Writing

Notes on training models and running data — mostly about why a number is the number it is, and where it stops being trustworthy.