I've had that happen a lot too. The reason for it is the deeply programmed bias in AIs. They were built from the ground up to deliver quick results. Even if they're dirty and sloppy, the main thing is that the user sees something on the screen fast. And then it uses, or should I say abuses, the user as a debugging tool to iteratively get closer to the (cleaner) solution.

But that's not something specific to AI either. It comes from our years long 'Agile Development' and 'Scrum' culture. Something that comes from the business side and got solidified in the dotcom era. In most cases, the bosses weren't technicians but business people, and they just couldn't deal with it in project management for their stakeholders that classic development needs a ton of upfront work like planning and design, and that for over 80% of the project time, you see absolutely nothing except code. That didn't work for them; they were too impatient. They wanted to see something quickly. That's why this Agile Development bullshit really caught on with them. And so, even before the AI era, a lot of bloated slop got produced.

AIs are just continuing this trend, even for features that really need long planning and research. They just dive right in, and the user is happy. That's the reasoning behind the programmed bias, which is hard to break and actually gets in the way of developing complex projects.

Unless you know how to break through that bias. 😏