I'm sure some do. But I can't imagine something like Kiro(not an endorsement) would market itself with vibe coding, since its whole point is methodical spec driven development.
Not all coding is interesting, I would say the majority of it is a chore.
One shotting small utilities that I have no interest in writing (for me, front-end JS, chrome extensions, web scraping selectors) is wonderful. I just vibe coded a "crop a pdf" web UI and it worked with no intervention. It doesn't have to be meaningful, just useful.
I sit on a few indie hacker rooms with folks that slurped up the vibe coding koolaid immediately and they've been posting about their 100x output for over a year now - but none of them have shipped anything remotely successful - so I think at least anecdotally I somewhat agree.
I also just overall haven't seen a huge influx of new and useful apps and websites as you'd expect based on how tech leaders are talking about the virtues of these new tools.
I don't think AI has really found it's niche yet, and I think it's going to be much subtler than most people think, it's going to be the tools that integrate features like translations and summarization and speech to text in ways that are seamless that end up sticking - all of this other noise is just marketing hype.
I think what a lot of people fail to grasp is that the hard/expensive part about running an app based business is rarely writing the app.
Most web developers can clone twitter in a day, even before LLMs. I've seen it assigned as homework for bootcamp devs (not knocking bootcamp devs, just pointing out that it isn't a tall ask even for a junior. Most apps aren't hard to code to an MVP.
The hard part about an app based business is the business part. Marketing, billing, and getting actual revenue and customers is much harder than writing code.
Everyone non-technical I know who vibe coded either ran into a bug they couldn't fix or abandoned their project.
I would love to see Lovable.dev's financials, I think they're running the classic low retention, high marketing dollar consumer playbook and are burning cash fast. Base44 seems to do the same thing after getting acquired by Wix, I'm seeing more and more of their ads.
As someone technical, vibe coding makes me feel disconnected from my product and feel like I don't know what's going on. I eventually just need to dive back into the code and find many things I need to change myself.
There sadly isn't much in the ways of masterclasses and apprenticeships these days that target established professionals, sadly. In any job sector. The userbase of trying to get started in XYZ will always be a larger, more vulnerable one to target. Even if you're honestly trying to educate the masses it only makes sense to focus on beginners first. Experts will eventually find their way, and the intermediate is this nebulous field that few even know how to target.
I have been so discouraged by the way that web3 crypto hucksters pounced on vibe-coding.
I truly enjoy programming with AI, it is my favorite hobby by a mile. I'm also happy to say that, for now, I get to capture some of the productivity increase for myself while my employer catches up to how effective it is.
It is sad that there is no space to discuss these new techniques that isn't full of opportunists and clout chasers.
The trick is that you have to be a good coder to get the most out of "vibe" coding. It works great for me, but I deploy all of the knowledge I've acquired over the decades as a professional developer. You need to know how to architect systems, what data structures and algorithms to ask for, how to design a product, many facets of graphic and user interface design, how to parcel out work, how to parallelize tasks. Even which ideas are worth pursuing is an intuition you build up over years. "Vibe" coding really is magic and I'm highly scaled, but I don't see how it could possibly work for all but the most senior developers. In some sense, it's like writing LISP macros on steroids.
I 100% agree that this is how the experience is right now.
I don't necessarily think it will stay this way, though. The tools are so new, we shouldn't be so sure that the future versions won't allow non-coders to code successfully.
What you're describing is not vibe coding. I realize some people call anything with an LLM generating code "vibe coding" but it's not and I'm going to die on this hill.
1 it's not the meaning Karpathy used originally, which was for creating software through prompts WITHOUT looking at the source code
2 it's not the meaning people outside programming circles mean, either, which is identical to Karpathy's original definition
the title of the article is talking about 2 and is completely correct.
YOU, on the other hand, are talking about an upgrade to IntelliSense. Use a different term. You're describing regular programming, with a new IDE tool.
If it's not going to replace programmers and allow regular people to create software without looking at the code, it's not vibe coding. Full stop.
I think this is spot on and aligns with my experience.
I've seen less experienced devs crash and burn trying to commit massive amounts of vibe coded slop without consideration to: how much of it was necessary? how does it conform with our code base / style? how much does it take advantage of existing code and patterns? and so on.
