Its likely operations, logistics, distribution, etc.. If your business is writing a better PDF parser for other businesses to use to extract data they need, your point of differentiation is a technical one. Which one do you think gives the person writing the code better job security?
You're making it seem like the point of differentiation is categorically different, but it really is just a matter of difficulty from a technical point of view. Basically, PDF parser - hard / Scheduling app - easy. Although, honestly, a full-featured scheduling app is no joke. I worked on one and the collision detection algorithm is not something that you can ask an AI to develop for you, at least not without a mountain of prompts and patience.
@@parlor3115i probably didnt use the best examples, but i think were in the same idea. The point is to position yourself to solve engineering problems, not apply premade libraries for your day job.
As for point of differentiation, i think it us categorically different. If your company views your job as a programmer as a cost sink (i.e. maintaining a web app for say a real-estate company) verses views your job as a value generator (i.e. maintaining a proprietary, custom app which is your value proposition to the customer), your job security will be very different.
You want to be what makes the company money. Not what is just overhead on a company balance sheet.
@@hamm8934Fair enough, but trust me, current AI is capable of solving very specific business problems. It's just the overhead of providing the adequate prompts to get it to do so does not justify the end result. Basically it's a situation of "I'll just do it myself, instead of hand holding through the entire process". However, when AI evolves to be more mature / the right AI startup creates one purely for the sake of engineering software (already the case just needs more work and polishing), it will actively ask businesses and technical question and make proposals and act on them autonomously.
Personally, I feel the only way we can be safe as SE is if AI development reaches a dead end and they can't develop more sophisticated AI for a few more decades or at least until a point where the hype and investment dies out or slows down considerably.
i get what you are saying but if ai can do all the work where is the joy of writing code anymore it's boring . Unless you are a sadistic and you like obedient ai
"flex", great... i have to rewrite intern code again
Going to blast StartUps NOW!!!🥳
Meh, i’d try it, but it will take a lot of time to switch from webstorm to cursror(vscode)
This seems very convenient but I dont think this can work on complex desktop to tablet to mobile designs. Cause designers are a pain in the a$$😂😂
I love tailwind and how easy it to style with it, but OMGG the time i need to spend to make the pages responsive is crazzy. If you say it is that easy then my life would get soo much better. I feel like it's with every ai tool, it helps u to be more productive, would it replace me completely... well.... that's an heated topic 😅
tune in tomorrow folks for cody's next video: "Why AI Sucks!"
That’s not a bad idea
You said it in the beginning. I created a live stream, and it's catching lots of views... You are a content creator, aspiring for content engagement, so here we are.
we got cursed.
Does it work with react-native?
RESPONSIVE IS EVERYONE'S PROBLEM
It just feels wrong to do everything with AI, you're learning nothing.
I use codeium, which is very minimal assistance and also free. Sure, you're slower but at least you understand what your codebase does.
at this point, I can identify what problems are worth solving by hand, and what I would consider a "boiler plate problem", meaning something I could find on stack overflow and doesn't require much heavy thinking to implement. AI helps write the boilerplate and I'm ok with that.
I think nothing can make learning easier.....
If the ultimate goal of a dev is to solve coding problems all by your brain, don't use AI. But if the goal is to deliver apps that solve end users' problems, use whatever tools help you reach your goal!
I just feel like conservative devs try to romanticise and pride themselves too much.
@@kanasva if I want to use my brain I’ll solve leetcode all day 😜
IT'S OVER
RICK JAMES B.
Good job babe!!! First!
like always! love you babe!
@@WebDevCody i need some love too :(
@@WebDevCodylove u too
❤❤❤
@@WebDevCody❤❤❤
What keyboard are you using?
klack.app
@@WebDevCody Oh, lol.
inspired me to use cursor more! Thanks
Let's always do alot of good ❤
Nam myoho renge kyo
Job title: read only developer 😅
Future JavaScript developers will be hired to review AI generated code
Waiting on the primeagen roast 😂
getting this shit asap
Nice demo of CursorAI. I could do without the clickbait title, but I get ya. Dang that YT algorithm 😂
The denial in the comments is real lol
Not to be mean, but those are pretty simple instructions any llm can do this.
fully integrated with my code base?
Typical frontend dev overreacting
Yeah probably
yea but dude you just saw a small feature added in 20 seconds. that would take me like 10-15 minutes
and this a typical coper. this wont replace devs. but it will replace most. only few will be hired to write the prompts and know enough to say it works or not.
its the same as continue extantion in vscode lol
Never heard of it, so this gets the promotion, simple as that