Just a little more compute bro, we just need a little more context window. trust me bro Claude 5 will be AGI
post
Have you noticed how every LinkedIn bro is talking about their vibe coding workflows, but no one is showing what they’ve made with vibe coding?
They said it can write code. They never said it can fix code or debug issues.
That's like saying you can drive, bit can't navigate or park.
I have met people that dont know how to navigate with Google maps and its painful to be a passenger with them.
Driving with a nav system is a part of the driver license test here.
I think the largest failure is bad architecture, or that is the inability to understand architecture at all. After your vibe coded project gets larger than a prototype and you "zoom out" to try and grasp what is going on, is when you see how bad it is.
Given the potential damage from it, it's more like saying it can drive while being confident the brake pedal is always the left-most one, because it'd seen it on reddit.
I had to give this a really good think because I can't remember the last time I've seen a car with a manual transmission so I was fully like
What do you mean? The brake is the left most pedal?
Did you vibegraph those lines?
I remember there being a CORS problem in a team project.
Perhaps I should ask the webdev at the time whether they had made that thing with AI or they really just made the whole thing themselves and somehow overlooked CORS.
Cors was usually not part of any tutorials. To new people it was more of an afterthought, just set policy to get you page to work.
CORS has always been a problem. I keep having to remind the backend team to send proper headers. They say it works in curl so why not in the browser. It’s not very intuitive.
When I was learning, CORS was a pain in my ass. It’s not taught well, and often glossed over.
So I guess AI and hence, vibe coders are having the same problems as normal programmers.
It could be either way, I haven’t tried “vibe coding” but I imagine you’re right if you don’t tell it to explicitly deal with CORS, AI probably doesn’t
And to think all those times I cursed CORS under my breath. Doing God's work.
I would prefer to title it "Gravity's Rainbow Of Vibe Coding" as this is hardly a cycle, more like a denial committed with force that will eventually give out and cause regression into a shattering of identity.
Slightly off topic, buddy of mine decided to switch to Debian and A.i his way through everything, waiting for the day his system fails to boot because of some obscure command it tells him to run.
Idk. I think using ai to learn Linux as you switch to it is fair ground. In the end they’re free from Microsoft. It’s a win. Just make sure they have data backups.
As the great Linus Torvald said:
It's why I strongly want this to be that "just a tool".
The problem I’ve seen is the lack of knowledge retention when AI feeds you stuff, buddy wouldn’t even bother to read nor memorize what it’s telling him and just copy-paste commands thinking it’ll fix whatever obscure issue he is encountering.
I’ve been using Debian for the last ~3ish years now relying on documentation from others so I’ve seen how fragile it can get.
In terms of usage of AI, I'm thinking "doing something a million people already know how to do" is probably on more secure footing than trying to go out and pioneer something new. When you're in the realm of copying and maybe remixing things for which there are lots of examples and lots of documentation (presumably in the training data), I'd bet large language models stay within a normal framework.

all 42 comments