I think if you are using one of these tools and experienced, it should be difficult to tell you are using one at all. The code I produce looks like stuff I would write, and I understand all it wrote, I don't want to outsource to producing tons of code that I don't even understand.
This has been my experience. I only have 12 years under my belt, but I've had a lot of good results using Claude Code/Cursor agent since this May. You have to know what questions to ask, how to mold the requirements with the agent, what to tell it to do. I treat it like an employee I'm pairing with. My productivity is at a new level.
It really depends what you're coding. I'm a former PM who can't code, and I'm currently using GPT-5 in Cursor to write an internal application for my business of acquiring and operating e-commerce brands that sell on Amazon. I'm really vibe coding (just clicking accept without ever reading any of the code changes), and it's working great! I've got a whole dashboard that's retrieving inventory and sales data from multiple Amazon stores, projecting future sales and reorder point, etc. The code might be total dogshit, but it's incredibly useful for me (and also really enjoyable).
That said I might be sort of a weird case, since I am accustomed to designing and documenting product requirements, the fact that it's just for me means architecture doesn't really matter, and I'm competent enough to help it resolve some design problems like the fact that it was obviously hitting the wrong Amazon API that was pulling a report of every single sale as a line item rather than just a report with total sales numbers, etc.
I was able to teach my interns more about architectural designs instead of coding. Teaching them more about DDD instead of going through what’s broken with this function. We might be close to a point where you can teach product people about these basic concepts, packages and saas tools, and have them vibe code a whole app.
“You need to know…” do you need to know though? I agree I need to know the things to work in a mature codebase or make something maintainable in the long run but to bang out a get rich quick project? Probably not.
Completely agreed. I’ve got about 20 years of professional software development experience. Tools like Claude Code let me build incredibly quickly when I use a stack I know intimately.
But when I try it with a stack I don’t know, I quickly find myself needing to learn the new stack to get anything done.
Extremely experienced engineers will be able to move quickly with these tools, but specialization will still exist, and I’m not clear on how juniors are going to ramp up on all of this.
It’s funny you should mention LISP macros. Five years ago, I’d have told you that the most efficient way to build something is to get an extremely experienced dev with mastery over a flexible language like LISP, and just let them go nuts. I constantly have to hold myself back from metaprogramming tricks because my coworkers won’t be able to maintain it. But on my personal projects, I really get to flex my muscles.
It occurred to me a few months ago that that’s exactly the same spot in my toolbox where Claude Code is. It’s a tool that allows one person with mastery of a tech stack to build things incredibly quickly. Look at Cursor where a tiny team has over $100M in annualized revenue.
Big companies often find themselves spreading ownership across too many hands. You know exactly how to build what you need, but you’re not allowed to touch 80% of the code involved because it belongs to a different team. But coordinating all of that is a massive amount of overhead that will derail most projects. This is why so little happens at big companies.
But these tools enable a smaller number of experts to do a job that used to take many. I expect the speed improvements to be super-linear as a result.
Of course, that brings concerns about what happens to the employment situation for devs if that happens. I hope it means that 10,000 new startups can build more ambitious products and absorb all the people who will probably be laid off by the huge corps in the next few years.
But is it really vibe coding if you’re carefully building step by step and checking everything along the way? I feel like the kind of vibe coding people usually mean is more about blindly iterating until things work and patching bugs as they pop up—where you eventually get an app that runs, but it’s so messy that even a senior dev would struggle to audit or fully understand it.
I have no idea how it's marketed - if it's as a replacement for developers then yeah it's BS.
I regularly use Claude code (and Claude) to write code for
- throwaways data processing or visualization
- connecting to an API I have some documentation for
- demoing some concept
I think it's great for data science and product stuff, and on those merits will be a multi-billion dollar industry. I would never let it near a real codebase.
> at first i was very hopeful i can finally 'build' now with my minimal tech skills
This is the problem. If you couldn't have coded it slowly in the old world, you will have problems coding it in AI world.
However if you have a lot of coding experience, you can now compress the time it would have taken you be an enormous amount. My experience is that I can now make extensive changes with very little effort, and very few dead ends. I've been able to take on entire secondary projects where I was just replication existing knowledge with slightly different tools.
