AI 일타강사가 알려주는 클로드 4 완벽 활용법, 놓치면 후회!

You've come to the right place now.

 Wow, it  looks like Lazy technology has improved a lot this time.

 My appearance when choosing delivery Hamlet This is really this legend, this happens every evening, I turn on the baeming and watch it for a long time, really like a person talking, it's really funny, I think they're bringing in the speech of the Z generation .

 Let's try Python data analysis.

 yes.

 Wow, you've done a great job analyzing the data.

 In this way, we have now visualized the data.

  When you  click on Analysis, detailed statistics for the Lunar New Year  are displayed in [Music].

 The era of Claude 4 has arrived.

 It has been officially announced as the world's best coding AI .

 yes.

 So if you watch the video until the end, I   will explain exactly how I use this Cloud 4 from a practitioner's perspective.

 This time, something called Extool Use came out.

 So both models can use tools like extended thinking and webcolor, and Claude says that they can use them interchangeably with inference tools to improve the response .

 So, what we currently see is that Claude has Opus 4 and Sonnet 4.

 Next, here too, of course, it was there originally, but now, if you press deep thinking mode, it allows you to think more complexly .

 Both models are said to be used in parallel and follow the instructions more accurately.

 Next, we're talking about how memory capabilities have been greatly improved, which should now allow for a bit more reduction in halogenation .

 Okay, so the Claude code is finally officially released.

 So, Claude's code is said to support background tasks and basic integration with VS Code, yes, Rebrain, through gitup actions .

 Now, finally, the API feature.

      Entropic is adding support for four additional features to the API, including code execution, MCP connector, file API, and caching prompting for up to an hour, to enable developers to build more powerful AI agents.

 Of course, you have to pay to  use all the current expanded features.

 So after you pay like this, if you turn on this deep thinking mode, you can now search the web in more detail, and if you  go into various sources like Google Drive, Calendar, Gmail, etc.

 and connect to MCP, you can now get more context and come up with answers.

 This time, the updated version that has been made even more powerful is called the system memory update.

 So, when we asked AI questions before, it was not very good at taking in the previous context, but now we've made it so that it can take in this long-term context better .

 Now,  let’s take a look at how to use it in earnest.

 First of all, the first thing we'll do is  test your long-form writing skills.

  Please write a funny essay using everyday material from Generation Z.

 These days, college students talk like this: putting off homework, talking about delivery food, etc.

 Based on fun content, create something that sounds like it was written by an actual college student and make it fun .

 I met a really funny college student .

 Real Real Daily Talk About My College Student Doesn't It Make Me Laugh? Today, I 'm still putting off my homework, and suddenly I feel like my life is like a mess, so I'm organizing it.

 My appearance when I choose delivery Hamlet This is really this legend, this happens every evening and I watch it for a long time, really like a person talking like this, it's really funny, I think they're bringing in some of that Jet Generation speech.

 yes.

 It seems that the long-form skills have definitely improved and become more fun and funny than the previous models .

 Okay, are you able to find the document content correctly this time? Since you said the token size has increased, I will try to test it.

 So, here's this paper.

 The problem with these models is that because of this phenomenon called the lost in the middle,   they don't know the content in the middle of contracts and other documents that are hundreds or even tens of pages long .

 There is a lot of halogenation in the answer, so  I will test it to see if it really doesn't happen .

 Yes.

 I will upload this paper and then ask questions.

 Review the paper section by section.

 Let me give you a quick rundown of each section, the gist, the content, the summary, and then a quick breakdown of the main points before moving on to the next section .

 Then, I will now do an introduction and cut it out one by one like this and have you do it .

 yes.

 yes.

 So, it seems like the current thesis is being summarized very well .

 I asked him to explain the image that was currently displayed here  .

 yes.

 So in this process for piggy 1, now the current input data is 0 and then feedback 1 and then go to refine 2 and so on  until the current 1 and 2 are satisfied the stopping condition.

 Oh, it's coming out really well .

 Okay, so this time , I'm going to move on to something a little bit more mid-level .

 Yes, so right now, I'm looking into it with JavaScript .

 It's definitely different than when I was looking for content before .

 Before, I just simply searched for RH , but now, I use JavaScript to search for it , and the search function  seems to have improved greatly.

 So now you've come to the right place .

 It seems like YG technology has improved a lot this time.

 Well then, this time I will try to review the contract .

 yes.

 We usually send out a lot of contracts in Word, right ? So I have this contract in Word uploaded now.

 Usually, a contract has a table of contents like this, and then the standard contract contents are listed like this.

 Let me put a previously saved prompt here .

 Now, let's review each clause of the uploaded contract sequentially .

 From the first clause like this.

 Okay, let's analyze it.

 Okay, so  let's put in a prompt like this and review the contract.

