I agree with this in principle and it also mirrors my own experience during a PhD program: time is cheap, tinkering is learning, etc. This is especially true in

I agree with this in principle and it also mirrors my own experience during a PhD program: time is cheap, tinkering is learning, etc. This is especially true in academia research, which is, in most disciplines, helped by wiring things up in ingenious ways. Practicing those skills comes very useful. Robert Wood probably would not become such a great experimental physicist if he did not tinker with hardware as a youngster. The problem is that this approach does not scale: you can build everything from scratch, but beyond certain complexity you need existing subcomponents. Nuclear physicists do not need to learn gear tooth shapes and build their own servos. And in my experience people are bad at making the decision when to stop building and start buying. Maybe a better approach is something like 80/20 split: spend 20% of time building own designs and instrumentation and 80% on target research. And those who are really good at building from scratch and love it can spend more time on it.