It's also useful for exercising and perhaps improving ones ability to postulate based upon limited information. In a situation where you can subsequently check the result.
In the "real world", people end up doing a lot of such postulation. Or, they end up being pushed around by those who do or who have inside information.
P.S. Addressing the OP article itself, my version is/was "don't memorize what you can deduce in 30 seconds". For example, in physics, many of my classmates seemed to spend a lot of time memorizing all the formulae in the three chapters being tested. I would be familiar with them but also recall the few, um, "fundamental" formulae. Or, put another way, those formulae or expressions in which the relevant fields were usually summarized. During a test, it was trivial to start with these and manipulate to what I needed. But many of my classmate either had that variant memorized cold, or they got stuck.
The problem in the computer/tech world is that much of the detail we need to get things done, is partially or entirely arbitrary. What flag(s) did author/programmer X choose for their utility that I'm using? Maybe it was purely a personal choice. Or maybe they were influenced by environment/mentor/community Y. (In turn influenced by Z...)
Where you can track those influences, or simple memonics ("r" for "recurse" -- but wait, is it "r" or "R"...), there may be some rhyme and reason that builds up particularly over time and with increasing familiarity and breadth.
In other cases, though, X seemingly simply liked to use e.g. "s", and there may be no way to "deduce" or "infer" this.