the first few paragraphs lay out the precise extent to which this matters. there's a snippet of code (`x^x*/:x:2_!100`) in a language (k7) using a feature ("rank-sensitive" set difference) not present in previous langs (before 2019); the prev langs' workaround (razing `,/` the table into a list) is bothersome (to chrispsn) so his article examines some alternatives.

chrispsn is sharing the map he drew of the rabbit hole he explored. it involves a lot of haggling with "depth", which is when you nest a list in another list: good for organizing your data, but bad for producing long vectors over which your programs could vectorize. so if you want this to matter more broadly, you can consider it a meditation on some of the mechanics of that, i.e. how to flatten some deep constructions in algorithms, and how to generalize shallow operations to apply at depth