A working diary, not a publication. Short entries while a build is in motion — date, what happened, what I learned. Filter by project to follow one thread.
Staring at a result and thinking "that does not look safe" turned out to be a better instinct than the maths sitting behind it.
Twice I had a confident diagnosis. Twice, actually measuring it took ten minutes and proved me wrong.
The results were mathematically correct for a long time before they were any use to anybody.