I can stop my doomscrolling now. You are completely correct, and you win the internet for today. Congratulations, and enjoy your prize of “one internet.”
(Thank you for phrasing it this way. I was having a hard time explaining what “vibe coding” meant to someone who asked me, and your explanation is exactly the right way to understand it, IMO.)
All of the demos I’ve seen so far of “vibe coding” have all been videogames. In games there’s more wiggle room for being “almost correct”: the game’s still playable if your sprite is bigger than you want. How well does vibe coding work on a large, complex task with stricter requirements?
(By “vibe coding”, I’m sharing Simon’s definition of not even looking at the code. Using it as an augment doesn’t count!)
It occurred to me that in my view, “software development” is for products I’m most aware of: database backed line-of-business applications with a ton of hooks for APIs, reporting, mobile views etc.
But for a lot of people, “software development” probably means: create a mobile app and hope it goes viral on an app store. Very different lifetime and support requirements, as well as different revenue streams.
Whether GenAI is appropriate for either is outside my expertise.
I think it could be useful for one-off scripts, e.g. transformations of data where you want a rough correctness, just as if you build your pipeline with sed, awk, grep and not writing proper tests, just this “ah it looks good enough”.
And no, I can’t imagine this for proper well-engineered software that should be bug free.
I’m as much annoyed by over-hype as I am by knee-jerk reactions to hype. But I am irritated that I clicked on this article based on the question in the headline only to find that it makes zero attempt to answer the question!
The title is clickbait in that the question it actually tries to answer is “should I use vibe coding to produce maintainable code for production settings today?” It says nothing about the future or about software development in general—the author even hedges by only talking about code that is both (1) maintainable and (2) production. So maybe they think vibe coding will be OK for non-maintainable production code? Who knows! I would not have clicked the article if it had a more accurate title.
He said (1) vibe coding for production is a “terrible idea”, for reasons that will not change in the future; (2) vibe coding for prototypes is fine, “go wild”. Obviously, unmaintainable production code is not OK, so no need to state that explicitly.
tl;dr: vibe coding is the code equivalent of writing a sentence by just tapping the next suggested autocomplete word on your phone’s keyboard
I can stop my doomscrolling now. You are completely correct, and you win the internet for today. Congratulations, and enjoy your prize of “one internet.”
(Thank you for phrasing it this way. I was having a hard time explaining what “vibe coding” meant to someone who asked me, and your explanation is exactly the right way to understand it, IMO.)
All of the demos I’ve seen so far of “vibe coding” have all been videogames. In games there’s more wiggle room for being “almost correct”: the game’s still playable if your sprite is bigger than you want. How well does vibe coding work on a large, complex task with stricter requirements?
(By “vibe coding”, I’m sharing Simon’s definition of not even looking at the code. Using it as an augment doesn’t count!)
It occurred to me that in my view, “software development” is for products I’m most aware of: database backed line-of-business applications with a ton of hooks for APIs, reporting, mobile views etc.
But for a lot of people, “software development” probably means: create a mobile app and hope it goes viral on an app store. Very different lifetime and support requirements, as well as different revenue streams.
Whether GenAI is appropriate for either is outside my expertise.
I think it could be useful for one-off scripts, e.g. transformations of data where you want a rough correctness, just as if you build your pipeline with sed, awk, grep and not writing proper tests, just this “ah it looks good enough”.
And no, I can’t imagine this for proper well-engineered software that should be bug free.
I’m as much annoyed by over-hype as I am by knee-jerk reactions to hype. But I am irritated that I clicked on this article based on the question in the headline only to find that it makes zero attempt to answer the question!
The title is clickbait in that the question it actually tries to answer is “should I use vibe coding to produce maintainable code for production settings today?” It says nothing about the future or about software development in general—the author even hedges by only talking about code that is both (1) maintainable and (2) production. So maybe they think vibe coding will be OK for non-maintainable production code? Who knows! I would not have clicked the article if it had a more accurate title.
He said (1) vibe coding for production is a “terrible idea”, for reasons that will not change in the future; (2) vibe coding for prototypes is fine, “go wild”. Obviously, unmaintainable production code is not OK, so no need to state that explicitly.
I didn’t (and won’t) read the article but you should be aware of Betteridge’s Law. It saves time!