소프트웨어 엔지니어링: 효과적인 디버깅 및 AI 활용 기법

### 소프트웨어 엔지니어링 마스터하기: 디버깅 및 학습을 위한 필수 단계
* 🛠 특히 **코드의 오류를 정확히 찾아내는 데 도움을 주는 AI 도구**가 통합되면서 디버깅에 대한 접근성이 더욱 높아졌습니다.
* 🤖 문제 해결의 첫 번째 단계는 AI에 문의하기 전에 **코드를 직접 수정하여 잠재적인 오류를 파악**하는 것으로, 이를 통해 질문 기술을 향상시킬 수 있습니다.
* 📜 ChatGPT 또는 Bing의 AI와 같은 AI와 상담할 때는 **명확하고 구체적인 지침과 문맥**을 제공하고, AI가 잘못된 정보를 제공할 수 있는 **'환각'에 유의**하세요.
* 📚 AI가 문제를 해결하지 못하는 경우 문서를 읽는 것이 중요하며, **ChatGPT와 같은 도구를 사용하면 문서 발췌 내용을 맥락에 맞게 명확하게 파악**하는 데 도움이 될 수 있습니다.
* 🌐 도움을 요청할 때는 **스택 오버플로 또는 피라냐와 같은 포럼**을 활용하고, 질문의 가시성과 응답률을 높이기 위해 **마크다운을 사용하여 질문의 형식을 잘 갖추세요.**
* 🔄 학습과 디버깅 과정은 반복적인 과정이므로 **질문하고, 오류를 검색하고, 커뮤니티 리소스를 활용**하여 접근 방식을 지속적으로 개선하세요.
* 👥 GitHub에 이슈를 게시하여 오픈 소스 커뮤니티에 참여하는 것은 **개인의 성장과 사용하는 도구의 개선**에 기여하는 데 모두 중요합니다.
* 📈 잠재적 고용주에게 자신의 기술을 보여주고 소프트웨어 엔지니어링에서 커뮤니티 상호 작용의 중요성을 강조하려면 **프로젝트의 강력한 GitHub 포트폴리오**를 구축하는 것이 필수적입니다.

소프트웨어 엔지니어는 AI, 포럼 및 문서를 효과적으로 사용하여 코딩 문제를 해결함으로써 자신의 기술을 향상시킬 수 있습니다.

All right, my code's done, time to ship it. Delivery free, boxed, you dog water, zero earn, copr in the age of AI. Debugging and getting into software engineering has never been easier. You're the exact six steps you need to take to unblock yourself from any software engineering

area you'll ever get. Number one, tinker. Try to pinpoint your error. You can usually use an AI buddy to help you out here. Pinpoint parts in your code will potentially solve your problem before you even go into an AI and allow you to craft a better question to an AI once

you've pinpointed the issue. You can move to step two, which is ask your AI. You can use ChatGPT, find Bing's AI or if you want to get wrong answers, Google's Bard. There are six principles to prompt engineering so you can get the

best out of your AI. Write clear and specific instructions. Give as much context as possible. Use delimiters to clearly indicate distinct parts of the input and especially look out for something called hallucinations. Hallucinations are when your AI gives you an output that it thinks is right

but is completely wrong. For example, if I write about writing solidity and varying tests and Foundry, OBT disgraces us by saying with the npm install it from Open Zeppelin. These can be tough to spot, but once you try it out, you'll see it

doesn't work. And finally, you want to understand the limitations of the AI you're working with and iterate constantly. Large learning language models are trained on human conversations, so you can interact with them as if you're having a conversation, but it's important to know the

limitations of these AIS as most AIS have a limit on how many tokens or words they keep in context at one time. AI is trained off human language so if you're good at asking other humans questions you'll probably be good at asking robots questions too. Asking questions is a

skill so keep practicing. I've got a link in the description to learn.deeplearning.ai which is a free course to help software engineers be better prompt engineers. Now when the AIS can't help you you'll have to go back to the old standbys, actually doing work

yourself, and one of the first pieces of work is reading the documentation. You probably should have done this already, however we can still use ChatGPT because a strategy that I constantly use is I'll copy paste sections of the documentation added to ChatGPT's context and

say something like, "The above or the docs for Tool X, based on those docs how do I do Y?" So Google might be crying because ChatGPT is eating its lunch, but Google still has what AI doesn't have: the entire internet. Previously anytime I ran

into an issue I prayed someone else had run into it before I made a post on it so I could Google search that exact issue. There's a new tool called Find that combines web search with AI, as it does a web search and it crawls through

all the data of the sites, reads them all and then gives you an answer based off of what it reads. 5. Asking a forum. Sometimes the information just isn't out there and we need to ask human beings. We always want to ask her a question on a

webbed indexed forum like Stack Exchange, Stack Overflow, Hirana, or Reddit. This way web crawlers and more likely AIs can scrape the data from these sites and learn from us. That way the next time we have this question, we can get our answers quickly. Asking on Discord and

