> And what I have found is that most – say, 90%, of what I learn at one job is completely useless for the next one. Then the problem isn't with what's being learned. The problem is that the author lacks a framework for generalizing and internalizing the lessons learned. Letting 90% of lessons learned on the job go to waste later in life is going to lead to a very difficult career path. Later... > Compounding is a pre…
I came up with the 90% number not because I have any real data on this, but as a way to provoke myself to challenge some assumptions. Earlier on in my career, it was easier to tell myself the story that everything I learned, all the hours I spent doing stuff, were all part of a bigger narrative. That it would all build upon itself in a one-directional way. Even if I forgot specific facts, I'd still always be learning and growing in one way or another. I'd be developing critical thinking, learning how to learn, moving to higher levels of abstraction, pruning out useless knowledge and strengthening core insights. That kind of thing.
That was actually a pretty useful mentality, and still is in a lot of ways. It gave me the confidence to do a bunch of career changes (like I think I mentioned in that section of the blog) because I did believe that I was drawing lessons from each successive career area to the next. And there was a lot of truth in that.
However.
At the end of the day, as a developer, I'm certainly not doing Leetcode or Kaggle all day. The majority of what I've spent my time learning has been very specific knowledge: learning the components and business logic in specific systems, getting comfortable with internal tools and processes, getting to become very familiar with some subset of all the company code, working with clients and dependent teams and internal customers. I do believe that being a tech giant can make the situation worse in that regard since the specific problems tend to be so narrow. But my hunch is that it's still the norm for developers.
I'll put things in a different perspective. According to a lot of studies, doctors get worse at their jobs on average over time (link: https://hbr.org/2017/05/do-doctors-get-worse-as-they-get-old...). Most of what they learn on a day-to-day basis is very specific, time-bound knowledge (specific patient info, office info) rather than general-purpose lessons about medicine. By going for the 90% number I was trying to push myself to consider that that sort of effect is the norm (in programming and elsewhere) in contrast to my earlier notions about "constant generalization."
I think this number can be changed but like you mention, it requires an amount of deliberate effort like writing that just hasn't been baked into my usual routines.