miniblog.

There's a class of performance bugs that you can feel. Today I inadvertently implemented "Schlemiel the Painter's algorithm", and it's obvious: performance was great at first, and then it gradually deteriorates.

'Fishing' for p<0.05 is an unfortunate term:it implies ill-intent rather than inadvertently finding spurious results https://www.stat.columbia.edu/~gelman/research/unpublished/p_hacking.pdf

Turns out I've inadvertently reimplemented GNU Stow in my dotfiles repo. Nice to know I can switch to a widely used tool.