It's a good point, it can be true, if done well. You have to have an abstract/productive view on tinkering. Not just bike shedding.. otherwise yeah it's just constant costs.

ps: one talk that I find a pretty strong example is "the unix chainsaw" by Gary Bernhardt (of wat js fame). He shows how to using tools `against` themselves, as data, to help your work. It's not rocket science, but it's 1) something I rarely do truly 2) easy to fall back as tools as silos instead of .. `objects` collaborating.