Both are important. You do need to work on your own projects and own concrete case, find tricky cases, look up suggestions online, etc. It can be a great learning opportunity. For example, back in the day I used to do a lot of video recoding, ripping DVDs to SVCD, so I had to learn what the terms mean in the programs and why results weren't as I expected. This led me to learn basic concepts and a mental model of coding theory, color spaces and signal processing. Or when I built a website, I had to learn about networking, about programming some Javascript, etc. just to get the website to display. Similar with game modding, learning about graphics, textures, scripts, writing code to decode, modify and encode binary files containing text to translate them from English to my language. This meant I had to learn about character encodings etc. If you have a personal purpose in mind, learning can "just happen". I didn't intend to learn the things, I was just striving to make the thing work.
I have no idea how I'd make someone interested in these things if their mind doesn't pop up such tinkering goals for them.
On the other hand, the above is very inefficient in itself. It is hard to overstate the importance of reading good books, besides the practical learning of the above paragraph.
You need to be able to relate to the topics to some degree, but once you have the hang of it, you can benefit from the years and years of experience of experts and hone your mental models, discover new topics and relations across topics and understand things more holistically.
Learning only from books/courses is not sufficient, you won't be able to apply what you learned if you don't constantly try to map from what the book says to something you are already familiar with a little bit.
But you have to pick good books. Bad books can be more confusing and discouraging. I think American CS books especially from MIT Press are really high-quality. They may be expensive (but there's LibGen).
You may also have problems with delayed gratification. The benefits of an hour spent on reading a textbook will be vague, uncertain and delayed and often hard to attribute to the book. One hour of watching Netflix or Youtube or browsing through 20 genuinely interesting links from HN feels more rewarding. This is a big problem for us all in today's attention economy. You need to be very conscious of this to break out of it.
I have no idea how I'd make someone interested in these things if their mind doesn't pop up such tinkering goals for them.
On the other hand, the above is very inefficient in itself. It is hard to overstate the importance of reading good books, besides the practical learning of the above paragraph.
You need to be able to relate to the topics to some degree, but once you have the hang of it, you can benefit from the years and years of experience of experts and hone your mental models, discover new topics and relations across topics and understand things more holistically.
Learning only from books/courses is not sufficient, you won't be able to apply what you learned if you don't constantly try to map from what the book says to something you are already familiar with a little bit.
But you have to pick good books. Bad books can be more confusing and discouraging. I think American CS books especially from MIT Press are really high-quality. They may be expensive (but there's LibGen).
You may also have problems with delayed gratification. The benefits of an hour spent on reading a textbook will be vague, uncertain and delayed and often hard to attribute to the book. One hour of watching Netflix or Youtube or browsing through 20 genuinely interesting links from HN feels more rewarding. This is a big problem for us all in today's attention economy. You need to be very conscious of this to break out of it.