First of all, you can rarely write down in English, what you want in 15 minutes… It’s even common to have longer specification, than its implementation. Just look at tests. Especially, if you want to do something which was never done before, the disparity can be staggering.

Claude Code for example is also not that quick at all. It produces some code quickly, but even scaffolding three hello world level example projects together definitely takes more than an hour. And that’s with zero novelty. The first version of code is done quickly, but the continuous loop of self corrections after that takes a long time. Even with Serena, Context7, and other MCPs.

And, of course, without real code review. That’s easily hours even with just few thousands lines of code, if it uses something which you don’t know. But I know that almost everybody gave up understanding “their” “own” code, during vibe coding. Even before AIs, it was a well known fact, that real code reviewing is hard, and people rarely did it.

AI can make you quicker in certain situations, but these “15 minutes” claims are totally baseless. This is one reason why many people are against AIs, vibe coding, etc. These stupid claims which cannot hold even the smallest scrutiny.