Comments by "Christian Baune" (@programaths) on "What Self-Taught Developers NEVER Learn" video.

Partially right. You need to know what is available to you so you are able to gauge if it's worth to investigate it more. You would never consider hooking an FPGA to your server for heavy processing (e.g. cryptographic hashes), because you don't know it's a possibility. As an SE, your job is not to reply to some request, but help your client/employer do the right thing. That may even means no programming at all. (Changing processes) So, no need to learn in-depth, but surely required to learn as much as possible. Where it's nuanced is that you can do it in an intelligent way. If you know Angular, Dagger is not be be learn as it's basically the same principles at play. If you know a C family language, C, C# or Java shouldn't be that complicated etc. But if you totally ignore one aspect, then you'll never learn it when required, unless it's part of some specifications you got. It's why in the academic track I got, we did a grand tour. I even programmed robots you find in pharmaceutical companies. Did also some German assembly, designed an build pcbs for interfacing, analyses a library processes etc. Stuff I would have not done on my own and that gives me an edge. In one of my job, my interview was cut with "Ok, you are hired". I asked how they came to it and they replied that I was the only one who pulled some UML and asked the right questions. They asked me to make a small program which was an ETL and before digging into the code, I did a small design. The created only one class to have the "feel" of it, then called them to discuss the solution before churning through code. In the end, though, there is ample room for self taught people. They can do basic algorithms, which are 90% of the jobs (mine is a bad example as I mainly works with trees transformations). Then you need only one guy who know how to tie everything together and being efficient while keeping it readable and maintainable. That's also that guy who can tell what is worth learning. Still, there is a huge plus-value to have some IT culture.
11
In one company I worked, they didn't know how to migrate databases, because they were not indexing and indexing them would take days (at least). At some point I passed by, discussed and the issue popped. So, I replied that it is easily solved with dichotomic search. It took me 5 minutes to solve an issue they were wrapping their head around. When asked how I came with that so fast I replied that it is literally something you learn in the first quarter of the first year at university. So my solution came straight from the book. (of course, I had to recall it) Same with my current boos to whom I can swiftly answer, just because I learned the material. As a stupid example, if we need to do some intensive process, I'll see if it is possible to hand off the problem to an FPGA. Self taught programmer do not even know that it exists and can't design hardware at all. Before the academic track, I was a self taught programmer and I was already good. Learning made me even better in the sense I have more horizons.
3
@collinsa8909 Lol, I never looked at diploma and certificates when I interviewed. Just ran though the protocol. Still, I do not recall any self taught doing good...neither those who did put forward their credentials. It's mostly those who thought they didn't do that good who passed. Self taught have issues you'll rarely find int those who go though "professional school" (So, not university, but alike where you learn to "do".). Mind that university usually means not actionable. They get the theory, but the practice is not there yet.
2
Example related to it, I needed to apply a regular expression to a series of streams of characters (not a String) and stop as soon the pattern fails (no point in reading 100MB worth of character if after the two first, the pattern fails). Yep, Pattern and Matcher can go to the bin. Yet, there is a much faster solution: FSM. FSM is not a library, it's a concept and you've to tailor it to your solution (literally designing your FSM -- graphviz-- then coding it)
2