End user doesn’t care about code quality if the thing works.
nostr:note1yemxmytglpk3q0cn5rslq4azvvqgnftjjt5mlhyz8vd26ypfswnqj7v0qc
Discussion
Right, businesses do. As long as this is the state of vibe coding I am not worried one bit about job security lmao
I didn’t say anything about job security. I actually think it’s a great tool for great devs and makes them that much more invaluable.
Yes I agree. The issue I see is that modern juniors might not even catch that there is a big problem with the code snippet generated in the picture above that might cost the business a lot of money. But if they would instead tried to read up how to build a message queue that would have immediately avoided that.
This is the very single issue I have with vibe coding. You might take away the opportunity to learn from yourself
Yeah, I understand. I never planned on learning so it works for me but I can see how it may harm those who do want to be full time devs and do it correctly.
You're comparing a bad coder with AI against a good coder without AI. What you should be worried about if at all is a good coder with AI.
Good or bad, everyone will use AI. Once it exists there is no going back. If anyone denies it and tries to get by without it, they will be uncompetitive with the rest.
This is not about whether seniors should use AI (they should). The only thing I am worried about is juniors who use AI instead of learning.
A friend of mine told me that most of his classmates “cheated” through their DSA class using AI. I see no future in software for those people.
If you don't use AI as a dev, your job isn't secure.
How long before the dev isn't needed at all?
If you use AI as a dev, your code isn't secure.
Out of interest, what don't you like about the code?
The time complexity of dequeue in this snippet is O(n), while if it were implemented correctly it would be O(1). A queue is usually implemented using a singly linked list and not an array.
I feel like this examples illustrates the dangers of vibe coding best. This code will not throw an error, it will run just fine and there is nothing that will make someone who does not know better look twice, but it will cost you.
besides the obvious ai smell of having comments that describe what's obvious by just reading the code, and no checking of anything whatsoever, it's not implementing a queue in any way, that's just an array with extra steps, it loses all the benefits of a real queue.
It probably stole that code from someone who doesn't understand data types either.
A queue has a pointer to a start and end; it has a limited size, and depending on the implementation and what you're trying to achieve it should either circularly wrap if it achieves the size limit, or it should throw an error.
Enqueuing and dequeueing should be constant time (O(1)), that means that adding and removing elements should always take around the same time.
Arrays wrappers like what chatgpt there created get progressively slower over time, as they work in linear time (O(n)).
if the thing does one thing, never needs to change or add features and never needs a security update - then sure. but once you have users and requests and edge cases and security fixes, the quality becomes really important really fast. and it's not something you can easily catch up on.
it's called "tech **debt**" for a reason
assuming this isn't going to majorly improve in the next few years is a fallacy
remember where it started and where we are now
nobody is disputing that, **one day** it will be good.
But we're 10 years into the "self driving cars will replace driving in 1 or 2 years" claim
or for a more direct comparison that the primeagen likes to give, we're already 24 months into 6 months away from ai stealing your jobs
well.. have you been in a Waymo..?
I haven't, but the claim was that human driving would be obsolete, in 1 or 2 years, 10 years ago; this has not happened.
yes, advances in self driving cars have been made, and it's amazing to see how much progress we have made.
and in all honesty there's probably no company I would turst less than google with my **actual life**, haha
oh yea, no disagreement there at all. one day my previous note will be entirely obsoleted.
but there's kind of an escape velocity thing here - like the human longevity space. will today's vibe-coded app survive long enough in the short term to live forever?
Also, it improves literally every week, so a year from now this conversation will be over.
True on the surface, but if it's really bad code quality, it's guaranteed to break, or at least break the moment you make the tiniest change.
Although I'm with you on the larger vibe coded apps taking over. If apps can be tailored to individual comfort, who cares who coded it.
I wouldn't necessarily want to have thousands of users on a fully vibe coded app!
But it is absolutely amazing for super personal highly specific productivity apps, whic is more than enough to kill off a bunch of bull shit saas vendors 😉
actually this mentality is what caused the software industry to degrade so much; objectively speaking software is of worse quality today than it's ever been