>what does it feel like to have 90% code/abstraction with 10% of visual sketch input

I was playing around with automating some aspects of architectural model making a few years back. I was using grasshopper (because I am an architect and can't really code). I found it interesting at the time, that the program encourages you to play with the ratio of visual-to-programmatic metric, but that ratio also reinforces how you interact with the project.

It was pretty simple stuff.. automatically unfolding meshes for paper cutting, adding tabs and rebates to sheet material for more precise joins, tagging elements for correct assembly, sorting elements by size and shape and using some basic genetic algorithms to optimise orientation and position on sheets to minimise time and material wastage.

But it really got me that when you push the programmatic elements to be more precise and repeatable, you really close down the problem space. When you increase the visual feedback, you tend to do it at the expense of precision and repeatability - your process becomes more tightly coupled to the specific project at hand.

Anyway, I am def going to have to make some time to try some of these out.