Just this week I had a litmus test. I had an existing database that I'm pushing huge amount of data to. I decided to try a different underlying database. This would have taken me a full week of looking at documentation and writing supporting scripts, now I've done it in the spare time I had in two days of my actual work.
And it's not like the AI just did it all unsupervised. It threatened to do down the wrong path a few times, but each time I spotted it and steered it the way I wanted. I also asked it a few questions about curiosities I discovered in the emitted code, and that led to fixes as well.
If I didn't know how to code before, I would still be coding this alternative database.
I have about 6 months of actual coding knowledge via bootcamp. But I’ve seen enough repos that I know what good practices look like.
That’s about enough for me to build out fairly complex products very fast with AI (Claude Code). Claude Code often makes some fundamental mistakes that you wouldn’t be able to catch if you didn’t have any coding experience (like today, it was trying to save large images directly in the database instead of using file storage).
> However if you have a lot of coding experience, you can now compress the time it would have taken you be an enormous amount. My experience is that I can now make extensive changes with very little effort, and very few dead ends. I've been able to take on entire secondary projects where I was just replication existing knowledge with slightly different tools.
This has not been my experience and it is very frustrating. I've been programming almost 20 years, I'm pretty good at it
I don't know where the disconnect is, but no matter how I try (and I am trying, I don't want to get left behind) I cannot get remotely good results from LLM coding tools
They always, always, always take longer to build what I want than just doing it myself
Vibe coding is fine. It can make junior coders work like a team of junior coders, and senior coders work like they have a team. All the problems associated with having a team of juniors you didn't vet, can't trust, and must always review come with it... but that's an economic trade off. The value is real.
But to be fair, this technology is still at an early stage, and we don't know it's limits.
It's scary to imagine a future where the development process in companies is fully handled by AI agents, which are the only ones who can read and maintain the code.
- "Vibe coding" is a state of action and being, a mental design paradigm for how you approach a problem.
- Generative/LLM based coding is the technology used to enable vibe coding.
I'm not going to fully dismiss the latter because I can't predict the future, and if we're being frank: no one "cares" about the underlying code when you launch a product. Only that it works. Art/Design is king.
Vibe Coding will inevitably fail once you try to scale to make bigger projects or reason with harder problems. Engineering is about understanding a problem space and figuring out a solution to it with the tools/techniques you acquire and derive; black boxing that aspect of engineering will only take you so far.
I made a strategic choice to focus on ‘vibe coding’ rather than using platforms like n8n, aiming to pick up something new with each project I build. From familiarity with terminal, github, heroku + postgres, etc.
I'm still vibe coding in terms of syntax and logic, but I do understand my codebase, and have made some excellent automations at work.
I think there is one thing that most people miss, obviously!
=> You can use AI to double-check (not single check), avoid overengineering, fill knowledge gaps, scan for inconsistencies and many more things. The main benefit it has for me is not that it codes me stuff real fast. It is that my learning curve improved, drastically, whenever I have a missunderstanding. I dive down and test it. Sometimes AI is wrong and I have to go to docs, sometimes docs are wrong and I have to test and I have to open a issue.
If you combine AI capabilities with debugging skills and testing you have a lot of power as developer nowadays. Its a lot of fun.
To me it seems like most things tend to develop to: It’s more fair. If you are lazy and try to do a shorcut you’ll be punished with wasted time and a lot of frustration. If you try to do thing conscientiousnessly and take ownership for the code you push, publish and run. You’ll learn faster, improve faster, ship faster and have way less headache.
AI can let you build crappier code more quickly. I find this useful for throwaway code and unit tests. But more importantly, it lets me write better code with a significant but small time penalty. I can try multiple approaches, do experiments, bounce ideas off the AI, et cetera.
I was at a coffee shop a little while ago with some friends and a stranger was sitting near us working on his laptop. Someone from our group decided to strike up a conversation with him and the stranger started telling us about how now they have these AI tools that can have you make an app in a few hours and without knowing how to code. He was not a software person but look at what he is able to do with it. The implication was that he has an idea and was going to make it a reality as quickly as possible using whatever AI tools he was using.
I am not sure what the moral of the story is but it reminds me a bit of that parable about the investor getting his shoes shined and the shoe shine kid giving him investment advice.
