Working on the Witness – Semantic Compression

I find this to be an interesting and very pragmatic approach to programming, and I've never really seen anyone talk about this style before. It seems like it would require quite a bit of experience though, since the idea is basically to constantly make judgements about what level of compression, or structure, will give the best cost/benefit ratio now and for the future. So I wonder how well a less experienced programmer, such as myself, could make those decisions, with less knowledge about the consequences of the different options.