Twitter because your knowledge will get lost in the unsearchable conversations that Discord is, and web crawlers don't index them. The super secret alpha is to ask a question on Stack Exchange and then post your Stack Exchange link to Discord. You should 100% always ask your questions and

format them with Markdown. And if you're not sure how to do Markdown, you can ask chat GPT to help you format your questions and Markdown. Ask on the support GitHub or forum if the tool you're working on is open source. Awesome. Ask a well-formatted question on their

Git page. Not open source? Never use that tool again. This is how we improve our tools by creating issues, engaging with each other, and even taking on issues and contributing to the open source code. And then finally, iterate. Do all these same steps over again. And as always, keep

hopping through the code. And until next time, stay riveting, my fellow blockchain ears. Foreign [Music] Fantastic. So now we know a little bit more about the different tools and techniques we can use when we do in fact get stuck. We knew something like chat

GPT, forums, ask questions. We obviously want to Google search, and especially for this course, you definitely want to make sure you use the discussions channel. But I wanted to give you some very specific tips about working with this course. Number one, limit yourself to triaging to 15

or 20 minutes for example if we're trying to compile our priceconverter.sol and maybe we forgot to add the address in here right and we're going to go ahead and compile and we're getting this error this is exactly one argument expected for explicit type conversion now this should be pretty obvious what the error is hey it's expecting one argument and we obviously just removed that argument but maybe you're having a really hard time and you're trying all this different stuff and you can't seem to figure it out

limit yourself to 15 or 20 minutes we do not want to be messing around for so long that we can't figure something out right something like this would be very easy for us to compile copy this error go to chat gbt and say something like

hey I'm getting an error compiling my solidity here is the error the three back ticks paste it in here like that three back ticks is markdown and like we said in the video we want to use delimiters to clearly identify what is code versus what is not oops and I hit

and I hit enter whoops the daisy we're actually going to hit stop generating and we're gonna copy paste the question again now we're going to say here is my solidity it's wrong then we'll copy the solidity line once again three back ticks we'll say

solidity here we'll paste it in three back ticks to end it hit enter and it'll probably give us the answer here and turn off the error message suggests that you need to provide an address when using explicit type conversion so we go ah okay whoops we need to actually add

an address in here it gives us an example and we can go back to our code and you know paste that address in here and be good to go so limit self triage to 15-20 minutes that's gonna be the first step number two don't be afraid to ask AI but don't

skip learning the purpose of this course is for you to be able to learn everything here AI is going to get things wrong and it's only the people who actually have learned what's going on under the hood who are going to be able to call out when AI gets something

wrong when AI gets something wrong it does something called hallucinate and we talked a little bit about that in that AI video we just watched so hallucinations are when the AI makes something up that it thinks is right but it's completely wrong and that's

something we absolutely need to watch out for when we're coding and when we're learning a perfect example of this is with Foundry which is a tool we're about to learn very soon in order for us to install Foundry we actually need to run

this command right get founder.sh it's the first thing that it tells us to do at the time of recording chat gbt has not been trained on The Foundry tool so if I ask it a question like I'd like to initialize a Foundry project with

solidity can you show me how to do that and will that enter so the first sentence out the gate is wrong it's not aware of Foundry so it tried to guess what Foundry was and it got it wrong it is not a rust-paced framework for

building blockchains at all and it is directly related to Solidity so it's really important that we use AI to augment our learning, but we don't skip over the learning with AI. It's really good to bounce questions and good ideas off of it, but in order for us to be even able

to recognize when Foundry is hallucinating and making stuff up, we need to have that knowledge ourselves. Simply, of course, use the forums, web3education.dev will have a place for you to ask questions, but in the meantime, definitely be sure to use The Foundry

full course discussions, use this to your advantage and use this to help other people out who have questions. This is a community project, we're all going to be helping each other out, learning. Number three, Google the exact error. If AI doesn't know, the forums aren't there, Google

the exact error. Maybe somebody else has come across this error previously, asked the question online, and you can just Google it. And then of course, post your question in Stack Exchange or Piranha, and like we said in that AI video, you want to use Markdown and format your

questions as best as possible. So if you're unfamiliar with how to use Markdown, if you're unfamiliar with how to format your questions, again, please be sure to ask ChatGPT or some other AI how to best format your questions. You can even simply say, ask ChatGPT, 'Could

you give me an example of a well-formatted Stack Overflow question with code or code?' And it'll actually give us a really good formatted question. So it gives us the question, it formats the code for us, it's using this with the three backticks that I showed you in

in order to get this code block here, it's using three backticks and it's being very verbose in giving all the details of the question, so be sure to ask questions like "Chat should be T awesome again." Let's say we're in here, compile, we get this error: "exactly one argument expected for explicit type conversion". If we copy this error, go to Google, we paste just the error in, we'll remove this our specific code, looks like the first thing we get at Google is actually contract inheritance, and we can see that this is actually exactly the error that we're

getting. It looks like this does actually explain the answer, but it's a little bit confusing to understand here, so maybe you would copy paste this into ChatGPT's context. Right, there's a lot of creative ways you can approach debugging some of your issues. Obviously when you're out in

the wild and you're actually building things yourself, obviously the education site and the discussions channel of the GitHub repo, the people in these aren't going to have context for the new creative thing that you're working on, so I want you to get used to doing that