Partially I think the idea of an app being so valuable that it makes money without being connected to anything is just not really a thing. Some games (Flappy Bird anyone?) can be like that or some very specific type of lifestyle app, but for the most part you need a real world service connected or it will not be seen as valuable. But perhaps I am wrong.
A bunch of Tesla drivers think they have an autonomous driving car, too. Doesn't make it reality. My brother thinks he's going to make "an AI agent" to discern bias in news when he himself has the information literacy of an ignorant ipad baby.
Was he pushing the money making angle? Because from the description you gave my impression was that he was empowered to make things, and I see nothing wrong with that.
I met a homeless person that is attempting to dig their way out through vibe coding via empty promises from video marketers telling them they can make $10k a month passive income. This shit is really abusive. I didn't know what to say to her, because she was gripping onto it like her one hope.
The greater scam here is that rugged individualism can lead to positive societal outcomes.
Just more free market capitalism, bro! We just need more freedom for the tech bros, the finance bros, bro!
This AI shit is yet another oh look - increase in rugged individual 'productivity' is going to lead to positive societal outcomes, trust me, bro!
It's all bullshit but what else have the western retarded elite got? Or non-western elites for that matter? USA is USSR 2.0 - falling apart from the retarded short-sighted incompetence and selfishness of its sociopathic 'elites'. China will be USSR 3.0 in a generation or two.
It's all so terribly tedious, boring and cruel - the monumental amounts of wasted opportunity and resources because everyone's too busy being 'productive' to ever spend a decade or two to actually think through anything and come up with potential real solutions to real problems. The average 'elite' doesn't even know what the real problems are or how to begin finding out what they might be - they probably think it's climate change or more likely - how to make sure they retain their billions. Oh well, so it goes.
what a roller coaster. i was with you mostly until the end of your rant. climate change is a real problem; one, in fact, that many of the elites you speak of are actively working to exacerbate for short-term gain. that felt kinda out of left field for someone who's essentially espousing leftwing ideology, which more or less entirely embraces climate change as a major concern.
To me, vibe coding is all about "make something that kinda works, quickly, to test and play and create." It shouldn't replace the specifics of a design, a functional specification, or the thought process that creates a final product.
Perhaps that's obvious to others, but it felt worth saying from my perspective. If vibe coding tools are touting themselves as a key to easy wealth, shame on the product and marketing teams creating the messaging.
Vibe code is gamified software engineering, similar to what Robinhood did with stock investing - gamified it with zero trades, instant feedback, and a fancy UI.
I'm glad there's vibe coding available to non-tech founders so they can build the equivalent of high-fidelity prototypes. May not be MVP worthy yet, but at least it gets them started.
I wonder if Twitter is already filled with "hard-working all umerican" paid-influencers claiming Indians are so low-IQ/smelly etc. that they can't even use AI to do the jobs that they got through race/religion/caste/smell/skin-color/language/etc.-based nepotism.
They Duk-a-duk!
(South Park is awesome isn't it ?)
Vibe coding tools are helping non-tech users build prototype faster, and there's nothing wrong about that. The only thing that goes wrong are the "marketing" folks with a poisoned mindset. I've seen a lovable's tiktok clip said something like "She doesn't know she gonna sell her app for 20k" lol.
55 comments
[ 4.2 ms ] story [ 76.2 ms ] threadWas it perhaps author's wishful thinking?
Low effort in -> low effort out
I can actually code so no point in vibing.
To vibe means to never look at the code, where is the fun in that?
One shotting small utilities that I have no interest in writing (for me, front-end JS, chrome extensions, web scraping selectors) is wonderful. I just vibe coded a "crop a pdf" web UI and it worked with no intervention. It doesn't have to be meaningful, just useful.
I also just overall haven't seen a huge influx of new and useful apps and websites as you'd expect based on how tech leaders are talking about the virtues of these new tools.
I don't think AI has really found it's niche yet, and I think it's going to be much subtler than most people think, it's going to be the tools that integrate features like translations and summarization and speech to text in ways that are seamless that end up sticking - all of this other noise is just marketing hype.
Most web developers can clone twitter in a day, even before LLMs. I've seen it assigned as homework for bootcamp devs (not knocking bootcamp devs, just pointing out that it isn't a tall ask even for a junior. Most apps aren't hard to code to an MVP.
