Comment by Jensson

I think the rat-race idea got baked deep into SV largely because of the setup of the post-1970 economy, and specifically California's rules: every startup had to be "fast" and "hit the ground running" not just because of literal business pressures, but because implicitly, everyone intends to exit and move on to the next thing immediately, so a slow-moving part-time effort would be a no-go. That derived just from mundane realities about the market, legal and financial environment, and protections afforded towards workers and businesses - everyone knows that we moved in the direction of corporations squeezing the workforce because they "couldn't afford" to build them up. Therefore career mindsets filtered to accommodate only those with the gold rush viewpoint, because the only way to win that game was to strike it rich and retire.

Your question: There is no finite set. There is no end to it. This is a process. You do it over and over. You don‘t invest into a set and be done with it (and likely over-invested in some). One step at a time. And once in a while reevaluate the territory.

I offer nothing definitive, but one of the greatest benefits of StackOverflow has been to quantifiably classify most of the common problems in computer science, programming, analytics (and to a lesser extent the other StackExchange communities ...)

If you go there, search for a tag, sort by popularity, etc. - most of those techniques and approaches and patterns are, if not timeless, then mostly timeless and relevant.

Master the basic building blocks of programming. Frameworks comes and go but the basics are the same. Master as in "I do X correct the first time even without thinking" not "I know how X works". The "do without thinking" is when your subconscious is doing it for you so your conscious doesn't have to. You want to put as much as your work there as possible, it is like putting work on a GPU instead of your CPU, so every reusable programming trick and pattern should be there. This also helps you reason about code since now your sub consciousness can build code on its own and therefore help you reason in much larger chunks.

A person who hasn't mastered loops and conditionals will fail FizzBuzz. That is bad, you recognise that I hope. But it doesn't end there, you can master so many more powerful parts of programming and get as fast as you were at FizzBuzz att many other much more complicated tasks as well. Ultimately programming takes time due to all the parts you didn't master, but when you have mastered enough parts then you can fit just about any problem into parts you have mastered and code them up extremely quickly.

So the speedup you see is roughly the speedup you saw from at first struggling a lot with loops and conditionals early on when learning programming, to today when you do them effortlessly and instantly. Now the other things to master aren't as simple patterns as loops or conditionals, but the end result is the same improvement in speedup and effort.

For example, competitive programming is about mastering a ton of such chunks. You don't memorize solutions to get fast, you master thousands of such chunks and compose them to solve problems in 5 minutes that would take regular programmers 5 hours if they can solve it at all. I went that route and it is similar to how people master chess, a grandmaster makes better moves after a few seconds of thinking than a typical chess player who has played for years thinking for many minutes.

It took about a year and then I could solve hard problems at similar pace as the best in the world with basically no errors, compare that with spending 4 years in college, I think the practice is worth it, I have never needed to practice that again as the chunks I mastered doesn't go away, it is like learning how to ride a bike. You forget the details, but your subconscious wont forget those chunks. Of course I can't solve hard competitive programming problems in 5 minutes today, but I can solve most leetcode "hard" problems I haven't seen before in under 15 minutes and basically all of them in under 30 even though I haven't practiced these things for many years.

Now, this wont solve all your programming woes, you still need to understand requirements, talk to customers, learn API's etc. But at least you are no longer bottle necked by composing programs. Instead you can throw together MVP's quickly, show prototypes to make discussions more fruitful etc. Also most jobs doesn't requires this level of mastery, only do it if you aim to join top teams using your technical skills if you really get good at this then many teams that wouldn't care about you before now gets very interested in you since so few are this fast, or if you want to build your own products on your own, or if you find mastering things fun.

Edit: Sorry for edits, but here is last bit. When I worked at a high performing team at Google I averaged over 500 lines of code a day going through code review and running in production when I wasn't constrained on non-programming bits like understanding the problem etc. To do that you write 10 changes, each about 50 lines and trivial to review since everything is super clean, otherwise you can't get through code reviews fast enough. Code reviews, unit tests, fixing edge cases etc, none of that are excuses for being slow. Now it is still fine to be slow, I don't blame people, I'm just saying it is possible to be fast if you deliberately practice a lot to be fast.

I feel the same way, but I also think that in all of the time I've spent thinking about it, I could have learned a bit of Vim or better typing, or something.

Interesting POV. Can you elaborate on an alternative to the way of industry programming? Or more simply, list some of the ways in which industry brand programming is flawed?