61–70 of 168 posts
This article is not about vibe coding per se, it's about not having strong boundaries between you as the developer, and your client. You should not be allowing the client to dictate how you work, much less them having the permissions to merge in code. This was true before AI too, where clients might say, do X this way, and you should simply say no, because they are paying for your expertise*. It's like hiring a plumb…
Yeah, its bad out there. At my company, we have a team of security professionals that focus on keeping our systems (and others') secure. AI for them has gone from "using it for scripting together nmap" to "we really need the platform your team is working on to do X, Y, and Z, so we vibed up this PR". On the engineering side, I don't have the political power to tell them no, because we don't really have senior leaders…
The people who think FizzBuzz is a leetcode programmer question are now vibecoding the same trash as always, except now they think they are smart x10 developers for forcing you to review and clean up their trash.
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 lethar…
I had to rewrite several vibe coded projects from scratch due to this effect. It's useful as a prototyping tool but not a complete productionizing tool.
Earlier quoted context omitted.
I have yet to find the niche where it is "good at the beginning". So far I've mostly tried asking to build C tools that use advanced linux API. Me: hey make this, detailed-spec.txt AI: okidoki (barfs 9k lines in 15 minutes) all done and tested! Me looks at the code, that has feature-sounding names, but all features are stubs, all tests are stubs, and it does not compile. Me: it does not compile. AI: Yes, but the code…
I'm not sure what you're using. I've used Claude in agent mode to port a very complex and spaghetti coded C application to nicely structured C++. The original code was so intertwined that I didn't want to figure out so I had shelved the project until AI came along. It wasn't super bad at converting the code but even it struggled with some of the logic. Luckily, I had it design a test suite to compare the outputs of t…
Been writing software for like 20 years now and I love it. I am also a fan of AI-assisted coding, but I only just started using Cursor. Gosh I do not like it at all for a simple reason: since I didn't write the code, in order to understand it I have to read it. But gaining understanding that way takes longer than writing it myself does. When you write the code, you understand it. When you read the code produced by an…
> since I didn't write the code, in order to understand it I have to read it. But gaining understanding that way takes longer than writing it myself does.
Been writing software for like 20 years now and I love it. I am also a fan of AI-assisted coding, but I only just started using Cursor. Gosh I do not like it at all for a simple reason: since I didn't write the code, in order to understand it I have to read it. But gaining understanding that way takes longer than writing it myself does. When you write the code, you understand it. When you read the code produced by an…
Yep. I had a few vibe coded projects that were fairly far along and then things broke. The code was so convoluted and it took me so long to understand that I just opted to rewrite everything from scratch without AI. Sure, it took longer but I understood all of it.
All the consulting practice arguments aside, this is fundamentally a gatekeeping argument about clients staying in their lane. I'm sure doctors feel the same way about patients with weirdly specific questions about HFpEF diagnoses. Doctors have always hated "Doctor Google", and now they have to contend with "Doctor GPT". It's up to you how much sympathy to have for them.
I treat my doctor as a subject matter expert/collaborator, which means that if I come to him with (for example) "what if it's lupus?" and he says "it's probably not lupus", I usually let the matter drop.
Earlier quoted context omitted.
I have yet to find the niche where it is "good at the beginning". So far I've mostly tried asking to build C tools that use advanced linux API. Me: hey make this, detailed-spec.txt AI: okidoki (barfs 9k lines in 15 minutes) all done and tested! Me looks at the code, that has feature-sounding names, but all features are stubs, all tests are stubs, and it does not compile. Me: it does not compile. AI: Yes, but the code…
I'm not sure what you're using. I've used Claude in agent mode to port a very complex and spaghetti coded C application to nicely structured C++. The original code was so intertwined that I didn't want to figure out so I had shelved the project until AI came along. It wasn't super bad at converting the code but even it struggled with some of the logic. Luckily, I had it design a test suite to compare the outputs of t…
I have seen AI agents fall into the exact loop that GP discussed and needed manual intervention to fall out of.
Also blindly having the AI migrate code from "spaghetti C" to "structured C++" sounds more like a recipe for "spaghetti C" to "fettuccine C++".
Sometimes its hidden data structures and algorithms you want to formalize when doing a large scale refactor and I have found that AIs are definitely able to identify that but it's definitely not their default behaviour and they fall out of that behaviour pretty quickly if not constantly reminded to do so.
Earlier quoted context omitted.
Yep. I had a few vibe coded projects that were fairly far along and then things broke. The code was so convoluted and it took me so long to understand that I just opted to rewrite everything from scratch without AI. Sure, it took longer but I understood all of it.
No joke: Maybe that is the current value of vibe coding. It helps you get started with a crappy version. In your experience, which one do think would take longer? (1) Vibe code until it breaks, then you rewrite everything from scratch or (2) Write everything from scratch. I don't vide code (yet?), but I do use LLMs to get ideas about how to solve problems and look at same code, especially when I don't what library fu…
> Okay, so this non-technical person is sending me codes now. I started wondering if this person was actually a developer here. Maybe just a typo, or maybe a dialect thing, but does anyone actually use "codes" as a plural?
It’s not just that, there are other things in the article pointing to the person being a non-native English speaker. Which is fine, I’m one too. It’s somehow ironic though that his written output could’ve been improved by running it through an AI tool.
I mean, it could've been homogonized by running it through an AI tool. I don't think there's a guarantee that it would've been an improvement. Yes, it probably could've helped refine away phrases that give away a non-native English speaker, but it also would've sanded down and ground away other aspects of the personality of the author. Is that an improvement? I'm not so sure.