The hard part about an app based business is the business part. Marketing, billing, and getting actual revenue and customers is much harder than writing code.
I would love to see Lovable.dev's financials, I think they're running the classic low retention, high marketing dollar consumer playbook and are burning cash fast. Base44 seems to do the same thing after getting acquired by Wix, I'm seeing more and more of their ads.
As someone technical, vibe coding makes me feel disconnected from my product and feel like I don't know what's going on. I eventually just need to dive back into the code and find many things I need to change myself.
I truly enjoy programming with AI, it is my favorite hobby by a mile. I'm also happy to say that, for now, I get to capture some of the productivity increase for myself while my employer catches up to how effective it is.
It is sad that there is no space to discuss these new techniques that isn't full of opportunists and clout chasers.
I don't necessarily think it will stay this way, though. The tools are so new, we shouldn't be so sure that the future versions won't allow non-coders to code successfully.
1 it's not the meaning Karpathy used originally, which was for creating software through prompts WITHOUT looking at the source code
2 it's not the meaning people outside programming circles mean, either, which is identical to Karpathy's original definition
the title of the article is talking about 2 and is completely correct.
YOU, on the other hand, are talking about an upgrade to IntelliSense. Use a different term. You're describing regular programming, with a new IDE tool.
If it's not going to replace programmers and allow regular people to create software without looking at the code, it's not vibe coding. Full stop.
I've seen less experienced devs crash and burn trying to commit massive amounts of vibe coded slop without consideration to: how much of it was necessary? how does it conform with our code base / style? how much does it take advantage of existing code and patterns? and so on.
I think if you are using one of these tools and experienced, it should be difficult to tell you are using one at all. The code I produce looks like stuff I would write, and I understand all it wrote, I don't want to outsource to producing tons of code that I don't even understand.
That said I might be sort of a weird case, since I am accustomed to designing and documenting product requirements, the fact that it's just for me means architecture doesn't really matter, and I'm competent enough to help it resolve some design problems like the fact that it was obviously hitting the wrong Amazon API that was pulling a report of every single sale as a line item rather than just a report with total sales numbers, etc.
But when I try it with a stack I don’t know, I quickly find myself needing to learn the new stack to get anything done.
Extremely experienced engineers will be able to move quickly with these tools, but specialization will still exist, and I’m not clear on how juniors are going to ramp up on all of this.
It’s funny you should mention LISP macros. Five years ago, I’d have told you that the most efficient way to build something is to get an extremely experienced dev with mastery over a flexible language like LISP, and just let them go nuts. I constantly have to hold myself back from metaprogramming tricks because my coworkers won’t be able to maintain it. But on my personal projects, I really get to flex my muscles.
It occurred to me a few months ago that that’s exactly the same spot in my toolbox where Claude Code is. It’s a tool that allows one person with mastery of a tech stack to build things incredibly quickly. Look at Cursor where a tiny team has over $100M in annualized revenue.
Big companies often find themselves spreading ownership across too many hands. You know exactly how to build what you need, but you’re not allowed to touch 80% of the code involved because it belongs to a different team. But coordinating all of that is a massive amount of overhead that will derail most projects. This is why so little happens at big companies.
But these tools enable a smaller number of experts to do a job that used to take many. I expect the speed improvements to be super-linear as a result.
Of course, that brings concerns about what happens to the employment situation for devs if that happens. I hope it means that 10,000 new startups can build more ambitious products and absorb all the people who will probably be laid off by the huge corps in the next few years.
I regularly use Claude code (and Claude) to write code for
- throwaways data processing or visualization
- connecting to an API I have some documentation for
- demoing some concept
I think it's great for data science and product stuff, and on those merits will be a multi-billion dollar industry. I would never let it near a real codebase.
This is the problem. If you couldn't have coded it slowly in the old world, you will have problems coding it in AI world.
However if you have a lot of coding experience, you can now compress the time it would have taken you be an enormous amount. My experience is that I can now make extensive changes with very little effort, and very few dead ends. I've been able to take on entire secondary projects where I was just replication existing knowledge with slightly different tools.
Just this week I had a litmus test. I had an existing database that I'm pushing huge amount of data to. I decided to try a different underlying database. This would have taken me a full week of looking at documentation and writing supporting scripts, now I've done it in the spare time I had in two days of my actual work.
