소프트웨어 개발: 심층 지식의 도전

### 소프트웨어 개발에서 심층 지식의 도전
* 💻 "세상에서 보고 싶은 변화가 되세요"라는 문구는 **다른 사람들이 따를 수 있도록 높은 코딩 표준**을 유지하는 것이 중요하다는 점을 강조하며 깊은 울림을 줍니다.
* 🛠 소프트웨어 개발의 진입 장벽이 낮기 때문에 **좋은 코드보다 나쁜 코드가 더 많이 사용**되어 양질의 라이브러리와 추상화를 만드는 것이 어렵다는 점이 부각됩니다.
* 📉 책임감과 전문성이 가장 중요한 토목 공학 분야와 달리 **'존경할 만한 엔지니어'를 구성하는 요소에 대한 명확성**이 부족합니다.
* 🤖 많은 개발자가 코드를 이해하거나 검토하지 못해 향후 잠재적인 문제를 야기할 수 있는 **AI 생성 코드의 증가**는 위험을 초래할 수 있습니다.
* 📚 프로그래밍 문화에서 중요한 문제는 깊은 지식과 얕은 지식을 구분하는 것인데, 많은 개발자가 **기본 개념에 대한 기초적인 이해가 부족**합니다.
* 🔍 개발자들은 종종 자바스크립트와 같은 핵심 프로그래밍 기초를 배우지 않고 **React와 같은 프레임워크**를 선호하기 때문에 **가상 DOM과 같은 기본 메커니즘에 대한 이해**가 부족할 수 있습니다.
* 🌊 2005년경 프로그래밍 붐 이후 신규 개발자가 유입되면서 많은 개발자가 **API의 심층적인 원리를 이해하지 못한 채 API 학습에만 집중**하면서 얕은 지식이 범람하고 있습니다.
* 🔄 궁극적으로 소프트웨어 산업은 계속 발전하고 있지만, 코드의 전반적인 품질을 개선하기 위해 **개발자들 사이에 깊은 지식과 이해의 문화**를 조성하는 것이 과제로 남아 있습니다.

현재 소프트웨어 엔지니어들은 종종 깊은 지식이 부족하여 얕은 코딩 관행이 널리 퍼져 있습니다.

so our friend Jonathan Blow is back, and he's going to blow us all away again with his hot take: current software engineers have no deep knowledge. I think there's a pretty serious problem that started happening in programming culture. I mean, it's been in programming for as long as I've been in, and this is beyond games even; this is everywhere. People don't have a clear picture of what knowledge is deep versus what knowledge is shallow, right? So, if you go to school and you're a good student and you learn the difference between n log n and n squared, right? That's deep knowledge that will apply in many cases, and it's something fundamental about computers, right? Um, on the other hand, if you go on Hacker News and read an article about somebody learning a specific JavaScript API and tutorial-ing it, right? That's very shallow knowledge because it's just, you're learning about arbitrary decisions that somebody made about how to interface with a thing that you're not even trying to understand, right? And

that but that knowledge. So, I have a couple of points here. One, everybody hates algorithms and data structures, but I think one thing you can't deny is that those topics are very fundamental. And that's why you see them applied in so many areas, like sliding window is not just something used to solve leak code problems, it's used in many places, data processing, rate limiting, a lot of places. But the bigger point, what he said, nowadays I see it manifested a lot when it comes to React developers, to be honest, because the first thing somebody does when they want to learn front-end development is they see React, that's the number one framework. But it's almost like they skip a step, they don't even learn JavaScript fundamentals. I don't know how you can possibly call yourself a React Dev and not know anything about the virtual DOM, for example, and why it helps like improve the performance of larger front-end apps. Sometimes I feel like most people don't even know why they're using a framework anymore, especially beginners, they just do it.

