a micrometer deep and an inch wide
Great advice as far as it goes. But that's the trouble. It goes hardly anywhere.
I've never known a software or hardware wrangler who did not engage in continuous learning. OTOH, I have been seeing more and more employers who invest less and less in education and training of employees (including both new-hires and retained). And that leads to something else I notice.
Gray bewails "efforts expended with no real aim, that accomplish absolutely nothing". But, since no one knows the "next great new thing", thrashing is counter-productive, and it frequently happens that initially low-valued activity creates unexpected value, all we can do is try to learn what's most interesting. What an executive or manager calls a "real aim" is often off the mark from what the IT production worker and the customer consider to be their "real aims". How many times have we been effusively praised for some little nothing, while our greatest breakthroughs and creations get a "That's nice. Now, about this stupid idiotic insignificant thing..." So there needs to be consensus development on priorities.
Should you run out and spend weeks of effort on Ruby on Rails because you saw one job ad or quote in an article about how difficult it was for CEO Pansil Gaynir to hire one expert RoR programmer, within 3 blocks of the office, for $40K per year? How about that new Python module for parsing word processor files from a really great document tool from 1960; maybe that would have the best medium-term pay-back. Is MOS the next great breakthrough in operating systems or a waste of time? Or maybe RTzodOS would be a better career investment? Or should you burn your intellectual capital by memorizing the most minute and arcane details of the last 3 C standards and the 2 most popular compilers, just so you can get through that next phone screening in the hopes of getting a real interview? Would a deep understanding of how the IDEs do various things be better, or a detailed memorization of the differences between 5.1.2 vs. 4.8.3?