methodology that we're teaching here, right? Asking your AI buddy, if your AI buddy doesn't know, Googling, asking questions on Stack Exchange, Ethereum, asking questions on Piranha. Piranha is arguably better because you're asking your questions in a decentralized

location as opposed to a centralized server, but I definitely want you to practice asking questions and go to Piranha, go to Stack Exchange, make accounts for these so that when you do run into an issue, because you will, you'll know where to post them. I'll

encourage you to pause and post a question on one of these forms right now just so that you get used to it just so that you get out of your shell and you're going to suck at asking questions in the beginning and that's okay you

will get better at asking questions as you continue with this course asking good questions is a skill in itself and if you learn how to ask very good questions that's the secret sauce to being amazing at AI prompting as well now the final step in all of this is

actually going to be posting an issue on GitHub Git a surprisingly incredibly important part of being a software engineer and being a developer in web3 is interacting with the community the vast majority of these tools that we're working with are going to be something called open source meaning

that the code associated with them is actually available for anybody to view this tool Foundry that we're going to be working with is one such example and if you see it has this issues tab where there's just a ton of issues people are posting while using this

tool this is how code is able to be improved and move forward all this code is public for people to use and work with and if there's a question that isn't answered you can post here and tell the developers of this tool hey I'm

running into an issue maybe we should make a project improvement I would say be very respectful with the way that you make issues you definitely want to do some searching on issues to make sure that nobody else has asked the question that you're working with and be sure to

use these other resources first like Piranha, like Stack Exchange, but in the future, once you leave this cohort, it's going to be incredibly important for you to participate in the ecosystem by making issues, by making pull requests, and actually improving some of the tools that we're going to be working with.

yourself. Even Ethereum has GitHub, and if there's any improvements that you want to make, you can add an EIP to actually improve Ethereum. We'll talk about EIPs more later. You can improve some of the clients that Ethereum runs, so interacting with each other is

incredibly important, and in fact, if you don't have a GitHub, we are going to get you set up with GitHub right now because GitHub is a platform that most developers use to share code and write code and interact with code in between each other. So if you don't have a GitHub,

let's sign up right now, and we're going to use the GitHub to create our own portfolios of all the different cool coding projects that we've created. This way, when you go to apply for a job, you can say, "Hey, go look at my GitHub. I've

got all these really cool projects." So we're going to make you a GitHub right now. If you don't have an email, you're going to need an email. I made a burner account just for this video, and I'm also going to walk you through formatting a

question. I know I'm belaboring the point here, but I really, really want to stress the importance of formatting these questions really well. So we're actually going to show a clip from my past video, so we're going to be using the full blockchain facility course JS, but again,

for this video just be sure to use the GitHub repo associated with this course and not the other old GitHub repo all right let's watch this video

Lesson 5 Part 1 | AI Prompting & Forums - Solidity & Foundry Full Course

Book a smart contract audit here 👇
https://www.cyfrin.io/


Cyfrin Audits

소프트웨어 엔지니어링의 디버깅 전략

코드를 수정하여 오류를 정확히 찾아내어 AI를 사용하기 전에 잠재적으로 문제를 해결할 수 있습니다.

신속한 엔지니어링 원칙에 따라 AI 도구(ChatGPT, Bing AI, Google Bard)에 문의하여 도움을 요청하세요.

사실인 것처럼 표시되는 잘못된 결과인 AI 환각에 주의하고 AI 제안을 테스트하세요.

AI가 실패하는 경우, 복잡한 부분을 해석하기 위해 AI를 사용하여 문서를 참조하세요.

검색 엔진(Google, Find)을 활용하여 이전에 발생한 문제에 대한 해결책을 찾아보세요.

검색 및 커뮤니티 학습을 위해 인덱싱된 포럼(Stack Exchange, Stack Overflow, Hirana, Reddit)에 질문을 게시하세요.

명확성과 검색 가능성을 위해 마크다운을 사용하여 질문의 형식을 지정하세요.

이슈를 게시하고 코드에 기여하여 오픈 소스 프로젝트에 참여하세요.

외부 도움을 받기 전에 초기 디버깅 시도를 15~20분으로 제한하세요.

AI를 학습을 대체하는 것이 아니라 학습을 보강하는 데 사용하고, AI 환각에 주의하세요.

질문이 있는 경우 코스 포럼 및 web3education.dev를 활용하세요.

Google에서 정확한 오류 메시지를 검색하여 기존 해결책을 찾아보세요.

마크다운을 사용하여 잘 정리된 형식의 질문을 스택 익스체인지 또는 피라냐에 게시하세요.

AI를 사용하여 복잡한 오류 메시지나 문서를 해석하세요.

GitHub와 오픈소스 기여의 중요성

오픈 소스 도구를 개선하고 개발 생태계에 참여하기 위해 GitHub에 이슈를 게시하세요.

새 이슈를 게시하기 전에 기존 이슈를 검색하고 정중하게 소통하세요.

프로젝트를 소개하고 입사 지원용 포트폴리오를 구축할 수 있는 GitHub 프로필을 만드세요.