In reply to @gabreal "These models are getting": I have the suspicion, that
despite ecosystem diversion, the average Nim code base pretty stable is.

Means, consistent.
That brings me to the concept of type stability, that is a type inference
related topic, that happens to take place in Julia circles.

AI is the same: It works best, if you give it consistent pattern, and you as the
user can use them to actually improve the performance of the product
significantly.

I am currently writing an Ada project, and they are training AI specifically for
it, in fine-tuning sessions, hobbyists do actually.

And that works really, really good.