because everyone else is doing it. If you have a really small one-page app and you need a little bit of interactivity, why don't you just write vanilla JavaScript? Oh wait, they can't. They don't know how. They've never directly manipulated the DOM before, and it's kind of embarrassing, and I mean that in the nicest way possible. If you're a beginner, I'm not trying to pick on you. I'm just telling you, you should learn to walk before you can run, 'cause it's going to catch up with you in the long run. Same thing with open-source development. Before, it used to be you need to do LeetCode to get a job. Nowadays, it's you need to do open-source software development to get a job. And so, you'll see beginners asking, and people have asked me this question, How do I get into open source? and they barely know how to code. They haven't done personal projects. They learned one language, never learned Git. God help the projects that they end up contributing to, because it's going to be a mess. Now, I do see the

other side obviously understanding the virtual Dom is not going to necessarily allow you to actually build an application, but like high-level knowledge of JSX, how React works, that stuff actually will allow you to build an app. So that's like the first thing people are going to go for, which is understandable. So even though I'm ranting, usually there is a reason for everything. It became very popular since, let's say, since 2005, once this second boom of programming happened, and everybody realized, Oh, you can make a lot of money in programming, or whatever. All these people showed up, and they just want the job where they could do JavaScript, and so they just learn the API, right? And so there's been a deluge of shallow knowledge. I wonder if Jonathan Blow considers front-end developers lower tier, and for the record, I don't. I did front-end development. Front-end development can go surprisingly deep, but I thought that was an interesting take. Seems like he's calling quite a few people out. Um, it's a little bit more

complicated than that because part of the way that you learn deep knowledge is by looking at shallow things and reading the pattern underneath. Like if there's an API to do something, if it's a good API, then by looking at the API choices, you can understand the problem and the mechanics, even if you can't see the code. It's better if you could see the code inside, but if you can't, you can still see some things about the mechanics. And you know, one example of that is a graphics API. It's like, okay, they want me to put the vertex data into a buffer and lock the buffer and then not touch it. Why is that? Oh, it's going to the GPU and all this stuff, right? You can figure this out. On the other hand, if it's a bad... I'm not familiar with GPU processing, but there's a similar pattern when it comes to big data processing. After each step, the data itself gets locked. One of the reasons is it's happening in parallel. So I think what he says

definitely applies. Patterns are so much more important than like the actual smaller details. API, which most APIs are, it'll be full of stuff that's stupid or even cargo culting where they're copying a good API. So, my favorite example of this was when Microsoft did Direct 2D, which was taking Direct 3D, which was successful, and then trying to add font rendering and all that to that. They made an ARGB color, right? So, red, green, blue, and alpha. A lockable resource. So, to create one of these, you had to lock it and fill the color data and unlock it. And I'm like, you have no idea what you're doing. And I actually told Microsoft this, and they did not like my opinion on that. But they were just copying, Oh, Direct 3D has you lock things and fill them, and so let's do that here. And it was completely absurd. So, I think he's absolutely right that this kind of stuff happens at big tech companies where

you wouldn't necessarily expect bad abstractions to exist 100% even at Google. I've talked probably enough about that. For the record, good abstractions exist as well, and that's what most people will hear about. You don't really hear about the bad abstractions, but they're everywhere, unfortunately. So how to fix it? There just isn't much of an idea right now of what an actual respectable engineer is. You know, like if you think back to civil engineering, like people who build bridges or whatever, you think of people who are responsible and know things well and make sure they're right because it's important to be right. We don't really have that in software at all. It's completely dead as far as... and the funny thing is AI is probably not going to help that because, okay, got a bunch of AI-generated code, thousands of lines, well, it works. I'm not going to read it. If you ever need to have an actual developer go in there, I mean, unless the AI can just literally do

everything unless it's AGI. At some point, you might need a developer to go in there and fix the mess that the AI made. It's going to be pretty damn hard if it's nonsensical code, as I can tell. And part of the reason it's dead is because nobody can tell what's right because it's just a bunch of people arguing in a loud room, and like no real information gets through. And if anybody has an idea how to solve that, I would love to hear that idea. I don't know. The only way that I know to solve it is to just do better in an obvious way. But even that maybe doesn't win, right? Because all these just do better JavaScript Frameworks are still popular even though they're all, yeah, the same as each other, right? You have npms to just to pad things because no one wants to write code anymore. In the same way, the market can remain irrational longer than you could remain solvent. Um, I think software industry culture could remain stupid longer than

