Quote:

While most of the code is somewhat ok (ie: it is valid C), it is still full of stuff that will simply not work. For example, the MUI event processing loop is wrong...

Also, it (and not he, this is not a real person here...) failed to use MorphOS specifics like our About window class for example.

Anyway, this clearly shows that:

- AI isn't a magic wand that is capable of producing 'production' code
- You still need to understand/write code
- You still need to know how the API stuff is supposed to work (in this case MUI).

The result is still interesting nonetheless. This AI has been able to produce some kind of MUI construct which is probably really uncommon in its training set (ie: there isn't large code base of MUI code example on the web compared to GTK, KDE, or whatever GUI toolkit)

Which brings the question of the use of such data to train their AI model. Has copyright been respected for ex? :)

Having an AI capable of doing that means that they *scraped* deeeeep in the web for having a decent (ie: very very very large) training dataset.

Anyway. This code needs to be human corrected in all cases.

Next step: request a makefile for building it.

Good luck for building something with Claude. And you'll need a fair amout of :)