 Now let's take a look at how you can bring negotiation points into each chapter of the contract .

 Okay, then let's look at the next match .

 In particular, we must look at it from B's perspective.

 Let me take a look at number 2 now.

 2- 1.

 As to 2- 1, the borrower absolutely transfers all profits to the trustee .

 yes.

 So, I think they are doing a good job of summarizing the key points of this very long content .

 So, the current problems are serious problems, and there is now excessive profit.

 So, I am  talking about excessive profit .

 yes.

 So the problem here is that these problems are occurring now.

 So, since it is impossible to secure operating funds by unconditionally transferring all profits, it is   said that a wide range of collateral is currently being secured here.

 So how can we politely request a correction to our wording ?  I'm going to  leave it as a note in Word.

 So, for 2-2 2-1, I'm making it so that I can leave notes on the clauses about the conditions that I'm currently unsatisfied with  .

 The current provision is unconditionally transferable, but this is not possible.

 Well, we can   see that the article is written very well about how to write correction comments like this .

 Request for correction comment.

 yes.

 Now it asks me to make a correction comment like this  .

 If you edit it like this and leave a comment like this, it should be fine.

  I think it might be a good idea to try something like this now.

 Okay, so next, let's try the next thing, and this time, let's try data analysis .

 Here, I'm   going to use Extend ThinkKing to analyze data and create a dashboard.

 So what kind of data is this data? Here are the data for Shoe Otani's current score .

 yes.

 So, I'll input the Otani Showay data and try this prompt .

 Okay, so let me try typing a prompt like this.

  Please analyze the uploaded baseball player data from the perspective of this player's coach.

 Okay, first of all, let's check the total number of player names for the season, evaluate the pitcher's performance, and then analyze the average speed and frequency of use for each major pitch, and then analyze the spin rate of the three most effective pitches .

 Okay, so now we have the pitch type here, and then we have the ball pitching type here, and then we have the current speed, and then we have the angle of the current pose, and things like that.

 So, I will input the following: Please       analyze this player's current weaknesses and provide some practical advice on how to improve them, such as identifying low-performing pitches or inconsistent situations, and then discussing specific training restrictions and game management strategies.

 I have Claude Opus 4 and currently have Extended ThinkKing Deep Thinking Mode turned on.

 Okay, so when I ask you to do data analysis, you're currently doing data analysis with JavaScript, and you're currently splitting the CSV I uploaded and checking the data structure, and it  seems to be doing a very good job of doing what I asked you to do.

  What's really interesting here is that we intentionally insert this type of data.

 Currently, we only have data on pitcher Sho Ohtani .

 There is no data as a typist, but if you ask GPT or LM about the data as a typist, they will  analyze it .

 You can see that it is well found from the current data values, so there is no exact answer .

 So now we have completed the data analysis.

 I think it took about 2 minutes.

 So, if you look at the player analysis report like this, for the  2023 season, Sho Ohtani threw a total of 2,094 pitches, and the average speed and frequency of use by main pitch type, the sweeper usage frequency is currently 35.

2%, the average speed and spin, the passed ball 33%, and the cup cutter, which is currently used 329 times .

 So right now the top 3 most effective pitches are curveball, sweeper, and 4-shin festival ball.

 Currently, the most effective pitch is the splitter curve.

 If we look at the passball, which is currently being used most effectively and then identify weaknesses and areas for improvement, the pitch that is  currently  lacking in inventory is the passball.

 Looking at the current data values, it seems to be low and somewhat inconsistent.

  The splitter's inventory is particularly low, which tends to separate ball counters, and the curveball is effective but has a low strike rate.

 If we do impose a rescue training limit, we are currently making suggestions on how we can increase it further  .

 Wow, it seems like I'm doing the same amount of work as a data analyst at any decent company .

 So, in the mountains, we can also use AI to analyze sales data in a similar way, or to set KPIs within the company and use  that as a basis for analyzing data to improve things  .

 Wow, Claude 4 seems to be really performing well right now .

 yes.

 Well, now they say we've reached our current usage limit.

  If you look at the current usage, it doesn't seem like it's that huge.

 After using it about 10 to 20 times, it seems to be getting a bit clogged now.

 I guess it's because the token usage cost is so high? So if you click on the Max subscription, it's a bit expensive.

 It 's $100 a month, but in reality, we've now spent $100, so it's a little bit insufficient for our analysis.

 However, at first, I could only use GPT3 about 20 to 30 times, but now, as time goes by, usually after about 3 weeks to a month, it starts to solve a lot of problems.

 GPT O3 was initially only usable 20 to 30 times a week, but now it can be used up to 100 times a week .

 So,  I think this too might open up a bit more as we use it.