you're alive, so I guess what he's saying is, be the change you want to see in the world. Just make your code a high standard, and then others can kind of follow that. So, just don't create garbage packages. Given that it's like a free world and anybody can become a developer, it's pretty easy. The barrier to entry is relatively low. I don't think there is a solution to what he's saying. There's probably more bad code out there than good code, and it's not easy to create good code. To be fair, creating really, really good libraries that can be extended and used, and providing good abstractions, is hard. That's the problem. It's not that developers are dumb and they're just always shipping really bad code; it's hard to do what he's saying. Like, that's just possible. So, I'm doing what I can do, but I'm very conscious that I have a limited ability to affect this giant thing.

Programmers in 2024 have no Deep Knowledge

NeetCodeIO

소프트웨어 엔지니어의 지식에 대한 조나단 블로우의 비판

조나단 블로우는 현재의 소프트웨어 엔지니어들이 깊은 지식이 부족하여 프로그래밍의 문화적 문제가 발생한다고 주장합니다.

그는 알고리즘의 복잡성을 이해하는 것과 같은 깊은 지식과 특정 API 튜토리얼을 학습하는 것과 같은 얕은 지식을 구분합니다.

Blow는 대기업들이 근본적인 원리를 이해하지 못한 채 성공적인 패턴을 모방하여 비효율적이고 불합리한 결정을 내린다고 비판합니다.

코딩 및 오픈 소스 개발 학습의 어려움

많은 초보자가 바닐라 자바스크립트로 DOM을 조작하는 등 기본적인 코딩 기술에 어려움을 겪고 있으며, 이는 발전의 걸림돌이 될 수 있습니다.

업무 요구 사항이 LeetCode에서 오픈 소스 개발로 바뀌면서 초보자는 탄탄한 코딩 기반 없이 기여를 시작하게 되었습니다.

오픈소스 프로젝트의 혼란스러운 특성과 유사한 JavaScript 프레임워크의 압도적인 인기는 소프트웨어 업계의 지속적인 도전 과제에 기여하고 있습니다.

프런트엔드 개발의 인기와 인식

프론트엔드 개발은 프로그래밍 붐과 수익성 높은 일자리에 대한 열망으로 인해 2005년부터 인기를 얻었습니다.

많은 개발자가 API 학습에 집중하면서 해당 분야에 대한 얕은 지식이 급증했습니다.

연사는 프론트엔드 개발자가 하위 계층이라는 인식에 의문을 제기하며 프론트엔드 개발은 의외로 깊이가 있을 수 있다고 주장합니다.

프로그래밍에서 기초 이해의 중요성

API 선택과 같은 얕은 것을 분석하여 근본적인 메커니즘을 이해함으로써 깊은 지식을 얻을 수 있습니다.

알고리즘과 데이터 구조는 기본적이고 광범위하게 적용 가능하며 코딩 문제를 넘어 다양한 분야에 영향을 미칩니다.

많은 지망생 프론트엔드 개발자가 자바스크립트 기초와 가상 DOM 학습을 건너뛰기 때문에 자신이 사용하는 프레임워크에 대한 이해가 부족합니다.

연사는 다른 사람들에게 영감을 주기 위해 코드에 대한 높은 기준을 설정하는 것이 중요하다고 강조하지만, 좋은 코드를 지속적으로 생산하기는 어렵다는 점을 인정합니다.

개발자의 진입 장벽이 낮기 때문에 좋은 코드보다 나쁜 코드가 더 널리 퍼져 있습니다.

소프트웨어 분야에서 존경할 만한 엔지니어에 대한 명확한 정의가 없는 것은 토목 공학 같은 다른 분야와 대조적이며, AI가 생성한 코드는 추상성이 떨어지는 문제를 해결하지 못할 수도 있습니다.