[kragen]

Yeah, although maybe mentorship can scale reasonably well (or at any rate much faster than we are scaling it at present), I agree with you about collaboration: ten times as many people working on the LHC (or HEP in general) probably wouldn't be very effective. Now that we have Sci-Hub, the General Index, Wikipedia, Stack Exchange, and Google Scholar, we can probably collaborate a little more effectively than before, but not enough to cram orders of magnitude of people into a given subfield.

There might be a path forward in the work on making computational work easily reproducible, by people like Konrad Hinsen, Yihui Xie, Jeremiah Orians, Eelco Dolstra, Ludovic Courtès, Shriram Krishnamurthi, Ricardo Wurmus, and Sam Tobin-Hochstadt, but clearly it hasn't been a panacea so far. Speedrunning results are in many cases reproducible by virtue of nailed-down console hardware and bit-identical game images, but that's harder to achieve even for FEM simulations of turbulent MHD systems, much less actual experimental MHD systems like a Farnsworth fusor.

How do people initially get up to speed on speedrunning? Are there tutorials, the equivalent of a textbook with problem sets, some other onramp? Can we gamify learning quantum mechanics? (I've tried QiskitBlocks but so far haven't been impressed.)