While interesting, it does sound a bit like the concept of UML: "the model is the code". That was supposed to revolutionize how software was written, as people would only design their application and UML would automagically create the code.
The result was messy code and far from optimized or optimizable. And that will have disastrous effects on scalability...in a similar way as people creating a solution in Excel for their department and then wonder why it doesn't work fast enough when deployed for the whole company. Access in Microsoft Office is also such an example of poor scalability.
The concept behind the essay I do like, but having seen the results of generated code and being subjected to it, color me very unimpressed. Now this was some 10 years ago and as always, with the passing of time almost everything improves, so maybe now this concept might work.