Like I don't use Lovable myself because I can make my own software already, but I definitely changed my opinion after I saw it int he wild and compared to all the internal software I saw in the past 20 years.
For me Vibe coding is mostly making apps via AI by people who have no domain knowledge in what they are doing or no technical expertise.
If you technical or do have tons of domain expertise in the are(like those warehouses worker that did things for themselves) - I wouldn't call it vibe coding personally.
It just regular product development in my eyes and i would try to separate two cases.
I think the main connotation of vibe coding is cluelessly producing code that's notably poor quality. But at this point I think the term is becoming a mixture of gatekeeping and denial. The reality is that the latest models are reliably producing competent to excellent quality code and architecture, even without particularly insightful prompting. And I see no reason to think they won't continue to improve for the foreseeable future.
Recently I had a bug in a program I'm writing with Claude. I manually figured out the issue and gave it the steps to solve it. But the thing that surprised me is that the newest model immediately pointed out what the bug was (which I did not say) and it was not at all obvious, even from the steps I told it to take to fix it. Since then I've been being much 'lazier' with the models and having just as good of results.
No idea what Lovable adds to the picture. I always thought the site was a bit sketchy because of that. Prompt engineering has always been overblown, and they're just working as a middleman between people and Claude. But hey they now have $400M and I do not, so props to them.
Vibecoding is using AI to produce code. This separation of producing poor quality or great quality is an delusion that many vibecoders are going through. There is the old way to produce code when you do it all and there is vibecoding that is asking AI to generate code. There is no middle ground.