 Okay, so now let's test how good he is at coding when he grows up with AI .

     I think there might be some people who don't know what a cursor is, so you can think of this cursor as an AI programming that automatically writes code when you literally just type in a command.

 Okay, so let me turn on the cursor and give it a try.

 Okay, so if you turn on the cursor like this .

 First, click on Open Project here .

 Here on my desktop I have an empty folder like this.

 Okay, let's click on the empty folder here and check.

 Then just click on the web here.

 So if you press open, you'll see a blank screen like this right now.

 yes.

 Then, on the right side, enter the prompt.

 I am currently using Cloud 4 Sonnet.

 Since it's a bit expensive here, you can see that they've measured it so that it's not a monthly subscription, but the cost is incurred every time you input it  .

 Okay, so let me try making a website request here .

 yes.

 I typed the prompt like this: Create a modern, fully functional eCommerce store website using HTML, CSS, and JavaScript .

 They asked me to create a shopping mall with a responsive design, a logo, a navigation bar, etc.

 I did n't enter a single line of code.

  If you just input the prompt like this, it will automatically write the code for you.

 Okay, so I created an index called ITML.

 If you look here, I can check it here .

 If you click on the index HTML, you will see this : yes.

 I think the website is really well made right now.

 Right now, it's been created like this, but if I  give it a command to make it more trendy, it can of course do ticket checks with me and quickly create websites like this .

 But  rather than saying that I will do all the development right now just because this is possible, I think it would be good if developers, planners, or current designers, at the POC stage, do interaction design, or say, “How about going in this design direction?” or use it to verify their ideas .

  Okay, now I'm going to try Python data analysis, which is the most used by office workers .

 Let's analyze the data of Sho Otani that I analyzed earlier using Python here again .

 Now, press data and open.

 Okay, so now you  can see that the data is in Excel.

 Now, at this time, you must install Python .

 Yes, there's Python, right? Python.

 You can try downloading this 3.

13 yogurt.

 I'm currently using Mac, and those who use Windows can just use it the same way.

 yes.

 So I'm going to install Python and enter some data like this.

  Please analyze the performance of the two and add it to a new sheet .

 I'll ask you to input the average restraint spin rate analysis frequency by New Year and do an analysis .

 I just read about Shoi Otani.

 So, right now, we are  checking the data currently in this folder and analyzing it .

 As I told you to install Python earlier, once you install Python, you can run Python scripts and actually run data analysis here.

 This will allow for  much more accurate data analysis than using GPT or other cloud web services directly.

 yes.

 So, when doing data analysis, we just used the most basic Pandas .

 Then, when it comes out like this, just click on the run here.

 If it turns orange, then yes.

 If you look at it, we have   currently completed data visualization by analyzing the data through feature analysis.

 If you look at it now, the fastball seems to be the fastest in terms of speed and velocity.

 Next, for spin ray, slider is the best right now, and for the  frequency by pitch type that is used the most right now, sweeper is the best right now.

 Next, when you look at the current speed, constraint, and current spin ray, you can see where the distribution is now, and you can even see a scatter plot of the distribution .

 Okay, so now  let's take a look at the Excel file that has just been created.

 Yes, the data analysis has been done very well.

 The analysis was completed by calculating the average  number of uses, average spin, spin rate, average 5th spin, standard deviation, and then the frequency of use .

 Then this time I will request something like this .

 Let me make an additional request .

 Create an interactive dashboard for coaches and analysts .

  Now that we've analyzed the data with HTML, Javascript, CSS, and Javascript, we're   creating an HTMLCS Javascript so that anyone can come in and see it .

 yes.

 yes.

 Then, if it comes out like this, there is a dashboard, and it seems that the effectiveness is definitely huge when you turn it on the cursor .

 If you look at the current selection, there are pitchers such as a fastball, sweeper, cutter, splitter, sinker, and slider, all of which are made separately for each type of pitch .

 Next, we visualized the average speed by city in this way .

 Wow, now interactive data has also been created right now.

 We've created an interactive dashboard that lets you see the current usage frequency by term by  hovering your mouse over it, and it also   provides a summary of the major indicators.

  If you click on Analysis, detailed statistics will appear.

 Now, let's talk about what happens to the next insight, the analysis insight .

 So now what should I do? Do you have any data here? All you have to do now is update the data .

 If you update and return the cursor, you will now have a websiboard like this that you can come in and view  .

 Of course, this requires hosting right away, but    I think it's really useful now that I'm using it as a term for testing alone locally or sharing it with team members without hosting.

 yes.

 Okay, so we've taken a look at Claude 4's coding skills so far .

 If you found this video helpful, please give it a thumbs up.

 Like, please subscribe.

 Then, I will cover more detailed and specific usage next time .

 yes.

 thank you [music]