I came back to an old project this spring and, to my surprise, it still built on the first try. Here are the few boring habits that made that possible — pinned toolchains, a vendored lockfile, and a README that assumes nothing.
The cheapest performance win is rarely in the code. More often it is the time between making a change and seeing the result. A short log of small experiments to shorten that gap.
This is a personal site. I write here irregularly, mostly to remember what I figured out so I do not have to figure it out again. No newsletter, no tracking, no comments.