I wonder what kind of edge cases you deal with when blas is your bottleneck in R. Stan code aside, I’ve seen few problems that are neither instant (i.e. sub hours) nor impossible (I.e years of compute).

Since you mentioned Stan, feels relevant to mention https://stan-playground.flatironinstitute.org/, which lets you run Stan in WASM and analyze the results using WebR

Isn’t the linear algebra conventional wisdom that matrix ops are ALWAYS the bottleneck?

I’m sure this is true in scientific computing.

In R maybe a bunch of resampling would be expected to dominate?