How Simple Are Optimisations in Machine Learning?

One of the arguments that you make o the paper is this idea that like, well, mas optimises are just like, more simple. So i'm wondering, in what languages do you think it's the case that optimises are simple? And in particular, i'm interested in the sort of quadunquit languages of, like the eltenorm of your neurol network and the cordincourt language for a multi layer preceptra oritiana. E, so that makes sense. But how hard is it to actually, like, write an ation procedure in sort of nirl, that wer rates how difficult that would be to do? I don't know. My guess