And it's not like the AI just did it all unsupervised. It threatened to do down the wrong path a few times, but each time I spotted it and steered it the way I wanted. I also asked it a few questions about curiosities I discovered in the emitted code, and that led to fixes as well.
If I didn't know how to code before, I would still be coding this alternative database.
That’s about enough for me to build out fairly complex products very fast with AI (Claude Code). Claude Code often makes some fundamental mistakes that you wouldn’t be able to catch if you didn’t have any coding experience (like today, it was trying to save large images directly in the database instead of using file storage).
This has not been my experience and it is very frustrating. I've been programming almost 20 years, I'm pretty good at it
I don't know where the disconnect is, but no matter how I try (and I am trying, I don't want to get left behind) I cannot get remotely good results from LLM coding tools
They always, always, always take longer to build what I want than just doing it myself
But to be fair, this technology is still at an early stage, and we don't know it's limits.
It's scary to imagine a future where the development process in companies is fully handled by AI agents, which are the only ones who can read and maintain the code.
- "Vibe coding" is a state of action and being, a mental design paradigm for how you approach a problem.
- Generative/LLM based coding is the technology used to enable vibe coding.
I'm not going to fully dismiss the latter because I can't predict the future, and if we're being frank: no one "cares" about the underlying code when you launch a product. Only that it works. Art/Design is king.
Vibe Coding will inevitably fail once you try to scale to make bigger projects or reason with harder problems. Engineering is about understanding a problem space and figuring out a solution to it with the tools/techniques you acquire and derive; black boxing that aspect of engineering will only take you so far.
I'm still vibe coding in terms of syntax and logic, but I do understand my codebase, and have made some excellent automations at work.
=> You can use AI to double-check (not single check), avoid overengineering, fill knowledge gaps, scan for inconsistencies and many more things. The main benefit it has for me is not that it codes me stuff real fast. It is that my learning curve improved, drastically, whenever I have a missunderstanding. I dive down and test it. Sometimes AI is wrong and I have to go to docs, sometimes docs are wrong and I have to test and I have to open a issue.
If you combine AI capabilities with debugging skills and testing you have a lot of power as developer nowadays. Its a lot of fun.
To me it seems like most things tend to develop to: It’s more fair. If you are lazy and try to do a shorcut you’ll be punished with wasted time and a lot of frustration. If you try to do thing conscientiousnessly and take ownership for the code you push, publish and run. You’ll learn faster, improve faster, ship faster and have way less headache.
I am not sure what the moral of the story is but it reminds me a bit of that parable about the investor getting his shoes shined and the shoe shine kid giving him investment advice.
Partially I think the idea of an app being so valuable that it makes money without being connected to anything is just not really a thing. Some games (Flappy Bird anyone?) can be like that or some very specific type of lifestyle app, but for the most part you need a real world service connected or it will not be seen as valuable. But perhaps I am wrong.
The greater scam here is that rugged individualism can lead to positive societal outcomes.
Just more free market capitalism, bro! We just need more freedom for the tech bros, the finance bros, bro!
This AI shit is yet another oh look - increase in rugged individual 'productivity' is going to lead to positive societal outcomes, trust me, bro!
It's all bullshit but what else have the western retarded elite got? Or non-western elites for that matter? USA is USSR 2.0 - falling apart from the retarded short-sighted incompetence and selfishness of its sociopathic 'elites'. China will be USSR 3.0 in a generation or two.
It's all so terribly tedious, boring and cruel - the monumental amounts of wasted opportunity and resources because everyone's too busy being 'productive' to ever spend a decade or two to actually think through anything and come up with potential real solutions to real problems. The average 'elite' doesn't even know what the real problems are or how to begin finding out what they might be - they probably think it's climate change or more likely - how to make sure they retain their billions. Oh well, so it goes.
Perhaps that's obvious to others, but it felt worth saying from my perspective. If vibe coding tools are touting themselves as a key to easy wealth, shame on the product and marketing teams creating the messaging.
the actual AI companies marketing hype seems to claim:
1) you can make fun non-serious toy apps
2) any day now massive productivity increases for software companies. this is different than anyone can launch a product with no software skills.
They Duk-a-duk! (South Park is awesome isn't it ?)