RISC-V is a family of instruction sets (which have various chips implementing them). Think "X86-64".
It looks like the baochip-1x is using the VexRiscv CPU. The HDL is available here under MIT: https://github.com/SpinalHDL/VexRiscv
It is much like a game might use a physics engine, or a new language might use the LLVM backend. To overly simplify, a CAD kernel will keep a list of operations (make a cube of this size here, drill a hole of this depth here, round these edges but not those). And combine that into a final volume. These responsibilities only get more and more complex as a part gets more complex - so using a pre-built engine allows CAD software to focus on tools and workflows to translate human instructions into lower-layer kernel geometry: the UI/UX.
It also crosses into compatibility, if you use the same Kernel as another CAD it is much simpler to export/import from them. Otherwise you would have to reimplement their kernel (or enough of it), or be stuck exporting triangulated versions of the final volume - sort of like converting an image from vector to raster.
We even use aluminium on "dumb" transformers for power transmission. Dry-type transformers tend to be physically larger because they use air and resin (rather than a tank of oil) to insulate, and so the major downside to aluminium conductors (needing a larger cross-section to carry the same current for the same loss) is no longer a limiting factor performance-wise. In most substations, and extra 20-30% physical size of the transformer is a fine trade-off for cheaper construction.
I think the issue might stem from the fact that (as I read it) the essay is talking about "for the people who are moderate (in the middle of the left/right axis), some are distributed higher on your graph, while some are lower". Which says nothing about "for the people who are distributed higher on the graph, how many are in the middle of the left/right axis". Your graph makes explicit an answer to the second question which the essay avoids. (There is a bit of an implication in the last two paragraphs, but PG is explicit it's only about people he knows).
I missed it on the first read-through but there is a link to the code used to run the simulations in the first appendix.
Homegrown python code (i.e. not a library), very nicely laid out. And would form a good basis for more experiments for anyone interested. I think I'll have a play around later and try and train my intuition.