You missed the point. The original post is about not reading code.
You actually missed the point in two ways, because my response had little or nothing to do with speed of producing code. I'm not sure why you felt the need to express that irrelevant objection.
Whenever I start reviewing code produced by Claude I find hundreds of ways to improve it.
I don't know if code quality really matters to most people or to the bottom line, but a good software engineer writes better code than Claude. It is a testament to library maintainers that Claude is able to code at all, in my opinion. One reason is that Claude uses API's in whacky ways. For instance by reading the SDL2 documentation I was able to find many ways that Claude writes SDL2 using archaic patterns from the SDL days.
I think there are a lot of hidden ways AI booster types benefit from basic software engineering practices that they actively promote damaging ideas about. Maybe it will only be 10 years from now that we learn that having good engineers is actually important.
You're writing Python and Javascript right? Those languages are extremely easy to write in (which conversely means the legibility is likely to be poor). People maintaining legacy systems in systems level languages aren't going to be able to produce as much code as people writing Python and Javascript.
I think a surprisingly good AI will come into existence and one of the lessons will be that we greatly overvalue intelligence over basic distribution. Giving a single kid millions and billions is symbolic of the actual problem (the distribution one).
Also, there's basically 0% chance this kid is one of the top 1000 most knowledgeable people in the world on this technology.
Can somebody explain to me how giving a 28 year old kid 250 million (or was it 1 billion) to run your AI lab is a good idea? Or is it actually a dumb idea? I think it is a dumb idea, but maybe somebody can make it make sense.
Everyone in this conversation talks about different activities. One version of vibe coding happens with Netflix open and without ever opening a text editor, and another happens with thoroughly reviewing every change.
AI generated code is pretty far from passing professional human generated code, unless you're talking about snippets. Who should be writing mission critical code in the next 10 years, the people currently doing it (with AI assistance), or e.g. some random team at Google using AI primarily? The answer is obvious.
I've tried getting claude to set up testing frameworks, but what ends up happening is it either creates canned tests, or it forgets about tests, or it outright lies about making tests. It's definitely helpful, but feels very far from a robust thing to rely on. If you're reviewing everything the AI does then it will probably work though.
A lot of the drama here is due to the ambiguity of what the word 'think' is supposed to mean. One camp associates 'thinking' to consciousness, another does not. I personally believe it is possible to create an animal-like or human-like intelligence, without consciousness existing in the system. I personally would still describe whatever processing that system is doing as 'thinking'. Others believe in "substrate independence"; they think any such system must be consciousness.
(Sneaking a bit of belief in here, to me "substrate independence" is a more extreme position than the idea that a system could be made which is intelligent but not conscious, hence I find it implausible.)
That was exactly my experience. It was either extremely friendly and fun, or extreme anger. It was terrifying. I would usually stay up 24 hours in a row preparing for meetings because I would be too anxious not knowing what I was going to get.
There's a scifi short story about a janitor who knows how to do basic arithmetic and becomes the most important person in the world when some disaster happens. Of course after things get set up again due to his expertise, he becomes low status again.
My 'boss' was an absolute pyschopath who was a direct cause in my, let's just say, poor mental health over a period of years. I'd rather say not say anything to them ever again, but horrible interactions with them replay in my mind constantly. Imagine spending thousands of hours doing your best to make someone happy just to be treated like the worst piece of garbage who ever existed. That's what my 'boss' was like.
"Continuing to use this feature is a moral failure on your part and you should be ashamed, your parents should be ashamed, and dishonor will follow you wherever you go."
It's an interesting concept, but inapplicable here because I don't trust the media reporting on LLMs and I personally believe expert programmers are never going to be replaced. My concept of the value of LLMs is that they are good for generating throwaway test code to assess the use of a library or to prototype a feature.
I don't think it's expert, I just don't think being expert is necessary to get some value out of it if you aren't an expert. The trap is letting the charade go on longer than it should though. I personally only see the main value in using it to create test projects or to get the gist of what a library can do. I do think that's pretty valuable, and I also think real expertise is more valuable.
Or you can do like some of the others suggest and eliminate pure vibecoding. Just use it as a back and forth where you understand along the way and make well-reasoned changes. That looks a lot more like real engineering, so it's not surprising the other commenters report better results.
It's terrible at the niches I actually have expertise in, which are in mathematics. I'd guess an expert is going to find the flaws in anything it's doing in their field. That being said, if you're just trying to e.g. see what some GUI library can do then it's pretty useful to get something going. I personally would prefer not using it in anything that's not very much a throwaway test project though, but that is my luxury as a jobless bum.
For me vibecoding has a similar feeling to a big bag of Doritos. It's really fun at first to slap down 10k lines of code in an afternoon knowing this is just an indulgence. I think AI is actually really useful for getting a quick view of some library or feature. Also, you can learn a lot if you approach it the right way. However, every time I do any amount of vibecoding eventually it just transitions into pure lethargy mode; (apparently lethargia is not a word, by the way). Once you eat half a bag of Doritos, are you really not going to eat the second half... do you really want to eat the second half? I don't feel like I'm benefitting as a human just being a QA tester for the AI, constantly shouting that X thing didn't work and Y thing needs to be changed slightly. I think pure vibecode AI use has a difficult to understand efficiency curve, where it's obviously very efficient in the beginning, but over time hard things start to compound such that if you didn't actually form a good understanding of the project, you won't be able to make progress after a while. At that point you ate the whole bag of Doritos, you feel like shit, and you can't get off the couch.