> You can learn by doing, or by theory and then doing That's true, but I felt like he was getting at something else, though. I've met (and worked with) progra

> You can learn by doing, or by theory and then doing That's true, but I felt like he was getting at something else, though. I've met (and worked with) programmers who tried to stumble through everything by looking for a solution to the problem immediately in front of them, solving that immediate problem as basically as possible, and then working on to the next problem. They don't really understand how things like, say TCP or SSL or even HTTP work. They don't understand how their IDE relates to their compiler. They don't understand machine learning well enough to recognize when or how to normalize data. They usually expect somebody else to come along and "help" them with all of this stuff in a "just-in-time" way. I remember attending a Java user group one time when Hadoop was relatively new and the speaker was presenting it. He was talking about some of the challenges they faced getting it up and running and a woman interrupted with "but whenever you had a problem, couldn't you just google the error message?" He said, "there usually weren't any error messages and even when there were, google didn't find anything". She seemed shocked and actually argued with him a bit in disbelief. I was dumbstruck that here was a woman who was actually working as a professional programmer deep enough that she was attending Java user groups who couldn't conceive that there even could be problems that required enough fundamental understanding that "googling an error message" wasn't going to help.