The definition of insanity is doing the same thing over and over again and expecting a different result. – Albert Einstein (probably apocryphal)

This quote came to mind the other day when I was trying to get some code past QA. Before code goes live (e.g. before a user sees my changes to a website), it needs to pass a series of automated tests. My tests failed. I clicked retry. It failed again. After poring over details of my updates for hours, I used the time-honored practice of just trying it again, and it worked.

I work on a team with a pretty large number of software engineers. This means that, when I click the retry button, things might have changed behind the scenes. In other words, doing the same thing over and over again was, in this case, not the definition of insanity. In the world of science, the state of the system is generally known completely, so Einstein’s mentality makes sense. Unfortunately, in the real world (e.g. SE world), there is state that is not easily available to us (This distinction is analagous to the difference between standard Markov Decision Processes (MDPs), where state is totally known, and Partially Observable Markov Decision Processes (POMDPs), where there is only a belief in the current state). In the real world, when we face scenarios with this massive unknown state, the optimal action, the sane action, is to just do it again. So rest easy, and click retry.