I am getting bold now that I like Observable more and more
Mathematica is a universal Turing machine, so it can execute any valid String input as a program. See the above link.
For procedural language programmers that is a whoopy dang do! But in AI and Machine Learning self-writing programming environments this is a Divine Gift so to say.
By Self-writing I mean a program that could modify itself or write a new program in runtime.
So we can design and code notebooks that design and code notebooks.
Q: Why this is so important?
A: In such self-writing programming environments we can input Strings which can be Textually Transformed to a new String to achieve a grammatical or heuristic alteration, and if this input String represents a valid program then we can have notebooks which Textually Transform notebooks into new ones with new instances of desired functionality.
Where did we used these?
Self-writing Neural Nets when we do not know the original connectivity of the nodes in the network
Non-standard Neural Networks which have no know no training algorithm
Symbolic Fit or Symbolic solvers where approximate better and better symbolic solutions to solve a problem
An example of a practical such application for Observable notebook:
Data Driven Self-modifying or Self-writing Notebook which calls Classifiers and predict functions to modify questionnaires or forms which are actually Observable notebooks!
We did some prototypes of these in Wolfram Cloud Forms:
But sadly Wolfram Cloud Forms are too primitive for most applications.
Now we like to investigate and build some prototypes for Observable …